!Kobold.cpp Logo{align=right}
@@ -102,3 +102,4 @@ Our best-case criteria represents what we would like to see from the perfect pro
- Should not require JavaScript.
- Providers should only have open-source models.
- No rate limit
+- Doesn't require expensive hardware
From 28ed9a8aac23b2413b8920e8b4337565a67715f1 Mon Sep 17 00:00:00 2001
From: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
Date: Wed, 24 Apr 2024 12:19:51 +0000
Subject: [PATCH 004/160] Add warnings, fix typos and missing links
Signed-off-by: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
---
docs/ai-chatbots.md | 39 +++++++++++++++++++++++----------------
1 file changed, 23 insertions(+), 16 deletions(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index b6f359d86a..6248f2dfed 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -9,18 +9,16 @@ Since the release of ChatGPT by OpenAI in December 2022, talking with Largue Lan
## Recommended Providers
-The recommendations here do not train their models using your chats, do not retain your chats for more than 30 days, and are accessible privately. However, you should not submit sensitive PII to an online provider.
-
-### DuckDuckGo AI Chat
+The recommendations for online providers here do not train their models using your chats and do not retain your chats for more than 30 days,based on each service's privacy policy. There is no guarantee that these privacy policies are honored.. and are accessible privately. Only submit personally identifiable information using a local model.
{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 the search engine. It is available on a Tor hidden site.
-[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-home-16: Homepage](https://duckduckgo.com/aichat)
+[:simple-torbrowser:](https://duckduckgogg42xjoc72x3sjasowoarfbgcmvfimaftt6twagswzczad.onion/aichat){ .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}
@@ -34,7 +32,11 @@ Availability : Any browser with JavaScript. Works with Tor.
Rate limit : none
Privacy : DuckDuckGo has an agreement with the companies ensuring your prompts and outputs aren't used for improving their models. The agreement also states the providers must delete all chats within 30 days.
-
+
+
Your chats are kept for 30 days
+DuckDuckGo uses OpenAI anc CLaude servers and they will retain your chats for 30 days. Use a local model if your prompts contains PII.
+
+
### Brave Leo AI
@@ -57,6 +59,11 @@ 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.
+
+
For one model, chats are kept for 30 days
+We recommend against using Anthropic's Claude Instant becuase they will keep a chat history for 30 days.
+
+
### Kobold.cpp
!Kobold.cpp Logo{align=right}
@@ -65,11 +72,11 @@ 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 }
+[:octicons-home-16: Homepage](https://github.com/LostRuins/koboldcpp)
+[:octicons-server-16:](https://github.com/LostRuins/koboldcpp/blob/2f3597c29abea8b6da28f21e714b6b24a5aca79b/SECURITY.md){ .card-link title="Security Policy"}
+[:octicons-info-16:](https://github.com/LostRuins/koboldcpp/wiki){ .card-link title=Documentation}
+[:octicons-code-16:](https://github.com/LostRuins/koboldcpp){ .card-link title="Source Code" }
+[:octicons-heart-16:](){ .card-link title=Contribute }
@@ -79,7 +86,7 @@ To run performant models, you will need at least 16GB of RAM. You do not need a
Models : Llama 1,2,3, Mistral, Phi-2, and a few others.
Type : Local model
-Availability : Linux,Windows.
+Availability : Linux, Windows.
Rate limit : None.
Privacy : Total
@@ -90,8 +97,8 @@ Please note we are not affiliated with any of the projects we recommend. In addi
### Minimum Requirements
- 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 be accessible over a VPN
+- Must provide transparency about models operators and privacy policies.
- Must have a reasonable rate limit (*ask feedback*).
### Best-Case
@@ -102,4 +109,4 @@ Our best-case criteria represents what we would like to see from the perfect pro
- Should not require JavaScript.
- Providers should only have open-source models.
