Text Prediction and Information Retrieval Assistant

Overview

This project integrates a text prediction model with a Retrieval-Augmented Generation (RAG) model using an LLM (Large Language Model) to provide an intelligent assistant capable of predicting text in real-time and retrieving relevant information from PDF documents. The goal is to enhance user productivity by combining predictive typing with document-based information retrieval.

Features

• Contextual Text Prediction: The assistant offers word and phrase suggestions as you type, based on the content of uploaded documents.
• Information Retrieval from PDFs: The RAG model can extract relevant information from a corpus of PDF documents, answering user queries contextually.
• AI-Powered Response Generation: By using a hybrid of retrieval and generation, the assistant can both fetch exact information from the PDFs and generate natural responses based on LLMs.

Workflow and Functionality

The application functions in two major workflows:

1. Text Prediction:

   • Uses a pre-trained language model for providing word suggestions.
   • Takes into account the content of the uploaded PDFs to give context-aware suggestions.

2. Retrieval-Augmented Generation (RAG):

   • Leverages a RAG model that combines a retriever and generator.
   • The retriever fetches relevant text passages from PDFs, and the generator (LLM) produces a response based on both the retrieved information and the user’s input.

Dependencies

Ensure the following dependencies are available inside your AI Workbench environment:

• Python 3.8+
• Libraries:
  • Tkinter
  • Tensorflow
  • faiss-cpu==1.8.0.post1
  • huggingface-hub==0.25.1
  • langchain==0.3.1
  • langchain-community==0.3.1
  • langchain-core==0.3.6
  • langchain-text-splitters==0.3.0
  • langsmith==0.1.129
  • numpy==1.26.4
  • PyPDF2==3.0.1
  • threadpoolctl==3.5.0
  • tokenizers==0.20.0
  • torch==2.4.1
  • transformers==4.45.1

Restrictions

• Models: The application requires pre-trained language models from Hugging Face (microsoft/Phi-3.5-mini-instruct) and a retrieval model for RAG.

Project Structure

• text_editor.py: Contains the logic for the text prediction module.
• rag_pre.py: Implements the RAG model for document-based information retrieval and response generation.
• requirements.txt: Lists all dependencies necessary to run the project.
• Dockerfile: Defines the container image for the application to run on NVIDIA AI Workbench.

Getting Started

1. Clone the Repository: you can use github to easily clone or get started with a workspace. Install all the requirements.
2. Download Requirements: By using requirements.txt
3. Python files: Ensure all files are in the same directory including the pdf you want to read for this version followed by running only the texteditor.py file 4: Usage: Start typing as normal and use ':' and a space to trigger the RAG model. After you get the suggestions, you can select it to autofill. (Check for API rate limit if text editor prompts error in finding suggestions)