simhadris17/NEELASTACK ? reverse-engineered prompt

Build me a local first AI engineering platform that runs mostly on my own machine and does not require a paid API key for the default setup.

I want a FastAPI backend, a React and TypeScript web app, and Docker Compose so I can get the whole stack running quickly. It should support chatting through a local Ollama model, with fallback to other providers if keys are added, plus simple voice features for transcription and text to speech. Please include a job system for queued tasks like chat, speech to text, and text to speech, with progress, retries, cancellation, and results. If possible, add basic RAG style local embeddings, shared rate limiting, and a clean API docs page. Make the setup easy to follow, with an example env file and sensible defaults for local development. Use current docs online if you need to check anything.