Update GitHub action for v0.2.0 release

.github/workflows/ci.yml

@@ -17,62 +17,53 @@ on:
 
 jobs:
   pre-commit:
-    name: Pre-commit CI
+    name: Pre-commit Hooks CI
     runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: ["3.12"]
-    timeout-minutes: 15
+    timeout-minutes: 10
     steps:
-      - name: Checkout
-        uses: actions/checkout@v3
+      - uses: actions/checkout@v6
 
-      - name: Setup Python
-        uses: actions/setup-python@v4
+      - name: "Set up Python"
+        uses: actions/setup-python@v6
         with:
-          python-version: ${{ matrix.python-version }}
+          python-version-file: "pyproject.toml"
 
-      - name: Install the latest version of uv
-        uses: astral-sh/setup-uv@v5
+      - name: Install uv
+        uses: astral-sh/setup-uv@v7
         with:
-          enable-cache: true
-          cache-dependency-glob: ".pre-commit-config.yaml"
+          version: "0.10.9"
 
-      - name: Install dependencies
-        run: |
-          uv sync --group ci
-          alias pip="uv pip"  # Trick pre-commit to use uv
+      - name: Install the project
+        run: uv sync --locked --all-extras --dev
 
       - name: Run Pre-commit
         run: |
           uv run pre-commit run --show-diff-on-failure --color=always --all-files
+
   unit_tests:
     name: Unit Tests
     runs-on: ubuntu-latest
+    timeout-minutes: 10
 
     steps:
-      - name: Checkout
-        uses: actions/checkout@v3
+      - uses: actions/checkout@v6
 
-      - name: Setup Python
-        uses: actions/setup-python@v4
+      - name: "Set up Python"
+        uses: actions/setup-python@v6
         with:
-          python-version: 3.12
+          python-version-file: "pyproject.toml"
 
-      - name: Install the latest version of uv
-        uses: astral-sh/setup-uv@v5
+      - name: Install uv
+        uses: astral-sh/setup-uv@v7
         with:
-          enable-cache: true
-          cache-dependency-glob: ".pre-commit-config.yaml"
+          version: "0.10.9"
 
-      - name: Install dependencies
-        run: |
-          uv sync --group ci
-          alias pip="uv pip"  # T
+      - name: Install the project
+        run: uv sync --locked --all-extras --dev
 
       - name: Run Tests
         run: |
-          uv run pytest tests/unit_tests
+          uv run pytest tests
 
   # Llama firewall is not yet supported
   # security_tests:

.github/workflows/scan-dependencies.yml

@@ -9,32 +9,32 @@ on:
       - main
       - develop
     paths:
-      - '**/poetry.lock'
+      - '**/uv.lock'
   push:
     branches:
       - main
       - develop
     paths:
-      - '**/poetry.lock'
+      - '**/uv.lock'
 
 jobs:
-  safety_scan:
-    name: Safety Scan
+  pip_audit:
+    name: pip-audit
     runs-on: ubuntu-latest
+    timeout-minutes: 10
 
     steps:
-      - name: Checkout code
-        uses: actions/checkout@v3
+      - uses: actions/checkout@v6
 
-      - name: Set up Python
-        uses: actions/setup-python@v4
+      - name: "Set up Python"
+        uses: actions/setup-python@v6
         with:
-          python-version: '3.10'
+          python-version-file: "pyproject.toml"
 
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install safety
-      - name: Run safety check
-        run: |
-          safety check --full-report
+      - name: Install uv
+        uses: astral-sh/setup-uv@v7
+
+      - name: Install the project
+        run: uv sync --locked --all-extras --dev
+
+      - uses: pypa/gh-action-pip-audit@v1.1.0

.pre-commit-config.yaml

@@ -37,7 +37,6 @@ repos:
         language: system
         types: [python]
         args: ["--config-file=mypy.ini"]
-        exclude: ^legacy/
 
