langchain-reference-architecture

# LangChain Reference Architecture

## Overview
Production-ready architectural patterns for building scalable, maintainable LangChain applications.

## Prerequisites
- Understanding of LangChain fundamentals
- Experience with software architecture
- Knowledge of cloud infrastructure

## Architecture Patterns

### Pattern 1: Layered Architecture
```
src/
├── api/                    # API layer (FastAPI/Flask)
│   ├── __init__.py
│   ├── routes/
│   │   ├── chat.py
│   │   └── agents.py
│   └── middleware/
│       ├── auth.py
│       └── rate_limit.py
├── core/                   # Business logic layer
│   ├── __init__.py
│   ├── chains/
│   │   ├── __init__.py
│   │   ├── chat_chain.py
│   │   └── rag_chain.py
│   ├── agents/
│   │   ├── __init__.py
│   │   └── research_agent.py
│   └── tools/
│       ├── __init__.py
│       └── search.py
├── infrastructure/         # Infrastructure layer
│   ├── __init__.py
│   ├── llm/
│   │   ├── __init__.py
│   │   └── provider.py
│   ├── vectorstore/
│   │   └── pinecone.py
│   └── cache/
│       └── redis.py
├── config/                 # Configuration
│   ├── __init__.py
│   └── settings.py
└── main.py
```

### Pattern 2: Provider Abstraction
```python
# infrastructure/llm/provider.py
from abc import ABC, abstractmethod
from langchain_core.language_models import BaseChatModel
from langchain_openai import ChatOpenAI
from langchain_anthropic import ChatAnthropic

class LLMProvider(ABC):
    """Abstract LLM provider."""

    @abstractmethod
    def get_chat_model(self, **kwargs) -> BaseChatModel:
        pass

class OpenAIProvider(LLMProvider):
    def get_chat_model(self, model: str = "gpt-4o-mini", **kwargs) -> BaseChatModel:
        return ChatOpenAI(model=model, **kwargs)

class AnthropicProvider(LLMProvider):
    def get_chat_model(self, model: str = "claude-3-5-sonnet-20241022", **kwargs) -> BaseChatModel:
        return ChatAnthropic(model=model, **kwargs)

class LLMFactory:
    """Factory for creating LLM instances."""