mudiit4/ocr_and_rag

OCR and RAG implementation

Handwriting recognition and Medical chatbot

This repository provides a thorough summary of my contributions to Noora Health. Utilizing Google's Gemini ProVision and DocumentAI technologies, I developed strong methods to pull tabular data from various images of paper documents. Furthermore, I led the creation of a RAG model that can effectively answer user questions in various languages, with the help of an advanced translation interface.

Project Overview:

OCR

The repository Streamlit_OCR is for the OCR part.

Prerequisites

• Python version >= 3.9 installed on your local machine

Installation

1. Clone the repository to your local machine: git clone <repository_url>

2. Navigate to the directory "Streamlit_OCR":

3. Install all the dependencies: pip install -r requirements.txt

4. Also install Google Cloud CLI installer from here: https://cloud.google.com/sdk/docs/install

5. The Google Cloud CLI is used for authorisation.

6. To know how to use Google Cloud CLI for verification watch this: https://www.youtube.com/watch?v=gpAiUerUdEA

Running Locally

Follow these steps to run the project locally:

1. Make sure gemini_chained.py is in the same folder as app_trial.py.
2. Open gemini_chained.py and:

• Enter your API key in line 282.
• Update other details from line 17 to 20 and 263 to 265 as needed.

3. In your terminal, run the following command: streamlit run app_trial.py

RAG Model

The repository RAG is for RAG Model using the Q-A bank given to us.

• Final app is in app.py

Installation

For running the code first install all the dependencies by

pip install -r requirements.txt

Then run the app.py file by running the following command on the command propmt to get a stremlit UI

streamlit run app.py

To change the Query please navigate the code in app.py and change the message variable with whatever Query you want to give in.