Operational handbook for AI agents working in this repository.
Automated pipeline that generates architecture documentation for OpenDataHub (ODH) and Red Hat OpenShift AI (RHOAI). The pipeline fetches component repos, analyzes them with Claude Code skills, and produces structured architecture summaries, diagrams, and overlays.
See README.md for pipeline details and AGENT_USAGE.md for consuming the generated architecture data.
This project uses the Agentic Work Ledger methodology. See docs/notes/agentic_work_ledger.md for the full spec.
Key rules:
- Read PLAN.md before starting work.
- Tasks live in
docs/tasks/— pick frompending/, move tocurrent/, thendone/orblocked/. - Record architectural decisions as ADRs in
docs/decisions/. - File bugs immediately in
docs/bugs/open/. - Append session activity to
docs/notes/session-log.md.
- Python 3.13+ — pipeline code in
main.pyandlib/. Managed withuv. - Go 1.25 —
arch-queryCLI insrc/arch-query/. Linted withgolangci-lint. - Claude Code skills —
.claude/skills/for component analysis, aggregation, and diagram generation.
- Linter:
ruffwith rules E, F, W, I. Runmake lint-python. - Tests:
pytestwithasyncio_mode = "strict". Tests intests/. - No type stubs required but type hints are preferred.
- Linter:
golangci-lintwith errcheck, govet, staticcheck, unused. Runmake lint-go. - Tests:
make test.
Run all checks:
make lint # Python + Go + overlays + platforms + architecture docs
make test # Go tests
uv run pytest # Python tests- architecture/ — generated output, organized by platform and version. Do not hand-edit.
- overlays/ — manually authored architecture updates between regeneration cycles. See AGENT_USAGE.md for format.
- checkouts/ — gitignored, exists only during pipeline runs.
- scripts/ — pipeline helper scripts and linters.
- platforms.yaml — platform/version configuration for the pipeline.
| Target | Purpose |
|---|---|
make lint |
Run all linters |
make test |
Run Go tests |
make build |
Build arch-query CLI |
make build-embedded |
Build arch-query with embedded data |