Skip to content

CLAUDE.md

Latest commit

 

History

History
54 lines (44 loc) · 1.68 KB

CLAUDE.md

File metadata and controls

54 lines (44 loc) · 1.68 KB

Project context

  • Agentic CLI that extracts fix findings from an inspiration Git repo and checks applicability to a target repo, producing structured JSON reports.
  • Good output = reviewable findings/assessments; it does not auto-patch repos.

Repo map

  • src/different_agent/: CLI + core logic (agents, config, git/github tools, reporting).
  • tests/: pytest suite.
  • outputs/: generated reports (local artifacts).
  • different.toml: default runtime config.

Tech stack

  • Python 3.11.
  • DeepAgents + LangChain OpenAI.
  • Tooling: uv, ruff, ty, pytest.

Common commands

uv sync --all-groups

different-agent --inspiration /path/to/inspiration --target /path/to/target
different-agent --inspiration /path/to/inspiration --extract-only
uv run ruff format .
uv run ruff check .
uv run ty check src/
uv run pytest

Code style

  • Ruff line length: 100; target version: py311.
  • E501 is ignored in src/different_agent/agents.py and src/different_agent/report.py (long prompt/template strings).

Workflow rules

  • Explore code before changes; keep edits minimal and focused.
  • Run format + lint + typecheck + tests before finishing.

Testing

  • Pytest with coverage and warnings-as-errors (see pyproject.toml).
  • Single test example:
uv run pytest tests/test_config.py -k test_name

Tooling notes

  • Pre-commit hooks: prek run --all-files (ruff, ty, shellcheck).
  • Runtime config is different.toml; CLI uses it by default. Use --model to override per run.

Gotchas

  • GitHub enrichment needs a token to avoid rate limits (see README).
  • Outputs are written under outputs/<project_name>/ with time-based suffixes.

More docs

  • @README.md