A practice repository implementing examples from the official LangChain documentation to learn about building AI agents, chat models, and semantic search engines.
This project demonstrates key LangChain concepts through three example scripts:
agent.py: A ReAct (Reasoning + Acting) agent with memory that can perform web searches using Tavily and process user querieschat_model.py: Chat model implementation showing prompt templates, streaming, and different input formatssemantic_search_engine.py: Document processing pipeline with PDF loading, text splitting, embedding generation, and vector search
langchain_community==0.3.20
langchain_core==0.3.49
langchain_mistralai==0.2.10
langchain_postgres==0.0.13
langchain_text_splitters==0.3.7
langgraph==0.3.21
python-dotenv==1.1.0
Additional dependencies: pypdf for PDF loading, PostgreSQL with pgvector extension (for vector storage)
- Install dependencies:
pip install -r requirements.txt - Create a
.envfile with required API keys:MISTRAL_API_KEY=your_key LANGSMITH_API_KEY=your_key # Optional for tracing LANGSMITH_PROJECT=default # Optional project name TAVILY_API_KEY=your_key # For web search capabilities
Each script demonstrates different LangChain capabilities:
# Create a ReAct agent with memory
agent_executor = create_react_agent(model, tools, checkpointer=memory)
# Run with conversation persistence
config = {"configurable": {"thread_id": "conversation_id"}}
response = agent_executor.invoke({"messages": [HumanMessage(content="What is the weather in Delhi?")]}, config)# Using prompt templates
prompt = prompt_template.invoke({"language": "French", "text": "Hello world"})
response = model.invoke(prompt)# Search documents
results = vector_store.similarity_search("How many distribution centers does Nike have?")For detailed explanations of these concepts, refer to the official LangChain documentation.
Note: Don't try to mess with the postgres connection URL - the DB is already removed so don't act smart.