juspay/neurolink — reverse-engineered prompt
Reverse engineered prompt
Build me a TypeScript library and command line tool called NeuroLink that acts like one simple pipe for AI features. I want developers to call one clean API and stream results back token by token, no matter if the request goes to OpenAI, Anthropic, Google, AWS Bedrock, Azure, local models, or other common providers.
It should make provider switching feel easy, support text generation, tool calls, voice, documents, and other stream based AI outputs where possible. Include a professional CLI for testing prompts, checking credentials, picking providers, and running simple examples. Add support for MCP servers and built in tools so apps can connect AI to outside actions without every project writing custom glue.
Please include sensible config, examples, error handling, provider fallback, per request credentials, and memory support such as Redis for production use. Make the developer experience polished, with clear docs, quick start examples, and tests. Look up current provider docs online if you need to.
Want more depth? Deep Reverse