AskDoc

AskDoc
Application that allows users to upload PDF documents and ask questions regarding the content of these documents. The backend will process these documents and utilize natural language processing to provide answers to the questions posed by the users.

Software Architecture
Overview
The application enables users to upload PDF documents and ask questions about their content. The backend, built with FastAPI, processes these documents using LangChain for natural language processing and a Hugging Face model to provide answers to the users' questions. The architecture includes endpoints for uploading PDFs, submitting questions, and processing documents to create and store vector embeddings locally.

Architecture Components
Frontend

User Interface: Web-based interface for users to upload PDFs and ask questions.
File Upload Form: Allows users to select and upload PDF files.
Question Submission Form: Allows users to input questions related to the uploaded PDFs.
Response Display: Displays the answers provided by the backend.
Backend

Framework: FastAPI for creating API endpoints and handling HTTP requests.
NLP Processing: LangChain for processing and managing language tasks.
Model Integration: Hugging Face models for generating responses to user questions.
Storage: Local file system for storing PDFs and vector embeddings.
Endpoints

PDF Upload Endpoint (/upload-pdf): Accepts PDF files and saves them locally.
Question Submission Endpoint (/question): Accepts user questions related to the uploaded PDFs.
Document Processing Endpoint (/answer): Processes the uploaded PDF, extracts text, creates vector embeddings, and stores them locally.
Detailed Architecture
Frontend

ReactJs: web technologies for creating the user interface.
File Upload Component: A form with file input to allow PDF uploads.
Question Input Component: A form for users to submit questions like a chatapp.
Answer Display Component: Displays the answers retrieved from the backend.
Backend (FastAPI)

Endpoints:
Upload PDF (/upload-pdf):
Method: POST
Function: Accepts PDF file upload, saves the file to a local directory.
Submit Question (/question):
Method: POST
Function: Accepts questions from the user, retrieves the relevant document, processes it, and returns the answer.
Process Document (/answer):
Method: POST
Function: Reads the uploaded PDF, extracts text, breaks it into chunks, embeds the text using LangChain, and stores the embeddings locally.
Processing Workflow

PDF Upload:
User uploads a PDF via the frontend.
The file is sent to the /upload-pdf endpoint and saved locally.
Document Processing:
Triggered by the /answer endpoint.
Reads the PDF from the local storage.
Extracts text and processes it into chunks.
Uses LangChain to create embeddings of these chunks.
Stores the embeddings in a local vector store.
Question Answering:
User submits a question via the frontend.
The question is sent to the /question endpoint.
The backend retrieves the relevant document embeddings.
LangChain processes the question and retrieves relevant text chunks.
Hugging Face model generates an answer based on the retrieved text.
The answer is returned to the frontend and displayed to the user.
Storage

Local File Storage:
PDFs are stored in a specific directory.
Vector embeddings are stored in a local vector store.
Technologies Used
FastAPI: Framework for building API endpoints.
Reactjs: Framework for building Web UI.
LangChain: For text processing, chunking, and creating embeddings.
Hugging Face: For utilizing pre-trained NLP models to generate answers.
Local Storage: For saving PDFs and vector embeddings.