.github/workflows/benchmark.yml

name: CI

on:
  push:
    branches: [master, main]
  pull_request:
    branches: [master, main]

jobs:
  # Fast, no-Ollama guardrails. Pin the two-deployment-mode contract (hosted
  # free web vs local Docker) and the prefilter's zero-false-negative invariant
  # so a change for one deployment can't silently break the other.
  guardrails:
    name: Guardrail tests (deployment modes + prefilter)
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v6

      - name: Set up Python 3.11
        uses: actions/setup-python@v6
        with:
          python-version: "3.11"

      - name: Install dependencies
        run: pip install -r requirements.txt -r requirements-web.txt httpx

      - name: Deployment-mode guardrail (hosted web vs local Docker)
        run: python3 tests/test_deployment_modes.py

      - name: Prefilter — zero false negatives
        run: python3 tests/test_prefilter.py

      - name: End-to-end pipeline (real routes; LLM + geocoder mocked)
        run: python3 tests/test_e2e.py

      - name: Persistence — CSV cache/restore round-trip is lossless
        run: python3 tests/test_persistence.py

      - name: Merge — living-library merge keeps curation
        run: python3 tests/test_merge.py

      - name: Capture — single-post extraction (Shortcut path)
        run: python3 tests/test_capture.py

  benchmark:
    name: Benchmark smoke test (Ollama)
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v6

      - name: Set up Python 3.11
        uses: actions/setup-python@v6
        with:
          python-version: "3.11"

      - name: Install Python dependencies
        run: pip install -r requirements.txt

      - name: Install Ollama
        run: |
          curl -fsSL https://ollama.com/install.sh | sh
          ollama serve &
          # Wait for the server to be ready
          for i in $(seq 1 30); do
            curl -sf http://localhost:11434/api/tags && break
            sleep 2
          done

      - name: Pull llama3.2
        run: ollama pull llama3.2

      - name: Run benchmark (Ollama / llama3.2)
        run: python3 tests/benchmark.py --provider ollama --model llama3.2

      - name: Check quality regression
        run: |
          python3 - <<'EOF'
          import json, sys
          from pathlib import Path

          results_path = Path("tests/benchmark_results.json")
          baseline_path = Path("tests/benchmark_baseline.json")

          if not results_path.exists():
              print("No results file found — benchmark must have failed.")
              sys.exit(1)

          with open(results_path) as f:
              results = json.load(f)

          score = results.get("llama3.2", {}).get("composite_score")
          if score is None:
              print("llama3.2 result missing from benchmark_results.json")
              sys.exit(1)

          print(f"llama3.2 composite_score: {score}")

          if not baseline_path.exists():
              print("No baseline file yet — storing this run as the baseline.")
              baseline_path.write_text(json.dumps({"llama3.2": score}, indent=2))
              sys.exit(0)

          with open(baseline_path) as f:
              baseline = json.load(f)

          baseline_score = baseline.get("llama3.2", 0)
          drop = baseline_score - score
          print(f"Baseline: {baseline_score}  Current: {score}  Drop: {drop:.1f}")

          if drop > 5:
              print(f"FAIL: composite_score dropped {drop:.1f} points (> 5 allowed)")
              sys.exit(1)

          print("PASS: quality within acceptable range.")
          EOF

      - name: Upload benchmark results
        if: always()
        uses: actions/upload-artifact@v7
        with:
          name: benchmark-results
          path: tests/benchmark_results.json