Rhea

Rhea is a local assistant built around a llama.cpp-compatible chat server.

Project layout:

• main.py starts the HTTP server and exists as the stable backend entrypoint.
• utils/chat_server.py owns the FastAPI /chat endpoint, LLM streaming, tool execution loop, session history, memory context, and keepalive.
• terminal_client.py owns the terminal chat UI.
• utils/settings.py loads .env values and shared model-selection config.
• utils/tool_call_strategy.py builds the system prompt and parses THINK / CALL tool requests.
• tools/ contains callable assistant tools and their argument schemas.

Run:

uv run rhea-server
uv run rhea-chat

Configuration lives in .env. Copy .env.example when setting up a new checkout.