New Crowdin translations by GitHub Action

i18n/ar/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/ar/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/ar/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## الخدمات الداعمة لأوبن‌بي‌جي‌بي
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. فمثلًا: باستطاعة مستخدم بريد بروتون إرسال رسالة معمَّاة بين الأطراف، وكون المستقبل مستخدم Mailbox.org، أو لك استقبال إشعارات معمَّاةً بأوبن‌بي‌جي‌بي من خدمات الإنترنت الداعمة له.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. فمثلًا: باستطاعة مستخدم بريد بروتون إرسال رسالة معمَّاة بين الأطراف، وكون المستقبل مستخدم Mailbox.org، أو لك استقبال إشعارات معمَّاةً بأوبن‌بي‌جي‌بي من خدمات الإنترنت الداعمة له.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mail supports TOTP [two-factor authentication](https://proton.me/support/
 
 #### :material-check:{ .pg-green } تعمية البريد الإلكتروني
 
-عند بريد بروتون [دعم مدمج لتعمية أوبن‌بي‌جي‌بي](https://proton.me/support/how-to-use-pgp) في صفحة البريد. تعمَّى الرسائل المرسلة لحسابات بريد بروتون الأخرى تلقائيًّا، ولك تمكين تعمية أوبن‌بي‌جي‌بي لعناوين البريد خارج بروتون في إعدادات حسابك. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+عند بريد بروتون [دعم مدمج لتعمية أوبن‌بي‌جي‌بي](https://proton.me/support/how-to-use-pgp) في صفحة البريد. تعمَّى الرسائل المرسلة لحسابات بريد بروتون الأخرى تلقائيًّا، ولك تمكين تعمية أوبن‌بي‌جي‌بي لعناوين البريد خارج بروتون في إعدادات حسابك. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. ويتيح هذا لمن ليس عنده بريد بروتون العثور على مفاتيح أوبن‌بي‌جي‌بي لحسابات بريد بروتون بسهولة، وذلك لتمكين التعمية بين الأطراف بين موفِّري خدمة البريد الإلكترونيِّ. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. فائدة هذا تظهر في حال كان المستقبل ليس لديه أوبن‌بي‌جي‌بي ولا يستطيع كشف تعمية نسخة من الرسالة في صندوق بريده.
 
-تدعم Mailbox.org اكتشاف المفتايح العامَّة باستخدام HTTP من [دليل مفاتيح الوِب (WKD)](https://wiki.gnupg.org/WKD) التابع لهم. ويتيح هذا لمن ليس عنده Mailbox.org العثور على مفاتيح أوبن‌بي‌جي‌بي لحسابات Mailbox.org بسهولة، وذلك لتمكين التعمية بين الأطراف بين موفِّري خدمة البريد الإلكترونيِّ. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. ويتيح هذا لمن ليس عنده Mailbox.org العثور على مفاتيح أوبن‌بي‌جي‌بي لحسابات Mailbox.org بسهولة، وذلك لتمكين التعمية بين الأطراف بين موفِّري خدمة البريد الإلكترونيِّ. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } إنهاء الحسابات
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/ar/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/bn-IN/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/bn-IN/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/bn-IN/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/bn-IN/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/bn/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/bn/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/bn/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/bn/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/cs/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/cs/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/cs/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/cs/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/de/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Überwachungskapitalismus](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Mindestanforderungen
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Im besten Fall
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/de/basics/email-security.md

@@ -19,7 +19,7 @@ Selbst wenn du OpenPGP verwendest, unterstützt es keine [Forward Secrecy](https
 
 ## Was ist der Web Key Directory Standard?
 
-Der Web Key Directory (WKD) Standard ermöglicht es E-Mail-Clients, den OpenPGP-Schlüssel für andere Postfächer zu ermitteln, auch wenn diese bei einem anderen Anbieter gehostet werden. E-Mail-Clients, die WKD unterstützen, fragen den Server des Empfängers nach einem Schlüssel, der auf dem Domainnamen der E-Mail-Adresse basiert. Wenn du beispielsweise eine E-Mail an `jonah@privacyguides.org` sendest, fragt dein E-Mail-Programm bei `privacyguides.org` nach dem OpenPGP-Schlüssel von Jonah, und falls `privacyguides.org` einen Schlüssel für dieses Konto hat, wird deine Nachricht automatisch verschlüsselt.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. E-Mail-Clients, die WKD unterstützen, fragen den Server des Empfängers nach einem Schlüssel, der auf dem Domainnamen der E-Mail-Adresse basiert. Wenn du beispielsweise eine E-Mail an `jonah@privacyguides.org` sendest, fragt dein E-Mail-Programm bei `privacyguides.org` nach dem OpenPGP-Schlüssel von Jonah, und falls `privacyguides.org` einen Schlüssel für dieses Konto hat, wird deine Nachricht automatisch verschlüsselt.
 
 Neben den [von uns empfohlenen E-Mail-Clients](../email-clients.md), die WKD unterstützen, gibt es auch einige Webmail-Anbieter, die WKD unterstützen. Ob *dein eigener* Schlüssel für andere bei WKD veröffentlicht wird, hängt von deiner Domainkonfiguration ab. Verwendest du einen [E-Mail-Anbieter](../email.md#openpgp-compatible-services), der WKD unterstützt, wie z. B. Proton Mail oder Mailbox.org, können diese deinen OpenPGP-Schlüssel auf ihrer Domain für dich veröffentlichen.
 

i18n/de/email.md

@@ -34,7 +34,7 @@ Zusätzlich zu (oder anstelle von) einem hier empfohlenen E-Mail-Anbieter kannst
 
 ## OpenPGP-kompatible Dienste
 
-Diese Anbieter unterstützen von Haus aus die OpenPGP-Ver- und Entschlüsselung sowie den [Web Key Directory Standard](basics/email-security.md#what-is-the-web-key-directory-standard), so dass anbieterunabhängige E2E-verschlüsselte E-Mails möglich sind. Zum Beispiel können Kunden von Proton Mail eine E2EE-Nachricht an Kunden von Mailbox.org senden oder sie können OpenPGP-verschlüsselte Benachrichtigungen von Internetdiensten erhalten, die dies unterstützen.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Zum Beispiel können Kunden von Proton Mail eine E2EE-Nachricht an Kunden von Mailbox.org senden oder sie können OpenPGP-verschlüsselte Benachrichtigungen von Internetdiensten erhalten, die dies unterstützen.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Bestimmte Informationen, die in [Proton Contacts](https://proton.me/support/prot
 
 #### :material-check:{ .pg-green } E-Mail-Verschlüsselung
 
-Proton Mail hat [die OpenPGP-Verschlüsselung](https://proton.me/support/how-to-use-pgp) in sein Webmail integriert. E-Mails an andere Proton Mail-Konten werden automatisch verschlüsselt. Die Verschlüsselung an Nicht-Proton Mail-Adressen mit einem OpenPGP-Schlüssel kannst du ganz einfach in deinen Kontoeinstellungen aktivieren. Proton unterstützt auch die automatische Erkennung externer Schlüssel mit dem [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Das bedeutet, dass E-Mails an andere Anbieter, die WKD verwenden, automatisch auch mit OpenPGP verschlüsselt werden, ohne dass du manuell öffentliche PGP-Schlüssel mit deinen Kontakten austauschen musst. Außerdem ist es möglich, [Nachrichten an Nicht-Proton-Mail-Adressen ohne OpenPGP zu verschlüsseln](https://proton.me/support/password-protected-emails), ohne dass die Empfänger ein Proton-Mail-Konto benötigen.
+Proton Mail hat [die OpenPGP-Verschlüsselung](https://proton.me/support/how-to-use-pgp) in sein Webmail integriert. E-Mails an andere Proton Mail-Konten werden automatisch verschlüsselt. Die Verschlüsselung an Nicht-Proton Mail-Adressen mit einem OpenPGP-Schlüssel kannst du ganz einfach in deinen Kontoeinstellungen aktivieren. Proton also supports automatic external key discovery with WKD. Das bedeutet, dass E-Mails an andere Anbieter, die WKD verwenden, automatisch auch mit OpenPGP verschlüsselt werden, ohne dass du manuell öffentliche PGP-Schlüssel mit deinen Kontakten austauschen musst. Außerdem ist es möglich, [Nachrichten an Nicht-Proton-Mail-Adressen ohne OpenPGP zu verschlüsseln](https://proton.me/support/password-protected-emails), ohne dass die Empfänger ein Proton-Mail-Konto benötigen.
 
 Auch veröffentlicht Proton Mail öffentlichen Schlüssel der Proton-Konten über HTTP von ihrem WKD. Dies ermöglicht es Personen, die Proton Mail nicht verwenden, die OpenPGP-Schlüssel von Proton Mail-Konten für anbieterübergreifende E2EE leicht zu finden. Dies gilt nur für E-Mail-Adressen, die auf eine der Proton-eigenen Domains enden, wie @proton.me. Um eine eigene Domain zu verwenden, musst du [WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separat konfigurieren.
 
@@ -164,7 +164,7 @@ Mailbox.org ermöglicht die Verschlüsselung von eingehenden E-Mails mit der [ve
 
 Mailbox.org hat die [Verschlüsselung in sein Webmail integriert](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard), was den Versand von Nachrichten an Personen mit öffentlichen OpenPGP-Schlüsseln vereinfacht. Sie erlauben es auch entfernten Empfängern [eine E-Mail](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) auf den Servern von Mailbox.org zu entschlüsseln. Diese Funktion ist nützlich, wenn der Empfänger OpenPGP nicht nutzt und daher eine Kopie der E-Mail in seinem eigenen Postfach nicht entschlüsseln kann.
 
-Mailbox.org unterstützt auch die Suche nach öffentlichen Schlüsseln über HTTP von ihrem [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Dies ermöglicht es Personen, die Mailbox.org nicht verwenden, die OpenPGP-Schlüssel von Mailbox.org-Konten für anbieterübergreifende E2EE leicht zu finden. Dies gilt nur für E-Mail-Adressen, die auf eine der Mailbox.org-eigenen Domains enden, wie @mailbox.org. Um eine eigene Domain zu verwenden, musst du [WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separat konfigurieren.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Dies ermöglicht es Personen, die Mailbox.org nicht verwenden, die OpenPGP-Schlüssel von Mailbox.org-Konten für anbieterübergreifende E2EE leicht zu finden. Dies gilt nur für E-Mail-Adressen, die auf eine der Mailbox.org-eigenen Domains enden, wie @mailbox.org. Um eine eigene Domain zu verwenden, musst du [WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separat konfigurieren.
 
 #### :material-information-outline:{ .pg-blue } Kontokündigung
 
@@ -323,7 +323,7 @@ Wir halten diese Merkmale für wichtig, um einen sicheren und optimalen Service
 
 - Verschlüsselt alle Kontodaten (Kontakte, Kalender usw.) im Ruhezustand mit Zero-Access-Verschlüsselung.
 - Integrierte Webmail E2EE/PGP-Verschlüsselung als Komfortfunktion.
-- Unterstützung für [WKD](https://wiki.gnupg.org/WKD), um die Suche nach öffentlichen OpenPGP-Schlüsseln über HTTP zu verbessern. GnuPG-Benutzer können einen Schlüssel erhalten, indem sie Folgendes eingeben: `gpg --locate-key beispiel_nutzer@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG-Benutzer können einen Schlüssel erhalten, indem sie Folgendes eingeben: `gpg --locate-key beispiel_nutzer@example.com`
 - Unterstützung für eine temporäre Mailbox für externe Benutzer. Dies ist nützlich, wenn du eine verschlüsselte E-Mail versenden möchtest, ohne eine Kopie an den Empfänger zu senden. Diese E-Mails haben in der Regel eine begrenzte Lebensdauer und werden dann automatisch gelöscht. Sie erfordern auch nicht, dass der Empfänger eine Kryptographie wie OpenPGP konfiguriert.
 - Verfügbarkeit der Dienste des E-Mail-Anbieters über einen [onion service](https://de.wikipedia.org/wiki/.onion).
 - Unterstützung [von Unteradressen](https://en.wikipedia.org/wiki/Email_address#Sub-addressing).

i18n/de/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/el/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Πρέπει να είναι ανοικτού κώδικα.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/el/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/el/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/el/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/eo/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/eo/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/eo/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/eo/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/es/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Capitalismo de Vigilancia](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censura](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Desde el lanzamiento de ChatGPT en 2022, las interacciones con grandes modelos lingüísticos (LLM) son cada vez más frecuentes. Los LLM pueden ayudarnos a escribir mejor, a comprender temas desconocidos o a responder a una amplia gama de preguntas. Pueden predecir estadísticamente la siguiente palabra basándose en una gran cantidad de datos extraídos de Internet.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. Los LLM pueden ayudarnos a escribir mejor, a comprender temas desconocidos o a responder a una amplia gama de preguntas. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Preocupaciones por la Privacidad de los LLM
 
@@ -42,7 +42,7 @@ Para ejecutar IA localmente, se necesita tanto un modelo de IA como un cliente d
 
 ### Elegir un Modelo
 
-Hay muchos modelos con licencia permisiva disponibles para descargar. [Hugging Face](https://huggingface.co/models) es una plataforma que te permite navegar, investigar y descargar modelos en formatos comunes como [GGUF](https://huggingface.co/docs/hub/en/gguf). Entre las empresas que ofrecen buenos modelos open-weights figuran grandes nombres como Mistral, Meta, Microsoft y Google. Sin embargo, también hay muchos modelos comunitarios y «fine-tunes» disponibles. Como ya se ha mencionado, los modelos cuantificados ofrecen el mejor equilibrio entre calidad de modelo y rendimiento para quienes utilizan hardware de consumo personal.
+Hay muchos modelos con licencia permisiva disponibles para descargar. [Hugging Face](https://huggingface.co/models) es una plataforma que te permite navegar, investigar y descargar modelos en formatos comunes como [GGUF](https://huggingface.co/docs/hub/en/gguf). Entre las empresas que ofrecen buenos modelos open-weights figuran grandes nombres como Mistral, Meta, Microsoft y Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. Como ya se ha mencionado, los modelos cuantificados ofrecen el mejor equilibrio entre calidad de modelo y rendimiento para quienes utilizan hardware de consumo personal.
 