- No rate limit
-- Doesn't require expensive hardware
+- Shouldn't require expensive hardware
From f9db9955d3b91c7949394a0035884f7ddb4e338d Mon Sep 17 00:00:00 2001
From: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
Date: Wed, 24 Apr 2024 12:35:43 +0000
Subject: [PATCH 005/160] Update docs/ai-chatbots.md
Co-authored-by: rollsicecream
Signed-off-by: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
---
docs/ai-chatbots.md | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index 6248f2dfed..bb3a1a46a9 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -1,7 +1,7 @@
---
meta_title: "Recommended AI Chatbots : Private ChatGPT Alternatives - Privacy Guides"
title : "AI Chatbots"
-icon: file/file
+icon: :material-chat-processing-outline:
description: Unlike OpenAI ChatGPT and it's successors, those AI Chatbots do not train their models using your conversations.
cover:file/file
---
From 5d1aa46c4c171b2da6185051157f589e969c0246 Mon Sep 17 00:00:00 2001
From: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
Date: Wed, 24 Apr 2024 12:48:15 +0000
Subject: [PATCH 006/160] Update ai-chatbots.md
Signed-off-by: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
---
docs/ai-chatbots.md | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index bb3a1a46a9..60e98d6ff5 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -11,6 +11,7 @@ Since the release of ChatGPT by OpenAI in December 2022, talking with Largue Lan
The recommendations for online providers here do not train their models using your chats and do not retain your chats for more than 30 days,based on each service's privacy policy. There is no guarantee that these privacy policies are honored.. and are accessible privately. Only submit personally identifiable information using a local model.
+##DuckDuckGo AI Chat
{align=right}
@@ -66,7 +67,7 @@ We recommend against using Anthropic's Claude Instant becuase they will keep a c
### Kobold.cpp
-!Kobold.cpp Logo{align=right}
+!Kobold.cpp Logo(https://discord.com/invite/XuQWadgU9k){align=right}
Kobold.cpp is an AI client that runs locally on your computer.
From bd11c1a55d6e04af6889f7725f80c59cb4526110 Mon Sep 17 00:00:00 2001
From: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
Date: Sat, 27 Apr 2024 12:18:10 +0800
Subject: [PATCH 007/160] Fix some issues
Note: icon might need to be fixed but having "icon: :material-chat:" was breaking the code due to the double colons (:)
Signed-off-by: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
---
docs/ai-chatbots.md | 63 ++++++++++++++++++++++-----------------------
1 file changed, 31 insertions(+), 32 deletions(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index 60e98d6ff5..b2d22fdf42 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -1,24 +1,24 @@
---
meta_title: "Recommended AI Chatbots : Private ChatGPT Alternatives - Privacy Guides"
-title : "AI Chatbots"
-icon: :material-chat-processing-outline:
-description: Unlike OpenAI ChatGPT and it's successors, those AI Chatbots do not train their models using your conversations.
-cover:file/file
+title: "AI Chats"
+icon: material-chat-processing-outline
+description: Unlike OpenAI ChatGPT and it's successors, those AI Chats do not train their models using your conversations.
+cover: file/file
---
Since the release of ChatGPT by OpenAI in December 2022, talking with Largue Language Models (LLMs) has become common. It can help us write better, understand unfamiliar subjects or answer a wide range of questions. However, the developpers of those models need data to improve their model. Unfortunately, this include your conversations with the AI.
## Recommended Providers
-The recommendations for online providers here do not train their models using your chats and do not retain your chats for more than 30 days,based on each service's privacy policy. There is no guarantee that these privacy policies are honored.. and are accessible privately. Only submit personally identifiable information using a local model.
+The recommendations for online providers here do not train their models using your chats and do not retain your chats for more than 30 days, based on each service's privacy policy. There is no guarantee that these privacy policies are honored. Only submit personally identifiable information using a local model.
-##DuckDuckGo AI Chat
+### DuckDuckGo AI Chat
{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. It is available on a Tor hidden site.
-[octicons-home-16: Homepage](https://duckduckgo.com/aichat)
+[:octicons-home-16: Homepage](https://duckduckgo.com/aichat)
[:simple-torbrowser:](https://duckduckgogg42xjoc72x3sjasowoarfbgcmvfimaftt6twagswzczad.onion/aichat){ .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}
@@ -27,15 +27,15 @@ The recommendations for online providers here do not train their models using yo
-Models: Anthropic Claude 1.2 and Open AI GPT 3-5
-Type : third-party providers proxied trough DDG servers
-Availability : Any browser with JavaScript. Works with Tor.
+-Models: Anthropic Claude 1.2 and Open AI GPT 3-5
+-Type : third-party providers proxied trough DDG servers
+-Availability : Any browser with JavaScript. Works with Tor.
Rate limit : none
-Privacy : DuckDuckGo has an agreement with the companies ensuring your prompts and outputs aren't used for improving their models. The agreement also states the providers must delete all chats within 30 days.
+-Privacy : DuckDuckGo has an agreement with the companies ensuring your prompts and outputs aren't used for improving their models. The agreement also states the providers must delete all chats within 30 days.
