Skip to main content
Discover how LangChain’s libraries are organized across four layers—Core, Community, Official Implementations, and the Main LangChain package—and learn how they build upon one another to provide powerful LLM-powered applications.

​
Table of Contents

  1. LangChain Core
  2. LangChain Community
  3. Official Provider Implementations
  4. Main LangChain Library
  5. Putting It All Together
  6. Further Reading

​
1. LangChain Core

The langchain_core package defines the abstract base classes, types, and interfaces that power every higher-level component. 
# LangChain Core: base definitions
from langchain_core.messages import AIMessage, HumanMessage
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.tools import tool
  • AIMessage & HumanMessage: generic chat message classes
  • ChatPromptTemplate & MessagesPlaceholder: prompt templating primitives
  • tool: decorator for defining custom tool interfaces
Install the core package before anything else:
pip install langchain-core

​
2. LangChain Community

Under the langchain_community namespace, you’ll find community-contributed models, tools, loaders, and callbacks. Each provider lives in its own module for better organization. 
# LangChain Community: chat models and Gmail tools
from langchain_community.chat_models import ChatOpenAI, ChatDatabricks
from langchain_community.tools.gmail import (
    GmailCreateDraft,
    GmailGetMessage,
    GmailGetThread,
    GmailSearch,
    GmailSendMessage,
)
  • ChatOpenAI & ChatDatabricks: community-maintained wrappers around respective chat APIs
  • Gmail tools: comprehensive Gmail API integrations

​
3. Official Provider Implementations

Concrete classes for popular LLM providers are maintained in their own packages (e.g., langchain_openai, langchain_anthropic). These implement the methods defined by the core abstractions. 
# LangChain OpenAI: concrete LLM implementation
from langchain_openai import ChatOpenAI
  • ChatOpenAI: wraps the OpenAI API, extending BaseChatModel
  • Similar packages: langchain_anthropic, langchain_azure, etc.

​
4. Main LangChain Library

The primary langchain package ties together core, community, and official implementations to offer ready-to-use agents, chains, tools, and parsers. 
# Main LangChain: agents, tools, and parsers
from langchain.agents import AgentExecutor
from langchain.agents.format_scratchpad import format_to_openai_function_messages
from langchain.output_parsers import OpenAIFunctionsAgentOutputParser
from langchain.tools.render import format_tool_to_openai_function
ComponentDescription
AgentExecutorExecutes agent loops with tools and prompts
format_to_openai_function_messagesConverts scratchpads into OpenAI function-calling format
OpenAIFunctionsAgentOutputParserParses responses from function-calling agents
format_tool_to_openai_functionRenders a tool’s signature into OpenAI JSON schema
Ensure that langchain, langchain_core, and your provider packages share compatible versions to avoid API mismatches.

​
5. Putting It All Together

In a typical application, you’ll import from all four layers: 
# 1. Core components
from langchain_core.messages import AIMessage, HumanMessage
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.tools import tool

# 2. Community modules
from langchain_community.chat_models import ChatOpenAI, ChatDatabricks
from langchain_community.tools.gmail import (
    GmailCreateDraft, GmailGetMessage, GmailGetThread,
    GmailSearch, GmailSendMessage,
)

# 3. Official LLM provider
from langchain_openai import ChatOpenAI

# 4. Main LangChain abstractions
from langchain.agents import AgentExecutor
from langchain.agents.format_scratchpad import format_to_openai_function_messages
from langchain.output_parsers import OpenAIFunctionsAgentOutputParser
from langchain.tools.render import format_tool_to_openai_function
You will always need:
  • langchain_core for base types
  • At least one provider package (e.g., langchain_openai) to call an LLM
  • langchain for high-level agents, tools, and chains
  • langchain_community (optional) for extended integrations

​
6. Further Reading

Explore these resources to deepen your understanding of LangChain’s modular design and accelerate your LLM application development.

Watch Video