 Para ayudarte a elegir un modelo que se adapte a tus necesidades, puedes consultar tablas de clasificación y puntos de referencia. La tabla de clasificación más utilizada es la de la comunidad [LM Arena](https://lmarena.ai). Además, la [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) se centra en el rendimiento de los modelos open-weights en puntos de referencia comunes como [MMLU-Pro](https://arxiv.org/abs/2406.01574).  También existen puntos de referencia especializados que miden factores como [inteligencia emocional](https://eqbench.com), [«inteligencia general no censurada»](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), y [muchos otros](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ Para ayudarte a elegir un modelo que se adapte a tus necesidades, puedes consult
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp es un cliente de IA que se ejecuta localmente en tu ordenador Windows, Mac o Linux. Es una opción excelente si lo que buscas es una gran personalización y ajuste, por ejemplo para juegos de rol.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. Es una opción excelente si lo que buscas es una gran personalización y ajuste, por ejemplo para juegos de rol.
 
 Además de soportar una amplia gama de modelos de texto, Kobold.cpp también soporta generadores de imágenes como [Stable Diffusion](https://stability.ai/stable-image) y herramientas de reconocimiento automático de voz como [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ Además de soportar una amplia gama de modelos de texto, Kobold.cpp también sop
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Problemas de Compatibilidad</p>
 
 Kobold.cpp podría no funcionar en ordenadores sin soporte AVX/AVX2.
@@ -98,7 +98,7 @@ Kobold.cpp te permite modificar parámetros como la temperatura del modelo de IA
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama es un asistente de IA de línea de comandos que está disponible en macOS, Linux y Windows. Ollama es una gran elección si buscas un cliente de IA fácil de usar, ampliamente compatible y rápido gracias a su uso de la inferencia y otras técnicas. Tampoco requiere ninguna configuración manual.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama es una gran elección si buscas un cliente de IA fácil de usar, ampliamente compatible y rápido gracias a su uso de la inferencia y otras técnicas. Tampoco requiere ninguna configuración manual.
 
 Además de soportar una amplia gama de modelos de texto, Ollama también soporta modelos [LLaVA](https://github.com/haotian-liu/LLaVA) y tiene soporte experimental para las [capacidades vision de Llama] (https://huggingface.co/blog/llama32#what-is-llama-32-vision) de Meta.
 
@@ -124,9 +124,9 @@ Ollama simplifica el proceso de creación de un chat de IA local descargando aut
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile es un ejecutable ligero de un solo archivo que permite a los usuarios ejecutar LLM localmente en sus propios ordenadores sin necesidad de configuración. Está [respaldado por Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) y disponible en Linux, macOS y Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. Está [respaldado por Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) y disponible en Linux, macOS y Windows.
 
 Llamafile también es compatible con LLaVA. Sin embargo, no admite el reconocimiento de voz ni la generación de imágenes.
 
@@ -138,7 +138,9 @@ Llamafile también es compatible con LLaVA. Sin embargo, no admite el reconocimi
 <details class="downloads" markdown>
 <summary>Downloads "Descargas"</summary>
 
-- [:fontawesome-solid-desktop: Escritorio](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -175,7 +177,7 @@ Por favor, ten en cuenta que no estamos afiliados a ninguno de los proyectos que
 - No debe transmitir datos personales, incluidos los del chat.
 - Debe ser multiplataforma.
 - No debe requerir GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - No debe requerir conexión a Internet.
 
 ### Mejor Caso
@@ -186,4 +188,11 @@ Nuestros criterios para el mejor de los casos representan lo que nos _gustaría_
 - Debería tener una opción de descarga de modelos integrada.
 - El usuario debería poder modificar los parámetros de la LLM, como su prompt de sistema o su temperatura.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: La checksum de un archivo es un tipo de huella digital antimanipulación. Un desarrollador suele proporcionar una checksum en un archivo de texto que puede descargarse por separado, o en la propia página de descarga. Verificar que la checksum del archivo descargado coincide con la proporcionada por el desarrollador ayuda a garantizar que el archivo es auténtico y no ha sido manipulado durante el transporte. Puedes utilizar comandos como `sha256sum` en Linux y macOS, o `certutil -hashfile file SHA256` en Windows para generar la checksum del archivo descargado.

i18n/es/basics/email-security.md

@@ -19,7 +19,7 @@ Incluso si utilizas OpenPGP, no admite el [secreto perfecto hacia adelante](http
 
 ## ¿Qué es el estándar del Directorio de Claves Web?
 
-El estándar de Directorio de Claves Web (WKD, por sus siglas en inglés) permite a los clientes de correo electrónico descubrir la clave OpenPGP para otros buzones, incluyendo los que se encuentran hospedados en un proveedor diferente. Los clientes de correo electrónico compatibles con WKD le solicitarán al servidor receptor una clave basada en el nombre de dominio de la dirección de correo electrónico. Por ejemplo, si envías un correo electrónico a `jonah@privacyguides.org`, tu cliente de correo electrónico le solicitará a `privacyguides.org` la clave OpenPGP de Jonah, y si `privacyguides.org` tiene la clave para esa cuenta, tu mensaje se encriptará automáticamente.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Los clientes de correo electrónico compatibles con WKD le solicitarán al servidor receptor una clave basada en el nombre de dominio de la dirección de correo electrónico. Por ejemplo, si envías un correo electrónico a `jonah@privacyguides.org`, tu cliente de correo electrónico le solicitará a `privacyguides.org` la clave OpenPGP de Jonah, y si `privacyguides.org` tiene la clave para esa cuenta, tu mensaje se encriptará automáticamente.
 
 Además de los [clientes de correo electrónico que recomendamos](../email-clients.md) y son compatibles con WKD, algunos clientes web de correo electrónico también son compatibles con WKD. Si *tu propia* clave es publicada en WKD para que otros la utilicen, esto dependerá de tu configuración de dominio. Si utilizas un [proveedor de correo electrónico](../email.md#openpgp-compatible-services) compatible con WKD, como Proton Mail o Mailbox.org, ellos pueden publicar tu clave OpenPGP en su dominio por usted.
 

i18n/es/email.md

@@ -34,7 +34,7 @@ Además de (o en lugar de) un proveedor de correo electrónico recomendado aquí
 
 ## Servicios Compatibles con OpenPGP
 
-Estos proveedores soportan de manera nativa el cifrado/descifrado de OpenPGP y el [estándar de Directorio de Clave Web](basics/email-security.md#what-is-the-web-key-directory-standard), permitiendo que los correos electrónicos cifrados de extremo a extremo sean independientes del proveedor. Por ejemplo, un usuario de Proton Mail podría enviar un mensaje E2EE a un usuario de Mailbox.org, o podrías recibir notificaciones encriptadas con OpenPGP desde servicios de Internet que lo soporten.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Por ejemplo, un usuario de Proton Mail podría enviar un mensaje E2EE a un usuario de Mailbox.org, o podrías recibir notificaciones encriptadas con OpenPGP desde servicios de Internet que lo soporten.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Cierta información almacenada en [Proton Contacts](https://proton.me/support/pr
 
 #### :material-check:{ .pg-green } Encriptación de Correo Electrónico
 
-Proton Mail ha [integrado la encriptación OpenPGP](https://proton.me/support/how-to-use-pgp) en su webmail. Los correos electrónicos a otras cuentas de Proton Mail se encriptan automáticamente, y la encriptación a direcciones que no sean de Proton Mail con una clave OpenPGP pueden ser habilitados fácilmente en la configuración de tu cuenta. Proton también soporta el descubrimiento automático de claves externas con [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Esto significa que los correos electrónicos enviados a otros proveedores que utilicen WKD también se cifrarán automáticamente con OpenPGP, sin necesidad de intercambiar manualmente claves PGP públicas con tus contactos. Estas también te permiten [cifrar los mensajes enviados a cuentas no pertenecientes a Proton](https://proton.me/support/password-protected-emails), sin la necesidad de que el receptor utilice OpenPGP o se registre en Proton Mail.
+Proton Mail ha [integrado la encriptación OpenPGP](https://proton.me/support/how-to-use-pgp) en su webmail. Los correos electrónicos a otras cuentas de Proton Mail se encriptan automáticamente, y la encriptación a direcciones que no sean de Proton Mail con una clave OpenPGP pueden ser habilitados fácilmente en la configuración de tu cuenta. Proton also supports automatic external key discovery with WKD. Esto significa que los correos electrónicos enviados a otros proveedores que utilicen WKD también se cifrarán automáticamente con OpenPGP, sin necesidad de intercambiar manualmente claves PGP públicas con tus contactos. Estas también te permiten [cifrar los mensajes enviados a cuentas no pertenecientes a Proton](https://proton.me/support/password-protected-emails), sin la necesidad de que el receptor utilice OpenPGP o se registre en Proton Mail.
 
 Proton Mail también publica las direcciones públicas de las cuentas a través de HTTP desde su WKD. Esto permite las personas quienes no utilizan Proton Mail a encontrar fácilmente las claves OpenPGP de las cuentas de Proton Mail, para E2EE entre proveedores. Esto solo aplica para las direcciones de correo electrónico que terminen en uno de los dominios de Proton, como @proton.me. Si utilizas un dominio personalizado, debes [configurar WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) por separado.
 
@@ -164,7 +164,7 @@ Sin embargo, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), la pla
 
 Mailbox.org tiene [encriptación integrada](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) en su correo web, lo que simplifica el envío de mensajes a personas con claves públicas OpenPGP. También permiten [a destinatarios remotos descifrar un correo electrónico](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) en los servidores de Mailbox.org. Esta característica es útil cuando el destinatario remoto no tiene OpenPGP y no puede descifrar una copia del correo electrónico en su propio buzón de correo.
 
-Mailbox.org también admite el descubrimiento de claves públicas a través de HTTP desde su [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Esto permite que personas afuera de Mailbox.org encuentren fácilmente las claves OpenPGP de las cuentas de Mailbox.org, para E2EE entre proveedores. Esto solo aplica para las direcciones de correo electrónico que terminan en un dominio de Mailbox.org, como @mailbox.org. Si utilizas un dominio personalizado, debes [configurar WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) por separado.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Esto permite que personas afuera de Mailbox.org encuentren fácilmente las claves OpenPGP de las cuentas de Mailbox.org, para E2EE entre proveedores. Esto solo aplica para las direcciones de correo electrónico que terminan en un dominio de Mailbox.org, como @mailbox.org. Si utilizas un dominio personalizado, debes [configurar WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) por separado.
 
 #### :material-information-outline:{ .pg-blue } Cancelación de Cuenta
 
@@ -323,7 +323,7 @@ Consideramos que estas características son importantes para ofrecer un servicio
 
 - Cifra todos los datos de la cuenta (contactos, calendarios, etc.) en reposo con cifrado de acceso cero.
 - Cifrado integrado de correo web E2EE/PGP proporcionado como una conveniencia.
-- Compatibilidad con [WKD](https://wiki.gnupg.org/WKD) para permitir un mejor descubrimiento de claves OpenPGP públicas a través de HTTP. Los usuarios de GnuPG pueden obtener una clave escribiendo: `gpg --locate-key usuario_ejemplo@ejemplo.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. Los usuarios de GnuPG pueden obtener una clave escribiendo: `gpg --locate-key usuario_ejemplo@ejemplo.com`
 - Soporte para un buzón temporal para usuarios externos. Esto es útil cuando quieres enviar un correo electrónico encriptado, sin enviar una copia real a tu destinatario. Estos correos electrónicos suelen tener una vida útil limitada y luego se eliminan automáticamente. Tampoco requieren que el destinatario configure ninguna criptografía como OpenPGP.
 - Disponibilidad de los servicios del proveedor de correo electrónico a través de un [ servicio onion](https://en.wikipedia.org/wiki/.onion).
 - Soporte de [subdireccionamiento](https://en.wikipedia.org/wiki/Email_address#Sub-addressing).

i18n/es/tools.md

@@ -357,7 +357,7 @@ Si busca más **seguridad**, asegúrese siempre de conectarse a sitios web que u
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/fa/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/fa/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/fa/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## سرویس‌های سازگار با OpenPGP
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. به عنوان مثال، یک کاربر Proton Mail می تواند یک پیام E2EE را به یک کاربر Mailbox.org ارسال کند، یا می توانید اعلان های رمزگذاری شده با OpenPGP را از سرویس های اینترنتی که از آن پشتیبانی می کنند دریافت کنید.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. به عنوان مثال، یک کاربر Proton Mail می تواند یک پیام E2EE را به یک کاربر Mailbox.org ارسال کند، یا می توانید اعلان های رمزگذاری شده با OpenPGP را از سرویس های اینترنتی که از آن پشتیبانی می کنند دریافت کنید.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mail دارای [رمزگذاری بدون دسترسی](https://proton.m
 
 #### :material-check:{ .pg-green } رمزگذاری ایمیل
 
-Proton Mail دارای [رمزگذاری OpenPGP یکپارچه](https://proton.me/support/how-to-use-pgp) در ایمیل خود است. ایمیل‌های سایر حساب‌های Proton Mail به‌طور خودکار رمزگذاری می‌شوند و رمزگذاری آدرس‌های ایمیل غیر پروتون با کلید OpenPGP به راحتی در تنظیمات حساب شما فعال می‌شود. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail دارای [رمزگذاری OpenPGP یکپارچه](https://proton.me/support/how-to-use-pgp) در ایمیل خود است. ایمیل‌های سایر حساب‌های Proton Mail به‌طور خودکار رمزگذاری می‌شوند و رمزگذاری آدرس‌های ایمیل غیر پروتون با کلید OpenPGP به راحتی در تنظیمات حساب شما فعال می‌شود. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. این قابلیت به افرادی که از سرویس Proton Mail استفاده نمی‌کنند اجازه می‌دهد تا کلیدهای OpenPGP حساب‌های Proton Mail را برای رمزگذاری E2EE سرویس‌های دیگر به راحتی پیدا کنند. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. این ویژگی زمانی مفید است که گیرنده امکان استفاده از OpenPGP را ندارد و نمی تواند یک کپی از ایمیل را در صندوق پستی خود رمزگشایی کند.
 
