Merge pull request #205 from HolobiomicsLab/propose/ci-pytest

ci(tests): run pytest on pull requests and pushes

Author: Fosowl

.github/known-failing-tests.txt

@@ -0,0 +1,32 @@
+# Tests that fail on mimosa_v2 as of ff41764 (2026-09-02) in a fresh
+# `uv sync --group dev` environment without API keys, one pytest node id
+# per line. The CI workflow deselects them so that it can flag regressions in
+# the rest of the suite. Each entry is a code/test drift, not an environment
+# issue except the last one; remove the line when the test passes again.
+#
+# csv_mode.EVAL_LOG_FILE is logs/evaluation_csv_mode.log; the test expects
+# logs/evaluation._csv_mode.log.
+tests/csv_mode_logging_test.py::test_log_file_lives_in_logs_directory
+# EvaluationCLI._validate_queue was removed in 2d480dc; the tests still call it.
+tests/evaluation_cli_test.py::test_validate_queue_allows_overlapping_ports
+tests/evaluation_cli_test.py::test_validate_queue_rejects_shared_workspace
+# _setup_connectivity returns [''] where the tests expect [].
+tests/evaluation_cli_test.py::test_setup_connectivity_reuses_previous_runs_port_range
+tests/evaluation_cli_test.py::test_setup_connectivity_empty_ip_keeps_current_range
+tests/evaluation_cli_test.py::test_setup_connectivity_accepts_overlapping_range_without_constraint
+# The failure fingerprint has seven dimensions; the tests pin six.
+tests/failure_fingerprint_test.py::test_dimension_is_six
+tests/failure_fingerprint_test.py::test_absent_source_uses_neutral_and_mask
+tests/failure_fingerprint_test.py::test_source_label_normalisation
+tests/failure_fingerprint_test.py::test_unknown_source_is_ignored
+# Default pricing changed (0.0105 per call where the tests expect 0.00082).
+tests/pricing_test.py::test_calculate_cost_handles_scalar_and_list_model_id
+tests/pricing_test.py::test_single_agent_memory_is_priced
+# Generated save_memories code and SmolAgentFactory.model_id drifted from the tests.
+tests/single_agent_factory_logprobs_test.py::test_generated_save_memories_writes_logprobs_and_strips_raw
+tests/single_agent_factory_logprobs_test.py::test_generated_save_memories_warns_only_when_logprobs_all_missing
+tests/smolagent_factory_logprobs_test.py::test_save_memories_writes_logprobs_and_strips_raw
+tests/smolagent_factory_logprobs_test.py::test_save_memories_warns_only_when_logprobs_all_missing
+# Calls the live LLM provider through VariationEngine.mutation_prompt; fails
+# with an AuthenticationError wherever no API key is configured, CI included.
+tests/variation_engine_test.py::test_mutation_prompt_does_not_touch_gradient_history

.github/workflows/tests.yml

@@ -0,0 +1,36 @@
+name: Tests
+
+on:
+  push:
+    branches: [main, mimosa_v2]
+  pull_request:
+
+permissions:
+  contents: read
+
+jobs:
+  pytest:
+    name: pytest (Python 3.11, ubuntu)
+    runs-on: ubuntu-latest
+    timeout-minutes: 40
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v6
+        with:
+          python-version: "3.11"
+          enable-cache: true
+
+      - name: Install project and test dependencies
+        run: uv sync --group dev
+
+      - name: Run the suite
+        # Tests listed in .github/known-failing-tests.txt already fail on the
+        # base branch for reasons unrelated to any pull request. They are
+        # deselected so that a regression elsewhere stays visible. Delete a
+        # line as soon as its test passes again.
+        run: |
+          DESELECT=$(grep -v '^#' .github/known-failing-tests.txt | grep -v '^[[:space:]]*$' | sed 's/^/--deselect /' | tr '\n' ' ')
+          echo "Deselected: ${DESELECT:-none}"
+          uv run pytest tests/ -q $DESELECT