codeprobe 0.2.1__tar.gz → 0.2.2__tar.gz

{codeprobe-0.2.1 → codeprobe-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codeprobe
-Version: 0.2.1
+Version: 0.2.2
 Summary: Benchmark AI coding agents against your own codebase. Mine real tasks from repo history, run agents, interpret results.
 Author: codeprobe contributors
 License-Expression: Apache-2.0

{codeprobe-0.2.1 → codeprobe-0.2.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "codeprobe"
-version = "0.2.1"
+version = "0.2.2"
 description = "Benchmark AI coding agents against your own codebase. Mine real tasks from repo history, run agents, interpret results."
 readme = "README.md"
 license = "Apache-2.0"

{codeprobe-0.2.1 → codeprobe-0.2.2}/src/codeprobe/core/executor.py

@@ -334,14 +334,16 @@ def execute_task(
 _BILLABLE_COST_MODELS = frozenset({"per_token"})
 
 
-def _git_reset_workdir(repo_path: Path) -> None:
+def _git_reset_workdir(
+    repo_path: Path, *, extra_excludes: tuple[str, ...] = ()
+) -> None:
     """Reset the working directory to a clean state between sequential tasks.
 
     Runs ``git restore .`` and ``git clean -fd`` to discard modifications
     and remove untracked files so task N's leftovers don't corrupt task N+1.
     """
     try:
-        git_restore_clean(repo_path)
+        git_restore_clean(repo_path, extra_excludes=extra_excludes)
     except subprocess.CalledProcessError as exc:
         logger.warning(
             "Git reset failed (exit %d): %s",
@@ -458,6 +460,20 @@ def execute_config(
     results are returned.  Tasks with ``unknown`` or ``subscription``
     cost models are skipped in accumulation.
     """
+    # Compute directories to exclude from git clean so that experiment
+    # artifacts (runs/, tasks/, experiment.json) survive between tasks.
+    # runs_dir is e.g. <repo>/mcp-comparison/runs/baseline — walk up to
+    # find the experiment root relative to repo_path.
+    _clean_excludes: tuple[str, ...] = ()
+    if runs_dir is not None:
+        try:
+            exp_root = runs_dir.resolve().parent.parent  # runs/<label> → exp dir
+            rel = exp_root.relative_to(repo_path.resolve())
+            # Exclude the top-level experiment directory name
+            _clean_excludes = (str(rel).split("/")[0],)
+        except ValueError:
+            pass  # experiment dir is outside the repo — nothing to exclude
+
     checkpointed_ids, results = _restore_checkpointed(checkpoint_store)
 
     # Filter checkpointed results to only include tasks in the current
@@ -562,7 +578,7 @@ def execute_config(
             # Reset working directory between tasks so leftovers from
             # task N don't corrupt task N+1's results.
             if idx > 0:
-                _git_reset_workdir(repo_path)
+                _git_reset_workdir(repo_path, extra_excludes=_clean_excludes)
             task_result = _run_one(task_dir, repeat_index=repeat_index)
             _handle_result(task_result)
     else:

{codeprobe-0.2.1 → codeprobe-0.2.2}/src/codeprobe.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codeprobe
-Version: 0.2.1
+Version: 0.2.2
 Summary: Benchmark AI coding agents against your own codebase. Mine real tasks from repo history, run agents, interpret results.
 Author: codeprobe contributors
 License-Expression: Apache-2.0

{codeprobe-0.2.1 → codeprobe-0.2.2}/tests/test_executor.py

@@ -551,7 +551,7 @@ def test_execute_config_resets_workdir_between_sequential_tasks(tmp_path: Path):
         )
         # Reset should be called between tasks (not before first), so 2 times for 3 tasks
         assert mock_reset.call_count == 2
-        mock_reset.assert_any_call(Path("/repo"))
+        mock_reset.assert_any_call(Path("/repo"), extra_excludes=())
 
 
 def test_execute_config_no_reset_in_parallel_mode(tmp_path: Path):