This library provides a JavaScript interface for interacting with the Ollama API, enabling developers to easily integrate large language models into their web and Node.js applications. It simplifies tasks such as chat completion, text generation, model management, and embedding generation, offering a streamlined developer experience for LLM-powered features.

How It Works

The library acts as a client for the Ollama REST API, abstracting away HTTP requests and responses. It exposes methods that map directly to Ollama's functionalities, such as chat, generate, pull, and embed. The design supports both standard responses and streaming via AsyncGenerator, allowing for real-time LLM output. It also offers browser compatibility and customization for the Ollama host and fetch implementation.

Quick Start & Requirements

• Install: npm i ollama
• Browser Usage: import ollama from 'ollama/browser'
• Requires an Ollama server to be running.
• Official Docs: https://github.com/ollama/ollama-js

Highlighted Details

• Supports chat, generation, model pulling/pushing/creation/deletion/copying, and embedding.
• Enables streaming responses via AsyncGenerator for real-time interaction.
• Includes browser-compatible builds for client-side usage.
• Allows customization of Ollama host and fetch implementation.

Maintenance & Community

• Maintained by the Ollama team.
• Project repository: https://github.com/ollama/ollama-js

Licensing & Compatibility

• License: MIT
• Compatible with commercial and closed-source applications.

Limitations & Caveats

The files parameter for model creation is not currently supported. The abort method throws an AbortError on all listening streams, suggesting careful stream management or dedicated client instances per stream.