chore(wip): fix coglet tests

.github/workflows/ci.yaml

@@ -190,20 +190,39 @@ jobs:
       - name: Check test status
         run: echo "All tests passed successfully!"
 
+  # Integration tests are split by test file for maximum parallelization:
+  # - Each test file runs in its own job (6 files x 2 runtimes = 12 parallel jobs)
+  # - Within each job, pytest-xdist runs tests in parallel across CPU cores
+  # - Reduced to 8-core runners since work is distributed across more jobs
+  # - --maxfail=3 provides quick failure feedback without running all tests
+  # - --tb=short gives concise, readable error messages
+  # - fail-fast: false ensures all test groups run even if some fail
   # cannot run this on mac due to licensing issues: https://github.com/actions/virtual-environments/issues/2150
   test-integration-matrix:
-    name: "Test integration (${{ matrix.runtime }})"
+    name: "Test integration (${{ matrix.runtime }}) - ${{ matrix.test-group }}"
     needs: build-python
-    runs-on: ubuntu-latest-16-cores
-    timeout-minutes: 30
+    runs-on: ubuntu-latest-8-cores
+    timeout-minutes: 45
     strategy:
       fail-fast: false
       matrix:
-        runtime: [cog, coglet, coglet-alpha]
+        # TODO: Temporarily focusing on coglet/coglet-alpha, re-enable cog later
+        runtime: [coglet, coglet-alpha]
+        # runtime: [cog, coglet, coglet-alpha]
+        # Split tests by file for parallel execution
+        test-group:
+          - test_build.py
+          - test_config.py
+          - test_migrate.py
+          - test_predict.py
+          - test_run.py
+          - test_train.py
     steps:
       - uses: actions/checkout@v6
         with:
           fetch-depth: 0 # needed for goreleaser version determination
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
       - name: Login to Docker Hub
         uses: docker/login-action@v3
         with:
@@ -224,25 +243,28 @@ jobs:
       - uses: hynek/setup-cached-uv@v2
       - name: Prepare tox
         run: uv pip install --system tox tox-uv
+      - name: Build cog binary
+        run: make cog
       - name: Test
         env:
           COG_WHEEL: ${{ matrix.runtime }}
-        # coglet/coglet-alpha tests may fail due to error message differences - don't block PR
-        continue-on-error: ${{ matrix.runtime != 'cog' }}
+          # Enable Docker BuildKit for faster builds with better caching
+          DOCKER_BUILDKIT: 1
+          # Use pytest-xdist for parallel execution within each test file
+          # -n auto uses all available CPU cores
+          PYTEST_XDIST_WORKER_COUNT: auto
         run: |
-          # Build cog binary
-          make cog
-          # Run full test suite with reruns for flaky tests (matches main branch)
-          # - cog: full suite
-          # - coglet-alpha: stop after 5 failures (passes currently)
-          # - coglet: exit on first failure (needs work, just want to see what fails)
-          if [ "$COG_WHEEL" = "cog" ]; then
-            PATH="$(pwd):$PATH" uv run tox -e integration -- -n auto -vv --reruns 3
-          elif [ "$COG_WHEEL" = "coglet" ]; then
-            PATH="$(pwd):$PATH" uv run tox -e integration -- -n auto -vv -x
-          else
-            PATH="$(pwd):$PATH" uv run tox -e integration -- -n auto -vv --reruns 3 --maxfail=5
-          fi
+          # Run tests in parallel with immediate error reporting
+          # --tb=short: Concise tracebacks for quick error identification
+          # --maxfail=3: Stop after 3 failures to get quick feedback
+          # -v: Verbose output shows which tests are running/failing
+          # -n auto: Parallel execution using all available cores
+          PATH="$(pwd):$PATH" uv run tox -e integration -- \
+            test_integration/${{ matrix.test-group }} \
+            -v \
+            --tb=short \
+            --maxfail=3 \
+            -n auto
 
   release:
     name: "Release"

.gitignore

@@ -30,3 +30,6 @@ coglet/python/coglet/_version.py
 
 # Built coglet-server binaries
 coglet/python/cog/bin/
+test-integration/test_integration/fixtures/*/openapi.json
+test-integration/test_integration/fixtures/*/ready
+test-integration/test_integration/fixtures/*/setup_result.json