Introduction

Framework to build app that use LLMs.

• As a toolkit for making LLMs “smarter” in real-world apps.
• Not only sending and receiving text, it structuring interaction of LLM with other data, APIs, or tools.
  • Searching a database
  • Calling an API
  • Writing to or reading from files
  • Reasoning step by step (chains of thought)
  • Keeping context in multi-step conversations

Modern LangChain Architecture

Components

1. LLMs

   • The “brain” that generates text.
   • Can be OpenAI GPT, Hugging Face models, or others.

   from langchain_openai import ChatOpenAI
   
   llm = ChatOpenAI(
       model="gpt-4.1-mini",
       temperature=0
   )
   
   response = llm.invoke("Explain recursion simply")
   
   print(response.content)
   

2. Prompts Templates

   • Structure what you send to the model.
   • Can include dynamic variables, instructions, or context.

   from langchain_core.prompts import ChatPromptTemplate
   
   prompt = ChatPromptTemplate.from_template(
       "Explain {topic} like I am 10 years old"
   )
   
   formatted = prompt.invoke({
       "topic": "Docker"
   })
   
   print(formatted)
   

3. Chains

   • Sequences of actions or prompts.
   • Example: Prompt → LLM → Output
   • Example: Extract info → summarize → answer.

   from langchain_core.output_parsers import StrOutputParser
   
   chain = prompt | llm | StrOutputParser()
   
   result = chain.invoke({
       "topic": "Kubernetes"
   })
   
   print(result)
   

4. AI Agents

   • Not just reply users’ question.
   • It is special chains that decide actions dynamically.
   • Can choose which tool to call based on user input.
   • Example: If the user asks for weather, the agent decides to call a weather API.

5. Memory

   • Keeps track of previous interactions.
   • Useful for chatbots that remember context.

6. Data Connectors

   • Connect LLMs to external data: databases, Google Sheets, APIs, PDFs, websites, etc.

Typical Use Cases

• Chatbots with context or long-term memory
• Document QA: Ask questions about PDFs or websites