Ollama Provider

September 20, 2025 ยท View on GitHub

Ollama Logo

Ollama Provider

Ollama enables running large language models locally, providing complete privacy and no API costs.

Setup

1. Install Ollama

macOS/Linux:

curl -fsSL https://ollama.ai/install.sh | sh

Windows:

Download installer from ollama.ai

2. Pull a Model

ollama pull codellama:7b

3. Configure MegaLinter

LLM_ADVISOR_ENABLED: true
LLM_PROVIDER: ollama
LLM_MODEL_NAME: codellama:7b
LLM_MAX_TOKENS: 1000
LLM_TEMPERATURE: 0.1

Official Model List

For the most up-to-date list of models available for Ollama, see the official Ollama model library:

Configuration Options

Basic Configuration

LLM_PROVIDER: ollama
LLM_MODEL_NAME: codellama:7b        # Any installed Ollama model

Advanced Configuration

# Custom Ollama server
OLLAMA_BASE_URL: http://localhost:11434

# For remote Ollama servers
OLLAMA_BASE_URL: http://your-server:11434

Troubleshooting

Common Issues

  1. "Connection refused"

    • Ensure Ollama is running: ollama serve
    • Check the correct port (default: 11434)
    • Verify firewall settings

  2. "Model not found"

    • Pull the model first: ollama pull model-name
    • List available models: ollama list
    • Use exact model name with tag

  3. "Out of memory"

    • Use smaller models (7B instead of 13B)
    • Close other applications
    • Consider using quantized models

  4. Slow responses

    • Use smaller models for faster inference
    • Ensure sufficient RAM
    • Consider GPU acceleration if available