LangChain

1.0.x · active · verified Sat Feb 28

LangChain is an orchestration framework for building LLM-powered applications and agents. It provides abstractions for chaining model calls, tool use, memory, and multi-agent coordination. As of v1.0, LangGraph is the foundational runtime for stateful agent execution.

Warnings

breaking AgentExecutor is removed in v1.0. Use create_agent instead.
Fix: Replace AgentExecutor with create_agent from langchain
breaking LCEL pipe syntax (prompt | llm | parser) is deprecated. Use explicit chaining.
Fix: Use create_agent or direct llm.invoke() calls
breaking langchain-community does not follow semver. Breaking changes can occur in minor releases.
Fix: Pin langchain-community version explicitly in requirements.txt
gotcha pip install langchain alone is insufficient. Provider packages must be installed separately.
Fix: Always install langchain-openai or langchain-anthropic alongside langchain
deprecated astream_events(version='v1') uses vulnerable serialization. Use v2.
Fix: astream_events(version='v2')

Install

pip install langchain langchain-openai langchain-anthropic Python
npm install langchain @langchain/openai @langchain/anthropic Node.js

Imports

ChatOpenAI
```
from langchain_openai import ChatOpenAI
```
Moved to partner package in 0.2. Wrong import raises ImportError in 1.0.
ChatAnthropic
```
from langchain_anthropic import ChatAnthropic
```
Moved to partner package in 0.2. Wrong import raises ImportError in 1.0.
create_agent
```
from langchain import create_agent
```
create_react_agent still works in 1.0 but create_agent is now the idiomatic entry point.

Quickstart

Minimal LLM call using ChatOpenAI in LangChain 1.0.

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model='gpt-4o', temperature=0)
response = llm.invoke('What is the capital of France?')
print(response.content)

view raw JSON →