Metis (μῆτις) — a distributed cognitive layer over any LLM. Multi-agent reasoning orchestrator in the alexar76 AIMarket ecosystem.
A single LLM makes one interpretation of your task. Metis wraps any model with:
- Understanding Council — parallel diverse agents → structured
TaskSpec - DGPD — Disagreement-Gated Pipeline Depth (skip expensive layers on agreement)
- Layered MoA — propose → refine → aggregate
- Agent loop — plan, act, observe, reflect with tools + MCP
- Verifier — judge checks answer against task contract
- Economy — usage metering, cost attribution, budget gates
- OpenAI-compatible API —
POST /v1/chat/completionsfor IDE integration
| Page | Description |
|---|---|
| Architecture | System design and per-module model config |
| Quick Start | Install, CLI, first query |
| Configuration | config.yaml, env vars, modules |
| Docker Deployment | docker compose stack |
| IDE Integration | Continue, Cursor, model routing |
| Ecosystem | alexar76 / AIMarket integration |
| Research Evidence | Citations behind design choices |
| FAQ | Common questions |
| Troubleshooting | Errors and fixes |
| Command | Purpose |
|---|---|
metis |
Run a query through the cognitive stack |
metis-serve |
OpenAI-compatible API (/v1/chat/completions) |
metis-node |
Start a distributed worker node |
metis-coordinator |
Start the cluster coordinator |
metis-cluster |
Check cluster node health |
Full guides with diagrams live in the repository:
- docs/en/README.md — user guide
- docs/en/ARCHITECTURE.md — architecture deep dive
- docs/en/API.md — API reference
MIT