search

microsoft / RAG_Hack

Hack Together: RAG Hack | Register, Learn, Hack
MIT License
393 stars 81 forks source link

Project: AI-Powered Research Paper Assistant #91

Open iAMSagar44 opened 1 month ago

iAMSagar44 commented 1 month ago

Project Name

Paper Mentor AI

Description

Paper Mentor AI is an intelligent research assistant designed to help students, researchers, and professionals streamline their research process. This AI-powered tool leverages Retrieval-Augmented Generation (RAG) to help users search, analyze, and extract insights from vast databases of academic papers, journals, and research documents. By refining search queries based on user input and conversation history, the app ensures relevant, accurate information is quickly retrieved.

For this project, a PGvector VectorStore has been used to store document embeddings and perform similarity searches. This allows for efficient retrieval of relevant documents based on user queries.

Key Features - Data Ingestion Process:

  1. Document Analysis using Azure Document Intelligence: The Document Intelligence Layout model is used to extract paragraphs and tables from PDFs, creating specific chunks of the documents for better data processing.
  2. Extracting Document Keywords: A generative AI model is used to extract keywords and generate a summary from the first two pages of each document. The keywords improve the document's searchability and help in retrieving more accurate results.

    Key Features - Prompting and LLM Orchestration:

  3. AI Agents: The backend Java application uses two AI agents. The first agent refines the user's question for better search queries, while the second extracts unique keywords from the user’s input. These keywords are used to filter document titles, which are then applied in the vector search.
  4. Retrieval-Augmented Generation (RAG): RAG architecture is employed to provide precise and context-aware answers by fetching relevant documents and combining them with generative responses.
  5. Function (Tool) Calling: The application uses two functions. One retrieves a list of relevant research papers based on the search query, while the other fetches the summary of a specific paper for the user to review.

Paper Mentor AI enhances productivity and simplifies the research process for students, educators, researchers, and anyone involved in academia by providing an efficient and intuitive tool for academic research.

Tech stack:

The tech stack for this application includes:

Frontend: Next.js (React framework) Styling: Tailwind CSS for responsive and modern UI Backend: Java and Spring Boot for backend services and API endpoints LLM Orchestration: Spring AI for orchestrating calls with LLMs and Vector Store Vector Store: Integration with Postgres vector store for document indexing

Technology & Languages

Project Repository URL

https://github.com/iAMSagar44/paper-mentor-ai

Deployed Endpoint URL

No response

Project Video

Project Video

Team Members

iAMSagar44 (Sagar)

Yash-Kavaiya commented 1 month ago

@iAMSagar44 Great work

iAMSagar44 commented 1 month ago

Updates - provided a new video link in the project (as below).

Project Video

multispark commented 2 weeks ago

Hello @iamsagar44, thank you for participating in RAG Hack!

The team is working hard to distribute badges. Please have each team member fill out this form: aka.ms/raghack/badge-dist

Thank you!