Your chats are kept for 30 days
-DuckDuckGo uses OpenAI anc CLaude servers and they will retain your chats for 30 days. Use a local model if your prompts contains PII.
+DuckDuckGo act as proxy between you and OpenAI or Claude. However, those two companies will retain your chats for 30 days. Use a local model if your prompts contains PII, and Tor if it contains sensitive information.
### Brave Leo AI
@@ -44,52 +44,51 @@ DuckDuckGo uses OpenAI anc CLaude servers and they will retain your chats for 30
{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 the Brave browser, a browser which we recommend. It can analyse the context of the web page you are visiting, but we advise you to disable it on web pages with your PII.
-[octicons-home-16: Homepage](https://brave.com/leo)
+[: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}
-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.
-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.
+-Models : Antrophic Claude Instant, Mixtral 8x7b and Llama 2 13b.
+-Type : Claude is run by Anthropic but proxied trough Brave. Other models are hosted 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.
-
For one model, chats are kept for 30 days
+
Claude chats are kept for 30 days
We recommend against using Anthropic's Claude Instant becuase they will keep a chat history for 30 days.
### Kobold.cpp
-!Kobold.cpp Logo(https://discord.com/invite/XuQWadgU9k){align=right}
+{align=right}
Kobold.cpp is an AI client that runs locally on your computer.
[:octicons-home-16: Homepage](https://github.com/LostRuins/koboldcpp)
-[:octicons-server-16:](https://github.com/LostRuins/koboldcpp/blob/2f3597c29abea8b6da28f21e714b6b24a5aca79b/SECURITY.md){ .card-link title="Security Policy"}
+[:octicons-eye-16:](https://github.com/LostRuins/koboldcpp/blob/2f3597c29abea8b6da28f21e714b6b24a5aca79b/SECURITY.md){ .card-link title="Security Policy"}
[:octicons-info-16:](https://github.com/LostRuins/koboldcpp/wiki){ .card-link title=Documentation}
[:octicons-code-16:](https://github.com/LostRuins/koboldcpp){ .card-link title="Source Code" }
-[:octicons-heart-16:](){ .card-link title=Contribute }
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).
+Using quantitised model, you can except models to consume between 0.5GB and 2GB of RAM for every billion paramater. Since the minimum for a quality model is 8B, it is strongly recommended to have at least 12GB of RAM. You can download those quantised models in a format called GGUF, 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
+-Models : Llama 1,2,3, Mistral, Phi-2, and a few others.
+-Type : Local model
+-Availability : Linux, Windows
+-Rate limit : None
+-Privacy : Total
## Criteria
@@ -100,7 +99,7 @@ Please note we are not affiliated with any of the projects we recommend. In addi
- Must not use your chats for training.
- Must be accessible over a VPN
- Must provide transparency about models operators and privacy policies.
-- Must have a reasonable rate limit (*ask feedback*).
+- Must provide at least one model whith an high rate limit that allows you to use it for work.
### Best-Case
@@ -109,5 +108,5 @@ 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 only have open-source models.
-- No rate limit
-- Shouldn't require expensive hardware
+- Should not be rate-limited
+- Should not require expensive hardware
From 7540dc182e357b4833e322dd3726a39d388f4f04 Mon Sep 17 00:00:00 2001
From: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
Date: Tue, 21 May 2024 16:41:23 +0200
Subject: [PATCH 008/160] Make it readble for humans
Replaced the bullet points lists with actual constructed sentences.
Signed-off-by: TechFanTheo <78900789+TechFanTheo@users.noreply.github.com>
---
docs/ai-chatbots.md | 32 +++++++++++---------------------
1 file changed, 11 insertions(+), 21 deletions(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index b2d22fdf42..e5a64b19b6 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -16,7 +16,7 @@ The recommendations for online providers here do not train their models using yo
{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. It is available on a Tor hidden site.
+**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.
[:octicons-home-16: Homepage](https://duckduckgo.com/aichat)
[:simple-torbrowser:](https://duckduckgogg42xjoc72x3sjasowoarfbgcmvfimaftt6twagswzczad.onion/aichat){ .card-link title="Onion Service" }
@@ -27,15 +27,11 @@ The recommendations for online providers here do not train their models using yo
Your chats are kept for 30 days
-DuckDuckGo act as proxy between you and OpenAI or Claude. However, those two companies will retain your chats for 30 days. Use a local model if your prompts contains PII, and Tor if it contains sensitive information.
