ConEx

Concept Explainer Agent

Agentic RAG Streamlit Application — Multi-Document Concept Explainer

Making changes

📚 Purpose

This application provides a research-assistant-style platform where users can:

• Upload or link multiple PDF research papers
• Automatically index them into an in-memory FAISS vector store
• Ask concept-level questions
• Retrieve and summarize relevant context using LangChain + OpenAI embeddings
• Get a coherent, Agentic AI explanation synthesized via CrewAI agents

🧩 Key Features

• ✅ Upload multiple PDFs (up to 5)
• ✅ Index via OpenAI embeddings (FAISS vector store)
• ✅ Multi-document RAG retrieval
• ✅ CrewAI-based retriever + summarizer agents
• ✅ Streamlit UI with real-time interaction
• ✅ Session caching and reruns for seamless UX
• ✅ In-memory, privacy-preserving local processing

🚀 Core Functionality Overview

1. Streamlit Frontend

The app runs an interactive Streamlit interface with two main sections:

• Left Panel: Handles initialization, OpenAI API setup, and PDF uploads (or URL-based downloads)
• Right Panel: Enables the user to chat with the uploaded papers and explore scientific concepts

All application states (API key, docs, embeddings, chat history) are stored in Streamlit's session state for persistence during runtime.