langchain

LangChain framework for building AI agents and LLM applications with tools, memory, and streaming support

SKILL.md

GitHub

LangChain

LangChain is the easy way to build custom agents and applications powered by LLMs. With under 10 lines of code, you can connect to OpenAI, Anthropic, Google, and more.

Preferences

  • Use create_agent as the unified interface for creating agents
  • Prefer structured output for reliable data extraction
  • Always implement error handling for production agents
  • Use streaming for better user experience
  • Leverage middleware for cross-cutting concerns

Core Components

Topic Description Reference
Agents Core agent architecture with create_agent agents-basics.md
Models Model integration (OpenAI, Anthropic, etc.) models-integration.md
Messages Message formats and conversation structure messages-format.md
Tools Tool definition and dynamic tool selection agents-tools.md
Memory Short-term and conversation memory memory-short-term.md
Streaming Streaming output for real-time responses streaming-output.md
Structured Output Structured data extraction from LLMs structured-output.md

Middleware

Topic Description Reference
Middleware Overview Middleware architecture and patterns middleware-overview.md
Human-in-the-Loop Add human intervention to agents middleware-human-in-loop.md

Additional Features

Topic Description Reference
Prompt Templates Reusable prompt templates prompt-templates.md
RAG Basics Retrieval Augmented Generation rag-basics.md
Error Handling Production error handling patterns error-handling.md

Quick Reference

Basic Agent

from langchain.agents import create_agent

def get_weather(city: str) -> str:
    """Get weather for a given city."""
    return f"It's always sunny in {city}!"

agent = create_agent(
    model="claude-sonnet-4-6",
    tools=[get_weather],
    system_prompt="You are a helpful assistant",
)

result = agent.invoke(
    {"messages": [{"role": "user", "content": "what is the weather in sf"}]}
)

Agent with Model Instance

from langchain.agents import create_agent
from langchain_openai import ChatOpenAI

model = ChatOpenAI(
    model="gpt-4",
    temperature=0.1,
    max_tokens=1000,
    timeout=30
)

agent = create_agent(model, tools=[get_weather])

Streaming Agent

from langchain.agents import create_agent

agent = create_agent(
    model="gpt-4",
    tools=[get_weather],
    streaming=True
)

async for chunk in agent.stream({"messages": [...]}):
    print(chunk.content, end="", flush=True)

Structured Output

from pydantic import BaseModel
from langchain.agents import create_agent

class WeatherReport(BaseModel):
    city: str
    temperature: float
    condition: str

agent = create_agent(
    model="gpt-4",
    output_schema=WeatherReport
)

result = agent.invoke({"messages": [...]})
weather = result.structured_output

Key Imports

from langchain.agents import create_agent
from langchain.tools import tool
from langchain_openai import ChatOpenAI
from langchain_anthropic import ChatAnthropic
from langchain.agents.middleware import (
    wrap_model_call,
    SummarizationMiddleware,
    HumanInTheLoopMiddleware
)

Official Documentation