Gemini Scribe

Appearance

Getting Started ​

Welcome to Gemini Scribe, an Obsidian plugin that integrates Google's Gemini AI models as an intelligent agent that can actively work with your vault.

Gemini Scribe is agent-first — every conversation is powered by an AI agent with tool-calling capabilities. The agent can search files, create notes, research topics, and execute multi-step tasks autonomously while respecting your permissions.

Installation ​

  1. Open Obsidian Settings
  2. Navigate to "Community plugins"
  3. Ensure "Restricted mode" is OFF
  4. Click "Browse" and search for "Gemini Scribe"
  5. Click "Install" and then "Enable"

Manual Installation ​

  1. Download the latest release from the GitHub Releases page (you'll need main.js, manifest.json, and styles.css)
  2. Create a folder named gemini-scribe inside your vault's .obsidian/plugins/ directory
  3. Copy the downloaded files into the gemini-scribe folder
  4. In Obsidian, go to Settings → Community plugins and enable "Gemini Scribe"

Quick Start ​

  1. Get an API Key — Get your free key from Google AI Studio
  2. Connect a provider — Open Settings → Gemini Scribe → Providers, add the Gemini card, and paste your API key
  3. Route your features — Open the Features page and confirm Chat and agent (and anything else you want) is routed to Gemini — new installs default everything to Gemini automatically
  4. Initialize Context — Click "Initialize vault context" to help the agent understand your vault
  5. Start Chatting — Open Gemini chat with the ribbon icon or command palette and start giving the AI tasks!

Requires Obsidian 1.13.1 or later — the settings tab is built on Obsidian's declarative settings API.

Prefer running models locally, or already have an OpenAI or Anthropic key? ​

Gemini Scribe also supports Ollama as a provider so you can use local models such as Llama 3.2, Qwen 2.5, or Gemma 3 without an API key — see the Ollama Setup Guide. If you'd rather use your own OpenAI API key for chat or GPT Image generation, or point the plugin at an OpenAI-compatible server such as LM Studio or MLX, add the OpenAI card on the Providers page — see the OpenAI Setup Guide. To use Claude, add your Anthropic API key on the Anthropic card — see the Anthropic Setup Guide. Google Search, Google Maps, URL Context, Deep Research, and RAG remain Gemini-only, but each can be pointed at Gemini individually on the Features page while chat or image generation uses another provider.

Feature Overview ​

Agent mode (Core Feature) ​

An AI assistant that can actively work with your vault through tool calling.

  • Search and read files in your vault
  • Create, modify, and organize notes
  • Research topics with web search and URL fetching
  • Execute complex workflows autonomously
  • Respect your permissions with granular controls

Example tasks:

  • "Find all notes tagged with #important and create a summary"
  • "Research quantum computing and create a new note with your findings"
  • "Organize my meeting notes from this week into a weekly summary"

Full Agent mode Guide →

Custom Prompts ​

Create specialized AI behaviors for different workflows — technical documentation, creative writing, research formatting, and more.

Custom Prompts Guide →

AI-Assisted Writing ​

Work with selected text using AI — rewrite, explain, or ask questions about any selection.

AI Writing Guide →

IDE-Style Completions ​

Get real-time, context-aware text suggestions as you type, similar to code completion in programming IDEs.

Completions Guide →

Smart Summarization ​

Generate concise, one-sentence summaries stored in frontmatter for easy access and organization.

Summarization Guide →

Context System ​

Add specific notes as persistent context for your agent sessions using @ mentions.

Context System Guide →

Best Practices ​

  1. Start with the Agent — The agent is your primary interface. Be specific about what you want it to do and let it break down complex tasks into steps.
  2. Initialize vault context — Use "Initialize vault context" to help the agent understand your vault. Update it periodically as your vault grows.
  3. Use Context Files — Add relevant notes as context with @ mentions for focused sessions.
  4. Set Appropriate Permissions — Configure which operations require confirmation. Balance convenience with safety.
  5. Leverage Persistent Sessions — Continue conversations across Obsidian restarts and build on previous work.

After an Update ​

The first time Obsidian starts with a new version of Gemini Scribe, a release notes modal opens once, showing that version's highlights (or a short generic message when the version has no bundled notes) and a link to the full release notes on GitHub. It isn't shown on a fresh install.

To see it again later, run Gemini Scribe: View release notes from the Command Palette. It shows the notes for the version you have installed.

Troubleshooting ​

API Key Issues

If you see a message about a missing or inaccessible API key:

  1. Get a free Gemini API key — Visit Google AI Studio and click "Create API Key". A Google account is all you need.
  2. Enter your key in settings — Open Obsidian Settings, navigate to Gemini Scribe → Providers → Gemini card, and paste your key into the API Key field. The key is stored securely using Obsidian's secret storage.
  3. Key not working after entry? — If you see "Could not retrieve your API key from secure storage", try clearing the field and re-entering your key. This can happen if Obsidian's secure storage was reset (e.g., after a system update or vault migration).
  4. Check the developer console — Press Ctrl+Shift+I (Windows/Linux) or Cmd+Opt+I (macOS) to open the console and look for error messages prefixed with [Gemini Scribe].

Agent Not Responding

  • Check API key validity in Settings
  • Verify internet connection
  • Ensure your model supports tool calling (all current models support this)

Tools Failing

  • Check file permissions and paths
  • Verify files exist and are accessible
  • System folders (Obsidian's configuration folder — .obsidian by default, or a renamed one — and plugin folders) are protected from modifications

Poor Quality Output

  • Add specific notes as context files
  • Be more specific in your requests
  • Try Gemini 2.5 Pro for more capable responses

Performance Issues

  • Use Gemini Flash for faster responses
  • Reduce context file count for quicker processing
  • Break up large requests into smaller tasks

Session Issues

  • Try creating a new session
  • Check console (Ctrl/Cmd + Shift + I) for errors
  • Verify session files aren't corrupted

For detailed configuration, see the Settings Reference.

Further Reading ​