azllm: A Unified LLM Interface for Multi-Provider Access

azllm is a Python package that provides a unified interface to work with multiple LLM providers including OpenAI, DeepSeek, Grok, Gemini, Meta's LLaMA, Anthropic, Ollama, and more.

NOTE: For advanced usage, see the azllm documentation and/or examples.

Features

• One unified interface for all major LLM APIs
• Batch and parallel prompt generation
• Structured outputs (parsing) with Pydantic for models that support parsed outputs natively
• Structured outputs (parsing) with Pydantic for DeepSeek and Anthropic
• Per-model configurations and lazy initialization
• Clean error handling
• .env-based API key management

Supported Clients

• OpenAI
• DeepSeek
• Grok
• Anthropic
• Fireworks for Meta's LLaMA and others.
• Google's Gemini
• Ollama

NOTE: If you would like to request support for additional LLMs, please open an issue on our GitHub page.

Installation

You can install the azllm package via pip:

pip install azllm

Prerequisites

• Python 3.11+

• Create a .env file to store your API keys. For example:

  OPENAI_API_KEY=your_openai_api_key
  DEEPSEEK_API_KEY=your_deepseek_api_key
  XAI_API_KEY=your_xai_api_key
  GEMINI_API_KEY=your_gemini_api_key
  ANTHROPIC_API_KEY=your_anthropic_api_key
  FIREWORKS_API_KEY=your_fireworks_api_key
  

• Ollama must be installed and running locally to use Ollama models.

Quick Start

Basic Initialization

from azllm import azLLM
manager = azLLM()  # Instantiated with default parameters 

Generate Text from a Single Prompt

prompt = 'What is the captial of France?'
generated_text = manager.generate_text('openai', prompt)
print(generated_text)

Batch Generation

Generate responses for multiple prompts at once:

batch_prompts = [
    'What is the capital of France?',
    'Tell me a joke.'
    ]

results = manager.batch_generate('openai', batch_prompts)
for result in results:
    print(result)

Parallel Generation

Run a single prompt across multiple models simultaneously:

prompt = 'What is the capital of France?'
models = [
    'openai',
    'grok',
    'ollama']

results = manager.generate_parallel(prompt, models)
for model, result in results.items():
    print(f"Model: {model},\nResult: {result}\n")

License

MIT License

Citation

@misc{azLLM,
  title        = {azllm},
  author       = {Hanif Sajid and Benjamin Radford and Yaoyao Dai and Jason Windett},
  year         = {2025},
  month        = apr,
  version      = {0.1.6},
  howpublished = {https://github.com/hanifsajid/azllm},
  note         = {MIT License},
  abstract     = {azllm is a Python package designed to interface with various large language models (LLMs) from different AI providers. It offers a unified interface for interacting with models from providers like OpenAI, DeepSeek, Grok, Gemini, Meta's Llama, Anthropic, Ollama, and others. The package allows for customizable configurations, batch generation, parallel generation, error handling, and the ability to parse structured responses from different models.}
}