LangChainGang

A unified interface for LangChain chat model providers that simplifies working with multiple LLM providers.

Features

• Unified API: Single interface for multiple LangChain chat model providers
• Lazy Loading: Providers are only imported when needed
• Graceful Fallbacks: Missing dependencies don't break the entire package
• Provider Discovery: Easily list available providers
• Standardized Parameters: Consistent parameter handling across providers

Supported Providers

• OpenAI (openai) - ChatOpenAI
• Azure OpenAI (azure_openai) - AzureChatOpenAI
• AWS Bedrock (bedrock) - ChatBedrock
• Google Vertex AI (vertex) - ChatVertexAI
• Google Gemini (gemini) - ChatGoogleGenerativeAI
• Anthropic (anthropic) - ChatAnthropic
• Ollama (ollama) - ChatOllama

Installation

Install the base package:

uv add langchaingang

Install with specific provider support:

# OpenAI (and Azure OpenAI) support
uv add "langchaingang[openai]"

# AWS Bedrock support
uv add "langchaingang[aws]"

# Google Gemini and Vertex AI support
uv add "langchaingang[google]"

# Anthropic support
uv add "langchaingang[anthropic]"

# Ollama support
uv add "langchaingang[ollama]"

# Multiple providers
uv add "langchaingang[openai,anthropic,aws]"

# All providers
uv add "langchaingang[all]"

Quick Start

import langchaingang

# List available providers
providers = langchaingang.get_provider_list()
print(providers)  # ['openai', 'anthropic', 'ollama', ...]

# Get a chat model
model = langchaingang.get_chat_model(
    provider_name="openai", 
    model="gpt-4o-mini",
    api_key="your-api-key"
)

# Use the model
response = model.invoke("Hello, world!")
print(response.content)

Usage Examples

OpenAI

model = langchaingang.get_chat_model(
    provider_name="openai",
    model="gpt-4o-mini",
    api_key="your-openai-key"
)

Azure OpenAI

model = langchaingang.get_chat_model(
    provider_name="azure_openai",
    model="gpt-4o-mini",
    azure_endpoint="https://your-resource.openai.azure.com/",
    api_key="your-azure-key",
    api_version="2024-02-01"
)

AWS Bedrock

model = langchaingang.get_chat_model(
    provider_name="bedrock",
    model="meta.llama3-2-70b-instruct-v1:0",  # Will be converted to model_id
    region_name="us-east-1"
)

Google Vertex AI

model = langchaingang.get_chat_model(
    provider_name="vertex",
    model="gemini-2.0-flash-001",  # Will be converted to model_name
    project="your-gcp-project"
)

Anthropic

model = langchaingang.get_chat_model(
    provider_name="anthropic",
    model="claude-sonnet-4-0",
    api_key="your-anthropic-key"
)

Ollama

model = langchaingang.get_chat_model(
    provider_name="ollama",
    model="llama3",
    base_url="http://localhost:11434"  # Optional, defaults to localhost:11434
)

Parameter Handling

LangChainGang automatically handles provider-specific parameter differences:

• Bedrock: model parameter is converted to model_id
• Vertex AI: model parameter is converted to model_name
• All others: model parameter is passed through unchanged

Error Handling

The package gracefully handles missing dependencies:

# This won't fail even if langchain-openai isn't installed
providers = langchaingang.get_provider_list()

# This will raise ImportError if langchain-openai isn't installed
model = langchaingang.get_chat_model("openai", model="gpt-4o-mini")

Development

Install development dependencies:

uv add "langchaingang[dev]"

Run tests:

pytest

Format code:

black langchaingang/
isort langchaingang/

Type checking:

mypy langchaingang/

License

MIT License - see LICENSE file for details.