+DuckDuckGo and the providers can have access to your chats. Use a local model if your prompts contains PII, and Tor if it contains sensitive information.
### Brave Leo AI
@@ -44,22 +40,21 @@ DuckDuckGo act as proxy between you and OpenAI or Claude. However, those two com
{align=right}
-**Brave Leo** is an AI assistant available in the Brave browser, a browser which we recommend. It can analyse the context of the web page you are visiting, but we advise you to disable it on web pages with your PII.
-
+**Brave Leo** is an AI assistant available inside the Brave browser, a 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}
+Brave Leo supports a range of models, including open-source models from Meta and Mixtral and closed-source models from Anthropic. Brave host the open models and do not retain chat or train on the data from those models. For Claude models, Brave proxies your chats through their servers,and the agreement between Brave and Anthropic states that chats are to be deleted after 30 days and that 'personal data' isn't used for training. Leo AI is available on the Brave Browser on all platforms. There is a strict rate limit for most models, but there is 1 free model whith an high-rate limit.
--Models : Antrophic Claude Instant, Mixtral 8x7b and Llama 2 13b.
--Type : Claude is run by Anthropic but proxied trough Brave. Other models are hosted 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.
+
+
Pages content are sent by default
+Brave will send the page content to their severs by default, and there is no way to permanently disable it. However, we recommend manually switching off the "Use page context for response" toggle for pages with PII.
+
Claude chats are kept for 30 days
We recommend against using Anthropic's Claude Instant becuase they will keep a chat history for 30 days.
@@ -80,15 +75,10 @@ Kobold.cpp is an AI client that runs locally on your computer.
+-You can run models from Meta, Mistral, Microsoft (Phi-2) and others locally on your Windows or Linux PC. Since your data isn't leaving your PC, you don't have to worry about sharing sensitive PII.
+If you use a quantitised model, you can except models to consume between 0.5GB and 2GB of RAM for every billion paramater. Since the minimum parameters count for a satisfactory model is 8B, you should have at least 12gb of RAM. You can download quantised models in a format called GGUF, at [Hugging Face] (https://huggingface.co/models?sort=trending&search=.GGUF).
-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.
-Using quantitised model, you can except models to consume between 0.5GB and 2GB of RAM for every billion paramater. Since the minimum for a quality model is 8B, it is strongly recommended to have at least 12GB of RAM. You can download those quantised models in a format called GGUF, 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
From 83d004cf2fd4e56c6f13f4a495cdf1feda935656 Mon Sep 17 00:00:00 2001
From: Triple T <78900789+I-I-IT@users.noreply.github.com>
Date: Tue, 4 Jun 2024 20:17:07 +0200
Subject: [PATCH 009/160] Revamp of criterias and add Ollama
I changed the criteria so they are way more strict. Basically, the baseline is now only short retention of your chats, with a certainty that third-party will respect this. DuckDuckGo will be removed unless they clarify whether their open-source models satisfy criterias.
Signed-off-by: Triple T <78900789+I-I-IT@users.noreply.github.com>
---
docs/ai-chatbots.md | 55 ++++++++++++++++++++++++++++++++-------------
1 file changed, 40 insertions(+), 15 deletions(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index e5a64b19b6..aafba666b9 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -9,7 +9,7 @@ Since the release of ChatGPT by OpenAI in December 2022, talking with Largue Lan
## Recommended Providers
-The recommendations for online providers here do not train their models using your chats and do not retain your chats for more than 30 days, based on each service's privacy policy. There is no guarantee that these privacy policies are honored. Only submit personally identifiable information using a local model.
+The recommendations for online providers here do not train their models using your chats and do not retain your chats for more than 14 days, based on each service's privacy policy. There is no guarantee that these privacy policies are honored. Only submit personally identifiable information using a local model.
### DuckDuckGo AI Chat
@@ -27,7 +27,7 @@ The recommendations for online providers here do not train their models using yo
-DuckDuckGo offers proprietary models from by Anthropic and Open AI. To protect your privacy, DuckDuckGo proxies your chats through their servers, and the agreement between DuckDuckGo and OpenAI and Antrhopic states that your chats aren't to be used for training and be deleted after 30 days. It is available directly on the (DuckDuckGo)[duckduckgo.com] search engine, including on the Tor hidden site. The later guarantees that you have "unlimited" rate limits for the use of the chatbot.
