Ollama Integration Guide

mindX supports multiple ways to connect to Ollama servers:

1. Custom Implementation (api/ollama_url.py) - Default, includes rate limiting and metrics
2. Official Ollama Python Library (api/ollama_official.py) - Optional, better compatibility

Official Ollama Python Library

The official ollama-python library provides:

• Better compatibility with Ollama API updates
• Official support and maintenance
• AsyncClient for async operations
• Streaming support
• Cloud model support (ollama.com)

Installation

pip install ollama

Usage

from api.ollama_official import create_ollama_client, OfficialOllamaAdapter

Create client (auto-detects if official library is available)
client = create_ollama_client(base_url="http://localhost:11434")

if client:
    # Use official library
    models = await client.list_models()
    response = await client.generate_text(
        prompt="Hello, world!",
        model="llama3:8b"
    )
else:
    # Fallback to custom implementation
    from api.ollama_url import create_ollama_api
    api = create_ollama_api()
    models = await api.list_models()

Cloud Models

The official library supports Ollama Cloud models:

import os
from api.ollama_official import OfficialOllamaAdapter

Connect to Ollama Cloud
client = OfficialOllamaAdapter(
    base_url="https://ollama.com",
    api_key=os.environ.get("OLLAMA_API_KEY")
)

Use cloud models
response = await client.generate_text(
    prompt="Hello!",
    model="gpt-oss:120b-cloud"
)

Custom Implementation

The custom implementation (api/ollama_url.py) provides:

• Rate limiting
• Metrics tracking
• Better integration with mindX config system
• SettingsManager support

This is the default implementation and works without additional dependencies.

Configuration

Both implementations respect the same configuration:

Environment Variables

export MINDX_LLM__OLLAMA__BASE_URL=http://localhost:11434
export OLLAMA_API_KEY=your_key_here  # For cloud

SettingsManager

from webmind.settings import SettingsManager

settings = SettingsManager()
base_url = settings.get('ollama_base_url', 'http://localhost:11434')

Config System

from utils.config import Config

config = Config()
base_url = config.get('llm.ollama.base_url', 'http://localhost:11434')

References

• Official Ollama Python Library
• Ollama API Documentation
• Ollama Cloud Models