docs: add run_tests.py documentation and update README

README.md

@@ -17,7 +17,7 @@ Validation tools and CI/CD workflows for Autohive integrations.
 | `scripts/check_readme.py` | README update verification ([docs](scripts/docs/check_readme.md)) |
 | `scripts/check_version_bump.py` | Version bump verification with bump-level recommendations ([docs](scripts/docs/check_version_bump.md)) |
 | `scripts/check_config_sync.py` | Config-code sync checker ([docs](scripts/docs/check_config_sync.md)) |
-| `scripts/run_tests.py` | Unit test runner with coverage ([docs](#running-unit-tests)) |
+| `scripts/run_tests.py` | Unit test runner with coverage ([docs](scripts/docs/run_tests.md)) |
 | `scripts/get_changed_dirs.py` | Changed directory detection ([docs](scripts/docs/get_changed_dirs.md)) |
 | `.github/workflows/validate-integration.yml` | PR validation pipeline |
 | `.github/workflows/self-test.yml` | Regression guard for tooling scripts |
@@ -80,7 +80,7 @@ flowchart TB
 | Detect changes | `get_changed_dirs.py` | `git diff` → extract top-level dirs, filter out `.github`, `scripts`, `tests` |
 | Structure check | `validate_integration.py` | Folder name, required files, config.json schema, `__init__.py`, requirements.txt, tests/, icon size, unused scopes |
 | Code check | `check_code.py` | pip install, py_compile, check_imports, JSON validity, ruff check, ruff format, bandit, pip-audit, check_config_sync |
-| Tests | `run_tests.py` | Discovers and runs `test_*_unit.py` files with pytest, reports coverage. Warns (does not fail) if no unit tests exist |
+| Tests | `run_tests.py` | Installs each integration's dependencies, then discovers and runs `test_*_unit.py` files with pytest per-integration. Warns (does not fail) if no unit tests exist |
 | README check | `check_readme.py` | New integration files added → was README.md also updated? |
 | Version check | `check_version_bump.py` | Version in config.json incremented? Recommends major/minor/patch based on config and code changes |
 

scripts/docs/run_tests.md

@@ -0,0 +1,145 @@
+# run_tests.py
+
+Discovers and runs unit tests for integration directories, installing each integration's own dependencies first.
+
+## Overview
+
+Each integration pins its own SDK version in `requirements.txt`. The test runner installs dependencies and runs pytest **per-integration** to ensure each integration is tested against its own pinned SDK version.
+
+Integrations without `test_*_unit.py` files are skipped with a warning — they do not cause a failure.
+
+## Usage
+
+```bash
+python scripts/run_tests.py [dir ...]
+```
+
+### Arguments
+
+| Argument | Required | Description |
+|----------|----------|-------------|
+| `dir` | No (zero or more) | Integration directories to test. If omitted, auto-discovers all directories with a `config.json` |
+
+### Exit Codes
+
+| Code | Meaning |
+|------|---------|
+| `0`  | All tests passed (or no tests found) |
+| `1`  | One or more test failures |
+| `2`  | An error occurred (directory not found) |
+
+### Examples
+
+```bash
+# Run tests for a single integration
+python scripts/run_tests.py hackernews
+
+# Run tests for multiple integrations
+python scripts/run_tests.py hackernews bitly notion
+
+# Auto-discover and run all integrations with unit tests
+python scripts/run_tests.py
+```
+
+## How It Works
+
+```mermaid
+flowchart TD
+    A[Resolve directories] --> B{For each directory}
+    B --> C{Has test_*_unit.py files?}
+    C -->|No| D[Skip with warning]
+    C -->|Yes| E[Install requirements.txt]
+    E --> F[Run pytest]
+    F --> G{Tests passed?}
+    G -->|Yes| H[Record pass]
+    G -->|No| I[Record failure]
+    D --> B
+    H --> B
+    I --> B
+    B --> J[Print summary]
+    J --> K{Any failures?}
+    K -->|Yes| L[Exit 1]
+    K -->|No| M[Exit 0]
+```
+
+### Step-by-Step
+
+1. Resolve integration directories from CLI args or auto-detect (subdirectories with `config.json`)
+2. For each directory, check for `test_*_unit.py` files in `tests/`
+3. Skip directories without unit tests (warn but don't fail)
+4. For each testable integration:
+   a. Install the integration's `requirements.txt` (includes its pinned SDK version)
+   b. Run pytest with `--import-mode=importlib`, `-m unit`, coverage enabled
+   c. Record pass or failure
+5. Print summary of passed/failed integrations
+6. Exit 0 if all passed, 1 if any failed
+
+### Per-Integration Isolation
+
+Different integrations may pin different SDK versions. For example:
+
+- `bitly/requirements.txt` → `autohive-integrations-sdk~=1.0.2`
+- `notion/requirements.txt` → `autohive-integrations-sdk~=2.0.0`
+
+The test runner handles this by installing dependencies and running pytest separately for each integration. The last `pip install` overwrites the previous SDK version, which is why pytest must run immediately after each install.
+
+### Test Discovery
+
+Unit test files must follow the naming convention `test_*_unit.py` and be located in the integration's `tests/` directory:
+
+```
+my-integration/
+  tests/
+    test_my_integration_unit.py   ← discovered
+    test_my_integration.py        ← NOT discovered (legacy manual test)
+    conftest.py                   ← loaded by pytest automatically
+```
+
+## Output Format
+
+### Tests found and passing:
+
+```
+🧪 Running unit tests for: hackernews, bitly
+============================================================
+  hackernews
+============================================================
+... pytest output ...
+
+============================================================
+  bitly
+============================================================
+... pytest output ...
+
+============================================================
+✅ Tests passed: hackernews, bitly
+============================================================
+```
+
+### Some tests failing:
+
+```
+============================================================
+✅ Tests passed: hackernews
+❌ Tests failed: bitly
+============================================================
+```
+
+### No unit tests found:
+
+```
+⚠️  No unit tests (test_*_unit.py) found in: my-integration
+⚠️  No unit tests to run
+```
+
+## Integration with CI
+
+Called by the composite action in `action.yml`:
+
+```yaml
+- name: Tests
+  if: steps.detect.outputs.dirs != ''
+  run: python scripts/run_tests.py ${{ steps.detect.outputs.dirs }}
+```
+
+The test infrastructure (`pyproject.toml`, `conftest.py`, `requirements-test.txt`) lives in the integrations repo — see its `CONTRIBUTING.md` for how to write and run tests locally.