+DuckDuckGo offers proprietary models from Anthropic and Open AI, as well as open-source models from Meta and Mixtral. To protect your privacy, DuckDuckGo proxies your chats through their servers, and the agreement between DuckDuckGo and the providers states that your chats aren't to be used for training and be deleted after 30 days, with " limited exceptions for safety and legal compliance". Privacy Guides is seeking greater clarity on the exact agreements. It is available directly on the (DuckDuckGo)[duckduckgo.com] search engine, including on the Tor hidden site. The later guarantees that you have "unlimited" rate limits for the use of the chatbot.
Your chats are kept for 30 days
@@ -64,39 +64,64 @@ We recommend against using Anthropic's Claude Instant becuase they will keep a c
{align=right}
-Kobold.cpp is an AI client that runs locally on your computer.
+Kobold.cpp is an AI client that runs locally on your Windows or Linux computer.
[:octicons-home-16: Homepage](https://github.com/LostRuins/koboldcpp)
+
[:octicons-eye-16:](https://github.com/LostRuins/koboldcpp/blob/2f3597c29abea8b6da28f21e714b6b24a5aca79b/SECURITY.md){ .card-link title="Security Policy"}
+[:octicons-home-16:](https://github.com/LostRuins/koboldcpp?tab=readme-ov-file#notes) {.card-link title=" Models supported"}
[:octicons-info-16:](https://github.com/LostRuins/koboldcpp/wiki){ .card-link title=Documentation}
[:octicons-code-16:](https://github.com/LostRuins/koboldcpp){ .card-link title="Source Code" }
--You can run models from Meta, Mistral, Microsoft (Phi-2) and others locally on your Windows or Linux PC. Since your data isn't leaving your PC, you don't have to worry about sharing sensitive PII.
-If you use a quantitised model, you can except models to consume between 0.5GB and 2GB of RAM for every billion paramater. Since the minimum parameters count for a satisfactory model is 8B, you should have at least 12gb of RAM. You can download quantised models in a format called GGUF, at [Hugging Face] (https://huggingface.co/models?sort=trending&search=.GGUF).
+-You can run models from Meta, Mistral, Microsoft (Phi-2) and others locally on your Windows or Linux PC. You can even use speak directly to the AI, and it will transcribe it to text.
+Since your data isn't leaving your PC, you don't have to worry about sharing sensitive PII.
+You should use a quantitised model of at least 7-8B, which means your computer needs at least 8gb of RAM. You can download quantised models in a format called GGUF, at [Hugging Face] (https://huggingface.co/models?sort=trending&search=.GGUF).
+### Ollama
+
+{align=right}
+Ollama is a command line AI assitant that is available on Mac, Linux and Windows(preview).
+[:octicons-home-16: Homepage](https://github.com/ollama/ollama)
+[:octicons-home-16:](https://ollama.com/library) {.card-link title=" Models supported"}
+[:octicons-info-16:](https://github.com/ollama/ollama){ .card-link title=Documentation}
+[:octicons-code-16:](https://github.com/ollama/ollama){ .card-link title="Source Code"}
+
+
+You can run models from Meta, Mistral, Microsoft. You can also use Llava, a model which can analyse image. Since your data isn't leaving your PC, you don't have to worry about sharing sensitive PII.
+You should use a quantitised model of at least 7-8B, which means your computer needs at least 8gb of RAM. You can download quantised models in a format called GGUF, at [Hugging Face] (https://huggingface.co/models?sort=trending&search=.GGUF).
## Criteria
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
+ For cloud providers :
+ - Must not use your chats for training
+ - Must not retain your chats for more than 14 days
+ - Must be accessible privately (no account required, accepts request from VPN users)
+ - Must provide models they host themselves, or with a third-party that acts in their behalf.
+ - Must provide at least one model whith an high rate limit
+ For local models :
+ - Must not send personal data. Any opt-in mechanism must rely on free consent and avoid dark patern.
+ - Must not require a GPU
+ - Must have GPU support for fast inference
+ - Must not require internet connection
-- Must not use your chats for training.
-- Must be accessible over a VPN
-- Must provide transparency about models operators and privacy policies.
-- Must provide at least one model whith an high rate limit that allows you to use it for work.
### Best-Case
Our best-case criteria represents what we would like to see from the perfect project in this category. Our recommendations may not include any or all of this functionality, but those which do may rank higher than others on this page.
-
-- Should be accessible over Tor.
-- Should not require JavaScript.
