Phaedrix

Multi-agent AI system with a supervisor router that classifies each message and dispatches it to a specialized agent — Research, Code, Data, or Utility — backed by multi-provider LLM fallback (Groq → Gemini → OpenRouter).

Local setup

pip install -r requirements.txt
cp .env.example .env   # fill in at least one LLM key + TAVILY_API_KEY
uvicorn backend.main:app --reload

Visit http://127.0.0.1:8000.

Environment variables

Variable	Required	Used for
GROQ_API_KEY	one of the three	LLM provider
GEMINI_API_KEY	one of the three	LLM provider
OPENROUTER_API_KEY	one of the three	LLM provider
TAVILY_API_KEY	yes	Research Agent web search

At least one LLM provider key is required; more keys means more fallback resilience.

Deploy on Render

1. Push this repo to GitHub.
2. In Render: New → Blueprint, point at the repo — render.yaml configures the service automatically.
3. Set the env vars above in the Render dashboard (they're marked sync: false so Render will prompt for them).
4. Deploy. Health check: GET /health.

Note: chat history is stored as local JSON files under backend/memory/chats/ and the in-process session store is in memory — both reset on every redeploy or restart. Fine for a demo; swap in a database (Postgres/Redis) if you need persistence across deploys.