tashfeenahmed/freellmapi ? reverse-engineered prompt

Build me a local proxy server that combines the free tiers from around 14 AI providers like Groq, Google Gemini, Cerebras, SambaNova, Mistral, and the rest into one single OpenAI compatible endpoint.

I want to point any normal OpenAI client or library at it with one unified key and get chat completions that automatically route to the best available model. If a provider rate limits or errors it should failover to the next one in my list, keep track of usage per key so I never exceed the free caps, and support streaming responses plus basic tool calling.

Add a simple web dashboard where I can paste in my provider keys, reorder the fallback chain, see analytics on which ones are working, and test prompts in a playground. Store the keys encrypted, use a lightweight database, and make the whole thing easy to run with npm on my laptop or even a Raspberry Pi. Focus on personal use and keep everything clean and reliable.