RAG
FastAPI + embeddings + Qdrant or Chroma, with Docker and starter wiring.
Metis scaffolds composable development environments for RAG, Agents and MCP — with Docker, configuration, metadata, diagnostics and safe evolution built in.
$ metis init agent
Project name: my-agent
LLM: openai
Runtime: plain-sdk
State: redis
✓ FastAPI app created
✓ Redis configured
✓ Docker Compose generated
✓ Project metadata created
$ metis add rag
✓ Existing FastAPI reused
✓ Existing LLM reused
✓ Qdrant added
✓ Project updated safely
Choose the building blocks. Metis wires the local environment and gives you a clean starting point.
FastAPI + embeddings + Qdrant or Chroma, with Docker and starter wiring.
FastAPI + LLM runtime + optional Redis state, ready for your tools and behavior.
Python MCP server with stdio or streamable HTTP and a minimal example tool.
Create a working AI development environment from a recipe.
$ metis init agent
$ metis init rag
$ metis init mcp-server
Safely add a capability without rebuilding what already exists.
$ metis add rag
✓ Existing FastAPI reused
✓ Existing LLM reused
✓ Qdrant added
Understand what Metis sees in the current project.
$ metis inspect
Recipe: agent
Capabilities:
✓ fastapi
✓ llm
✓ redis
✓ rag
Validate local tooling, configuration and reachable services.
$ metis doctor
✓ Docker available
✓ Compose valid
✓ App reachable
✓ Qdrant reachable
Start with the environment you need now.
Add new capability without starting over.
See what is configured and detected.
Validate your local environment quickly.
Start from a working development environment instead of wiring infrastructure by hand.
Recipes create environments. Capabilities extend them.
Reuse compatible pieces and preserve what already exists.
Metadata, inspect and doctor make the environment easier to reason about.
metis init agent