BACKEND: AI Agent Proxy Layer

[tosoham]

Build the interface for LLM-powered agents:

• Integrate Ollama (local Llama 2, Mistral, etc.) via its HTTP/CLI API

• Expose POST /agent/response that accepts { prompt, roomId } and returns AI text

• Add retry/fallback logic if Ollama is unavailable

Deliverable: Reliable API to “talk” to your AI models.

See:

Ollama docs: https://ollama.com/docs

llama.cpp overview: https://github.com/ggerganov/llama.cpp