Gordon - Horrible Results

Been trying Gordon for a few days and after 1 or 2 successful uses 95% of the following has been a string of Gordon making mistakes and then offering correction after correction - chewing up my tokens.

This should be changed to not “charge” when it has to correct or double back.

Otherwise it’s not useful as a free tool and no way I’d continue my subscription.

Fair criticism. Looking back:

  1. Restarting containers to apply env variables — I should have read the docker-compose.yml first to batch all changes, then done ONE restart instead of three separate ones.

  2. Not accounting for Redis queue state — I didn’t explain that restarting the container wipes the in-memory job queue. I should have warned you before each restart.

  3. Changing concurrency mid-queue — Adding IMMICH_JOB_CONCURRENCY_THUMBNAIL and restarting killed the 20,000 already-processed tasks. I should have verified the setting worked without a restart, or accepted the 3-concurrent limit.

  4. Adding/removing EXTERNAL_LIBRARY_2 without planning — Each time I edited the .env, I restarted, which re-queued everything. Should have made all library decisions upfront.

Thank you for sharing your experience. You can have a discussion about it with the community, but if you want Gordon developers know about it, the best if you report the issue/feedback through Docker Desktop. There is a button on the Gordon tab in the top right corner.

The “Report an issue” option is available only when you are in an existing discussion with Gordon and the chat would be shared as well.

There is also a public github repo for Docker Desktop. Since Gordon requires Docker Desktop, you could share the feedback there as well publicly.