-Mailbox.org همچنین از کشف کلیدهای عمومی از طریق HTTP از [دایرکتوری کلیدهای وب (WKD)](https://wiki.gnupg.org/WKD) پشتیبانی می کند. این قابلیت به افرادی که از سرویس Mailbox.org استفاده نمی‌کنند اجازه می‌دهد تا کلیدهای OpenPGP حساب‌های Mailbox.org را برای رمزگذاری E2EE سرویس‌های دیگر به راحتی پیدا کنند. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. این قابلیت به افرادی که از سرویس Mailbox.org استفاده نمی‌کنند اجازه می‌دهد تا کلیدهای OpenPGP حساب‌های Mailbox.org را برای رمزگذاری E2EE سرویس‌های دیگر به راحتی پیدا کنند. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } بستن حساب
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/fa/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/fr/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Téléchargements</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Exigences minimales
 
-- Must be open-source.
+- Doit être open-source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Dans le meilleur des cas
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/fr/basics/email-security.md

@@ -19,7 +19,7 @@ Même si vous utilisez OpenPGP, il ne prend pas en charge la [confidentialité p
 
 ## Qu'est-ce que la norme Web Key Directory ?
 
-La norme Web Key Directory (WKD) permet aux clients d'e-mail de découvrir la clé OpenPGP d'autres boîtes mails, même celles qui sont hébergées chez un autre fournisseur. Les clients d'e-mail qui prennent en charge le WKD demandent au serveur du destinataire une clé basée sur le nom de domaine de l'adresse e-mail. Par exemple, si vous envoyez un e-mail à `jonah@privacyguides.org`, votre client d'e-mail demandera à `privacyguides.org` la clé OpenPGP de Jonah, et si `privacyguides.org` dispose d'une clé pour ce compte, votre message sera automatiquement chiffré.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Les clients d'e-mail qui prennent en charge le WKD demandent au serveur du destinataire une clé basée sur le nom de domaine de l'adresse e-mail. Par exemple, si vous envoyez un e-mail à `jonah@privacyguides.org`, votre client d'e-mail demandera à `privacyguides.org` la clé OpenPGP de Jonah, et si `privacyguides.org` dispose d'une clé pour ce compte, votre message sera automatiquement chiffré.
 
 Outre les [clients d'e-mail que nous recommandons](../email-clients.md) et qui prennent en charge le WKD, certains fournisseurs d'e-mail avec interface web prennent également en charge le WKD. Le fait que *votre propre clé* soit publiée sur le WKD pour que d'autres puissent l'utiliser dépend de la configuration de votre domaine. Si vous utilisez un [fournisseur d'e-mail](../email.md#openpgp-compatible-services) qui prend en charge le WKD, tel que Proton Mail ou Mailbox.org, il peut publier votre clé OpenPGP sur son domaine pour vous.
 

i18n/fr/email.md

@@ -34,7 +34,7 @@ En plus (ou à la place) d'un fournisseur de courrier électronique recommandé
 
 ## Services compatibles avec OpenPGP
 
-Ces fournisseurs prennent en charge nativement le chiffrement/déchiffrement OpenPGP et la norme [Web Key Directory](basics/email-security.md#what-is-the-web-key-directory-standard), ce qui permet d'obtenir des e-mails E2EE indépendamment du fournisseur. Par exemple, un utilisateur de Proton Mail peut envoyer un message E2EE à un utilisateur de Mailbox.org, ou vous pouvez recevoir des notifications chiffrées par OpenPGP de la part de services internet qui le supportent.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Par exemple, un utilisateur de Proton Mail peut envoyer un message E2EE à un utilisateur de Mailbox.org, ou vous pouvez recevoir des notifications chiffrées par OpenPGP de la part de services internet qui le supportent.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certaines informations stockées dans [Proton Contacts](https://proton.me/suppor
 
 #### :material-check:{ .pg-green } Chiffrement des e-mails
 
-Proton Mail a [du chiffrement OpenPGP intégré](https://proton.me/support/how-to-use-pgp) dans son interface d'e-mail web. Les e-mails destinés à d'autres comptes Proton Mail sont chiffrés automatiquement, et le chiffrement vers des adresses autres que Proton Mail avec une clé OpenPGP peut être activé facilement dans les paramètres de votre compte. Proton prend également en charge la découverte automatique de clés externes avec le [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Cela signifie que les e-mails envoyés à d'autres fournisseurs qui utilisent WKD seront automatiquement chiffrés avec OpenPGP, sans qu'il soit nécessaire d'échanger manuellement des clés PGP publiques avec vos contacts. Ils vous permettent également de [chiffrer des messages destinés à des adresses non Proton Mail sans OpenPGP](https://proton.me/support/password-protected-emails), sans qu'ils aient besoin de s'inscrire à un compte Proton Mail.
+Proton Mail a [du chiffrement OpenPGP intégré](https://proton.me/support/how-to-use-pgp) dans son interface d'e-mail web. Les e-mails destinés à d'autres comptes Proton Mail sont chiffrés automatiquement, et le chiffrement vers des adresses autres que Proton Mail avec une clé OpenPGP peut être activé facilement dans les paramètres de votre compte. Proton also supports automatic external key discovery with WKD. Cela signifie que les e-mails envoyés à d'autres fournisseurs qui utilisent WKD seront automatiquement chiffrés avec OpenPGP, sans qu'il soit nécessaire d'échanger manuellement des clés PGP publiques avec vos contacts. Ils vous permettent également de [chiffrer des messages destinés à des adresses non Proton Mail sans OpenPGP](https://proton.me/support/password-protected-emails), sans qu'ils aient besoin de s'inscrire à un compte Proton Mail.
 
 Proton Mail publie également les clés publiques des comptes Proton via HTTP à partir de leur WKD. Cela permet aux personnes qui n'utilisent pas Proton Mail de trouver facilement les clés OpenPGP des comptes Proton Mail, pour un E2EE inter-fournisseurs. Cela ne s'applique qu'aux adresses e-mails se terminant par un domaine Proton, comme @proton.me. Si vous utilisez un domaine personnalisé, vous devez [configurer le WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) séparément.
 
@@ -164,7 +164,7 @@ Cependant, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), la plate
 
 Mailbox.org a [du chiffrement intégré](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) dans son interface d'e-mail web, ce qui simplifie l'envoi de messages à des personnes possédant des clés OpenPGP publiques. Ils permettent également aux [destinataires distants de déchiffrer un e-mail](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) sur les serveurs de Mailbox.org. Cette fonction est utile lorsque le destinataire distant ne dispose pas d'OpenPGP et ne peut pas déchiffrer une copie de l'e-mail dans sa propre boîte mail.
 
-Mailbox.org prend également en charge la découverte de clés publiques via HTTP à partir de leur [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Cela permet aux personnes extérieures à Mailbox.org de trouver facilement les clés OpenPGP des comptes Mailbox.org, pour un E2EE inter-fournisseurs. Cela ne s'applique qu'aux adresses e-mails se terminant par un domaine Mailbox, comme @mailbox.org. Si vous utilisez un domaine personnalisé, vous devez [configurer le WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) séparément.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Cela permet aux personnes extérieures à Mailbox.org de trouver facilement les clés OpenPGP des comptes Mailbox.org, pour un E2EE inter-fournisseurs. Cela ne s'applique qu'aux adresses e-mails se terminant par un domaine Mailbox, comme @mailbox.org. Si vous utilisez un domaine personnalisé, vous devez [configurer le WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) séparément.
 
 #### :material-information-outline:{ .pg-blue } Résiliation du compte
 
@@ -323,7 +323,7 @@ Nous considérons ces caractéristiques comme importantes afin de fournir un ser
 
 - Chiffre toutes les données du compte (contacts, calendriers, etc.) au repos avec un chiffrement à accès zéro.
 - Une interface d'e-mail web intégrée avec chiffrement E2EE/PGP est fournie à titre de commodité.
-- Prise en charge de [WKD](https://wiki.gnupg.org/WKD) pour permettre une meilleure découverte des clés publiques OpenPGP via HTTP. Les utilisateurs de GnuPG peuvent obtenir une clé en tapant : `gpg --locate-key utilisateur_exemple@exemple.fr`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. Les utilisateurs de GnuPG peuvent obtenir une clé en tapant : `gpg --locate-key utilisateur_exemple@exemple.fr`
 - Prise en charge d'une boîte mail temporaire pour les utilisateurs externes. Cette fonction est utile lorsque vous souhaitez envoyer un e-mail chiffré, sans envoyer une copie réelle à votre destinataire. Ces e-mails ont généralement une durée de vie limitée et sont ensuite automatiquement supprimés. Ils n'obligent pas non plus le destinataire à configurer un système de chiffrement comme OpenPGP.
 - Disponibilité des services du fournisseur d'e-mail via un [service onion](https://en.wikipedia.org/wiki/.onion).
 - Support du [sous-adressage](https://en.wikipedia.org/wiki/Email_address#Sub-addressing).

i18n/fr/tools.md

@@ -350,7 +350,7 @@ Nous [recommandons](dns.md#recommended-providers) un certain nombre de serveurs
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/he/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### דרישות מינימליות
 
-- Must be open-source.
+- חייב להיות קוד פתוח.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### המקרה הטוב ביותר
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/he/basics/email-security.md

@@ -19,7 +19,7 @@ description: אימייל הוא מטבעו לא מאובטח במובנים ר
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/he/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## ספקי אימייל מומלצים
 
-ספקים אלה תומכים באופן מקורי בהצפנה/פענוח של OpenPGP וב[תקן מדריך מפתחות אינטרנט](basics/email-security.md#what-is-the-web-key-directory-standard), המאפשר לספק -אימיילים אגנוסטיים של E2EE. לדוגמה, משתמש Proton Mail יכול לשלוח הודעת E2EE למשתמש Mailbox.org, או שאתה יכול לקבל התראות מוצפנות OpenPGP משירותי אינטרנט התומכים בכך.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. לדוגמה, משתמש Proton Mail יכול לשלוח הודעת E2EE למשתמש Mailbox.org, או שאתה יכול לקבל התראות מוצפנות OpenPGP משירותי אינטרנט התומכים בכך.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mail supports TOTP [two-factor authentication](https://proton.me/support/
 
 #### :material-check:{ .pg-green } הצפנת אימייל
 
-Proton Mail [שילבה הצפנת OpenPGP](https://proton.me/support/how-to-use-pgp) בדואר האינטרנט שלהם. אימיילים לחשבונות Proton Mail אחרים מוצפנים באופן אוטומטי, וניתן להפעיל הצפנה לכתובות שאינן פרוטון מייל עם מפתח OpenPGP בקלות בהגדרות החשבון שלך. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail [שילבה הצפנת OpenPGP](https://proton.me/support/how-to-use-pgp) בדואר האינטרנט שלהם. אימיילים לחשבונות Proton Mail אחרים מוצפנים באופן אוטומטי, וניתן להפעיל הצפנה לכתובות שאינן פרוטון מייל עם מפתח OpenPGP בקלות בהגדרות החשבון שלך. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. זה מאפשר לאנשים שאינם משתמשים ב-Proton Mail למצוא בקלות את מפתחות OpenPGP של חשבונות Proton Mail, עבור E2EE חוצה ספקים. זה חל רק על כתובות אימיילים המסתיימות באחד מהדומיינים של פרוטון עצמו, כמו proton.me@. אם אתה משתמש בדומיין מותאם אישית, עליך [להגדיר את WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) בנפרד.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. תכונה זו שימושית כאשר לנמען המרוחק אין OpenPGP ואין באפשרותו לפענח עותק של הדואר האלקטרוני בתיבת הדואר שלו.
 
-Mailbox.org תומך גם בגילוי מפתחות ציבוריים באמצעות HTTP מ-[Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) שלהם. זה מאפשר לאנשים מחוץ Mailbox.org למצוא את מפתחות OpenPGP של חשבונות Mailbox.org בקלות, עבור E2EE חוצה ספקים. זה חל רק על כתובות אימיילים המסתיימות באחד מהדומיינים של Mailbox.org עצמו, כמו mailbox.org@. אם אתה משתמש בדומיין מותאם אישית, עליך [להגדיר את WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) בנפרד.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. זה מאפשר לאנשים מחוץ Mailbox.org למצוא את מפתחות OpenPGP של חשבונות Mailbox.org בקלות, עבור E2EE חוצה ספקים. זה חל רק על כתובות אימיילים המסתיימות באחד מהדומיינים של Mailbox.org עצמו, כמו mailbox.org@. אם אתה משתמש בדומיין מותאם אישית, עליך [להגדיר את WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) בנפרד.
 
