Skip to content

Configuration

[!NOTE] All configuration is managed via the Web UI at /settings. Environment variables can be used for initial bootstrapping, but values set in the database (via UI) take precedence.

Web UI Settings

The most convenient way to manage configuration is via the Settings page in the Web UI. Changes made here are applied instantly (triggering a soft restart).

Integrations

KOSync (KOReader)

  • Server: Your KOSync server URL (e.g., https://koreader.mydomain.com/api/koreader).
  • Username: Your KOSync username.
  • Password: Your KOSync password.
  • Save Hash Method: How KOReader calculates document integrity. Keep as default (content) unless you know what you're doing.

Storyteller

  • Storyteller URL: URL to your Storyteller instance.
  • Storyteller API Key: Your API key (if required).
  • Sync Mode: REST API (default).

Booklore

  • Booklore URL: URL to your Booklore/Calibre-Web instance.
  • API Key: For managing shelves/collections.

Transcription Settings

Configure the engine used for audio-to-text alignment.

Setting Default Description
Provider local local (faster-whisper), deepgram, or whisper_cpp (via server).
Whisper Model tiny Model size (tiny, base, small, medium, large).
Whisper Device auto auto, cpu, or cuda. See GPU Support below.
Compute Type auto Precision (int8, float16, float32). Use float16 for GPU.

Deepgram

  • API Key: Your Deepgram API Key.
  • Model: Specific Deepgram model tier (e.g., nova-2).

WhisperCPP

  • Server URL: URL to your running whisper.cpp server if using that provider.

Sync Tuning

Advanced settings to fine-tune the synchronization logic.

Setting Default Description
Sync Period (Minutes) 5 How often the background sync runs.
ABS Delta (Seconds) 30 Minimum progress change (in seconds) required to trigger an update from ABS.
KoSync Delta (%) 0.005 Minimum progress change (0.5%) required to trigger an update from KOReader.
Fuzzy Match Threshold 0.80 (0.0-1.0) Confidence required for text matching (80%).
Job Retries 5 How many times to retry failed transcription jobs.

Book Linker (Filesystem)

Required only if using the Book Linker tool to prepare files for Storyteller.

Variable Default Description
LINKER_BOOKS_DIR /linker_books Source directory for ebooks.
AUDIOBOOKS_DIR /audiobooks Source directory for audiobooks (usually read-only).
PROCESSING_DIR /processing Temporary working directory for copying files.
STORYTELLER_INGEST_DIR /linker_books Destination directory where Storyteller picks up new books.

Toggles

  • Sync ABS Ebook: If enabled, also syncs progress to the ebook item in ABS (if you have both mapped). This allows you to read the ebook in the ABS web reader and have that progress sync to KOReader.
  • XPath Fallback: Strategy for handling position lookups when exact paths fail.

GPU Support (Optional)

For significantly faster transcription (when using local provider), you can enable NVIDIA GPU acceleration.

1. Install NVIDIA Container Toolkit

Follow the official guide to install the NVIDIA Container Toolkit for your host OS.

2. Update Docker Compose

Uncomment/Add the deploy section to your docker-compose.yml:

services:
  abs-kosync:
    # ... other config ...
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

3. Configure Settings

In the Web UI Settings: 1. Set Whisper Device to cuda. 2. Set Whisper Compute Type to float16. 3. Set Whisper Model to small or medium (GPUs can handle larger models easily).