Framework Integration

AI API Gateway
for LangChain

Integrate AI API Gateway with LangChain framework. Build scalable LLM applications with unified API management, intelligent caching, and comprehensive monitoring.

Get Started View Features
LangChain Integration
📦
LLM Wrapper GatewayLLM
🔗
Chain Support Full Compatible
💾
Cache Layer Redis/Memory
📊
Tracing LangSmith Ready

Why Use Gateway with LangChain?

Enhance your LangChain applications with enterprise features

🔄
Unified LLM Interface
Access multiple LLM providers through a single interface. Switch between OpenAI, Anthropic, and others without code changes.
⚡
Intelligent Caching
Cache LLM responses to reduce costs and latency. Supports semantic similarity matching for cache hits.
🛡️
Rate Limiting
Protect your LLM budget with configurable rate limits. Prevent runaway chains from consuming all tokens.
📊
Observability
Full tracing integration with LangSmith. Monitor chain execution, token usage, and performance metrics.
🔐
Secure API Keys
Centralize API key management. Rotate keys without redeploying applications.
🎯
Fallback Support
Configure automatic fallbacks when primary LLM fails. Ensure high availability for critical chains.

Integration Steps

Get started in minutes with simple configuration

1
Install Dependencies
pip install langchain ai-gateway-client
2
Configure Gateway
Set up your gateway endpoint and API credentials
3
Create LLM Instance
Use GatewayLLM wrapper instead of direct LLM classes
4
Build Chains
Create LangChain chains as usual - gateway handles the rest

Code Examples

Quick start guide for LangChain integration

🦜 LangChain with Gateway Python 
# Import LangChain and Gateway components
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from ai_gateway.langchain import GatewayLLM

# Create Gateway LLM instance
llm = GatewayLLM(
    endpoint="https://gateway.example.com",
    api_key="your-gateway-key",
    model="gpt-4",
    
    # Enable caching
    cache=True,
    cache_ttl=3600,
    
    # Rate limiting
    max_requests_per_minute=60,
    
    # Fallback model
    fallback_model="gpt-3.5-turbo"
)

# Create a chain as usual
prompt = PromptTemplate(
    input_variables=["topic"],
    template="Write a short joke about {topic}."
)

chain = LLMChain(llm=llm, prompt=prompt)

# Run the chain - gateway handles everything
result = chain.run("programming")
print(result)

# Streaming support
for chunk in chain.stream("AI"):
    print(chunk, end="", flush=True)