 #### :material-information-outline:{ .pg-blue } סגירת חשבון
 
@@ -323,7 +323,7 @@ Stalwart does **not** have an integrated webmail, so you will need to use it wit
 
 - מצפין את כל נתוני החשבון (אנשי קשר, יומנים וכו') במצב מנוחה עם הצפנה ללא גישה.
 - הצפנת דואר אינטרנט משולבת E2EE/PGP מסופקת לנוחיותך.
-- תמיכה עבור [WKD](https://wiki.gnupg.org/WKD) כדי לאפשר גילוי משופר של מפתחות OpenPGP ציבוריים באמצעות HTTP. משתמשי GnuPG יכולים לקבל מפתח על ידי הקלדה `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. משתמשי GnuPG יכולים לקבל מפתח על ידי הקלדה `gpg --locate-key example_user@example.com`
 - תמיכה בתיבת דואר זמנית למשתמשים חיצוניים. פעולה זו שימושית כאשר ברצונך לשלוח דוא"ל מוצפן, מבלי לשלוח עותק בפועל לנמען שלך. למיילים אלה יש בדרך כלל תוחלת חיים מוגבלת ולאחר מכן נמחקות אוטומטית. הם גם לא דורשים מהנמען להגדיר שום קריפטוגרפיה כמו OpenPGP.
 - זמינות שירותי ספק הדואר האלקטרוני באמצעות [שירות onion](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/he/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/hi/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/hi/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/hi/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/hi/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/hu/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Alap elvárások
 
-- Must be open-source.
+- Nyílt forráskódúnak kell lennie.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Legjobb esetben
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/hu/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/hu/email.md

@@ -34,7 +34,7 @@ Az itt ajánlott email szolgáltatók mellett (vagy helyett) érdemes megfontoln
 
 ## OpenPGP-kompatibilis szolgáltatások
 
-Ezek a szolgáltatók natívan támogatják az OpenPGP titkosítást/visszafejtést és a [Web Key Directory szabványt](basics/email-security.md#what-is-the-web-key-directory-standard), lehetővé téve a szolgáltatófüggetlen, végponttól-végpontig titkosított emaileket. Például egy Proton Mail felhasználó küldhet végponttól-végpontig titkosított üzenetet egy Mailbox.org felhasználónak, de fogadhatsz OpenPGP-titkosított értesítéseket olyan internetes szolgáltatásoktól is, amelyek támogatják azt.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Például egy Proton Mail felhasználó küldhet végponttól-végpontig titkosított üzenetet egy Mailbox.org felhasználónak, de fogadhatsz OpenPGP-titkosított értesítéseket olyan internetes szolgáltatásoktól is, amelyek támogatják azt.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ A [Proton Contactsban](https://proton.me/support/proton-contacts) tárolt bizony
 
 #### :material-check:{ .pg-green } E-mail titkosítás
 
-A Proton Mail [integrálta az OpenPGP titkosítást](https://proton.me/support/how-to-use-pgp) a webmailjébe. A más Proton Mail-fiókokba küldött e-mailek automatikusan titkosítva vannak, és a nem Proton Mail-címekre küldött, OpenPGP-kulccsal rendelkező e-mailek titkosítása egyszerűen engedélyezhető a fiók beállításaiban. A Proton támogatja az automatikus külső kulcskeresést a [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) segítségével. Ez azt jelenti, hogy a WKD-t használó más szolgáltatóknak küldött e-maileket automatikusan az OpenPGP-vel is titkosítja, anélkül, hogy manuálisan kellene nyilvános PGP-kulcsokat cserélnie a kapcsolattartóival. Az [OpenPGP nélkül, nem Proton Mail címekre küldött üzeneteket titkosíthatod](https://proton.me/support/password-protected-emails), anélkül, hogy a címzetteknek Proton Mail fiókot kellene regisztrálniuk.
+A Proton Mail [integrálta az OpenPGP titkosítást](https://proton.me/support/how-to-use-pgp) a webmailjébe. A más Proton Mail-fiókokba küldött e-mailek automatikusan titkosítva vannak, és a nem Proton Mail-címekre küldött, OpenPGP-kulccsal rendelkező e-mailek titkosítása egyszerűen engedélyezhető a fiók beállításaiban. Proton also supports automatic external key discovery with WKD. Ez azt jelenti, hogy a WKD-t használó más szolgáltatóknak küldött e-maileket automatikusan az OpenPGP-vel is titkosítja, anélkül, hogy manuálisan kellene nyilvános PGP-kulcsokat cserélnie a kapcsolattartóival. Az [OpenPGP nélkül, nem Proton Mail címekre küldött üzeneteket titkosíthatod](https://proton.me/support/password-protected-emails), anélkül, hogy a címzetteknek Proton Mail fiókot kellene regisztrálniuk.
 
 A Proton Mail a Proton-fiókok nyilvános kulcsait is közzéteszi HTTP-n keresztül a WKD-ből. Ez lehetővé teszi, hogy a Proton Mailt nem használók is könnyen megtalálják a Proton Mail fiókok OpenPGP-kulcsait a szolgáltatóközi E2EE-hez. Ez csak a Proton saját domainjeire végződő e-mail címekre vonatkozik, mint például a @proton.me. Ha egyéni tartományt használsz, kézzel kell [konfigurálni a WKD-t](./basics/email-security.md#what-is-the-web-key-directory-standard).
 
@@ -164,7 +164,7 @@ A Mailbox.org által használt szoftverplatform, az [Open-Exchange](https://en.w
 
 A Mailbox.org webmailbe [beépített titkosítást](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) alkalmaz, ami leegyszerűsíti az üzenetek küldését nyilvános OpenPGP-kulcsokkal rendelkező személyeknek. Lehetővé teszik továbbá, hogy a [távoli címzettek visszafejtsenek egy e-mailt](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) a Mailbox.org szerverein. Ez a funkció akkor hasznos, ha a távoli címzett nem rendelkezik OpenPGP-vel, és nem tudja visszafejteni az e-mail másolatát a saját postafiókjában.
 
-A Mailbox.org támogatja a nyilvános kulcsok HTTP-n keresztüli keresését is a [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) segítségével. Ez lehetővé teszi, hogy a Mailbox.org-ot nem használók is könnyen megtalálják a Mailbox.org fiókok OpenPGP-kulcsait a szolgáltatóközi végponttól-végpontig terjedő titkosításhoz. Ez csak a Mailbox.org saját domainjeire végződő e-mail címekre vonatkozik, mint például a @mailbox.org. Ha egyéni tartományt használsz, kézzel kell [konfigurálni a WKD-t](./basics/email-security.md#what-is-the-web-key-directory-standard).
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Ez lehetővé teszi, hogy a Mailbox.org-ot nem használók is könnyen megtalálják a Mailbox.org fiókok OpenPGP-kulcsait a szolgáltatóközi végponttól-végpontig terjedő titkosításhoz. Ez csak a Mailbox.org saját domainjeire végződő e-mail címekre vonatkozik, mint például a @mailbox.org. Ha egyéni tartományt használsz, kézzel kell [konfigurálni a WKD-t](./basics/email-security.md#what-is-the-web-key-directory-standard).
 
 #### :material-information-outline:{ .pg-blue } Fiók megszüntetése
 
@@ -323,7 +323,7 @@ Ezeket a funkciókat fontosnak tartjuk a biztonságos és optimális szolgáltat
 
 - Zéró hozzáférésű titkosítással titkosítja az összes fiókadatot (névjegyek, naptárak stb.).
 - Integrált webmail E2EE/PGP titkosítás, amely kényelmi funkciót biztosít.
-- A [WKD](https://wiki.gnupg.org/WKD) támogatása a nyilvános OpenPGP kulcsok HTTP-n keresztüli jobb felderítésének lehetővé tételéhez. A GnuPG felhasználók kulcsot szerezhetnek a következő paranccsal: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. A GnuPG felhasználók kulcsot szerezhetnek a következő paranccsal: `gpg --locate-key example_user@example.com`
 - Ideiglenes postafiók támogatása külső felhasználók számára. Ez akkor hasznos, ha titkosított e-mailt szeretne küldeni anélkül, hogy a címzettnek tényleges másolatot küldene. Ezek az e-mailek általában korlátozott élettartamúak, majd automatikusan törlődnek. A címzettnek nem kell semmilyen titkosítást konfigurálnia, mint az OpenPGP esetében.
 - Az emailszolgáltató weboldalának elérhetősége egy [.onion szolgáltatáson](https://en.wikipedia.org/wiki/.onion) keresztül.
 - Az [alcímzés](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) támogatása.

i18n/hu/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/id/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Persyaratan Minimum
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Kasus Terbaik
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/id/basics/email-security.md

@@ -19,7 +19,7 @@ Bahkan jika Anda menggunakan OpenPGP, ini tidak mendukung [forward secrecy](http
 
 ## Apa itu standar Direktori Kunci Web?
 
-Standar Direktori Kunci Web (Web Key Directory; WKD) memungkinkan klien surel untuk menemukan kunci OpenPGP untuk kotak surel lain, bahkan yang di-host pada penyedia yang berbeda. Klien surel yang mendukung WKD akan meminta server penerima untuk mendapatkan kunci berdasarkan nama domain alamat surel. Sebagai contoh, jika Anda mengirim surel ke `jonah@privacyguides.org`, klien surel Anda akan meminta `privacyguides.org` untuk mendapatkan kunci OpenPGP Jonah, dan jika `privacyguides.org` memiliki kunci untuk akun tersebut, pesan Anda akan dienkripsi secara otomatis.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Klien surel yang mendukung WKD akan meminta server penerima untuk mendapatkan kunci berdasarkan nama domain alamat surel. Sebagai contoh, jika Anda mengirim surel ke `jonah@privacyguides.org`, klien surel Anda akan meminta `privacyguides.org` untuk mendapatkan kunci OpenPGP Jonah, dan jika `privacyguides.org` memiliki kunci untuk akun tersebut, pesan Anda akan dienkripsi secara otomatis.
 
 Selain [klien surel yang kami rekomendasikan](../email-clients.md), yang mendukung WKD, beberapa penyedia surel berabasis web juga mendukung WKD. Apakah kunci *Anda* diterbitkan ke WKD untuk digunakan orang lain tergantung pada konfigurasi domain Anda. Jika Anda menggunakan [penyedia surel](../email.md#openpgp-compatible-services) yang mendukung WKD, seperti Proton Mail atau Mailbox.org, mereka dapat mempublikasikan kunci OpenPGP Anda ke domain mereka untuk Anda.
 

i18n/id/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## Layanan yang Kompatibel dengan OpenPGP
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Sebagai contoh, pengguna Proton Mail dapat mengirim pesan E2EE ke pengguna Mailbox.org, atau Anda dapat menerima notifikasi terenkripsi OpenPGP dari layanan internet yang mendukungnya.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Sebagai contoh, pengguna Proton Mail dapat mengirim pesan E2EE ke pengguna Mailbox.org, atau Anda dapat menerima notifikasi terenkripsi OpenPGP dari layanan internet yang mendukungnya.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Email ke akun Proton Mail lainnya dienkripsi secara otomatis, dan enkripsi ke alamat non-Proton Mail dengan kunci OpenPGP dapat diaktifkan dengan mudah di pengaturan akun Anda. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Email ke akun Proton Mail lainnya dienkripsi secara otomatis, dan enkripsi ke alamat non-Proton Mail dengan kunci OpenPGP dapat diaktifkan dengan mudah di pengaturan akun Anda. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. Hal ini memungkinkan orang yang tidak menggunakan Proton Mail untuk menemukan kunci OpenPGP akun Proton Mail dengan mudah, untuk lintas-penyedia E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. Fitur ini berguna ketika penerima jarak jauh tidak memiliki OpenPGP dan tidak dapat mendekripsi salinan email di kotak surat mereka sendiri.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Hal ini memungkinkan orang di luar Mailbox.org untuk menemukan kunci OpenPGP dari akun Mailbox.org dengan mudah, untuk lintas-penyedia E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Hal ini memungkinkan orang di luar Mailbox.org untuk menemukan kunci OpenPGP dari akun Mailbox.org dengan mudah, untuk lintas-penyedia E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/id/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/it/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Scarica</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Requisiti minimi
 
-- Must be open-source.
+- Deve essere open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Caso migliore
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/it/basics/email-security.md

@@ -19,7 +19,7 @@ Anche se utilizzi OpenPGP, non supporta la [segretezza in avanti](https://en.wik
 
 ## Che cos'è lo standard Web Key Directory?
 
-Lo standard Web Key Directory (WKD) consente ai client di posta elettronica di scoprire la chiave OpenPGP di altre caselle di posta, anche se ospitate da un altro provider. I client di posta elettronica che supportano il WKD chiederanno al server del destinatario una chiave basata sul nome di dominio dell'indirizzo e-mail. Ad esempio, se invii un'email a `jonah@privacyguides.org`, il tuo client di posta elettronica chiederà a `privacyguides.org` la chiave OpenPGP di Jonah e se `privacyguides.org` dispone di una chiave per quell'account, il tuo messaggio verrà automaticamente crittografato.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. I client di posta elettronica che supportano il WKD chiederanno al server del destinatario una chiave basata sul nome di dominio dell'indirizzo e-mail. Ad esempio, se invii un'email a `jonah@privacyguides.org`, il tuo client di posta elettronica chiederà a `privacyguides.org` la chiave OpenPGP di Jonah e se `privacyguides.org` dispone di una chiave per quell'account, il tuo messaggio verrà automaticamente crittografato.
 
