Skip to content

Quick Start

This page walks you through a minimal, working setup so you can import a Zotero item and ask your first question.

Prerequisites

  • Obsidian (desktop)
  • Zotero 7, 8, or 9 (desktop)
  • Docker Desktop or Podman (for Redis Stack)
  • A local or cloud OpenAI-compatible model provider

1) Enable the Zotero local API (read-only)

In Zotero:

  1. Open Settings → Advanced.
  2. Enable Allow other applications on this computer to communicate with Zotero.

No API key is needed for the local API.

2) Install the plugin

Recommended:

  1. Open Settings → Community plugins in Obsidian.
  2. Turn off Restricted mode if needed.
  3. Click Browse and search for Zotero Research Assistant.
  4. Click Install, then Enable.

Alternatives:

  • Manual install:

  • Download main.js, manifest.json, and styles.css from the latest release.

  • Place them in your vault at:

    • <vault>/.obsidian/plugins/zotero-redisearch-rag/
    • Enable the plugin. Bundled helper files are written into tools/ on first load.

  • BRAT: Use BRAT to install jmiba/zotero-redisearch-rag.

  • Build from source:

bash npm install npm run build Then copy the plugin folder into your vault as above.

3) Start Redis Stack

In Obsidian, run:

  • Command palette → Start Redis Stack (Docker/Podman Compose)
  • Command palette → Recreate Redis Stack (Pull Configured Image) if you need to refresh the Redis container after a plugin update

If Docker or Podman isn’t running, start it first. First startup can be slow because Docker/Podman may need to pull images and build worker dependencies. On slower networks/machines this can take 10+ minutes.

4) Start a model provider

Use any OpenAI‑compatible local server or a cloud provider.

Local options:

  • LM Studio: start the server and note the model ID shown in LM Studio.
  • Ollama: start the daemon and use the OpenAI‑compatible endpoint at http://localhost:11434/v1.

Cloud options:

  • OpenAI or OpenRouter with your API key.

5) Set up Python for Docling

Default (recommended):

  1. Open Settings → Community plugins → Zotero Research Assistant → Prerequisites.
  2. Keep Advanced Python runtime options disabled (default).
  3. Click Start Redis stack now.
  4. Wait for startup to finish (first run may take 10+ minutes while images and worker dependencies are prepared).

Fallback (terminal, local runtime mode only):

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

If you need local mode, use Settings → Maintenance → Python Runtime → Use local runtime (legacy) first.

For full details, see Python Setup.

6) Configure the plugin

Open Settings → Community plugins → Zotero Research Assistant and set:

  • Prerequisites: Docker/Podman path, Redis URL, and (optional) advanced local runtime controls
  • LLM provider profiles (optional but recommended)
  • Embeddings and Chat models
  • Output folders (PDFs, notes, chats)

[!TIP] Templating: Customize note frontmatter/body with placeholders. See Templating reference.

7) Import your first item

Run:

  • Command palette → Import Zotero item and index (Docling → RedisSearch)

Pick an item with a PDF attachment. The plugin will create a note and index its chunks.

8) Ask your first question

Run:

  • Command palette → Ask my Zotero library (RAG via RedisSearch)

Answers include citations that link back to the exact chunk in your note or PDF.

If you want a deeper walkthrough, continue to How It Works and Daily Workflow.