test-package

# Run targeted tests — $ARGUMENTS Parse `$ARGUMENTS`: - If `$ARGUMENTS` is `changed`: 1. Run `git diff --name-only origin/main...HEAD` and `git diff --name-only` (working tree). 2. Union the paths. For each path, map to its owning package: - `libs/core/**` → `core` - `cli/**` → `cli` - `libs/connectors/<name>/**` → `<name>` - `libs/spec/**` → `core` (schemas are consumed by core) - `docs/**`, `examples/**`, root-level configs → no test run (report and stop) 3. Run tests for each affected package (see mapping below). Stop on first failure. - Otherwise treat `$ARGUMENTS` as one package name. Valid values: - `core`, `cli` - Connectors: `openai`, `google`, `runway`, `luma`, `decart`, `replicate`, `elevenlabs`, `stability-audio`, `lmnt`, `gmicloud`, `s3`, `langsmith` ## Test command mapping | Package | Command | |---|---| | `core` | `cd libs/core && pytest tests/ -v` | | `cli` | `cd cli && pytest tests/ -v` | | any connector `<name>` | `cd libs/connectors/<name> && pytest -v` | ## Quick subset (single file under core) If the user hints at one file (e.g. "just the pipeline test"), run `cd libs/core && pytest tests/unit/<file>.py -v` per the CLAUDE.md guidance. ## Report - Pass/fail per package, with pytest's summary line. - If any failure, surface the failing test name and the first traceback — do not proceed to other packages. - Remind the user that `make test` is still the full-suite gate before PR.