LangChain Chains

This repository demonstrates how to use Chains in LangChain to build modular AI pipelines.

What are Chains?

In LangChain, a Chain is a sequence of components connected together to perform a task. Each component in the chain does one job, and the output of one step becomes the input of the next.

For example: Prompt → LLM → Output Parser

LangChain chains help break down tasks into manageable steps. You can combine models, prompts, tools, and logic using chains to create flexible AI applications.

How This Works

This project uses LangChain's Runnable interfaces to build a conditional chain. Here's what it does:

1. Takes a feedback text input.
2. Uses a prompt and LLM to classify the feedback as Positive or Negative.
3. Parses the classification using a Pydantic model.
4. Based on the sentiment, routes the input to different prompts to generate an appropriate response.
5. Returns the final response.

The routing logic is handled using RunnableBranch, which acts like an if-else statement.

Key Components Used

• PromptTemplate: To define prompt templates for the classifier and response generator.
• ChatGroq: LLM wrapper for making API calls to Groq-hosted LLMs.
• PydanticOutputParser: To parse the output of the model into structured format.
• RunnableBranch: For conditional logic based on sentiment.
• RunnableLambda: For simple transformations inside chains.

Example Use Case

Input: "I love the new features in the app, they are amazing."

Step-by-step flow:

1. Classify sentiment → Positive
2. Choose the positive response prompt
3. Generate and return the appropriate reply

Why Use Chains

• Clear and modular code
• Easier debugging and testing
• Scalable logic using simple building blocks
• Useful for tasks like routing, classification, tool use, and complex workflows