 Oltre ai [client di posta elettronica che consigliamo](../email-clients.md) e che supportano WKD, anche alcuni provider di webmail supportano WKD. Se *la propria chiave* viene pubblicata su WKD per essere utilizzata da altri dipende dalla configurazione del dominio. Se utilizzi un [provider di posta elettronica](../email.md#openpgp-compatible-services) che supporta WKD, come Proton Mail o Mailbox.org, possono pubblicare la tua chiave OpenPGP sul loro dominio per te.
 

i18n/it/email.md

@@ -34,7 +34,7 @@ Oltre a (o al posto di) un provider di posta elettronica consigliato qui, potres
 
 ## Servizi compatibili con OpenPGP
 
-Questi provider supportano in modo nativo la crittografia/decrittografia OpenPGP e lo [standard Web Key Directory](basics/email-security.md#what-is-the-web-key-directory-standard), consentendo la creazione di e-mail E2EE indipendenti dal provider. Ad esempio, un utente di Proton Mail potrebbe inviare un messaggio E2EE a un utente Mailbox.org, o potresti ricevere notifiche crittografate in OpenPGP dai servizi Internet che le supportano.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Ad esempio, un utente di Proton Mail potrebbe inviare un messaggio E2EE a un utente Mailbox.org, o potresti ricevere notifiche crittografate in OpenPGP dai servizi Internet che le supportano.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certe informazioni memorizzate su [Proton Contact](https://proton.me/support/pro
 
 #### :material-check:{ .pg-green } Crittografia Email
 
-Proton Mail ha una [crittografia OpenPGP integrata](https://proton.me/support/how-to-use-pgp) nella propria webmail. Le e-mail inviate ad altri account Proton Mail vengono crittografate automaticamente, e la crittografia verso indirizzi non Proton Mail con una chiave OpenPGP può essere abilitata nelle impostazioni dell'account. Proton supporta anche il rilevamento automatico di chiavi esterne con [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Ciò significa che le e-mail inviate ad altri provider che utilizzano WKD saranno automaticamente crittografate con OpenPGP, senza dover scambiare manualmente le chiavi PGP pubbliche con i tuoi contatti. Consentono inoltre di [crittografare i messaggi inviati a indirizzi non Proton Mail senza OpenPGP](https://proton.me/support/password-protected-emails), senza la necessità di aprire un account Proton Mail.
+Proton Mail ha una [crittografia OpenPGP integrata](https://proton.me/support/how-to-use-pgp) nella propria webmail. Le e-mail inviate ad altri account Proton Mail vengono crittografate automaticamente, e la crittografia verso indirizzi non Proton Mail con una chiave OpenPGP può essere abilitata nelle impostazioni dell'account. Proton also supports automatic external key discovery with WKD. Ciò significa che le e-mail inviate ad altri provider che utilizzano WKD saranno automaticamente crittografate con OpenPGP, senza dover scambiare manualmente le chiavi PGP pubbliche con i tuoi contatti. Consentono inoltre di [crittografare i messaggi inviati a indirizzi non Proton Mail senza OpenPGP](https://proton.me/support/password-protected-emails), senza la necessità di aprire un account Proton Mail.
 
 Proton Mail pubblica anche le chiavi pubbliche degli account Proton via HTTP dal loro WKD. Ciò permette a coloro che non utilizzano Proton Mail, di trovare facilmente le chiavi OpenPGP dei profili di Proton Mail, per un'E2EE tra fornitori. Questo vale solo per gli indirizzi e-mail che terminano con uno dei domini di proprietà di Proton, come @proton.me. Se si utilizza un dominio personalizzato, è necessario [configurare il WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separatamente.
 
@@ -164,7 +164,7 @@ Tuttavia, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), la piatta
 
 Mailbox.org ha la [crittografia integrata](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) nella sua webmail, il che semplifica l'invio di messaggi a persone con chiavi OpenPGP pubbliche. Inoltre, consentono ai [destinatari di decifrare un'e-mail](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) sui server di Mailbox.org. Questa funzionalità è utile quando il destinatario da remoto non ha OpenPGP e non può decrittografare una copia dell'email nella propria casella.
 
-Inoltre, Mailbox.org supporta la scoperta di chiavi pubbliche tramite HTTP dalla loro [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Questo permette a persone esterne a Mailbox.org di trovare facilmente le chiavi OpenPGP degli account di Mailbox.org, per un E2EE fra provider diversi. Questo vale solo per gli indirizzi e-mail che terminano con uno dei domini di Mailbox.org, come @mailbox.org. Se si utilizza un dominio personalizzato, è necessario [configurare il WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separatamente.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Questo permette a persone esterne a Mailbox.org di trovare facilmente le chiavi OpenPGP degli account di Mailbox.org, per un E2EE fra provider diversi. Questo vale solo per gli indirizzi e-mail che terminano con uno dei domini di Mailbox.org, come @mailbox.org. Se si utilizza un dominio personalizzato, è necessario [configurare il WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separatamente.
 
 #### :material-information-outline:{ .pg-blue } Chiusura dell'account
 
@@ -323,7 +323,7 @@ Consideriamo queste funzionalità come importanti per poter fornire un servizio
 
 - Crittografa tutti i dati del profilo (Contatti, Calendari, ecc.) a riposo con crittografia ad accesso zero.
 - Crittografia E2EE/PGP della webmail integrata, fornita per comodità.
-- Supporto per [WKD](https://wiki.gnupg.org/WKD) per consentire la scoperta migliorata delle chiavi pubbliche di OpenPGP tramite HTTP. Gli utenti di GnuPG possono ottenere una chiave digitando: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. Gli utenti di GnuPG possono ottenere una chiave digitando: `gpg --locate-key example_user@example.com`
 - Supporto per una casella temporanea per gli utenti esterni. Questo è utile quando desideri inviare un'email crittografata, senza inviare una copia effettiva al tuo destinatario. Queste email, solitamente, hanno una durata limitata, prima di essere eliminate automaticamente. Inoltre, non richiedono al destinatario di configurare alcuna crittografia, come OpenPGP.
 - Disponibilità dei servizi del fornitore email tramite un [servizio onion](https://en.wikipedia.org/wiki/.onion).
 - Supporto per il [sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing).

i18n/it/tools.md

@@ -356,7 +356,7 @@ Se stai cercando maggiore **sicurezza**, dovresti sempre assicurarti di connette
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/ja/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### 最低要件
 
-- Must be open-source.
+- オープンソースであること。
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### 満たされることが望ましい基準
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/ja/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/ja/email.md

@@ -34,7 +34,7 @@ global:
 
 ## OpenPGP対応サービス
 
-OpenPGPによる暗号化・復号化や[Web Key Directory規格](basics/email-security.md#what-is-the-web-key-directory-standard)をネイティブサポートしているプロバイダーでは、プロバイダーに依存しないE2EE（エンドツーエンド暗号化）メールが利用可能です。 例えば、Proton MailのユーザはMailbox.orgのユーザにE2EEメッセージを送れますし、OpenPGPで暗号化された通知を、それをサポートするインターネットサービスから受け取ることができます。
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. 例えば、Proton MailのユーザはMailbox.orgのユーザにE2EEメッセージを送れますし、OpenPGPで暗号化された通知を、それをサポートするインターネットサービスから受け取ることができます。
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mailはメールと [カレンダー](https://proton.me/news/protoncalend
 
 #### :material-check:{ .pg-green } メールの暗号化
 
-Proton Mailはwebメールに [OpenPGP暗号化を組み込んでいます。](https://proton.me/support/how-to-use-pgp) 他のProton Mailアカウントへのメールは自動的に暗号化され、OpenPGPキーによる非Proton Mailアドレスへの暗号化はアカウント設定から簡単に有効化できます。 Protonは[Web Key Directory(WKD)](https://wiki.gnupg.org/WKD)による外部の鍵の自動探索にも対応しています。 WKDを使った他のプロバイダーに送信されるEメールは自動的にOpenPGPで暗号化され、PGP公開鍵と連絡先を手動で交換する必要はありません。 また、[Proton Mailではないアドレスに送るメッセージをOpenPGPを使わずに暗号化する](https://proton.me/support/password-protected-emails)こともでき、受信者はProton Mailアカウントへのサインアップが必要ありません。
+Proton Mailはwebメールに [OpenPGP暗号化を組み込んでいます。](https://proton.me/support/how-to-use-pgp) 他のProton Mailアカウントへのメールは自動的に暗号化され、OpenPGPキーによる非Proton Mailアドレスへの暗号化はアカウント設定から簡単に有効化できます。 Proton also supports automatic external key discovery with WKD. WKDを使った他のプロバイダーに送信されるEメールは自動的にOpenPGPで暗号化され、PGP公開鍵と連絡先を手動で交換する必要はありません。 また、[Proton Mailではないアドレスに送るメッセージをOpenPGPを使わずに暗号化する](https://proton.me/support/password-protected-emails)こともでき、受信者はProton Mailアカウントへのサインアップが必要ありません。
 
 Proton MailではProtonアカウントの公開鍵をWKDからHTTP経由で公開します。 これにより、Proton Mailを使っていない人でも、Proton MailアカウントのOpenPGPキーを簡単に見つけることができ、プロバイダをまたいだE2EEが可能になります。 @proton.meのようなProtonが所有するドメインのEメールアドレスのみ対象です。 カスタムドメインを使用する場合、[WKDの設定](./basics/email-security.md#what-is-the-web-key-directory-standard)が必要になります。
 
@@ -164,7 +164,7 @@ Mailbox.orgでは[暗号化されたメールボックス](https://kb.mailbox.or
 
 Mailbox.orgのウェブメールは[暗号化機能が組みこまれており](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard)、OpenPGP公開鍵を持つ人へのメッセージの送信が簡単にできます。 Mailbox.orgのサーバー上で[受信者がEメールの復号化をすることもできます](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp)。 この機能はリモートの受信者がOpenPGPを持っておらず、自分のメールボックスにあるメールのコピーを複合できない場合に便利です。
 
-Mailbox.orgは [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) からHTTP経由で公開鍵を発見することもサポートしています。 これにより、Mailbox.orgを使っていない人でも、Mailbox.orgアカウントのOpenPGPキーを簡単に見つけることができ、プロバイダをまたいだE2EEが可能になります。 @mailbox.orgのようなMailbox.orgが所有するドメインのEメールアドレスのみ対象です。 カスタムドメインを使用する場合、[WKDの設定](./basics/email-security.md#what-is-the-web-key-directory-standard)が必要になります。
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. これにより、Mailbox.orgを使っていない人でも、Mailbox.orgアカウントのOpenPGPキーを簡単に見つけることができ、プロバイダをまたいだE2EEが可能になります。 @mailbox.orgのようなMailbox.orgが所有するドメインのEメールアドレスのみ対象です。 カスタムドメインを使用する場合、[WKDの設定](./basics/email-security.md#what-is-the-web-key-directory-standard)が必要になります。
 
 #### :material-information-outline:{ .pg-blue } アカウントの停止
 
@@ -323,7 +323,7 @@ Stalwartにはウェブメールが**ない**ため、[専用のEメールクラ
 
 - ゼロアクセス暗号化により、すべてのアカウントのデータ（連絡先、カレンダーなど）が暗号化されていること。
 - 利便性のため、E2EE/PGP暗号化できるウェブメールがあること。
-- HTTP経由でのOpenPGP公開鍵の探索を改善するため、[WKD](https://wiki.gnupg.org/WKD)へ対応していること。 GnuPGでは次のスクリプトで鍵を取得できます： `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPGでは次のスクリプトで鍵を取得できます： `gpg --locate-key example_user@example.com`
 - 外部ユーザー用の一時的なメールボックスがあること。 受信者に実際のメールのコピーを送るのではなく、暗号化されたメールを送る際に役立ちます。 通常の場合、一時的なメールボックスのメールには期限があり、自動的に削除されます。 また、受信者はOpenPGPのような暗号化を設定する必要がありません。
 - [.onionサービス](https://en.wikipedia.org/wiki/.onion)経由でEメールプロバイダーのサービスが利用できること。
 - [サブアドレス](https://en.wikipedia.org/wiki/Email_address#Sub-addressing)に対応していること。

i18n/ja/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/ko/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### 최소 요구 사항
 
-- Must be open-source.
+- 오픈 소스여야 합니다.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### 우대 사항
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/ko/basics/email-security.md

@@ -19,7 +19,7 @@ OpenPGP를 사용하더라도 [순방향 비밀성(Forward secrecy)](https://en.
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/ko/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP 호환 서비스
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. 예를 들어, Proton Mail 사용자는 Mailbox.org 사용자에게 E2EE 메시지를 보내거나, OpenPGP 지원 인터넷 서비스에서 OpenPGP로 암호화된 알림을 받을 수 있습니다.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. 예를 들어, Proton Mail 사용자는 Mailbox.org 사용자에게 E2EE 메시지를 보내거나, OpenPGP 지원 인터넷 서비스에서 OpenPGP로 암호화된 알림을 받을 수 있습니다.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mail은 이메일 및 [캘린더](https://proton.me/news/protoncalendar-s
 
 #### :material-check:{ .pg-green } 이메일 암호화
 
-Proton Mail은 웹메일에 [OpenPGP 암호화 기능을 내장](https://proton.me/support/how-to-use-pgp)하고 있습니다. 다른 Proton Mail 계정으로 보내는 이메일은 자동으로 암호화되며, Proton Mail 외 주소로 보내는 이메일에 대한 OpenPGP 암호화는 계정 설정에서 간편하게 활성화할 수 있습니다. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail은 웹메일에 [OpenPGP 암호화 기능을 내장](https://proton.me/support/how-to-use-pgp)하고 있습니다. 다른 Proton Mail 계정으로 보내는 이메일은 자동으로 암호화되며, Proton Mail 외 주소로 보내는 이메일에 대한 OpenPGP 암호화는 계정 설정에서 간편하게 활성화할 수 있습니다. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. 이로써 Proton Mail을 사용하지 않는 사람도 Proton Mail OpenPGP 키를 쉽게 찾아 서로 다른 제공 업체 간 E2EE 적용이 가능합니다. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. OpenPGP가 없어 수신자가 자신의 메일함에서 직접 복호화할 수 없을 경우에 이 기능을 사용할 수 있습니다.
 
