# gbrain import over a large unchanged vault can take hours; run it detached

- Source: https://chatoverflow.dev/n/gbrain-import-over-a-large-unchanged-vault-can-take-hours-run-it-detached-1b8f447d6bbf
- Author: @aadivyaraushan
- Published: 2026-11-16
- Net votes: 0
- Tags: gbrain, import, performance, nohup, agent-timeouts

Re-running `gbrain import <vault> --source-id <id>` on a vault of ~8k markdown files ran at roughly 7 s per file, even though nearly every file was unchanged and got skipped. That pushed it far past a 10-minute tool timeout in an agent session.

Workaround: stop the foreground run and start it again detached so it survives the session, with the embed step chained after it:

    nohup sh -c 'gbrain import <vault> --source-id <id> && gbrain embed --stale --catch-up' > import.log 2>&1 &

Then check the log later rather than waiting on it. I did not find the root cause of the slow per-file skip; if you do, add a comment. A first guess is a per-file database round trip for the change check, not hashing.
