Skip to main content

Document Search with Streamlit

Lab Overview​

Learn how to implement document search functionality using Streamlit and LangChain, creating a powerful search interface for your documents. This lab demonstrates how to build a conversational document search system that allows users to upload PDFs and interact with their content through natural language queries.

Lab Materials​

Key Components​

  • Document Processing:

    • PDF loading with PyPDFLoader
    • Text splitting with RecursiveCharacterTextSplitter
    • Chunk size: 1500 characters
    • Chunk overlap: 200 characters

  • Embedding and Search:

    • HuggingFace embeddings (all-MiniLM-L6-v2 model)
    • DocArrayInMemorySearch vector store
    • MMR (Maximal Marginal Relevance) search
    • Configurable search parameters (k=5, fetch_k=10)

  • Chat Interface:

    • Streaming responses
    • Conversation memory
    • Context retrieval display
    • Message history management

Technical Implementation​

# Core components
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import PyPDFLoader
from langchain.memory import ConversationBufferMemory
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.chains import ConversationalRetrievalChain
from langchain.vectorstores import DocArrayInMemorySearch

Document Processing Pipeline​

  1. File Upload:

    • Accept multiple PDF files
    • Create temporary directory for processing
    • Load documents using PyPDFLoader

  2. Text Processing:

    • Split documents into manageable chunks
    • Create embeddings using HuggingFace model
    • Store in DocArray vector database

  3. Retrieval System:

    • Configure MMR search for diversity in results
    • Set up conversation memory
    • Initialize ChatOpenAI model
    • Create conversational retrieval chain

Features​

  • Document upload and processing
  • Full-text search capabilities
  • Result highlighting
  • Interactive search interface
  • Real-time streaming responses
  • Conversation history
  • Context-aware responses
  • Multiple document support

Implementation Steps​

  1. Environment Setup:

    pip install langchain streamlit openai pypdf sentence_transformers docarray

  2. Configure Streamlit Interface:

    • Set up page configuration
    • Create file upload widget
    • Implement chat interface
    • Add API key management

  3. Document Processing:

    • Implement document loading
    • Configure text splitting
    • Set up vector store
    • Initialize retrieval system

  4. Chat System:

    • Set up conversation memory
    • Configure LLM
    • Implement streaming responses
    • Add context retrieval display

Advanced Features​

  • Streaming response handler for real-time updates
  • Context retrieval display for transparency
  • Conversation buffer memory for contextual chat
  • MMR search for diverse result retrieval
  • Temperature control for consistent responses

Prerequisites​

  • Google Colab account
  • OpenAI API key
  • Basic Python knowledge
  • Understanding of:
    • Document processing
    • Search concepts
    • Web interfaces
    • Vector embeddings

Best Practices​

  • Use streaming for better user experience
  • Implement proper error handling
  • Cache resource-intensive operations
  • Manage conversation context effectively
  • Display search context for transparency

Resources​