Contextual compression in LangChain - Mini Project: Build & Apply

Contextual Compression with LangChain

📖 Scenario: You are building a simple text processing tool using LangChain to compress context before sending it to a language model. This helps reduce the amount of text while keeping the important meaning.

🎯 Goal: Create a LangChain pipeline that compresses a given text context using a compression chain and a language model.

📋 What You'll Learn

Create a variable text_context with the exact string: "LangChain helps build applications with language models."

Create a variable compression_prompt with the exact string: "Summarize the following text concisely:"

Use LangChain's LLMChain with a mock OpenAI model and the compression_prompt to create a compression chain called compressor.

Create a ContextualCompressionChain named compression_chain using the compressor.

Run the compression_chain on text_context and store the result in compressed_text.

💡 Why This Matters

🌍 Real World

Contextual compression helps reduce the size of text data sent to language models, saving costs and improving speed in applications like chatbots and summarizers.

💼 Career

Understanding how to build and use compression chains with LangChain is useful for developers working on AI-powered text applications, improving efficiency and user experience.

1

DATA SETUP: Create the initial text context

Create a variable called text_context and assign it the exact string "LangChain helps build applications with language models."

LangChain

# Your code here

Need a hint?

Use a simple assignment with the exact string inside double quotes.

2

CONFIGURATION: Define the compression prompt

Create a variable called compression_prompt and assign it the exact string "Summarize the following text concisely:"

LangChain

text_context = "LangChain helps build applications with language models."
# Your code here

Need a hint?

Assign the prompt string exactly as shown to the variable compression_prompt.

3

CORE LOGIC: Create the compression chain using LangChain

Import OpenAI and LLMChain from langchain, then create a variable compressor as an LLMChain using OpenAI() as the language model and compression_prompt as the prompt template.

LangChain

text_context = "LangChain helps build applications with language models."
compression_prompt = "Summarize the following text concisely:"
# Your code here

Need a hint?

Import the classes, then create compressor with LLMChain passing OpenAI() and compression_prompt.

4

COMPLETION: Use ContextualCompressionChain and compress the text

Import ContextualCompressionChain from langchain.chains, create a variable compression_chain using ContextualCompressionChain(compressor=compressor), then run compression_chain.compress(text_context) and assign the result to compressed_text.

LangChain

from langchain.llms import OpenAI
from langchain.chains import LLMChain

text_context = "LangChain helps build applications with language models."
compression_prompt = "Summarize the following text concisely:"

compressor = LLMChain(llm=OpenAI(), prompt=compression_prompt)
# Your code here

Need a hint?

Import ContextualCompressionChain, create compression_chain with compressor, then compress text_context.