Lynchpin keeps public documentation close to the layer that owns each contract. This page is the short map for readers; source docstrings, schemas, and generated catalogs provide the detailed API reference.
- Architecture explains owner-native inputs, typed source APIs, canonical materialization, DuckDB refreshes, evidence graphs, and product interfaces.
- Data sources maps configured roots and active adapter families to their roles and coverage boundaries.
- Analysis methodology defines the evidence, timeframe, denominator, calibration, and reporting rules for canonical analysis products.
- Observability model explains the machine telemetry substrate and the limits of resource-attribution claims.
- Lynchpin–Polylogue boundary records the AI-session boundary.
- Start with the README for the supported Nix-first environment, materialization commands, current-state packs, and MCP server.
- Use
python -m lynchpin.cli.materialize --helpto inspect dependency-ordered materialization and promotion options. - Use
python -m lynchpin.mcpfor the eight-tool stdio MCP surface; tool schemas and examples are available throughlynchpin_catalog. - Run
pytest -q,ruff check ., andnix flake checkfor the public quality gates.
Active work belongs to the committed Beads graph. Browse the
web board or use bd ready
locally.