Overview

7 days30 days

Latest CI Pipeline Executions

Fuzzy

Succeeded
818-flaky-e2e-multimodal-+-approval-flow-tests-intermittently-get-empty-responses-chatstream-fatal-from-aimock
e74551fe test(e2e): fix flaky multimodal + approval-flow specs (#818) The multimodal and approval-flow E2E specs intermittently failed with empty assistant responses (surfaced as `chatStream fatal`). Root cause was a test-harness race, not aimock or the library — every request the harness actually sent succeeded. Multimodal: `sendMessageWithImage` typed into a controlled React input and then attached the image, which auto-sends using that input's value. Under CPU load `pressSequentially` dropped leading characters, so the prompt reached aimock truncated (e.g. "cribe this image") and 404'd as "No fixture matched"; and React state could lag the committed DOM value so the auto-send fired with empty text and dispatched no request at all. - helpers: type until the full prompt is committed, then retry the typing + attach until the send actually fires (user bubble renders). - ChatUI: the image auto-send reads the live input DOM value instead of possibly-stale React state. Approval-flow: `runTest` treated the optimistic user-message bump as "run started", returning before any real stream activity — a stalled run then timed out waiting for an approval that never appeared. - runTest: require real stream activity (loading on, a tool call, completion, or an assistant message) before returning, and retry the click otherwise. Verified: multimodal 200/200 with 0 flaky at 20x (4 workers, retries=2) and at 6 workers/retries=0; approval-flow 450/450 at 25x/8 workers/retries=0; full E2E suite green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
by Tom Beckenham
Succeeded
818-flaky-e2e-multimodal-+-approval-flow-tests-intermittently-get-empty-responses-chatstream-fatal-from-aimock
e74551fe Merge 5b82deab6d515856dbef771538d013c76a6ff929 into df40512bda853f714b2913bee630ec599e664207
by Tom Beckenham