---

A unified, resilient, and lightweight client for seamless multi-LLM API integration.

unimodel is a lightweight HTTP client for LLM APIs. It adds a couple of useful, but unobtrusive abstractions to handle common scenarios.

The primary use case is to allow application developers to easily switch between the various vendors and models in their applications.

• Adapters
• Scopes

Usage

pip install unimodel

touch .env && "OPENAI_API_KEY=${OPENAI_API_KEY}" >> ".env"

Example

from unimodel import Client

client = Client()
# Use the Chat API, prefer OpenAI/GPT-4o, if it fails,
# fallback to Anthropic/Claude-3.5-Sonnet
response = client.chat.completions.create(
    models=[
        "openai/gpt-4o",
        "anthropic/claude-3-5-sonnet"
    ],
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Who is the best Formula 1 driver?"},
    ]
)

Features

• Supports many vendors (OpenAI compatible APIs like OpenAI and Azure, Anthrophic API, …)
• Fallback to different vendor/model when first one fails
• Auto-retry w/ exponential backoff
• Use Jinja2 for prompt templates
• Multi-step prompt grammar
• Helps you write better tests

Design decisions

• The API roughly follows the OpenAI Python SDK (most popular)

What is it not?

• Not an agent, guidance framework, etc. (e.g. DSPy, Guidance, LangChain, lmql, etc.)

Status

• Vendors
  • Anthropic
  • Azure
  • OpenAI
  • Replicate
• Chat completion
  • Support all params
  • JSON mode
  • Tools (Functions)
  • Streaming
  • Access original response
• Text completion
• Embeddings
• Images
• Audio
• Cost estimation (and limits)