> ## 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.

# Quickstart

> Get started with Local GPT in minutes - from installation to your first AI action

## Get your first AI action running

This guide walks you through using Local GPT for the first time, from installation to running your first AI-powered action.

<Note>
  Already installed? Skip to [Using the Context Menu](#step-3-run-your-first-action).
</Note>

## Prerequisites

Before you begin, you'll need:

* **Obsidian** 0.15.0 or higher
* **Local GPT** and **AI Providers** plugins installed ([Installation guide](/installation))
* At least one AI provider configured (Ollama, OpenAI, etc.)

## Step 1: Install and Configure

<Steps>
  <Step title="Install Required Plugins">
    Install both the **AI Providers** and **Local GPT** plugins from Obsidian's Community Plugins.

    See the [Installation guide](/installation) for detailed instructions.
  </Step>

  <Step title="Set Up an AI Provider">
    **For local/offline use (Recommended):**

    1. Install [Ollama](https://ollama.ai)
    2. Pull a model:
       ```bash theme={null}
       ollama pull gemma2
       ```
    3. In Obsidian, go to **Settings** → **AI Providers** → **Add Provider** → **Ollama**
    4. Set URL: `http://localhost:11434`
    5. Select model: `gemma2:latest`

    **For cloud use:**

    Configure OpenAI or another provider in **Settings** → **AI Providers**.
  </Step>

  <Step title="Configure Local GPT">
    1. Open **Settings** → **Local GPT**
    2. Under **Main Provider**, select the AI provider you just configured
    3. (Optional) Configure **Vision Provider** and **Embedding Provider**
  </Step>
</Steps>

## Step 2: Set Up Hotkeys (Recommended)

For the best experience, assign keyboard shortcuts:

<Steps>
  <Step title="Open Hotkeys">
    Go to **Settings** → **Hotkeys**
  </Step>

  <Step title="Assign Context Menu Hotkey">
    Search for `Local GPT: Show context menu` and assign a shortcut like:

    * Mac: `Cmd + M`
    * Windows/Linux: `Ctrl + M`
  </Step>

  <Step title="Assign Action Palette Hotkey">
    Search for `Local GPT: Action Palette` and assign a shortcut like:

    * Mac: `Cmd + J`
    * Windows/Linux: `Ctrl + J`
  </Step>
</Steps>

<Tip>
  The Action Palette is perfect for one-off prompts, while the context menu is ideal for repeated actions.
</Tip>

## Step 3: Run Your First Action

### Using the Context Menu

The context menu provides quick access to your saved actions:

<Steps>
  <Step title="Select Text">
    In any note, highlight some text you want to work with. For example:

    ```markdown theme={null}
    Local GPT is a plugin for Obsidian that lets you use AI
    ```
  </Step>

  <Step title="Open Context Menu">
    Right-click the selected text (or use your hotkey `Cmd/Ctrl + M`)
  </Step>

  <Step title="Choose an Action">
    Select an action from the menu:

    * **Continue writing** - Expand your thoughts
    * **Summarize** - Create a concise summary
    * **Fix spelling and grammar** - Proofread the text
    * **Find action items** - Extract tasks
    * **General help** - Use text as a prompt
  </Step>

  <Step title="See Results">
    Wait for the AI to process (you'll see a spinner). The result appears below your text:

    ```markdown theme={null}
    Local GPT is a plugin for Obsidian that lets you use AI

    Local GPT is a plugin for Obsidian that integrates AI-powered 
    language models directly into your note-taking workflow. It 
    supports both local models through Ollama and cloud providers 
    like OpenAI, giving you flexibility in how you process your notes.
    ```
  </Step>
</Steps>

<img className="rounded-lg border border-gray-200 dark:border-gray-800" src="https://github.com/user-attachments/assets/8d32dc1d-9431-4a16-9336-c45e853a3242" alt="Action Palette interface" />

### Using the Action Palette

The Action Palette is perfect for one-time prompts and quick questions:

<Steps>
  <Step title="Open Action Palette">
    Press your hotkey (`Cmd/Ctrl + J`) or use the command palette to run `Local GPT: Action Palette`
  </Step>

  <Step title="Type Your Prompt">
    Type a question or instruction directly:

    ```
    List 5 benefits of using local AI models
    ```
  </Step>

  <Step title="(Optional) Select Context Files">
    Click the folder icon to attach specific notes or PDFs as context
  </Step>

  <Step title="(Optional) Choose a System Prompt">
    Click the gear icon to apply a saved system prompt from your actions
  </Step>

  <Step title="Submit">
    Press `Enter` to run the prompt. The AI response appears at your cursor position.
  </Step>
</Steps>

<Tip>
  The Action Palette remembers your selected provider and model across sessions.
</Tip>

## Step 4: Try Enhanced Actions (RAG)

Enhanced Actions use RAG to pull context from your vault:

<Steps>
  <Step title="Configure Embedding Provider">
    1. Install an embedding model:
       ```bash theme={null}
       ollama pull nomic-embed-text
       ```
    2. Add it as a provider in **AI Providers**
    3. Set it as **Embedding Provider** in **Local GPT** settings
  </Step>

  <Step title="Create Linked Notes">
    Create a note with links to other notes:

    ```markdown theme={null}
    # Project Overview

    See [[Requirements]] and [[Timeline]] for details.

