Repository navigation
Troubleshooting and FAQ
This guide covers common questions, edge cases, and troubleshooting steps for Berry AI Studio.
- Cause: The folder may not be configured as an active Ingestion Pipeline, or the filesystem watcher quiet period is waiting for rendering to complete.
-
Solution:
- Check that the folder is registered in the left sidebar.
- If using AUTOMATIC1111 or ComfyUI, ensure the folder mode is set to Mode C: Pipeline so write-lock debouncing is active.
- Right-click the folder in the sidebar and click "Harvest Pipeline Outputs" or "Rescan Folder" to trigger an immediate check.
- Cause: On large initial imports, background Rayon threads downscale images in batches.
-
Solution:
- Click ⚡ Activity in the bottom status bar to check if the thumbnail worker pool is still indexing.
- In Settings > Display & Safety, ensure your Thumbnail Cache Budget is set to at least
2048 MB. - Avoid running heavy video renders or gaming simultaneously during the first large folder import.
-
Cause: You may be in Semantic Mode (
🧠) instead of Syntax Mode (🔍), or filtering on an unindexed field. -
Solution:
- Check the icon in the search bar: click the brain icon to toggle back to Syntax Search (
🔍). - If searching for exact phrases with spaces, wrap your query in quotes:
prompt:"cyberpunk city". - Open the Filter Drawer (
☰ Filter) and click "Reset All" to ensure no active filters (such as a 5-star rating filter) are hiding matching results.
- Check the icon in the search bar: click the brain icon to toggle back to Syntax Search (
- Cause: Stacks can become fragmented if member images were renamed or moved externally via your OS file explorer.
-
Solution: Select the affected cards in the gallery and press
Ctrl + Gto re-group them cleanly into a single unified stack.
- Cause: ComfyUI is not running locally, or is listening on a different port.
-
Solution:
- Open Settings > Generation Interop.
- Verify that the ComfyUI Base URL matches your terminal output (default:
http://127.0.0.1:8188). - Click "Test Connection" to verify the port is open and reachable.
Yes. Berry AI Studio is free, open-source software licensed under AGPL-3.0. There are no subscriptions, paywalls, or locked features.
Yes. Berry was engineered from the ground up to support massive collections. It uses:
-
Keyset cursor deep pagination (
search_files_cursor_page) that maintains sub-millisecond query performance regardless of library size. - SQLite Write-Ahead Logging (WAL) for high-throughput non-blocking reads.
- Dynamic DOM virtualization that only renders elements visible in your viewport.
No. All scanning, metadata extraction, database storage, and AI inference (CLIP and WD14) run 100% locally on your machine. No telemetry or analytics are collected.
Yes. Dragging an image card from the gallery directly into your web browser or external desktop application transmits standard OS file drop payloads, preserving embedded metadata chunks.
Removing a folder from Berry AI Studio removes the folder and its index records from Berry's database. It never deletes or moves your physical media files on disk.
Yes. By switching from SQLite to a shared MySQL 8.0+ or PostgreSQL 14+ database in Settings > Team & Collaboration, multiple artists can connect to a shared network library with real-time change synchronization and cross-platform path mapping.