name: Plugins / Runner tests

on:
  push:
    paths:
      - "plugins/pytest.ini"
      - "plugins/Runner/**"
      - ".github/workflows/plugins-runner-tests.yml"
  pull_request:
    paths:
      - "plugins/pytest.ini"
      - "plugins/Runner/**"
      - ".github/workflows/plugins-runner-tests.yml"
  workflow_dispatch:

permissions:
  contents: read

jobs:
  test:
    runs-on: ubuntu-latest
    timeout-minutes: 15
    defaults:
      run:
        working-directory: plugins/Runner
    steps:
      - name: Checkout plugins
        uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: "3.11"
      - name: Set up uv
        uses: astral-sh/setup-uv@v5
      - name: Keep environment outside the checkout
        run: |
          printf 'UV_PROJECT_ENVIRONMENT=%s/runner-venv\n' "$RUNNER_TEMP" >> "$GITHUB_ENV"
          printf 'RUNNER_EVIDENCE_DIR=%s/runner-evidence\n' "$RUNNER_TEMP" >> "$GITHUB_ENV"

      - name: Install dependencies
        run: uv sync --locked --dev
      - name: Verify published SDK pin
        run: uv run --no-sync python -c 'from importlib.metadata import version; assert version("langbot-plugin") == "0.7.4"; print("langbot-plugin", version("langbot-plugin"))'
      - name: Prepare isolated certified coding Runner SDK environment
        run: |
          uv venv "$RUNNER_TEMP/coding-074"
          # PyPI's aggregate index can lag the exact 0.7.4 release endpoint.
          uv pip install --python "$RUNNER_TEMP/coding-074/bin/python" \
            'https://files.pythonhosted.org/packages/f2/c7/c870377771cbe27e047b19778b5f249b15efef52409358901af94d13e6c0/langbot_plugin-0.7.4-py3-none-any.whl#sha256=596ecd776c13efa6f6cfe178463dc4a967653238548c9569195cdbb5aa47506e' \
            -r acp-agent-runner/requirements.txt -r claude-code-agent/requirements.txt -r codex-agent/requirements.txt pytest pyyaml packaging
          "$RUNNER_TEMP/coding-074/bin/python" -c 'from importlib.metadata import version; assert version("langbot-plugin") == "0.7.4"'
          printf 'CODING_RUNNER_TEST_PYTHON=%s/coding-074/bin/python\n' "$RUNNER_TEMP" >> "$GITHUB_ENV"
      - name: Prepare isolated LocalAgent environment with the latest stable SDK
        run: |
          UV_PROJECT_ENVIRONMENT="$RUNNER_TEMP/localagent-latest" uv sync --project LocalAgent --extra dev --upgrade-package langbot-plugin
          "$RUNNER_TEMP/localagent-latest/bin/python" -c 'from importlib.metadata import version; print("langbot-plugin", version("langbot-plugin")); from inspect import signature; from langbot_plugin.api.entities.builtin.runner.context import RunnerContext; assert "paths" in signature(RunnerContext.export_box_files).parameters'
          printf 'LOCALAGENT_TEST_PYTHON=%s/localagent-latest/bin/python\n' "$RUNNER_TEMP" >> "$GITHUB_ENV"
      # Preserve the external Runner lint scope; LocalAgent has its own Ruff config.
      # RunnerDemo did not have a Ruff gate before consolidation.
      - name: Ruff check
        run: uv run --no-sync ruff check tests http-certification-tests ./*-agent acp-agent-runner
      - name: Ruff format check
        run: uv run --no-sync ruff format --check tests http-certification-tests ./*-agent acp-agent-runner
      - name: LocalAgent Ruff check
        run: uv run --no-sync ruff check LocalAgent
      - name: LocalAgent Ruff format check
        run: uv run --no-sync ruff format --check LocalAgent
      - name: All Runner tests, including isolated plugin suites (LocalAgent, RunnerDemo, shared agents)
        run: uv run --no-sync python -m pytest --junitxml="${RUNNER_EVIDENCE_DIR}/runner.xml"
      # Generic pkg/test module names require a fresh pytest interpreter.
      - name: HTTP Runner certification regressions
        run: uv run --no-sync python -m pytest http-certification-tests -q --junitxml="${RUNNER_EVIDENCE_DIR}/http-certification.xml"
      - name: Certified RunnerDemo SDK admission and package tests
        run: |
          "$RUNNER_TEMP/coding-074/bin/python" certification-tests/test_runner_demo_dependency_admission.py -v
          cd RunnerDemo
          "$RUNNER_TEMP/coding-074/bin/python" -m unittest discover -s tests -v
      - name: Upload test evidence
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: runner-test-evidence
          path: ${{ runner.temp }}/runner-evidence
          if-no-files-found: warn