|
|
@@ -86,6 +86,7 @@ Repeated setup work, long sleeps/timeouts, serial integration tests, filesystem/
|
|
|
| Prompt tests without LLM calls do not need the test LLM server | Added a no-server runner and moved obvious non-LLM prompt/shell cases to it | 25.41s | 21.03s | keep | Full prompt file after simplify pass median from three reruns (20.66, 21.03, 21.64); LLM-backed tests stay on original runner. |
|
|
|
| CLI run subprocess cases can run independently | Marked `run-process.test.ts` subprocess cases concurrent | 11.87s | 4.13s | keep | Newest-dev single baseline; after median from three reruns (4.13, 4.17, 4.11). Each case has an isolated temp home and LLM port. |
|
|
|
| Snapshot initialization does not need to commit seeded files in the source repo | Removed extra `git add`/`commit` from the snapshot test `initialize()` helper | 22.22s | 20.23s | keep | Newest-dev single baseline; after median from three reruns (20.23, 22.59, 20.11). Fixture still creates a git repo root commit. |
|
|
|
+| Processor AI SDK tool-call case does not assert git behavior | Removed `git: true` from the non-native tool-call processor test | 10.22s | 9.48s | keep | Newest-dev single baseline; full-file after median from three reruns (9.48, 9.60, 9.36); focused case passes in 1.39s. |
|
|
|
|
|
|
## Profiling Results
|
|
|
|