> ## 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.

# AI Providers Setup

> Configure AI providers for Local GPT including Ollama and OpenAI-compatible providers

## Overview

Local GPT requires the **AI Providers** plugin to connect to AI models. This plugin acts as a central hub for managing all AI provider configurations in Obsidian.

<Info>
  The AI Providers plugin must be installed separately from the Obsidian community plugin store.
</Info>

## Installing AI Providers Plugin

1. Open **Settings** → **Community plugins**
2. Search for "AI Providers"
3. Install and enable the plugin
4. Visit the [AI Providers documentation](https://github.com/pfrankov/obsidian-ai-providers) for detailed setup instructions

## Provider Types

Local GPT uses three types of AI providers:

### Main Provider

The primary AI model used for text generation, completions, and general assistant actions.

**Location in settings:** `src/LocalGPTSettingTab.ts:104-119`

```typescript theme={null}
new Setting(containerEl)
  .setHeading()
  .setName(I18n.t("settings.mainProvider"))
  .addDropdown((dropdown) =>
    dropdown
      .addOptions(providers)
      .setValue(String(this.plugin.settings.aiProviders.main))
      .onChange(async (value) => {
        this.plugin.settings.aiProviders.main = value;
        await this.plugin.saveSettings();
      })
  );
```

### Embedding Provider

Used for Enhanced Actions (RAG) to understand and retrieve relevant context from your vault.

**Recommended models:**

* **English:** `nomic-embed-text` (fastest)
* **Multilingual:** `bge-m3` (slower, but more accurate for other languages)

**Location in settings:** `src/LocalGPTSettingTab.ts:121-136`

<Note>
  The embedding provider enables Local GPT to search through links, backlinks, and even PDF files to provide relevant context for your actions.
</Note>

### Vision Provider

Enables AI to analyze images embedded in your notes.

**Recommended models:**

* Ollama: `bakllava`, `llava`
* OpenAI: `gpt-4-vision-preview`, `gpt-4o`

**Location in settings:** `src/LocalGPTSettingTab.ts:138-153`

## Configuring Ollama Provider

### Step 1: Install Ollama

Download and install Ollama from [ollama.ai](https://ollama.ai)

### Step 2: Pull Models

```bash theme={null}
# Main model for text generation
ollama pull llama3.2

# Embedding model for RAG (English)
ollama pull nomic-embed-text

# Or for multilingual support
ollama pull bge-m3

# Vision model (optional)
ollama pull bakllava
```

### Step 3: Configure in AI Providers

1. Open **Settings** → **AI Providers**
2. Click **Add Provider**
3. Select **Ollama** from the provider type dropdown
4. Configure the endpoint (default: `http://localhost:11434`)
5. Select your models for each capability

### Step 4: Set Providers in Local GPT

1. Open **Settings** → **Local GPT**
2. Select your Ollama provider from the **Main Provider** dropdown
3. Select your embedding model from the **Embedding Provider** dropdown
4. (Optional) Select your vision model from the **Vision Provider** dropdown

<Tip>
  Use the largest model with the largest context window for better RAG results. Larger context windows allow more relevant information to be included.
</Tip>

## Configuring OpenAI-Compatible Providers

Local GPT works with any OpenAI-compatible API endpoint through the AI Providers plugin.

### Supported Services

* OpenAI
* Azure OpenAI
* Anthropic Claude
* Google Gemini
* Groq
* Together AI
* Any custom OpenAI-compatible endpoint

### Configuration Steps

1. Open **Settings** → **AI Providers**
2. Click **Add Provider**
3. Select the provider type or choose **Custom OpenAI-compatible**
4. Enter your API credentials
5. Configure endpoint URL (if custom)
6. Select available models

### Example: OpenAI Configuration

1. Provider Type: **OpenAI**
2. API Key: Your OpenAI API key
3. Models available:
   * Main: `gpt-4`, `gpt-3.5-turbo`
   * Embedding: `text-embedding-3-small`, `text-embedding-ada-002`
   * Vision: `gpt-4-vision-preview`, `gpt-4o`

<Warning>
  API keys are stored locally in Obsidian. Never commit your vault's `.obsidian` folder to public repositories if it contains API credentials.
</Warning>

## Model Selection Best Practices

### For Main Provider

* **Local development:** Use smaller, faster models like `llama3.2` or `mistral`
* **Production use:** Use larger models like `gpt-4` or `claude-3-opus` for better quality
* **Speed vs. quality:** Balance based on your use case

### For Embedding Provider

* **Must match your content language:** Use multilingual models like `bge-m3` for non-English content
* **Vault size matters:** Larger vaults benefit from better embedding models
* **Consistency:** Use the same embedding model for all indexing to maintain quality

### For Vision Provider

* **Image quality:** Higher resolution images require more capable models
* **Speed:** Vision models are generally slower; use only when needed
* **Cost:** Cloud vision APIs can be expensive; consider local alternatives like `llava`

## Troubleshooting

### Provider Not Appearing in Dropdown

1. Ensure AI Providers plugin is installed and enabled
2. Restart Obsidian
3. Check that the provider is properly configured in AI Providers settings

### Connection Errors

* **Ollama:** Verify Ollama is running (`ollama list` in terminal)
* **Cloud APIs:** Check API key validity and network connection
* **Custom endpoints:** Verify URL format and accessibility

### Model Not Available

* **Ollama:** Pull the model first using `ollama pull <model-name>`
* **Cloud APIs:** Ensure you have access to the model in your API account
* **Check AI Providers settings:** Model must be configured in the provider settings

## Advanced Configuration

### Temperature Settings

You can set default creativity levels in Local GPT settings:

```typescript theme={null}
// From src/defaultSettings.ts:49-62
export const CREATIVITY: { [index: string]: any } = {
  "": {
    temperature: 0,
  },
  low: {
    temperature: 0.2,
  },
  medium: {
    temperature: 0.5,
  },
  high: {
    temperature: 1,
  },
};
```

* **None (0):** Deterministic, consistent outputs
* **Low (0.2):** Slightly varied, good for factual tasks
* **Medium (0.5):** Balanced creativity and consistency
* **High (1.0):** Maximum creativity, best for creative writing

### Context Limits

Configure how much context to retrieve for Enhanced Actions:

```typescript theme={null}
// From src/interfaces.ts:8-13
defaults: {
  creativity: string;
  contextLimit?: string; // 'local' | 'cloud' | 'advanced' | 'max'
}
```

* **local:** Optimized for local models with limited context windows
* **cloud:** Balanced for cloud APIs
* **advanced:** More context for capable models
* **max:** Maximum context retrieval

## Related Resources

<CardGroup cols={2}>
  <Card title="Creating Custom Actions" icon="wand-magic-sparkles" href="/guides/creating-custom-actions">
    Learn how to create custom actions with specific prompts
  </Card>

  <Card title="Prompt Templating" icon="code" href="/guides/prompt-templating">
    Use template keywords for dynamic prompts
  </Card>
</CardGroup>