    Can you summarize the requirements and timeline?
    ```
  </Step>

  <Step title="Run Action with Context">
    1. Select the question text
    2. Open the context menu
    3. Choose an action like **General help**

    Local GPT will automatically:

    * Read the content of `Requirements.md` and `Timeline.md`
    * Include that context in the AI prompt
    * Generate a response based on your entire knowledge base
  </Step>
</Steps>

<Note>
  RAG works with wiki links (`[[note]]`), markdown links, backlinks, and even PDF files. Learn more in the [Enhanced Actions guide](/features/enhanced-actions).
</Note>

## Step 5: Work with Images (Vision)

If you have a vision-capable model configured:

<Steps>
  <Step title="Configure Vision Provider">
    1. Pull a vision model:
       ```bash theme={null}
       ollama pull llava
       ```
    2. Add it as a provider in **AI Providers**
    3. Set it as **Vision Provider** in **Local GPT** settings
  </Step>

  <Step title="Embed an Image">
    Add an image to your note:

    ```markdown theme={null}
    ![[screenshot.png]]
    What UI improvements would you suggest?
    ```
  </Step>

  <Step title="Select Image and Text">
    Highlight both the image embed and your question
  </Step>

  <Step title="Run Action">
    Choose **General help** from the context menu. The vision model will analyze the image and respond.
  </Step>
</Steps>

## Common Workflows

<AccordionGroup>
  <Accordion icon="pen" title="Expanding Draft Notes">
    1. Write a rough outline or bullet points
    2. Select the text
    3. Use **Continue writing** to expand it into full paragraphs
  </Accordion>

  <Accordion icon="clipboard-list" title="Extracting Tasks from Meeting Notes">
    1. Write or paste meeting notes
    2. Select the entire note
    3. Use **Find action items** to extract actionable tasks
  </Accordion>

  <Accordion icon="language" title="Fixing Grammar and Style">
    1. Write your content
    2. Select a paragraph or section
    3. Use **Fix spelling and grammar** to polish it
    4. The text is replaced with the corrected version
  </Accordion>

  <Accordion icon="sparkles" title="Quick Research Questions">
    1. Open the Action Palette (`Cmd/Ctrl + J`)
    2. Type a question like "Explain quantum entanglement in simple terms"
    3. Press Enter - the answer appears at your cursor
  </Accordion>

  <Accordion icon="book" title="Summarizing Linked Notes">
    1. Create a note that links to multiple other notes
    2. Select a question or prompt
    3. Run **General help** - Local GPT reads all linked notes for context
  </Accordion>
</AccordionGroup>

## Tips for Better Results

<CardGroup cols={2}>
  <Card title="Be Specific" icon="bullseye">
    Instead of "improve this", try "make this more concise" or "add technical details"
  </Card>

  <Card title="Use Context" icon="link">
    Link to relevant notes to give AI more context about your work
  </Card>

  <Card title="Iterate" icon="rotate">
    Run actions multiple times with different prompts to refine output
  </Card>

  <Card title="Create Custom Actions" icon="wand-magic-sparkles">
    Save frequently used prompts as custom actions for quick access
  </Card>
</CardGroup>

## Keyboard Shortcuts Reference

| Action                | Default        | Description                       |
| --------------------- | -------------- | --------------------------------- |
| **Show Context Menu** | `Cmd/Ctrl + M` | Open action menu on selected text |
| **Action Palette**    | `Cmd/Ctrl + J` | Open palette for one-time prompts |
| **Escape**            | `Esc`          | Cancel running AI request         |

<Tip>
  You can customize these shortcuts in **Settings** → **Hotkeys**
</Tip>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Nothing happens when I select text">
    * Make sure Local GPT is enabled in Community Plugins
    * Verify you have a Main Provider configured
    * Check if Ollama is running: `ollama list`
  </Accordion>

  <Accordion title="AI responses are slow">
    * Local models depend on your hardware (CPU/GPU)
    * Try a smaller model like `gemma2:2b` or `phi3`
    * Consider using a cloud provider for faster responses
  </Accordion>

  <Accordion title="RAG/Enhanced Actions not working">
    * Ensure you have an Embedding Provider configured
    * Pull an embedding model: `ollama pull nomic-embed-text`
    * Check that your notes actually contain links to other files
  </Accordion>

  <Accordion title="Vision features not working">
    * Configure a separate Vision Provider in settings
    * Use a vision model like `llava` or `bakllava`
    * Ensure images are embedded with `![[image.png]]` syntax
  </Accordion>
</AccordionGroup>

For more help, see the [Troubleshooting guide](/advanced/troubleshooting).

## Next Steps

<CardGroup cols={2}>
  <Card title="Create Custom Actions" icon="wand-magic-sparkles" href="/guides/creating-custom-actions">
    Build actions tailored to your specific workflows
  </Card>

  <Card title="Master the Action Palette" icon="palette" href="/features/action-palette">
    Learn advanced Action Palette features
  </Card>

  <Card title="Understand RAG" icon="brain" href="/advanced/rag-system">
    Deep dive into how Enhanced Actions work
  </Card>

  <Card title="Prompt Templating" icon="code" href="/guides/prompt-templating">
    Use advanced prompt templates with variables
  </Card>
</CardGroup>