> ## Documentation Index
> Fetch the complete documentation index at: https://docs.enconvo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Ollama

> Run local LLMs with Ollama in EnConvo

## Overview

Ollama allows you to run large language models locally on your Mac. Perfect for privacy-focused work and offline usage.

## Supported Models

Any model available in Ollama:

| Model             | Size    | Best For        |
| ----------------- | ------- | --------------- |
| **Llama 3.1 70B** | 40GB    | Complex tasks   |
| **Llama 3.1 8B**  | 4.7GB   | General purpose |
| **Mistral 7B**    | 4.1GB   | Fast responses  |
| **CodeLlama**     | 4.7GB   | Programming     |
| **Phi-3**         | 2.2GB   | Lightweight     |
| **Gemma 2**       | 5.4GB   | Balanced        |
| **Qwen 2**        | Various | Multilingual    |

## Setup

<Steps>
  <Step title="Install Ollama">
    Download from [ollama.ai](https://ollama.ai) or:

    ```bash theme={null}
    brew install ollama
    ```
  </Step>

  <Step title="Pull a Model">
    ```bash theme={null}
    ollama pull llama3.1
    ```
  </Step>

  <Step title="Verify Installation">
    ```bash theme={null}
    ollama list
    ```
  </Step>

  <Step title="Configure in EnConvo">
    1. Open **Settings** → **AI Provider**
    2. Select **Ollama**
    3. Go to **Credentials** module
    4. Set endpoint: `http://localhost:11434`
  </Step>

  <Step title="Select Model">
    Choose from your installed models
  </Step>
</Steps>

## Configuration

| Setting         | Description            | Default         |
| --------------- | ---------------------- | --------------- |
| **Credentials** | Endpoint configuration | localhost:11434 |
| **Model Name**  | Installed model        | llama2:latest   |
| **Temperature** | Creativity (0-2)       | Medium (1)      |

## Validate and Use

<Steps>
  <Step title="Start Ollama">
    Make sure Ollama is running before validating. Use the Ollama app or run `ollama serve` from a terminal.
  </Step>

  <Step title="Confirm models are installed">
    Run `ollama list` and make sure at least one model is available locally.
  </Step>

  <Step title="Validate the endpoint">
    Use `http://localhost:11434` for the default local endpoint. Do not include a model-specific path.
  </Step>

  <Step title="Test a small model first">
    Start with a smaller model to verify the setup, then switch to larger models when memory and latency are acceptable.
  </Step>
</Steps>

## Reasoning Mode

Enable thinking for compatible models:

| Option       | Description        |
| ------------ | ------------------ |
| **Disabled** | Standard responses |
| **Thinking** | Enable reasoning   |

## Popular Models

```bash theme={null}
ollama pull llama3.1        # General purpose
ollama pull codellama       # Coding
ollama pull mistral         # Fast
ollama pull phi3            # Lightweight
ollama pull gemma2          # Google's model
ollama pull qwen2           # Multilingual
```

## System Requirements

| RAM   | Recommended Models                   |
| ----- | ------------------------------------ |
| 8GB   | 7B models (Llama 3.1 8B, Mistral 7B) |
| 16GB  | Larger 7B models, some 13B           |
| 32GB  | 13B-30B models                       |
| 64GB+ | 70B models                           |

<Tip>
  Apple Silicon Macs with Metal acceleration provide excellent local LLM performance.
</Tip>

## Privacy Benefits

<CardGroup cols={2}>
  <Card title="Complete Privacy" icon="shield-halved">
    Data never leaves your Mac
  </Card>

  <Card title="Offline Access" icon="wifi-slash">
    Works without internet
  </Card>

  <Card title="No Usage Limits" icon="infinity">
    Unlimited local queries
  </Card>

  <Card title="Full Control" icon="sliders">
    Choose exactly which models to run
  </Card>
</CardGroup>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Connection refused">
    * Ensure Ollama is running: `ollama serve`
    * Check port 11434 is available
    * Verify endpoint in settings
  </Accordion>

  <Accordion title="Slow responses">
    * Use smaller models
    * Close memory-intensive apps
    * Consider quantized models
  </Accordion>

  <Accordion title="Out of memory">
    * Use smaller model
    * Reduce context length
    * Restart Ollama
  </Accordion>
</AccordionGroup>