Add Kobold.cpp as a local AI

And simplifications and clarifications

Signed-off-by: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
This commit is contained in:
TechFanTheo 2024-04-23 23:33:23 +00:00 committed by GitHub
parent 2c37ce1ec5
commit 301f75f0e4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 42 additions and 17 deletions

View File

@ -15,14 +15,14 @@ The recommendations here do not train their models using your chats, do not reta
<div class="admonition recommendation" markdown>
![DuckDuckGo logo](assets/img/search-engines/duckduckgo.svg){align=right}
!DuckDuckGo logo{align=right}
**DuckDuckGo AI Chat** is made by the popular [seach engine provider] (https://www.privacyguides.org/en/search-engines) and is available directly in the search engine.
DuckDuckGo AI Chat is made by the popular [seach engine provider] (https://www.privacyguides.org/en/search-engines) and is available directly in it.
[octicons-home-16: Homepage](https://duckduckgo.com/?q=Chat&ia=chat)
[:simple-torbrowser:](https://duckduckgogg42xjoc72x3sjasowoarfbgcmvfimaftt6twagswzczad.onion/?q=Chat&ia=chat){ .card-link title="Onion Service" }
[:octicons-eye-16:](https://duckduckgo.com/aichat/privacy-terms){ .card-link title="Privacy Policy" }
[:octicons-info-16:](https://help.duckduckgo.com){ .card-link title=Documentation}
octicons-home-16: Homepage
:simple-torbrowser:{ .card-link title="Onion Service" }
:octicons-eye-16:{ .card-link title="Privacy Policy" }
:octicons-info-16:{ .card-link title=Documentation}
</details>
@ -40,34 +40,59 @@ Privacy : DuckDuckGo has an agreement with the companies ensuring your prompts a
<div class="admonition recommendation" markdown>
![Brave Logo](assets/img/browsers/brave.svg){align=right}
!Brave Logo{align=right}
**Brave Leo** is an AI assistant available in the Brave browser, a browser which we recommend.
Brave Leo is an AI assistant available in Brave, a web browser which we recommend.
[octicons-home-16: Homepage](https://brave.com/leo)
[:octicons-eye-16:](https://brave.com/privacy/browser/#brave-leo){ .card-link title="Privacy Policy" }
[:octicons-info-16:](https://github.com/brave/brave-browser/wiki/Brave-Leo){ .card-link title=Documentation}
octicons-home-16: Homepage
:octicons-eye-16:{ .card-link title="Privacy Policy" }
:octicons-info-16:{ .card-link title=Documentation}
</details>
</div>
Models : Antrophic Claude Instant, Mixtral 8x7b and Llama 13b.
Type : Claude is run by Anthropic but proxied trough Brave. Other models are run by Brave directly.
Type : Claude is run by Anthropic but proxied trough Brave. Other models are run by Brave.
Availability: only available in the Brave Browser.
Rate limit : Unlimited for Llama 13b, and heavily limited for the rest.
Privacy : For in-house models, all chats are discarded after session is ended, and chats aren't used for training. For Claude, all data is retained for 30 days but 'personal data' isn't used for training.
### Kobold.cpp
<div class="admonition recommendation" markdown>
!Kobold.cpp Logo{align=right}
Kobold.cpp is an AI client that runs locally on your computer.
:octicons-home-16: Homepage
:octicons-server-16:{ .card-link title="Security Policy"}
:octicons-info-16:{ .card-link title=Documentation}
:octicons-code-16:{ .card-link title="Source Code" }
:octicons-heart-16:{ .card-link title=Contribute }
</details>
</div>
Kobold.cpp allows you to easily run an AI model right on your computer. You simply select your model files and click run. A chat interface will open in your browser.Since it is run locally, you can submit it sensitive information.
To run performant models, you will need at least 16GB of RAM. You do not need a GPU. You can download quantitised models, which are more efficient but slightly less performant. Those models are available in a GGUF format, and can be downloaded at [Hugging Face] (https://huggingface.co/models?sort=trending&search=.GGUF).
Models : Llama 1,2,3, Mistral, Phi-2, and a few others.
Type : Local model
Availability : Linux,Windows.
Rate limit : None.
Privacy : Total
## Criteria
**Please note we are not affiliated with any of the projects we recommend.** In addition to [our standard criteria](about/criteria.md), we have developed a clear set of requirements to allow us to provide objective recommendations. We suggest you familiarize yourself with this list before choosing to use a project, and conduct your own research to ensure it's the right choice for you.
Please note we are not affiliated with any of the projects we recommend. In addition to our standard criteria, we have developed a clear set of requirements to allow us to provide objective recommendations. We suggest you familiarize yourself with this list before choosing to use a project, and conduct your own research to ensure it's the right choice for you.
### Minimum Requirements
- Must not use your chats for training by default.
- Must be accessible with a VPN or be self-hostable.
- Must not use your chats for training.
- Must be accesible over a VPN
- Must provide transparency about which models is used and who operates it.
- Must have a reasonable rate limit (*ask feedback*).
- Online models must be accessible on a website. An exception is made if it require a [web browser](https://privacyguides.org/en/desktop-browsers/) that we recommend.
### Best-Case
@ -75,5 +100,5 @@ Our best-case criteria represents what we would like to see from the perfect pro
- Should be accessible over Tor.
- Should not require JavaScript.
- Providers should self-host open-source models.
- Providers should only have open-source models.
- No rate limit