Skip to content

Fix GitHub Action timeout during Meili indexation#725

Merged
mishig25 merged 2 commits intomainfrom
fix-meili-indexation-timeout
Jan 14, 2026
Merged

Fix GitHub Action timeout during Meili indexation#725
mishig25 merged 2 commits intomainfrom
fix-meili-indexation-timeout

Conversation

@mishig25
Copy link
Contributor

Summary

Fix the GitHub Action timing out while waiting for Meilisearch indexation.

Changes:

  1. Stop waiting for indexation tasks - Removed @wait_for_task_completion decorator from add_embeddings_to_db so it submits tasks without blocking

  2. Process one library per hour - Instead of indexing all libraries at once (which overloads the database), process one library at a time with a 1-hour delay between each. Added logging to show:

    • Which library is being processed (📚 PROCESSING LIBRARY n/total: name)
    • Progress on embedding generation and uploads
    • Next library in queue
  3. Double text chunk length - Increased chunk sizes to reduce the number of embeddings:

    • CHUNK_LEN_CHARS: 2000 → 4000
    • --excerpt-length default: 1000 → 2000

- Remove wait_for_task_completion from add_embeddings_to_db to avoid timeout
- Process libraries one at a time with 1 hour delay between each
- Add logging to show which library is being indexed
- Double text chunk length (2000→4000, 1000→2000) to reduce embeddings count
@mishig25 mishig25 merged commit 894c424 into main Jan 14, 2026
4 checks passed
@mishig25 mishig25 deleted the fix-meili-indexation-timeout branch January 14, 2026 11:12
mishig25 added a commit that referenced this pull request Jan 26, 2026
Reverts the processing logic changes from commit 894c424 that:
- Removed @wait_for_task_completion decorator from add_embeddings_to_db
- Changed to process one library at a time with 1-hour delays

Keeps the chunk length increases (2000->4000 and 1000->2000) which help
reduce the total number of embeddings.
mishig25 added a commit that referenced this pull request Jan 26, 2026
Reverts the processing logic changes from commit 894c424 that:
- Removed @wait_for_task_completion decorator from add_embeddings_to_db
- Changed to process one library at a time with 1-hour delays

Keeps the chunk length increases (2000->4000 and 1000->2000) which help
reduce the total number of embeddings.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant