PDF Query with LangChain and Cassandra Vector Store

Overview

This project demonstrates how to use LangChain to build a question-answering system that can process and query the content of a PDF document. It leverages OpenAI's API for language processing and embedding, and uses Apache Cassandra as a vector store for efficient similarity searching.

Requirements

• Python 3.7 or higher
• Required libraries:
  • cassio
  • datasets
  • langchain
  • openai
  • tiktoken
  • PyPDF2

Installation

First, ensure you have the required Python libraries. You can install them using the following commands:

!pip install -q cassio datasets langchain openai tiktoken
!pip install pyPDF2

Configuration

Ensure you have your Astra DB and OpenAI API credentials ready:

• ASTRA_DB_APPLICATION_TOKEN
• ASTRA_DB_ID
• OPENAI_API_KEY

Replace YOUR_ASTRA_DB_TOKEN, YOUR_ASTRA_DB_ID, and YOUR_OPENAI_API_KEY with your actual credentials within the code.

Usage

1. Import necessary libraries and set up configurations:

   • Importing relevant modules for processing and querying.

   from langchain.vectorstores.cassandra import Cassandra
   from langchain.indexes.vectorstore import VectorStoreIndexWrapper
   from langchain.llms import OpenAI
   from langchain.embeddings import OpenAIEmbeddings
   import cassio
   
   from PyPDF2 import PdfReader

2. Read the PDF content:

   • The PdfReader module is used to read 'Climate-report.pdf'.

   pdfreader = PdfReader('Climate-report.pdf')
   
   raw_text = ''
   
   for i, page in enumerate(pdfreader.pages):
       content = page.extract_text()
       if content:
           raw_text += content

3. Initialize the connection to Astra DB:

   • Set up the connection using your credentials.

   cassio.init(token=ASTRA_DB_APPLICATION_TOKEN, database_id=ASTRA_DB_ID)

4. Setup OpenAI and Embedding configuration:

   • Initialize OpenAI with API key and embeddings.

   llm = OpenAI(openai_api_key=OPENAI_API_KEY)
   embedding = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)

5. Create and populate the vector store:

   • Set up Cassandra as the vector store for the dataset.

   astra_vector_store = Cassandra(
       embedding=embedding,
       table_name='pdfqueryproj',
       session=None,
       keyspace=None
   )

6. Process the text into chunks:

   • Split text into manageable chunks for vector storage.

   from langchain.text_splitter import CharacterTextSplitter
   
   text_splitter = CharacterTextSplitter(
       separator="\n",
       chunk_size=800,
       chunk_overlap=200,
       length_function=len,
   )
   
   texts = text_splitter.split_text(raw_text)

7. Add text datasets into the vector store

   astra_vector_store.add_texts(texts[:50])
   
   astra_vector_index = VectorStoreIndexWrapper(vectorstore=astra_vector_store)

8. Query the data interactively:

   • Use an interactive loop to query the data with user input.

   first_question = True
   
   while True:
       if first_question:
           query_text = input("Enter your question..(type quit to exit): ").strip()
       else:
           query_text = input("\n What is your next question.. :").strip()
   
       if query_text.lower() == "quit":
           break
       if query_text.lower() == "":
           continue
       first_question = False
   
       print("\nQuestion: \"%s\"" % query_text)
       answer = astra_vector_index.query(query_text, llm=llm).strip()
       print("\nAnswer: \"%s\"\n" % answer)
   
       print("FIRST DOCUMENTS BY RELEVANCE: ")
       for doc, score in astra_vector_store.similarity_search_with_score(query_text, k=4):
           print("  [%0.4f] \"%s ...\"" % (score, doc.page_content[:84]))

Conclusion

This project provides a robust solution for querying and extracting information from PDF documents by leveraging LangChain, OpenAI, and Cassandra.

Architecture

• 

Notes

• Make sure to replace placeholders for the tokens and IDs with actual values.
• Adjust the dataset and model configurations as needed for your specific use case.
• Ensure the PDF document path is correct and accessible within your environment.