PDF-Query

A Lang-chain project using Cassandra DB supported by Google Gemini AI model and API. Showcase use RAG pipeline for usage of PDF source for answering the input queries

Dataflow Diagram

graph TD
    subgraph Ingestion_Flow
        A[PDF Document] --> B["Reading the Document (Text Extraction & Splitting)"]
        B --> C[Text Chunks]
        C --> D[Google Gemini Embeddings]
        D --> E["Vector Database (Cassandra/Astra DB)"]
    end

    subgraph Query_Flow
        F[Human] --> G["Text + Query"]
        G -- "Similarity Search" --> E
        E -- "DataStax Vector Search" --> H["Text embeddings \n(Relevant context)"]
        H --> I[Google Gemini LLM]
        G --> I
        I --> J[Final Answer]
    end

Loading

Setup

1. Install Dependencies:

   pip install -r requirements.txt

2. Environment Variables: Create a .env file based on .env.example and add your credentials:

   • ASTRA_DB_APPLICATION_TOKEN
   • ASTRA_DB_ID
   • GOOGLE_API_KEY

How to Run

Launch the Streamlit application:

streamlit run app.py

Features

• Dynamic PDF Upload: Upload any PDF to query its content.
• Astra DB Integration: Powered by DataStax for high-performance vector search.
• Google Gemini AI: Uses state-of-the-art models for embeddings and text generation.

Resources

• DataStax Astra DB
• Google AI Studio
• LangChain Documentation