Documentation & Video Recording HOW TO USE CHATGPT/CLAUDE/GEMINI SUBSCRIPTION WITH WEBBRAIN

[esokullu]

Use a CLI tool like:
https://github.com/AmazingAng/auth2api
https://github.com/router-for-me/CLIProxyAPI
https://github.com/Wei-Shaw/sub2api
which would create a local API

then use that in Providers settings of webbrain.

So we need to document this, or record a video explaining the process.

alteratively we can fork and create our on cli based on these too, if they're mit, and if it's genuinely make the integration smoother.

[alectimison-maker]

I think documenting this workflow is both necessary and interesting. Allowing users to use their existing ChatGPT, Claude, or Gemini subscriptions with WebBrain could significantly lower the barrier to entry, while also helping us understand the real-world compatibility of different subscription proxies.

My suggestion is to start with a small, provider-agnostic approach:

• Keep subscription authentication in a separate local proxy and connect it to WebBrain through an OpenAI-compatible API.
• CLIProxyAPI could be the primary example in the documentation because it is MIT-licensed and supports ChatGPT, Claude, and Gemini. I would avoid forking or bundling a proxy into WebBrain for now.
• WebBrain would only need the local proxy base URL, API key, and model. Upstream OAuth credentials should always remain on the proxy side.
• The documentation should use secure defaults: bind only to 127.0.0.1, disable remote management, generate a random local API key, and clearly warn users not to expose the proxy to a LAN or the public internet.
• We should test model listing, regular chat, streaming responses, and tool calling, and record compatibility differences across Chrome, Firefox, and the three providers.

I suggest making the documentation the source of truth first, then creating a short video from it. WebBrain could also consider adding a generic provider named "Local OpenAI-compatible Proxy" so users do not mistake this for an official subscription integration.

The documentation should also clarify that this is an experimental, community-supported integration that may stop working when provider authentication policies change. Official API keys remain the more stable integration path.

[esokullu]

Agreed @alectimison-maker
either we can add Local OpenAI-compatible Proxy as a new provider
or there are so many local LLM providers, like LM Studio, llama.cpp, ollama etc. they can just pick one that is unused for this task.