Where does the bug appear (feature/product)?
Background Agent (GitHub, Slack, Web, Linear)
BugBot
Describe the Bug
BugBot reviews are consistently very slow (~9–12 minutes), even on relatively small PRs.
We investigated locally and found that the latency is almost entirely in the “bug detection” phase rather than context gathering or rule loading.
This is occurring via the GitHub integration (not Cursor IDE or CLI).
Steps to Reproduce
- Open a PR (even with small diff size)
- Let BugBot run via GitHub checks
- Observe total runtime (~9–12 minutes)
- Expand BugBot details and inspect timing breakdown
Expected Behavior
BugBot reviews should complete in a few minutes, especially for small PRs.
Screenshots / Screen Recordings
Operating System
N/A (runs via GitHub integration / Background Agent)
Version Information
N/A (not using Cursor IDE or CLI — this is BugBot via GitHub)
For AI issues: which model did you use?
Unknown (BugBot internal model)
For AI issues: add Request ID with privacy disabled
Request ID:
serverGenReqId_1b1456ca-2af2-404d-a819-5b3ed8d43f01
Additional Information
We verified the following:
-
Only one BugBot file:
apps/octogen-voyager/.cursor/BUGBOT.md -
Contents are minimal (no links or external references):
- Use Pydantic models instead of dicts
- Look for encapsulation violations
- Look for DRY violations
- Look for duplicated code
-
No root-level .cursor/BUGBOT.md
-
We are not linking to any .md or .mdc files
-
.cursorignore is configured
-
Regular .cursor/rules/*.mdc files exist but are not referenced by BugBot
Timing breakdown from BugBot UI:
- Gathered PR context: 3s
- Completed bug detection: 7m 52s
- Posted results: 1s
Observation:
Nearly all latency is in the bug detection phase, not repo scanning or rule loading.
This suggests the bottleneck may be in the model/inference step rather than repository configuration.
Does this stop you from using Cursor?
Sometimes - I can sometimes use Cursor
