Indexing Stuck

What to do when an index doesn't complete or fails.

A first index usually takes a few minutes. Incremental indexes are much faster. The repository's status on the Repositories page is one of pending, indexing, indexed, or failed.

Status stays "pending"

The index is queued but hasn't started. Brief queueing is normal. If it stays pending for more than ten minutes, re-trigger the index from your editor (below). If it still won't start, contact support through feedback.

Status stays "indexing"

The index is running. A first index on a very large repository takes longer. An index that runs past 10 minutes times out and is marked failed.

Status is "failed"

When an index fails, Scrubby emails you with the error. The most common causes:

The repository can't be read. Confirm the GitHub App is installed and this repository is in its install scope, or that you're indexing a repository you can access on GitHub.
The index timed out. Very large repositories can exceed the 10-minute limit. Re-trigger the index; if it keeps timing out, contact support.
History changed mid-index. A force-push or rewind during indexing can invalidate the run. Re-trigger it.

Re-triggering an index

Ask your editor to re-index:

"Re-index this repo with Scrubby."

If the GitHub App is installed, a merge to the default branch also re-indexes automatically. See Indexing.

Forcing a full re-classification

Incremental indexes skip domain reclassification unless a dependency file changed or 20 or more new files were added. After a major restructure, force a full pass:

"Re-index this repo with Scrubby, full (not incremental)."

Some files are missing from the index

Certain files are intentionally left out of deep analysis:

Minified and generated files (.min.js, .min.css, .map, .d.ts).
Lockfiles and vendored dependencies (*.lock, vendor/, node_modules/, __pycache__/).
Files not committed and pushed to the indexed branch — Scrubby reads from GitHub, not your local working tree.

Domains look wrong after indexing

Domains and their connections keep refining as you use Scrubby. The connection graph adjusts weights from the reviews you run. If the initial domains look clearly off, send the repository name through feedback.