   - repo: https://github.com/crate-ci/typos
     rev: v1.42.1

bandit.yaml

@@ -1 +1 @@
-exclude_dirs: ['legacy']
+exclude_dirs: ['tests']

pyproject.toml

@@ -25,15 +25,19 @@ sre-agent-run-diagnosis-quality-eval = "sre_agent.eval.diagnosis_quality.run:mai
 [dependency-groups]
 dev = [
     "mypy>=1.19.1",
+    "pre-commit>=4.5.1",
+    "pytest>=9.0.2",
+]
+eval = [
+    "opik>=1.10.38",
 ]
 
 [build-system]
-requires = ["uv_build>=0.9.28,<0.10.0"]
+requires = ["uv_build>=0.9.28,<0.11.0"]
 build-backend = "uv_build"
 
 [tool.hatch.build.targets.wheel]
 packages = ["src/sre_agent"]
 
 [tool.pytest.ini_options]
-addopts = "--cov=sre_agent --cov-report term-missing"
 testpaths = ["tests"]

src/sre_agent/eval/diagnosis_quality/README.md

@@ -65,5 +65,6 @@ When the server is running, open [http://localhost:5173/](http://localhost:5173/
 Run command:
 
 ```bash
+uv sync --group eval
 uv run sre-agent-run-diagnosis-quality-eval
 ```

src/sre_agent/eval/diagnosis_quality/metrics/affected_services_match.py

@@ -5,7 +5,7 @@
 from opik.evaluation.metrics import base_metric, score_result
 
 
-class AffectedServicesMatch(base_metric.BaseMetric):
+class AffectedServicesMatch(base_metric.BaseMetric):  # type: ignore[misc]
     """Score overlap between predicted and expected affected services."""
 
     def __init__(self, name: str = "affected_services_match") -> None:

src/sre_agent/eval/diagnosis_quality/metrics/root_cause_correctness.py

@@ -5,7 +5,7 @@
 from opik.evaluation.metrics import GEval, base_metric, score_result
 
 
-class RootCauseCorrectness(base_metric.BaseMetric):
+class RootCauseCorrectness(base_metric.BaseMetric):  # type: ignore[misc]
     """Judge whether the root cause matches the expected issue."""
 
     def __init__(

src/sre_agent/eval/diagnosis_quality/metrics/suggested_fixes_quality.py

@@ -5,7 +5,7 @@
 from opik.evaluation.metrics import GEval, base_metric, score_result
 
 
-class SuggestedFixesQuality(base_metric.BaseMetric):
+class SuggestedFixesQuality(base_metric.BaseMetric):  # type: ignore[misc]
     """Judge whether suggested fixes are correct and actionable."""
 
     def __init__(

src/sre_agent/eval/tool_call/README.md

@@ -72,6 +72,7 @@ When the server is running, open [http://localhost:5173/](http://localhost:5173/
 Run command:
 
 ```bash
+uv sync --group eval
 uv run sre-agent-run-tool-call-eval
 ```
 

src/sre_agent/eval/tool_call/metrics/expected_tool_select_order.py

@@ -8,7 +8,7 @@
 from sre_agent.eval.tool_call.metrics.span_tools import extract_tool_names
 
 
-class ExpectedToolSelectOrder(base_metric.BaseMetric):
+class ExpectedToolSelectOrder(base_metric.BaseMetric):  # type: ignore[misc]
     """Validate the expected tool selection order."""
 
     def __init__(self, name: str = "expected_tool_select_order"):

src/sre_agent/eval/tool_call/metrics/expected_tool_selection.py

@@ -8,7 +8,7 @@
 from sre_agent.eval.tool_call.metrics.span_tools import extract_tool_names
 
 
-class ExpectedToolSelection(base_metric.BaseMetric):
+class ExpectedToolSelection(base_metric.BaseMetric):  # type: ignore[misc]
     """Validate the expected tool selection."""
 
     def __init__(self, name: str = "expected_tool_selection"):

tests/__init__.py

@@ -0,0 +1 @@
+"""Tests package for the SRE Agent."""

tests/test_dummy.py

@@ -0,0 +1,6 @@
+"""Dummy test file for pytest."""
+
+
+def test_dummy() -> None:
+    """A dummy test."""
+    assert True