Amm1rr/WebAI-to-API — reverse-engineered prompt

Build me a personal research tool that turns the Gemini web chat into a local API I can use from other apps. I want to run a Python web server on my machine, add my browser cookies in a config file, and then send normal chat requests to it like an OpenAI style API.

It should support chat completions, conversation history, system prompts, streaming replies, a models list, a simple Gemini chat endpoint, a translate endpoint, and a Google Generative AI compatible endpoint. Please include the current Gemini model names shown in the docs, with a fast default model and a stronger pro model.

I also want a fallback mode that can use gpt4free providers like ChatGPT, Claude, DeepSeek, Grok, and others if Gemini is not available. Add Docker support, example config files, clear setup instructions, and basic tests. Keep it modular so Gemini, Atlas Cloud, and fallback providers are separate and easy to maintain.