fix: resolve CI workflow issues

- Add ruff to dev dependencies
- Fix package installation in all workflows
- Add uv pip install -e . to install package in development mode
- Fix ruff paths to use src/fisher_scoring/ instead of fisher_scoring/
- Remove unused variables flagged by ruff
- All workflows now properly install and test the package

Fixes:
- ModuleNotFoundError: No module named 'fisher_scoring'
- Failed to spawn: ruff (No such file or directory)

Author: xRiskLab

.github/workflows/ci.yaml

@@ -38,6 +38,7 @@ jobs:
     - name: Install dependencies
       run: |
         uv sync --dev
+        uv pip install -e .
     
     - name: Run tests
       run: |
@@ -70,14 +71,15 @@ jobs:
     - name: Install dependencies
       run: |
         uv sync --dev
+        uv pip install -e .
     
     - name: Run ruff
       run: |
-        uv run ruff check fisher_scoring/ tests/
+        uv run ruff check src/fisher_scoring/ tests/
     
     - name: Run ruff format check
       run: |
-        uv run ruff format --check fisher_scoring/ tests/
+        uv run ruff format --check src/fisher_scoring/ tests/
 
   type-check:
     runs-on: ubuntu-latest
@@ -97,9 +99,10 @@ jobs:
     - name: Install dependencies
       run: |
         uv sync --dev
+        uv pip install -e .
     
     - name: Run mypy (if configured)
       run: |
         # Add mypy configuration and uncomment when ready
-        # uv run mypy fisher_scoring/
+        # uv run mypy src/fisher_scoring/
         echo "Type checking skipped - configure mypy when ready"

.github/workflows/compatibility.yml

@@ -62,6 +62,10 @@ jobs:
           uv add "scikit-learn"
         fi
     
+    - name: Install package in development mode
+      run: |
+        uv pip install -e .
+    
     - name: Run compatibility tests
       run: |
         echo "Testing Python ${{ matrix.python-version }} compatibility"

.github/workflows/release.yml

@@ -30,6 +30,7 @@ jobs:
     - name: Install dependencies
       run: |
         uv sync --dev
+        uv pip install -e .
     
     - name: Run tests
       run: |

pyproject.toml

@@ -41,6 +41,7 @@ dev = [
     "pytest-mock>=3.14.0,<4.0.0",
     "pydocstyle>=6.3.0,<7.0.0",
     "docformatter>=1.7.5,<2.0.0",
+    "ruff>=0.1.0,<1.0.0",
 ]
 
 [tool.setuptools]

src/fisher_scoring/fisher_scoring_multinomial.py

@@ -344,7 +344,6 @@ def predict_ci(self, X: np.ndarray, method: str = "logit") -> Dict[int, np.ndarr
 
         ci_results = {}
         for class_idx in range(self.beta.shape[1]):  # Iterate over each class
-            beta_k = self.beta[:, class_idx]
             logits_k = logits[:, class_idx]
             probabilities_k = probabilities[:, class_idx]
 

tests/test_fisher_scoring_focal.py

@@ -73,7 +73,7 @@ def test_fit_predict_proba(self):
     def test_predict_ci(self):
         """Test that predict_ci works correctly."""
         self.model.fit(self.X, self.y)
-        ci_logit = self._predict_ci(
+        self._predict_ci(
             "logit",
             "CI for logits should have shape (n_samples, 2).",
             "Lower CI should not exceed upper CI for logits.",

tests/test_fisher_scoring_logistic.py

@@ -70,7 +70,7 @@ def test_fit_predict_proba(self):
     def test_predict_ci(self):
         """Test the predict_ci method for both logit and proba."""
         self.model.fit(self.X, self.y)
-        ci_logit = self._predict_ci(
+        self._predict_ci(
             "logit",
             "CI for logits should have shape (n_samples, 2).",
             "Lower CI should not exceed upper CI for logits.",

tests/test_fisher_scoring_poisson.py

@@ -425,8 +425,6 @@ def test_negative_binomial_information_matrix_comparison(self):
 
         # Coefficients may differ significantly between expected and empirical methods
         # This is normal for count models where variance assumptions may not hold perfectly
-        coeff_diff = np.linalg.norm(model_expected.weights - model_empirical.weights)
-        
         # Just verify that both methods produce reasonable results (not NaN/Inf)
         self.assertTrue(np.all(np.isfinite(model_expected.weights)))
         self.assertTrue(np.all(np.isfinite(model_empirical.weights)))