Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Keep scrolling to the end of the reasoning transcript in the thread settings driver (#11575)

Daniel Hane8d6ef9bed6a
diff --git a/tests/studio/playwright_thread_scoped_settings.py b/tests/studio/playwright_thread_scoped_settings.py
index b4d6c5ea1bb..1ae5fd8cc24 100644
--- a/tests/studio/playwright_thread_scoped_settings.py
+++ b/tests/studio/playwright_thread_scoped_settings.py
@@ -356,7 +356,16 @@ def check_reasoning_transcript(page, token):
viewport = page.locator(".aui-thread-viewport")
box = viewport.bounding_box()
page.mouse.move(box["x"] + box["width"] / 2, box["y"] + box["height"] / 2)
- page.mouse.wheel(0, 100000)
+ # One wheel is not the end of a virtualized transcript. It stops at the bottom of the height
+ # ESTIMATED for the passages not yet mounted, and measuring the ones it mounts grows the
+ # transcript under it: measured against this build, the first wheel stopped at scrollTop 8468
+ # of 15637 and the second reached the end. On CI the retries then waited on a window that
+ # nothing was going to move. A reader keeps scrolling; so does this, bounded.
+ for _ in range(12):
+ page.mouse.wheel(0, 100000)
+ page.wait_for_timeout(250)
+ if "Step 0399." in body.inner_text():
+ break
expect(body).to_contain_text("Step 0399.")
assert len(body.inner_text()) < 20000, "scrolling mounted the entire trace"
expect(page.get_by_text("The final answer stays separate.", exact = True)).to_be_visible()