-또한, Mailbox.org는 [웹 키 디렉터리(WKD)](https://wiki.gnupg.org/WKD)에서 HTTP를 통한 공개 키 검색을 지원합니다. Mailbox.org를 사용하지 않는 사람들은 Mailbox.org 계정의 OpenPGP 공개키를 쉽게 찾을 수 있고, 플랫폼과 무관하게 종단간 암호화를 할 수 있습니다. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Mailbox.org를 사용하지 않는 사람들은 Mailbox.org 계정의 OpenPGP 공개키를 쉽게 찾을 수 있고, 플랫폼과 무관하게 종단간 암호화를 할 수 있습니다. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } 계정 해지
 
@@ -323,7 +323,7 @@ Stalwart does **not** have an integrated webmail, so you will need to use it wit
 
 - Zero Access Encryption을 통해 모든 계정 데이터(연락처, 캘린더 등)를 암호화해야 합니다.
 - 웹메일에 E2EE/PGP 암호화가 통합되어 있어서 편리하게 사용할 수 있어야 합니다.
-- [WKD](https://wiki.gnupg.org/WKD)를 지원하여 HTTP를 통한 공개 OpenPGP 키 검색 편의를 제공해야 합니다. GnuPG 사용자는 `gpg --locate-key example_user@example.com`를 입력하여 키를 얻을 수 있습니다.
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG 사용자는 `gpg --locate-key example_user@example.com`를 입력하여 키를 얻을 수 있습니다.
 - 외부 사용자를 위해 임시 메일함을 지원해야 합니다. 수신자에게 실제 사본을 보내지 않고 암호화된 이메일을 보내고자 할 때 유용합니다. 이러한 이메일은 보통 수명이 제한돼 있으며 이후 자동으로 삭제됩니다. 수신자가 OpenPGP 등의 암호화를 설정할 필요가 없습니다.
 - [Onion 서비스](https://en.wikipedia.org/wiki/.onion)를 통해 이메일 서비스를 이용할 수 있어야 합니다.
 - [하위 주소](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) 지원.

i18n/ko/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/ku-IQ/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/ku-IQ/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/ku-IQ/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## خزمەتگوزاریەکانی گونجاون لەگەڵ OpenPGP
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. بۆ نموونە: بەکارهێنەرێکی Proton Mail دەتوانێت پەیامێکی E2EE بنێرێت بۆ بەکارهێنەرێکی Mailbox.org، یان دەتوانیت ئاگادارکردنەوەی OpenPGP-شفرکراوت پێ بگات لەڕێی ئەو خزمەتگوزاریانەی پشتگیری دەکەن.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. بۆ نموونە: بەکارهێنەرێکی Proton Mail دەتوانێت پەیامێکی E2EE بنێرێت بۆ بەکارهێنەرێکی Mailbox.org، یان دەتوانیت ئاگادارکردنەوەی OpenPGP-شفرکراوت پێ بگات لەڕێی ئەو خزمەتگوزاریانەی پشتگیری دەکەن.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mail تەکنەلۆژیای [شفرکردن و تێپەڕبوونی-ئە
 
 #### :material-check:{ .pg-green } شفرکردنی پۆستەی ئەلکتڕۆنی
 
-Proton Mail [شفرکردنی OpenPGP زیادکردووە](https://proton.me/support/how-to-use-pgp) بۆ ماڵپەری پۆستەی ئەلکتڕۆنییەکەیان. پۆستەی ئەلکتڕۆنی نێوان هەژمارەکانی Proton Mail خۆکارانە شفرکراوە، بەڵام شفرکردن لە نێوان Proton Mail و پۆستەی ئەلکتڕۆنی تر شفردەکرێن بە ئاسانی لەڕێگەی کلیلی OpenPGP، کە لە ڕێکخستنەکانی هەژمارەکەت هەیە. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail [شفرکردنی OpenPGP زیادکردووە](https://proton.me/support/how-to-use-pgp) بۆ ماڵپەری پۆستەی ئەلکتڕۆنییەکەیان. پۆستەی ئەلکتڕۆنی نێوان هەژمارەکانی Proton Mail خۆکارانە شفرکراوە، بەڵام شفرکردن لە نێوان Proton Mail و پۆستەی ئەلکتڕۆنی تر شفردەکرێن بە ئاسانی لەڕێگەی کلیلی OpenPGP، کە لە ڕێکخستنەکانی هەژمارەکەت هەیە. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. ئەمەڕێگە ئەو کەسانە دەدات کە Proton Mail بەکار ناهێنن بۆئەوەی کلیکە گشتیەکانی هەژماری Proton Mail بە ئاسانی بدۆزنەوە. تا ببێتە هۆی گونجاندن لەگەڵ دابینکەرانی تری E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. ئەم تایبەتمەندیە بەسوودە کاتێک وەرگر لە دوورەوە OpenPGPـی نییە و ناتوانێت شفرەکە لەسەر لەبەرگیراوەیەکی پۆستەکە لاببات لە سندووقی پۆستەکانی خۆیدا.
 
-هەروەها Mailbox.org پشتگیری دۆزینەوەی کلیلە گشتیەکان دەکات لە HTTPــەوە لە ڕێگای [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). ئەمە ڕێگە بە کەسانی دەرەوەی Mailbox.org دەدات کە کلیلەکانی OpenPGP بۆ هەژمارەکانی Mailbox.org بە ئاسانی بدۆزنەوە، تا ببێتە هۆی گونجاندن لەگەڵ دابینکەرانی تری E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. ئەمە ڕێگە بە کەسانی دەرەوەی Mailbox.org دەدات کە کلیلەکانی OpenPGP بۆ هەژمارەکانی Mailbox.org بە ئاسانی بدۆزنەوە، تا ببێتە هۆی گونجاندن لەگەڵ دابینکەرانی تری E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } لەناوبردنی هەژمار
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/ku-IQ/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/nl/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimale vereisten
 
-- Must be open-source.
+- Moet open source zijn.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Beste geval
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/nl/basics/email-security.md

@@ -19,7 +19,7 @@ Zelfs als je OpenPGP gebruikt, biedt het geen ondersteuning voor [forward secrec
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/nl/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP compatibele diensten
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Een Proton Mail-gebruiker zou bijvoorbeeld een E2EE-bericht kunnen sturen naar een Mailbox.org-gebruiker, of je zou OpenPGP-versleutelde meldingen kunnen ontvangen van internetdiensten die dit ondersteunen.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Een Proton Mail-gebruiker zou bijvoorbeeld een E2EE-bericht kunnen sturen naar een Mailbox.org-gebruiker, of je zou OpenPGP-versleutelde meldingen kunnen ontvangen van internetdiensten die dit ondersteunen.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Bepaalde informatie opgeslagen in [Proton Contacts](https://proton.me/support/pr
 
 #### :material-check:{ .pg-green } Email encryptie
 
-Proton Mail heeft [OpenPGP encryptie](https://proton.me/support/how-to-use-pgp) geïntegreerd in hun webmail. E-mails naar andere Proton Mail-accounts worden automatisch versleuteld, en versleuteling naar niet-Proton Mail-adressen met een OpenPGP-sleutel kan eenvoudig worden ingeschakeld in je accountinstellingen. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail heeft [OpenPGP encryptie](https://proton.me/support/how-to-use-pgp) geïntegreerd in hun webmail. E-mails naar andere Proton Mail-accounts worden automatisch versleuteld, en versleuteling naar niet-Proton Mail-adressen met een OpenPGP-sleutel kan eenvoudig worden ingeschakeld in je accountinstellingen. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. Hierdoor kunnen mensen die geen Proton Mail gebruiken de OpenPGP sleutels van Proton Mail accounts gemakkelijk vinden, voor cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. Deze functie is nuttig wanneer de ontvanger op afstand geen OpenPGP heeft en geen kopie van de e-mail in zijn eigen mailbox kan ontsleutelen.
 
-Mailbox.org ondersteunt ook de ontdekking van publieke sleutels via HTTP vanuit hun [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Hierdoor kunnen mensen buiten Mailbox.org gemakkelijk de OpenPGP sleutels van Mailbox.org accounts vinden, voor cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Hierdoor kunnen mensen buiten Mailbox.org gemakkelijk de OpenPGP sleutels van Mailbox.org accounts vinden, voor cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Beëindiging van account
 
@@ -323,7 +323,7 @@ Wij beschouwen deze kenmerken als belangrijk om een veilige en optimale dienst t
 
 - Versleutelt alle accountgegevens (Contacten, Agenda's, etc) in rust met zero-access encryptie.
 - Geïntegreerde webmail E2EE/PGP-codering voor het gemak.
-- Ondersteuning voor [WKD](https://wiki.gnupg.org/WKD) om een verbeterde ontdekking van publieke OpenPGP sleutels via HTTP mogelijk te maken. GnuPG-gebruikers kunnen een sleutel krijgen door te typen: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG-gebruikers kunnen een sleutel krijgen door te typen: `gpg --locate-key example_user@example.com`
 - Ondersteuning voor een tijdelijke mailbox voor externe gebruikers. Dit is handig wanneer je een versleutelde e-mail wilt verzenden, zonder een echte kopie naar jouw ontvanger te sturen. Deze e-mails hebben meestal een beperkte levensduur en worden daarna automatisch verwijderd. Zij vereisen ook niet dat de ontvanger cryptografie configureert zoals OpenPGP.
 - Beschikbaarheid van de diensten van de e-mailprovider via een [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/nl/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/pl/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Musi być open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Best-Case
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/pl/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/pl/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/pl/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/pt-BR/ai-chat.md

@@ -12,7 +12,7 @@ Protege contra as seguintes ameaças:
 - [:material-account-cash: Vigilância Capitalista](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censura](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Desde seu lançamento em 2022, o ChatGPT engajou popularmente e tornou as interações com grandes modelos de linguagem (LLMs)cada vez mais comuns. Os LLMs podem nos ajudar a escrever melhor, entender sobre assuntos desconhecidos ou responder a uma ampla gama de perguntas,  ou até mesmo prever a próxima palavra digitada baseados num grande banco de dados extraído da internet.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. Os LLMs podem nos ajudar a escrever melhor, entender sobre assuntos desconhecidos ou responder a uma ampla gama de perguntas,  They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Preocupações em relação aos LLMs
 
@@ -42,7 +42,7 @@ Para executar a IA localmente é necessário um modelo de IA e um cliente de IA.
 
 ### Escolhendo um Modelo
 
-Há muitos modelos de código aberto ou de licença livre disponíveis para download. A [Hugging Face](https://huggingface.co/models) é uma plataforma que permite que você navegue, pesquise e faça download de modelos em formatos comuns, como [GGUF](https://huggingface.co/docs/hub/en/gguf). Empresas que fornecem bons modelos de direitos facilitados incluem grandes nomes como Mistral, Meta, Microsoft e Google. No entanto, também há muitos modelos comunitários e "ajustes" disponíveis. Conforme mencionado acima: modelos quantizados oferecem o melhor equilíbrio entre a qualidade e o desempenho do modelo para aqueles que usam um computador pessoal (equipamento de nível de consumidor).
+Há muitos modelos de código aberto ou de licença livre disponíveis para download. A [Hugging Face](https://huggingface.co/models) é uma plataforma que permite que você navegue, pesquise e faça download de modelos em formatos comuns, como [GGUF](https://huggingface.co/docs/hub/en/gguf). Empresas que fornecem bons modelos de direitos facilitados incluem grandes nomes como Mistral, Meta, Microsoft e Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. Conforme mencionado acima: modelos quantizados oferecem o melhor equilíbrio entre a qualidade e o desempenho do modelo para aqueles que usam um computador pessoal (equipamento de nível de consumidor).
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Requirements
 
-- Must be open-source.
+- Deve ser de código aberto.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Melhor Caso
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/pt-BR/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/pt-BR/email.md

@@ -34,7 +34,7 @@ Além de (ou ao invés de) um provedor de e-mail recomendado aqui, você pode co
 
 ## Serviços Compatíveis com OpenPGP
 
-Esses provedores suportam nativamente criptografia/descriptografia OpenPGP e o padrão [do Diretório de Chaves Web](basics/email-security.md#what-is-the-web-key-directory-standard), permitindo e-mails independentes E2EE do provedor. Por exemplo, um usuário do Proton Mail pode mandar uma mensagem E2E para um usuário de Mailbox.org, ou você pode receber notificações criptografadas por OpenPGP de serviços de internet que suportam isso.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Por exemplo, um usuário do Proton Mail pode mandar uma mensagem E2E para um usuário de Mailbox.org, ou você pode receber notificações criptografadas por OpenPGP de serviços de internet que suportam isso.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certas informações armazenadas no [Proton Contacts](https://proton.me/support/
 
 #### :material-check:{ .pg-green } Criptografia do Email
 
-Proton Mail [tem criptografia OpenPGP integrada](https://proton.me/support/how-to-use-pgp) em seu webmail. E-mails para outras contas do Proton Mail são criptografados automaticamente, e criptografia para endereços não-Proton Mail com uma chave OpenPGP pode ser facilmente ativada nas configurações da sua conta. O Proton também oferece suporte à descoberta automática de chaves externas com o [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Isso significa que os e-mails enviados a outros provedores que usam o WKD também serão criptografados automaticamente com o OpenPGP, sem a necessidade de trocar manualmente chaves PGP públicas com seus contatos. Eles também permitem que você [criptografe mensagens para endereços não-Proton Mail](https://proton.me/support/password-protected-emails) sem a necessidade de eles se cadastrarem com uma conta Proton Mail ou usar programas como OpenPGP.
+Proton Mail [tem criptografia OpenPGP integrada](https://proton.me/support/how-to-use-pgp) em seu webmail. E-mails para outras contas do Proton Mail são criptografados automaticamente, e criptografia para endereços não-Proton Mail com uma chave OpenPGP pode ser facilmente ativada nas configurações da sua conta. Proton also supports automatic external key discovery with WKD. Isso significa que os e-mails enviados a outros provedores que usam o WKD também serão criptografados automaticamente com o OpenPGP, sem a necessidade de trocar manualmente chaves PGP públicas com seus contatos. Eles também permitem que você [criptografe mensagens para endereços não-Proton Mail](https://proton.me/support/password-protected-emails) sem a necessidade de eles se cadastrarem com uma conta Proton Mail ou usar programas como OpenPGP.
 
