ci.yml

name: CI

on:
  push:
  pull_request:
  workflow_dispatch:

jobs:
  lint:
    name: Lint and type-check
    runs-on: ubuntu-latest

    steps:
      - name: Check out repository
        uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: "3.12"
          cache: "pip"

      - name: Install tooling
        run: |
          python -m pip install --upgrade pip
          pip install ruff black mypy

      - name: Ruff
        run: ruff check src app.py tests

      - name: Black
        run: black --check src app.py tests

      - name: Mypy
        run: mypy src app.py

  test-and-smoke:
    name: Test and smoke workflow on Python ${{ matrix.python-version }}
    runs-on: ubuntu-latest

    strategy:
      fail-fast: false
      matrix:
        python-version: ["3.10", "3.11", "3.12"]

    env:
      MPLBACKEND: Agg
      PYTHONWARNINGS: ignore

    steps:
      - name: Check out repository
        uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: ${{ matrix.python-version }}
          cache: "pip"

      - name: Install dependencies
        run: |
          python -m pip install --upgrade pip
          pip install -r requirements.txt

      - name: Check source files compile
        run: |
          python -m compileall src app.py tests

      - name: Run unit tests
        run: |
          python -m unittest discover -s tests -v

      - name: Generate harder synthetic fraud dataset
        run: |
          python -m src.generate_synthetic_data \
            --rows 3500 \
            --fraud-rate 0.08 \
            --label-noise 0.04 \
            --seed 42 \
            --output data/raw/synthetic_fraud_dataset.csv

      - name: Prepare data
        run: |
          python -m src.data_prep

      - name: Train model
        run: |
          python -m src.train_model

      - name: Evaluate model
        run: |
          python -m src.evaluate

      - name: Score held-out transactions
        run: |
          python -m src.score_new_transactions \
            data/processed/transactions_test.csv \
            --output_csv reports/metrics/test_scored.csv

      - name: Validate scored CSV schema
        run: |
          python - <<'PY'
          import pandas as pd
          from pathlib import Path

          path = Path("reports/metrics/test_scored.csv")
          if not path.exists():
              raise FileNotFoundError(path)

          df = pd.read_csv(path)
          required = {"fraud_probability", "fraud_flag"}
          missing = required - set(df.columns)
          if missing:
              raise AssertionError(f"Missing scored columns: {missing}")

          if not df["fraud_probability"].between(0, 1).all():
              raise AssertionError("fraud_probability must be between 0 and 1")

          if not set(df["fraud_flag"].dropna().unique()).issubset({0, 1}):
              raise AssertionError("fraud_flag must be binary")

          if not df["fraud_probability"].is_monotonic_decreasing:
              raise AssertionError("scored transactions should be sorted by descending fraud_probability")

          print("Scored CSV schema OK")
          PY

      - name: Verify expected outputs
        run: |
          test -f models/fraud_pipeline.joblib
          test -f models/threshold.json
          test -f reports/metrics/metrics.json
          test -f reports/metrics/evaluation_summary.json
          test -f reports/metrics/threshold_search.json
          test -f reports/metrics/threshold_search.csv
          test -f reports/metrics/threshold_policy.json
          test -f reports/metrics/threshold_policy.csv
          test -f reports/metrics/threshold_policy.md
          test -f reports/metrics/test_scored.csv
          test -f reports/figures/roc_curve.png
          test -f reports/figures/pr_curve.png
          test -f reports/figures/calibration_curve.png
          test -f reports/figures/confusion_matrix.png
          test -f reports/figures/threshold_cost_curve.png
          test -f reports/figures/threshold_tradeoffs.png

      - name: Validate evaluation metrics JSON
        run: |
          python - <<'PY'
          import json
          from pathlib import Path

          summary_path = Path("reports/metrics/evaluation_summary.json")
          data = json.loads(summary_path.read_text())

          required = {
              "roc_auc",
              "average_precision",
              "brier_score",
              "best_threshold",
              "baseline_metrics",
              "threshold_policy_artifacts",
          }
          missing = required - set(data)
          if missing:
              raise AssertionError(f"Missing evaluation summary keys: {missing}")

          for key in ["majority_class", "empirical_prior", "stratified_random"]:
              if key not in data["baseline_metrics"]:
                  raise AssertionError(f"Missing baseline metrics for {key}")

          policy_path = Path("reports/metrics/threshold_policy.json")
          policy = json.loads(policy_path.read_text())
          if "policy_candidates" not in policy or not policy["policy_candidates"]:
              raise AssertionError("threshold_policy.json must contain policy_candidates")

          print("Evaluation summary JSON OK")
          PY

      - name: Upload generated artifacts
        uses: actions/upload-artifact@v4
        with:
          name: fraud-risk-engine-artifacts-python-${{ matrix.python-version }}
          path: |
            models/
            reports/