Settings Reference¶
This page summarizes the main settings tabs and what they control.
Prerequisites¶
- Python runtime: Choose
Python worker container(recommended) orLocal interpreter/venv. - Python path: Local mode only. Interpreter used for Docling tools.
- Python env location: Local mode only. Store the venv in a shared cache or inside the plugin folder.
- Docker/Podman path: CLI path used to start Redis Stack and Python worker.
- Redis URL: Connection string for Redis Stack.
- Auto‑assign Redis port: Pick a free local port when starting Redis.
- Auto‑start Redis stack: Start Redis automatically when needed.
- Start Redis stack now: Manual start/restart button (also starts Python worker in worker mode).
Step-by-step setup guides: - Python Setup - Docker Setup
Zotero local API / web API¶
- Local API base URL: Usually
http://127.0.0.1:23119/api. - Web API settings: Optional fallback or write‑back. Set library type, library ID, and API key.
Output¶
- PDF and notes folders: Where PDFs and Zotero notes are stored in your vault.
- Frontmatter template: Editable template with
{{placeholders}}. - Note body template: Controls where annotations and Docling text appear.
- Tag sanitization: Normalize Zotero tags for Obsidian.
- Templating details: See Templating reference for the full placeholder catalog and template behavior.
Annotations¶
- Color map: Map Zotero highlight colors to callout types and headings.
- Include annotation images: Pull image‑based annotations when available.
- Zotero companion: Enable and configure the companion service.
OCR¶
- OCR engine: Auto, Tesseract, or Paddle options.
- OCR decision: When to OCR (auto, only if bad, always).
- Per‑page override: Force per‑page OCR for complex layouts.
- OCR quality threshold: Minimum text quality before OCR runs.
- OCR cleanup: Optional LLM cleanup for low‑quality OCR.
Paddle OCR (API)¶
If you select a Paddle OCR API engine, you’ll need an API key. See PaddleOCR-VL guide.
LLMs¶
- Provider profiles: Store base URL + API key once.
- Embeddings: Model selection, subchunking, context window, metadata inclusion.
- Chat: Model selection, temperature, history size, panel location.
- Cleanup: Model and thresholds for OCR cleanup.
- Tagging: Optional LLM‑generated chunk tags.
- Retrieval tuning: Agentic planner mode, query expansion, reranking, RRF, max‑per‑doc caps.
Maintenance¶
- Reindex Redis from cache: Rebuild index without re‑OCR.
- Recreate missing notes: Restore notes from cached data.
- Rebuild doc index: Refresh the index metadata cache.
- Logs: Enable file logging, open or clear the log file.
- Zotero companion tools: Install/health check shortcuts.