llm-factory

A flexible Python factory for working with multiple Large Language Model (LLM) providers (OpenAI, Anthropic, Gemini, Llama) using a unified interface, with robust configuration and extensibility.

Features

• ✅ Unified interface for multiple LLM providers (OpenAI, Anthropic, Gemini, Llama)
• ✅ Easy provider switching via configuration
• ✅ Pydantic-based response validation
• ✅ Environment variable-based secure configuration
• ✅ Extensible for new providers
• ✅ Supports model, temperature, max tokens, and retries per provider

Installation

pip install python-llm-factory

Configuration

The package uses environment variables for authentication and configuration. You can set these in a .env file or your environment:

# Required environment variables for each provider
OPENAI_API_KEY=your_openai_api_key
ANTHROPIC_API_KEY=your_anthropic_api_key
GEMINI_API_KEY=your_gemini_api_key

Examples

Basic Usage: Creating a Completion

from pydantic import BaseModel, Field
from python_llm_factory import LLMFactory
from python_llm_factory import Settings


class CompletionModel(BaseModel):
    response: str = Field(description="Your response to the user.")
    reasoning: str = Field(description="Explain your reasoning for the response.")


messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "If it takes 2 hours to dry 1 shirt out in the sun, how long will it take to dry 5 shirts?"},
]

llm = LLMFactory(
    settings=Settings().gemini.gemini_2_5_flash,
)
completion = llm.completions_create(
    response_model=CompletionModel,
    messages=messages,
)
print(f"Response: {completion.response}\n")
print(f"Reasoning: {completion.reasoning}")

🤝 Contributing

If you have a helpful tool, pattern, or improvement to suggest:

• Fork the repo
  
• Create a new branch
  
• Submit a pull request
  I welcome additions that promote clean, productive, and maintainable development.
  

🙏 Thanks

Thanks for exploring this repository!
Happy coding!