-- Providers should only have open-source models.
-- Should not be rate-limited
-- Should not require expensive hardware
+For cloud providers :
+ - Should not retain your chat
+ - Should be accessible anonymously (Tor Browser Safest security level)
+ - Providers should only have open-source models that they self-host.
+ - Should not be rate-limited
+For local models :
+ -Should provide one-click installer
+ -Should have a built-in "Download Models" option
+ -Should support advanced features like modifying the sytem prompt, temperature, etc.
From 32a2d17dd883a9ab88dd02fcd1dfc9cbe3718043 Mon Sep 17 00:00:00 2001
From: Triple T <78900789+I-I-IT@users.noreply.github.com>
Date: Tue, 4 Jun 2024 20:33:45 +0200
Subject: [PATCH 010/160] Update ai-chatbots.md
test good build page
Signed-off-by: Triple T <78900789+I-I-IT@users.noreply.github.com>
---
docs/ai-chatbots.md | 14 +++++---------
1 file changed, 5 insertions(+), 9 deletions(-)
diff --git a/docs/ai-chatbots.md b/docs/ai-chatbots.md
index aafba666b9..88aab33d57 100644
--- a/docs/ai-chatbots.md
+++ b/docs/ai-chatbots.md
@@ -23,8 +23,6 @@ The recommendations for online providers here do not train their models using yo
[: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}
-
-
DuckDuckGo offers proprietary models from Anthropic and Open AI, as well as open-source models from Meta and Mixtral. To protect your privacy, DuckDuckGo proxies your chats through their servers, and the agreement between DuckDuckGo and the providers states that your chats aren't to be used for training and be deleted after 30 days, with " limited exceptions for safety and legal compliance". Privacy Guides is seeking greater clarity on the exact agreements. It is available directly on the (DuckDuckGo)[duckduckgo.com] search engine, including on the Tor hidden site. The later guarantees that you have "unlimited" rate limits for the use of the chatbot.
@@ -40,12 +38,12 @@ DuckDuckGo and the providers can have access to your chats. Use a local model if
{align=right}
-**Brave Leo** is an AI assistant available inside the Brave browser, a browser which we recommend.
+**Brave Leo** is an AI assistant available inside the Brave browser, a 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}
-
Brave Leo supports a range of models, including open-source models from Meta and Mixtral and closed-source models from Anthropic. Brave host the open models and do not retain chat or train on the data from those models. For Claude models, Brave proxies your chats through their servers,and the agreement between Brave and Anthropic states that chats are to be deleted after 30 days and that 'personal data' isn't used for training. Leo AI is available on the Brave Browser on all platforms. There is a strict rate limit for most models, but there is 1 free model whith an high-rate limit.
@@ -62,20 +60,17 @@ We recommend against using Anthropic's Claude Instant becuase they will keep a c
### Kobold.cpp
+
{align=right}
Kobold.cpp is an AI client that runs locally on your Windows or Linux computer.
-
-
[:octicons-home-16: Homepage](https://github.com/LostRuins/koboldcpp)
-
[:octicons-eye-16:](https://github.com/LostRuins/koboldcpp/blob/2f3597c29abea8b6da28f21e714b6b24a5aca79b/SECURITY.md){ .card-link title="Security Policy"}
[:octicons-home-16:](https://github.com/LostRuins/koboldcpp?tab=readme-ov-file#notes) {.card-link title=" Models supported"}
[:octicons-info-16:](https://github.com/LostRuins/koboldcpp/wiki){ .card-link title=Documentation}
[:octicons-code-16:](https://github.com/LostRuins/koboldcpp){ .card-link title="Source Code" }
-
-You can run models from Meta, Mistral, Microsoft (Phi-2) and others locally on your Windows or Linux PC. You can even use speak directly to the AI, and it will transcribe it to text.
Since your data isn't leaving your PC, you don't have to worry about sharing sensitive PII.
@@ -86,11 +81,12 @@ You should use a quantitised model of at least 7-8B, which means your computer n
{align=right}
Ollama is a command line AI assitant that is available on Mac, Linux and Windows(preview).
+
[:octicons-home-16: Homepage](https://github.com/ollama/ollama)
[:octicons-home-16:](https://ollama.com/library) {.card-link title=" Models supported"}
[:octicons-info-16:](https://github.com/ollama/ollama){ .card-link title=Documentation}
[:octicons-code-16:](https://github.com/ollama/ollama){ .card-link title="Source Code"}
-
+