README.md

Backend Architecture and Workflow Overview

The backend system is designed to handle PDF uploads, process user questions, generate answers, and maintain a history of interactions.

Architecture Overview

1. FastAPI Framework:

   • Utilizes FastAPI for creating RESTful APIs with asynchronous capabilities, ensuring high performance and responsiveness.

2. Text Processing:

   • PDF uploads are processed to extract text using the PyPDF2 library.
   • Text is segmented into chunks using LangChain's RecursiveCharacterTextSplitter for efficient storage and analysis.

3. Embeddings and Similarity Search:

   • GoogleGenerativeAIEmbeddings from LangChain is used to create vector embeddings from text chunks.
   • These embeddings are stored and indexed using FAISS (Facebook AI Similarity Search) for efficient similarity search operations.

4. Question Answering Chain:

   • Utilizes a conversational question-answering model powered by ChatGoogleGenerativeAI and LangChain's load_qa_chain functionality.
   • User queries are processed against indexed text chunks to generate relevant and accurate answers.

5. Endpoints and Workflow:

   • PDF Upload Endpoint (/upload_pdf/):

     • Accepts PDF files for upload and processes them to extract text and create embeddings.
     • Stores the processed data for future queries.

   • Question Processing Endpoint (/ask_question/):

     • Accepts user questions and retrieves relevant document chunks from the indexed data.
     • Utilizes the question-answering chain to generate answers based on the context and user query.

   • History Retrieval Endpoint (/history/):

     • Provides access to the history of user queries and corresponding answers.
     • Enables users to review past interactions and responses.