jaturapornchai/bcproxyai — reverse-engineered prompt

Build me an OpenAI compatible LLM gateway for using only safe free tier AI providers. I want to add my provider API keys in a setup page, see a clear warning not to use accounts with credit cards attached, then call one local or hosted endpoint from normal OpenAI SDK clients.

The gateway should have a simple dashboard, a guide page with copy paste examples, and virtual models like auto, fast, tools, Thai, and consensus. Auto should pick a good free model, avoid dead or slow providers, respect rate limits, and fall back to another provider when something fails. It should support chat completions, models list, embeddings, and text to speech if the providers allow it.

Make it usable locally with no login, and also deployable on a VPS with password auth or Google login. Include Docker setup, environment examples, request validation, basic security headers, and logging so I can see which provider was used and whether it failed.