> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/pfrankov/obsidian-local-gpt/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> Meet Local GPT - your privacy-focused AI assistant for Obsidian with local LLM support, RAG capabilities, and vision features

## What is Local GPT?

Local GPT is the most casual AI assistant for Obsidian, designed with privacy and offline access in mind. It brings powerful AI capabilities directly into your note-taking workflow, allowing you to transform, enhance, and interact with your notes using local or cloud-based language models.

<img className="rounded-lg border border-gray-200 dark:border-gray-800" src="https://github.com/pfrankov/obsidian-local-gpt/assets/584632/724d4399-cb6c-4531-9f04-a1e5df2e3dad" alt="Local GPT demo showing AI assistance in action" />

## Why Local GPT?

<CardGroup cols={2}>
  <Card title="Maximum Privacy" icon="shield-halved" color="#906bda">
    Run AI models completely offline with Ollama or use OpenAI-compatible providers. Your notes stay on your machine - no data leaves unless you want it to.
  </Card>

  <Card title="Seamless Integration" icon="plug" color="#906bda">
    Works naturally within your Obsidian workflow. Select text and pick an action from the context menu, or use the Action Palette for quick one-time tasks.
  </Card>

  <Card title="Context-Aware (RAG)" icon="link" color="#906bda">
    Automatically pulls context from linked notes, backlinks, and even PDF files to give AI deeper understanding of your knowledge base.
  </Card>

  <Card title="Extensible" icon="puzzle-piece" color="#906bda">
    Create custom actions for your specific needs or browse and install community-created actions with automatic updates.
  </Card>
</CardGroup>

## Key Features

### Privacy-First Architecture

Local GPT is built with privacy as a core principle. You can:

* Run models completely offline with [Ollama](https://ollama.ai)
* Use any OpenAI-compatible API endpoint
* Keep your notes and AI interactions entirely local
* Choose different providers for different tasks

### Enhanced Actions with RAG

<Note>
  RAG (Retrieval-Augmented Generation) allows AI to understand context from your entire knowledge base, not just the selected text.
</Note>

Local GPT can automatically enhance AI responses by:

* Extracting context from wiki links (`[[linked note]]`) and markdown links
* Following backlinks to related notes
* Processing PDF files linked in your notes
* Traversing your knowledge graph up to 10 levels deep

<img className="rounded-lg border border-gray-200 dark:border-gray-800 w-96" src="https://github.com/user-attachments/assets/5fa2ed36-0ef5-43b0-8f16-07588f76d780" alt="Enhanced Actions with RAG context" />

### Vision Support

Analyze images embedded in your notes with vision-capable models:

```markdown theme={null}
![[image.png]]
What's in this image?
```

Select the text including the image reference, and Local GPT will send both to a vision model like `bakllava` or `llava`.

<img className="rounded-lg border border-gray-200 dark:border-gray-800 w-96" src="https://github.com/pfrankov/obsidian-local-gpt/assets/584632/a05d68fa-5419-4386-ac43-82b9513999ad" alt="Vision model analyzing an image in Obsidian" />

### Community Actions

Browse and install pre-built actions created by the community:

<img className="rounded-lg border border-gray-200 dark:border-gray-800 w-96" src="https://github.com/user-attachments/assets/5337e74c-864b-45cb-82e0-2c32bbbfa3ed" alt="Community Actions in settings" />

**Automatic Updates**: Community actions update automatically to the latest version. To preserve your changes or stick with a specific version, simply modify the **Prompt** or **System Prompt** - this disables auto-updates for that action.

<Tip>
  Changing only the **Name** of an action does not disable auto-updates.
</Tip>

## Default Actions

Local GPT comes with these built-in actions:

* **Continue writing** - Expand and complete your thoughts
* **Summarize text** - Create concise summaries
* **Fix spelling and grammar** - Proofread and correct text
* **Find action items** - Extract tasks and to-dos
* **General help** - Use selected text as a prompt for any purpose
* **New System Prompt** - Generate custom system prompts for new actions

## Multilingual Support

Local GPT supports 11 languages out of the box:

<CardGroup cols={3}>
  <Card icon="language">English</Card>
  <Card icon="language">Chinese</Card>
  <Card icon="language">Dutch</Card>
  <Card icon="language">French</Card>
  <Card icon="language">German</Card>
  <Card icon="language">Italian</Card>
  <Card icon="language">Japanese</Card>
  <Card icon="language">Korean</Card>
  <Card icon="language">Portuguese</Card>
  <Card icon="language">Russian</Card>
  <Card icon="language">Spanish</Card>
</CardGroup>

## How It Works

<Steps>
  <Step title="Select or Invoke">
    Highlight text in your note or open the Action Palette with a hotkey (`Cmd/Ctrl + J` by default).
  </Step>

  <Step title="Choose Action">
    Pick from default actions, custom actions, or type a one-time prompt in the Action Palette.
  </Step>

  <Step title="AI Processes">
    The plugin sends your text and any linked context to your configured AI provider.
  </Step>

  <Step title="Get Results">
    AI-generated content appears directly in your note, ready to use or refine.
  </Step>
</Steps>

## Next Steps

<CardGroup cols={2}>
  <Card title="Install Local GPT" icon="download" href="/installation">
    Install the plugin and required dependencies
  </Card>

  <Card title="Quick Start Guide" icon="rocket" href="/quickstart">
    Get your first AI action running in minutes
  </Card>

  <Card title="Configure AI Providers" icon="gear" href="/guides/ai-providers-setup">
    Set up Ollama, OpenAI, or other providers
  </Card>

  <Card title="Create Custom Actions" icon="wand-magic-sparkles" href="/guides/creating-custom-actions">
    Build actions tailored to your workflow
  </Card>
</CardGroup>