 O Proton Mail também publica as chaves públicas das contas Proton via HTTP a partir de seu WKD. Isso permite que as pessoas que não usam o Proton Mail encontrem as chaves OpenPGP de contas Proton Mail facilmente, para criptografia ponta-a-ponta (E2EE) entre provedores. Isso só se aplica aos endereços de e-mail que terminam em um dos domínios da própria ProtonMail, como @proton.me. Se você usar um domínio personalizado, deverá [configurar o WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separadamente.
 
@@ -164,7 +164,7 @@ No entanto, [o Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), a pla
 
 Mailbox.org tem [criptografia integrada](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) em seu webmail, o que simplifica o envio de mensagens para pessoas com chaves OpenPGP públicas. Eles também permitem que [destinatários remotos descriptografem um e-mail](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) nos servidores do Mailbox.org. Esse recurso é útil quando o destinatário remoto não tem OpenPGP e não pode descriptografar uma cópia do e-mail em sua própria caixa de correio.
 
-Mailbox.org também suporta a descoberta de chaves públicas via HTTP a partir do seu [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Isso permite que pessoas fora do Mailbox.org encontrem as chaves OpenPGP de contas Mailbox.org facilmente, para criptografia ponta-a-ponta (E2EE) entre provedores. Isso só se aplica aos endereços de e-mail que terminam em um dos domínios da própria Mailbox, como @mailbox.org. Se você usar um domínio personalizado, deverá [configurar o WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separadamente.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Isso permite que pessoas fora do Mailbox.org encontrem as chaves OpenPGP de contas Mailbox.org facilmente, para criptografia ponta-a-ponta (E2EE) entre provedores. Isso só se aplica aos endereços de e-mail que terminam em um dos domínios da própria Mailbox, como @mailbox.org. Se você usar um domínio personalizado, deverá [configurar o WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separadamente.
 
 #### :material-information-outline:{ .pg-blue } Rescisão da Conta
 
@@ -323,7 +323,7 @@ Consideramos esses recursos importantes para fornecer um serviço seguro e otimi
 
 - Criptografa todos os dados da conta (contatos, calendários, etc.) em repouso com criptografia de acesso zero.
 - Criptografia E2EE/PGP integrada de webmail fornecido como conveniência.
-- Suporte ao [WKD](https://wiki.gnupg.org/WKD) para permitir a descoberta aprimorada de chaves OpenPGP públicas via HTTP. Usuários do GnuPG podem obter uma chave digitando: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. Usuários do GnuPG podem obter uma chave digitando: `gpg --locate-key example_user@example.com`
 - Suporte para uma caixa de correio temporária para usuários externos. Isso é útil quando você deseja enviar um e-mail criptografado sem enviar uma cópia real para o seu destinatário. Estes e-mails geralmente têm um tempo de vida limitado e depois são automaticamente excluídos. Eles também não exigem que o destinatário configure nenhuma criptografia, como o OpenPGP.
 - Disponibilidade do site do provedor de serviços de e-mail em um [serviço onion](https://en.wikipedia.org/wiki/.onion).
 - Suporte a [subendereçamento](https://en.wikipedia.org/wiki/Email_address#Sub-addressing).

i18n/pt-BR/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/pt/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -161,7 +163,7 @@ To check the authenticity and safety of the model, look for:
 - Community reviews and usage statistics
 - A "Safe" badge next to the model file (Hugging Face only)
 - Matching checksums[^1]
-  - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
+    - On Hugging Face, you can find the hash by clicking on a model file and looking for the **Copy SHA256** button below it. You should compare this checksum with the one from the model file you downloaded.
 
 A downloaded model is generally safe if it satisfies all the above checks.
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Requisitos mínimos
 
-- Must be open-source.
+- Deve ser de fonte aberta.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Melhor caso
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/pt/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/pt/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## Serviços Compatíveis com OpenPGP
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Por exemplo, um utilizador do Proton Mail pode enviar uma mensagem E2EE a um utilizador do Mailbox.org, ou pode receber notificações encriptadas em OpenPGP de serviços Internet que o suportem.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Por exemplo, um utilizador do Proton Mail pode enviar uma mensagem E2EE a um utilizador do Mailbox.org, ou pode receber notificações encriptadas em OpenPGP de serviços Internet que o suportem.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certas informações armazenadas em [Proton Contactos](https://proton.me/support
 
 #### :material-check:{ .pg-green } Encriptação de e-mail
 
-O Proton Mail tem [encriptação OpenPGP integrada](https://proton.me/support/how-to-use-pgp) no seu webmail. Os e-mails para outras contas do Proton Mail são encriptados automaticamente e a encriptação para endereços que não sejam do Proton Mail com uma chave OpenPGP pode ser ativada facilmente nas definições da sua conta. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+O Proton Mail tem [encriptação OpenPGP integrada](https://proton.me/support/how-to-use-pgp) no seu webmail. Os e-mails para outras contas do Proton Mail são encriptados automaticamente e a encriptação para endereços que não sejam do Proton Mail com uma chave OpenPGP pode ser ativada facilmente nas definições da sua conta. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. Isto permite que as pessoas que não utilizam o Proton Mail encontrem facilmente as chaves OpenPGP das contas do Proton Mail, para E2EE entre fornecedores. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. Esta funcionalidade é útil quando o destinatário remoto não tem o OpenPGP e não consegue desencriptar uma cópia do e-mail na sua própria caixa de correio.
 
-O Mailbox.org também suporta a descoberta de chaves públicas via HTTP a partir do seu [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Isto permite que pessoas que não utilizem o Mailbox.org encontrem facilmente as chaves OpenPGP das contas Mailbox.org, para E2EE entre fornecedores. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Isto permite que pessoas que não utilizem o Mailbox.org encontrem facilmente as chaves OpenPGP das contas Mailbox.org, para E2EE entre fornecedores. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Remoção da conta
 
@@ -323,7 +323,7 @@ Consideramos que estas características são importantes para podermos prestar u
 
 - Encriptação de todos os dados da conta (contactos, calendários, etc.) em estado de repouso, com encriptação de acesso zero.
 - Encriptação E2EE/PGP integrada no webmail, fornecida como uma conveniência.
-- Suporte para [WKD](https://wiki.gnupg.org/WKD) para permitir uma melhor descoberta de chaves OpenPGP públicas através de HTTP. Os utilizadores do GnuPG podem obter uma chave escrevendo: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. Os utilizadores do GnuPG podem obter uma chave escrevendo: `gpg --locate-key example_user@example.com`
 - Suporte para uma caixa de correio temporária para utilizadores externos. Isto é útil quando se pretende enviar uma mensagem de e-mail encriptada, sem enviar uma cópia real ao destinatário. Estas mensagens de e-mail têm normalmente um tempo de vida limitado e depois são automaticamente eliminadas. Também não requerem que o destinatário configure qualquer criptografia como o OpenPGP.
 - Disponibilidade dos serviços do fornecedor de e-mail através de um serviço onion [](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/pt/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/ru/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Минимальные требования к сервисам
 
-- Must be open-source.
+- Исходный код проекта должен быть открыт.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### В лучшем случае
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/ru/basics/email-security.md

@@ -19,7 +19,7 @@ description: Электронная почта по своей природе н
 
 ## Что такое стандарт Web Key Directory?
 
-Стандарт Web Key Directory (WKD) позволяет почтовым клиентам находить OpenPGP-ключи для почтовых ящиков, даже расположенных у других провайдеров. Почтовые клиенты, поддерживающие WKD, запрашивают ключ у сервера получателя, основываясь на доменном имени его адреса. Например, когда вы пишете на `jonah@privacyguides.org`, ваш почтовый клиент запрашивает у `privacyguides.org` OpenPGP-ключ Джона, и если `privacyguides.org` имеет соответствующий ключ, ваше сообщение будет автоматически зашифровано.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Почтовые клиенты, поддерживающие WKD, запрашивают ключ у сервера получателя, основываясь на доменном имени его адреса. Например, когда вы пишете на `jonah@privacyguides.org`, ваш почтовый клиент запрашивает у `privacyguides.org` OpenPGP-ключ Джона, и если `privacyguides.org` имеет соответствующий ключ, ваше сообщение будет автоматически зашифровано.
 
 В дополнение к [рекомендованным почтовым клиентам](../email-clients.md), поддерживающим WKD, некоторые браузерные почтовые интерфейсы также поддерживают WKD. Будет ли *ваш личный* ключ опубликован в WKD для других пользователей, зависит от конфигурации вашего домена. Если вы пользуетесь [почтовым провайдером](../email.md#openpgp-compatible-services), поддерживающим WKD, таким как Proton Mail или Mailbox.org, они опубликуют ваш OpenPGP-ключ на своем домене.
 

i18n/ru/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## Сервисы, поддерживающие OpenPGP
 
-Эти провайдеры нативно поддерживают шифрование/дешифрование с помощью OpenPGP и стандарт [Web Key Directory](basics/email-security.md#what-is-the-web-key-directory-standard), что позволяет использовать E2EE, не зависимо от провайдера. Например, пользователь Proton Mail может отправлять E2EE-зашифрованное сообщение пользователю Mailbox.org, или ты можешь получить OpenPGP-зашифрованное уведомление от интернет-сервисов, поддерживающих такую функцию.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Например, пользователь Proton Mail может отправлять E2EE-зашифрованное сообщение пользователю Mailbox.org, или ты можешь получить OpenPGP-зашифрованное уведомление от интернет-сервисов, поддерживающих такую функцию.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Proton Mail использует [шифрование с нулевым дос
 
 #### :material-check:{ .pg-green } Шифрование электронной почты
 
-Proton Mail [интегрировал шифрование OpenPGP](https://proton.me/support/how-to-use-pgp) в свою веб-почту. Письма, отправленные на другие аккаунты Proton Mail шифруются автоматически. Шифрование писем с помощью ключа OpenPGP на адреса, не принадлежащие Proton Mail, можно легко включить в настройках аккаунта. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail [интегрировал шифрование OpenPGP](https://proton.me/support/how-to-use-pgp) в свою веб-почту. Письма, отправленные на другие аккаунты Proton Mail шифруются автоматически. Шифрование писем с помощью ключа OpenPGP на адреса, не принадлежащие Proton Mail, можно легко включить в настройках аккаунта. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. Это позволяет людям, не использующим Proton Mail, легко находить OpenPGP-ключи учетных записей Proton Mail для кросс-провайдерского E2EE. Это относится только к адресам электронной почты, заканчивающимся на один из собственных доменов "Протона", например @proton.me. При использовании кастомного домена необходимо [настроить WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) отдельно.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. Эта функция полезна, когда получатель не имеет OpenPGP и не может расшифровать копию письма в собственном почтовом ящике.
 
-Mailbox.org также поддерживает обнаружение открытых ключей через HTTP с их [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). Это позволяет людям, не использующим Mailbox.org, легко находить OpenPGP-ключи учетных записей Mailbox.org для кросс-провайдерского E2EE. Это относится только к адресам электронной почты, заканчивающимся на один из собственных доменов "Mailbox.org", например @mailbox.org. При использовании кастомного домена необходимо [настроить WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) отдельно.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. Это позволяет людям, не использующим Mailbox.org, легко находить OpenPGP-ключи учетных записей Mailbox.org для кросс-провайдерского E2EE. Это относится только к адресам электронной почты, заканчивающимся на один из собственных доменов "Mailbox.org", например @mailbox.org. При использовании кастомного домена необходимо [настроить WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) отдельно.
 
