Ollama Python Library

Use it when

• Your Python application needs chat, generation, or embeddings from models served by a local Ollama instance
• You are hand-writing httpx or requests calls against localhost:11434 and want a typed client with streaming, async, and error handling

What it solves

Not the fit when

• Does not run models itself; an Ollama server must be installed and running
• Not a multi-provider abstraction layer over OpenAI, Anthropic, and others
• Cloud models require an ollama.com sign-in or API key
• running or serving the model itself
• routing across multiple llm providers
• gpu memory capacity limits

Install

pip install ollama (requires Ollama installed and running, with a model pulled via ollama pull <model>)

Invoke

from ollama import chat; chat(model='gemma3', messages=[{'role':'user','content':'...'}]); stream=True for chunks, AsyncClient for async, ollama.embed for embeddings

Alternatives

No reviewed alternatives recorded yet.