LLM Provider Factory

A unified, extensible interface for multiple Large Language Model providers. Built with clean architecture principles and SOLID design patterns.

Features

• 🏭 Factory Pattern: Clean, consistent interface across providers
• 🔌 Extensible: Easy to add new LLM providers
• 🛡️ Type Safe: Full TypeScript-style typing support
• 🚀 Production Ready: Comprehensive error handling and logging
• 📦 Zero Dependencies: Only requires requests for HTTP calls
• 🎯 SOLID Principles: Clean, maintainable architecture

Supported Providers

• OpenAI (GPT-3.5, GPT-4, etc.)
• Anthropic (Claude models)
• Google Gemini (Gemini Pro, Flash, etc.)

Installation

pip install llm-provider

Quick Start

Method 1: Using Provider Instance (Recommended)

from llm_provider import LLMProviderFactory, OpenAI

# Initialize with provider instance
provider = LLMProviderFactory(OpenAI(api_key="your-openai-key"))

# Generate response
response = provider.generate(
    prompt="Hello, how are you?",
    history=[]
)

print(response.content)

Method 2: Using Factory Method

from llm_provider import LLMProviderFactory

# Create provider using factory method
provider = LLMProviderFactory.create_provider(
    "openai", 
    api_key="your-openai-key"
)

response = provider.generate(prompt="Hello", history=[])
print(response.content)

Advanced Usage

Working with Conversation History

from llm_provider import LLMProviderFactory, OpenAI, Message

provider = LLMProviderFactory(OpenAI(api_key="your-key"))

# Using Message objects
history = [
    Message(role="user", content="What's the capital of France?"),
    Message(role="assistant", content="The capital of France is Paris."),
]

response = provider.generate(
    prompt="What's its population?",
    history=history
)

# Or using dictionaries
history_dict = [
    {"role": "user", "content": "What's the capital of France?"},
    {"role": "assistant", "content": "The capital of France is Paris."},
]

response = provider.generate(
    prompt="What's its population?",
    history=history_dict
)

Custom Generation Parameters

response = provider.generate(
    prompt="Write a creative story",
    temperature=0.9,
    max_tokens=500,
    top_p=0.95
)

Using Different Providers

from llm_provider import LLMProviderFactory, OpenAI, Anthropic, Gemini

# OpenAI
openai_provider = LLMProviderFactory(OpenAI(api_key="openai-key"))

# Anthropic
anthropic_provider = LLMProviderFactory(Anthropic(api_key="anthropic-key"))

# Gemini
gemini_provider = LLMProviderFactory(Gemini(api_key="gemini-key"))

# Switch between providers
factory = LLMProviderFactory(openai_provider.provider)
factory.switch_provider(anthropic_provider.provider)

Error Handling

from llm_provider import (
    LLMProviderFactory, 
    OpenAI, 
    APIError, 
    RateLimitError, 
    AuthenticationError
)

try:
    provider = LLMProviderFactory(OpenAI(api_key="your-key"))
    response = provider.generate("Hello")
    
except AuthenticationError as e:
    print(f"Authentication failed: {e}")
    
except RateLimitError as e:
    print(f"Rate limit exceeded: {e}")
    
except APIError as e:
    print(f"API error: {e}")

Getting Available Models

provider = LLMProviderFactory(OpenAI(api_key="your-key"))
models = provider.get_available_models()
print(f"Available models: {models}")

Provider Information

provider = LLMProviderFactory(OpenAI(api_key="your-key"))
info = provider.get_provider_info()
print(f"Provider: {info['name']}")
print(f"Class: {info['class']}")

Extending with Custom Providers

from llm_provider import BaseLLMProvider, LLMResponse, LLMProviderFactory

class CustomProvider(BaseLLMProvider):
    def _validate_config(self):
        if not self.config.get('api_key'):
            raise ConfigurationError("API key required")
    
    def generate(self, prompt, history=None, **kwargs):
        # Your custom implementation
        return LLMResponse(
            content="Custom response",
            model="custom-model",
            metadata={"provider": "custom"}
        )
    
    def get_available_models(self):
        return ["custom-model-1", "custom-model-2"]

# Register the custom provider
LLMProviderFactory.register_provider("custom", CustomProvider)

# Use it
provider = LLMProviderFactory.create_provider("custom", api_key="test")

Configuration

Environment Variables

export LLM_PROVIDER_TIMEOUT=30
export LLM_PROVIDER_MAX_RETRIES=3
export LLM_PROVIDER_LOG_LEVEL=INFO

Custom Configuration

from llm_provider import Config, LLMProviderFactory, OpenAI

config = Config(
    default_timeout=60,
    default_max_retries=5,
    log_level="DEBUG"
)

provider = LLMProviderFactory(
    OpenAI(api_key="your-key"), 
    config=config
)

API Reference

LLMProviderFactory

• __init__(provider, config=None): Initialize with provider instance
• generate(prompt, history=None, **kwargs): Generate response
• get_available_models(): Get available models
• get_provider_info(): Get provider information
• switch_provider(provider): Switch to different provider
• create_provider(name, **kwargs): Class method to create provider
• register_provider(name, class): Class method to register custom provider

LLMResponse

• content: Generated text content
• model: Model used for generation
• usage: Usage statistics (tokens, etc.)
• metadata: Additional metadata

Message

• role: Message role ('user', 'assistant', 'system')
• content: Message content
• to_dict(): Convert to dictionary

Development

Setup Development Environment

git clone https://github.com/sadikhanecioglu/llm-provider
cd llm-provider
pip install -e ".[dev]"

Run Tests

pytest

Code Formatting

black src/ tests/
isort src/ tests/

Type Checking

mypy src/

Architecture

This package follows SOLID principles:

• Single Responsibility: Each class has one reason to change
• Open/Closed: Open for extension, closed for modification
• Liskov Substitution: Providers are interchangeable
• Interface Segregation: Minimal, focused interfaces
• Dependency Inversion: Depend on abstractions, not concretions

License

MIT License - see LICENSE file for details.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests
5. Run the test suite
6. Submit a pull request

Changelog

v1.0.0

• Initial release
• Support for OpenAI, Anthropic, and Gemini
• Factory pattern implementation
• Comprehensive error handling
• Full test coverage