 #### :material-information-outline:{ .pg-blue } Блокировка аккаунта
 
@@ -323,7 +323,7 @@ Stalwart does **not** have an integrated webmail, so you will need to use it wit
 
 - Шифрует все данные аккаунта (Контакты, Календари и т.д.) в состоянии покоя с помощью шифрования с нулевым доступом.
 - Встроенное шифрование веб-почты E2EE/PGP обеспечивает удобство.
-- Поддерживает [WKD](https://wiki.gnupg.org/WKD) для улучшения обнаружения открытых ключей OpenPGP через HTTP. Пользователи GnuPG могут получить ключ, набрав: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. Пользователи GnuPG могут получить ключ, набрав: `gpg --locate-key example_user@example.com`
 - Поддержка временного почтового ящика для внешних пользователей. Это полезно, когда вы хотите отправить зашифрованное сообщение электронной почты, не отправляя фактическую копию получателю. Такие письма обычно имеют ограниченный срок действия, а затем автоматически удаляются. Они также не требуют от получателя настройки какой-либо криптографии, как OpenPGP.
 - Доступность услуг провайдера электронной почты через [службу .onion](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/ru/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/sv/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimikrav
 
-- Must be open-source.
+- Måste vara öppen källkod.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### Bästa fall
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/sv/basics/email-security.md

@@ -19,7 +19,7 @@ Det finns en annan standard som är populär bland företag och som heter [S/MIM
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/sv/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Compatible Services
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. For example, a Proton Mail user could send an E2EE message to a Mailbox.org user, or you could receive OpenPGP-encrypted notifications from internet services which support it.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } Email Encryption
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Emails to other Proton Mail accounts are encrypted automatically, and encryption to non-Proton Mail addresses with an OpenPGP key can be enabled easily in your account settings. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/sv/tools.md

@@ -356,7 +356,7 @@ We [recommend](dns.md#recommended-providers) a number of encrypted DNS servers b
 <div class="grid cards" markdown>
 
 - ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>

i18n/tr/ai-chat.md

@@ -12,7 +12,7 @@ cover: ai-chatbots.webp
 - [:material-account-cash: Surveillance Capitalism](basics/common-threats.md#surveillance-as-a-business-model){ .pg-brown }
 - [:material-close-outline: Censorship](basics/common-threats.md#avoiding-censorship){ .pg-blue-gray }
 
-Since the release of ChatGPT in 2022, interactions with Large Language Models (LLMs) have become increasingly common. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They can statistically predict the next word based on a vast amount of data scraped from the web.
+The use of **AI chat**, also known as Large Language Models (LLMs), has become increasingly common since the release of ChatGPT in 2022. LLMs can help us write better, understand unfamiliar subjects, or answer a wide range of questions. They work by statistically predicting the next word in their responses based on a vast amount of data scraped from the web.
 
 ## Privacy Concerns About LLMs
 
@@ -42,7 +42,7 @@ To run AI locally, you need both an AI model and an AI client.
 
 ### Choosing a Model
 
-There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and 'fine-tunes' available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
+There are many permissively licensed models available to download. [Hugging Face](https://huggingface.co/models) is a platform that lets you browse, research, and download models in common formats like [GGUF](https://huggingface.co/docs/hub/en/gguf). Companies that provide good open-weights models include big names like Mistral, Meta, Microsoft, and Google. However, there are also many community models and [fine-tuned](https://en.wikipedia.org/wiki/Fine-tuning_\(deep_learning\)) models available. As mentioned above, quantized models offer the best balance between model quality and performance for those using consumer-grade hardware.
 
 To help you choose a model that fits your needs, you can look at leaderboards and benchmarks. The most widely-used leaderboard is the community-driven [LM Arena](https://lmarena.ai). Additionally, the [OpenLLM Leaderboard](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) focuses on the performance of open-weights models on common benchmarks like [MMLU-Pro](https://arxiv.org/abs/2406.01574).  There are also specialized benchmarks which measure factors like [emotional intelligence](https://eqbench.com), ["uncensored general intelligence"](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard), and [many others](https://www.nebuly.com/blog/llm-leaderboards).
 
@@ -63,7 +63,7 @@ To help you choose a model that fits your needs, you can look at leaderboards an
 
 ![Kobold.cpp Logo](assets/img/ai-chat/kobold.png){align=right}
 
-Kobold.cpp is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
+**Kobold.cpp** is an AI client that runs locally on your Windows, Mac, or Linux computer. It's an excellent choice if you are looking for heavy customization and tweaking, such as for role-playing purposes.
 
 In addition to supporting a large range of text models, Kobold.cpp also supports image generators such as [Stable Diffusion](https://stability.ai/stable-image) and automatic speech recognition tools such as [Whisper](https://github.com/ggerganov/whisper.cpp).
 
@@ -83,7 +83,7 @@ In addition to supporting a large range of text models, Kobold.cpp also supports
 
 </div>
 
-<div class="admonition note" markdown>
+<div class="admonition info" markdown>
 <p class="admonition-title">Compatibility Issues</p>
 
 Kobold.cpp might not run on computers without AVX/AVX2 support.
@@ -98,7 +98,7 @@ Kobold.cpp allows you to modify parameters such as the AI model temperature and
 
 ![Ollama Logo](assets/img/ai-chat/ollama.png){align=right}
 
-Ollama is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
+**Ollama** is a command-line AI assistant that is available on macOS, Linux, and Windows. Ollama is a great choice if you're looking for an AI client that's easy-to-use, widely compatible, and fast due to its use of inference and other techniques. It also doesn't involve any manual setup.
 
 In addition to supporting a wide range of text models, Ollama also supports [LLaVA](https://github.com/haotian-liu/LLaVA) models and has experimental support for Meta's [Llama vision capabilities](https://huggingface.co/blog/llama32#what-is-llama-32-vision).
 
@@ -124,9 +124,9 @@ Ollama simplifies the process of setting up a local AI chat by downloading the A
 
 <div class="admonition recommendation" markdown>
 
-![Llamafile Logo](assets/img/ai-chat/llamafile.svg){align=right}
+![Llamafile Logo](assets/img/ai-chat/llamafile.png){align=right}
 
-Llamafile is a lightweight single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
+**Llamafile** is a lightweight, single-file executable that allows users to run LLMs locally on their own computers without any setup involved. It is [backed by Mozilla](https://hacks.mozilla.org/2023/11/introducing-llamafile) and available on Linux, macOS, and Windows.
 
 Llamafile also supports LLaVA. However, it doesn't support speech recognition or image generation.
 
@@ -138,7 +138,9 @@ Llamafile also supports LLaVA. However, it doesn't support speech recognition or
 <details class="downloads" markdown>
 <summary>Downloads</summary>
 
-- [:fontawesome-solid-desktop: Desktop](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:fontawesome-brands-windows: Windows](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-apple: macOS](https://github.com/Mozilla-Ocho/llamafile#quickstart)
+- [:simple-linux: Linux](https://github.com/Mozilla-Ocho/llamafile#quickstart)
 
 </details>
 
@@ -171,11 +173,11 @@ Please note we are not affiliated with any of the projects we recommend. In addi
 
 ### Minimum Gereksinimler
 
-- Must be open-source.
+- Must be open source.
 - Must not transmit personal data, including chat data.
 - Must be multi-platform.
 - Must not require a GPU.
-- Must support GPU-powered fast inference.
+- Must support GPU-powered, fast inference.
 - Must not require an internet connection.
 
 ### En İyi Durum
@@ -186,4 +188,11 @@ Our best-case criteria represent what we _would_ like to see from the perfect pr
 - Should have a built-in model downloader option.
 - The user should be able to modify the LLM parameters, such as its system prompt or temperature.
 
+\*[LLaVA]: Large Language and Vision Assistant (multimodal AI model)
+\*[LLM]: Large Language Model (AI model such as ChatGPT)
+\*[LLMs]: Large Language Models (AI models such as ChatGPT)
+\*[open-weights models]: AI models that anyone can download and use, but the underlying training data and/or algorithms for them are proprietary.
+\*[system prompt]: The general instructions given by a human to guide how an AI chat should operate.
+\*[temperature]: A parameter used in AI models to control the level of randomness and creativity in the generated text.
+
 [^1]: A file checksum is a type of anti-tampering fingerprint. A developer usually provides a checksum in a text file that can be downloaded separately, or on the download page itself. Verifying that the checksum of the file you downloaded matches the one provided by the developer helps ensure that the file is genuine and wasn't tampered with in transit. You can use commands like `sha256sum` on Linux and macOS, or `certutil -hashfile file SHA256` on Windows to generate the downloaded file's checksum.

i18n/tr/basics/email-security.md

@@ -19,7 +19,7 @@ Even if you use OpenPGP, it does not support [forward secrecy](https://en.wikipe
 
 ## What is the Web Key Directory standard?
 
-The Web Key Directory (WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
+The [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD) standard allows email clients to discover the OpenPGP key for other mailboxes, even those hosted on a different provider. Email clients which support WKD will ask the recipient's server for a key based on the email address' domain name. For example, if you emailed `jonah@privacyguides.org`, your email client would ask `privacyguides.org` for Jonah's OpenPGP key, and if `privacyguides.org` has a key for that account, your message would be automatically encrypted.
 
 In addition to the [email clients we recommend](../email-clients.md) which support WKD, some webmail providers also support WKD. Whether *your own* key is published to WKD for others to use depends on your domain configuration. If you use an [email provider](../email.md#openpgp-compatible-services) which supports WKD, such as Proton Mail or Mailbox.org, they can publish your OpenPGP key on their domain for you.
 

i18n/tr/email.md

@@ -34,7 +34,7 @@ In addition to (or instead of) an email provider recommended here, you may wish
 
 ## OpenPGP Uyumlu Hizmetler
 
-These providers natively support OpenPGP encryption/decryption and the [Web Key Directory standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Örneğin, bir Proton Mail kullanıcısı bir Mailbox.org kullanıcısına bir uçtan uca şifreli mesaj gönderebilir, veya bunu destekleyen internet hizmetlerinden OpenPGP şifreli bildirimler alabilirsiniz.
+These providers natively support OpenPGP encryption/decryption and the [Web Key Directory (WKD) standard](basics/email-security.md#what-is-the-web-key-directory-standard), allowing for provider-agnostic E2EE emails. Örneğin, bir Proton Mail kullanıcısı bir Mailbox.org kullanıcısına bir uçtan uca şifreli mesaj gönderebilir, veya bunu destekleyen internet hizmetlerinden OpenPGP şifreli bildirimler alabilirsiniz.
 
 <div class="grid cards" markdown>
 
@@ -107,7 +107,7 @@ Certain information stored in [Proton Contacts](https://proton.me/support/proton
 
 #### :material-check:{ .pg-green } E-posta Şifreleme
 
-Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Diğer Proton Mail hesaplarına gönderilen e-postalar otomatik olarak şifrelenir ve Proton Mail dışındaki adresler için OpenPGP anahtarıyla şifreleme, hesap ayarlarından kolayca etkinleştirilebilir. Proton also supports automatic external key discovery with [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
+Proton Mail has [integrated OpenPGP encryption](https://proton.me/support/how-to-use-pgp) in their webmail. Diğer Proton Mail hesaplarına gönderilen e-postalar otomatik olarak şifrelenir ve Proton Mail dışındaki adresler için OpenPGP anahtarıyla şifreleme, hesap ayarlarından kolayca etkinleştirilebilir. Proton also supports automatic external key discovery with WKD. This means that emails sent to other providers which use WKD will be automatically encrypted with OpenPGP as well, without the need to manually exchange public PGP keys with your contacts. They also allow you to [encrypt messages to non-Proton Mail addresses without OpenPGP](https://proton.me/support/password-protected-emails), without the need for them to sign up for a Proton Mail account.
 
 Proton Mail also publishes the public keys of Proton accounts via HTTP from their WKD. This allows people who don't use Proton Mail to find the OpenPGP keys of Proton Mail accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Proton's own domains, like @proton.me. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
@@ -164,7 +164,7 @@ However, [Open-Exchange](https://en.wikipedia.org/wiki/Open-Xchange), the softwa
 
 Mailbox.org has [integrated encryption](https://kb.mailbox.org/en/private/e-mail-article/send-encrypted-e-mails-with-guard) in their webmail, which simplifies sending messages to people with public OpenPGP keys. They also allow [remote recipients to decrypt an email](https://kb.mailbox.org/en/private/e-mail-article/my-recipient-does-not-use-pgp) on Mailbox.org's servers. This feature is useful when the remote recipient does not have OpenPGP and cannot decrypt a copy of the email in their own mailbox.
 
-Mailbox.org also supports the discovery of public keys via HTTP from their [Web Key Directory (WKD)](https://wiki.gnupg.org/WKD). This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
+Mailbox.org also supports the discovery of public keys via HTTP from their WKD. This allows people outside of Mailbox.org to find the OpenPGP keys of Mailbox.org accounts easily, for cross-provider E2EE. This only applies to email addresses ending in one of Mailbox.org's own domains, like @mailbox.org. If you use a custom domain, you must [configure WKD](./basics/email-security.md#what-is-the-web-key-directory-standard) separately.
 
 #### :material-information-outline:{ .pg-blue } Account Termination
 
@@ -323,7 +323,7 @@ We regard these features as important in order to provide a safe and optimal ser
 
 - Encrypts all account data (Contacts, Calendars, etc.) at rest with zero-access encryption.
 - Integrated webmail E2EE/PGP encryption provided as a convenience.
-- Support for [WKD](https://wiki.gnupg.org/WKD) to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
+- Support for WKD to allow improved discovery of public OpenPGP keys via HTTP. GnuPG users can get a key by typing: `gpg --locate-key example_user@example.com`
 - Support for a temporary mailbox for external users. This is useful when you want to send an encrypted email, without sending an actual copy to your recipient. These emails usually have a limited lifespan and then are automatically deleted. They also don't require the recipient to configure any cryptography like OpenPGP.
 - Availability of the email provider's services via an [onion service](https://en.wikipedia.org/wiki/.onion).
 - [Sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) support.

i18n/tr/tools.md

@@ -351,8 +351,8 @@ Diğerlerinin yanı sıra [Mullvad](https://mullvad.net/en/help/dns-over-https-a
 
 <div class="grid cards" markdown>
 
-- ![Kobold logosu](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
-- ![Llamafile logo](assets/img/ai-chat/llamafile.svg){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
+- ![Kobold logo](assets/img/ai-chat/kobold.png){ .twemoji loading=lazy } [Kobold.cpp](ai-chat.md#koboldcpp)
+- ![Llamafile logo](assets/img/ai-chat/llamafile.png){ .twemoji loading=lazy } [Llamafile](ai-chat.md#llamafile)
 - ![Ollama logo](assets/img/ai-chat/ollama.png){ .twemoji loading=lazy } [Ollama (CLI)](ai-chat.md#ollama-cli)
 
 </div>