About vs. Demo: This demo folder now uses a standardized launch page. Start with About Demo, then choose the run mode that fits your class.

RAG & NLP • Browser • 35-45 min

University Knowledge Assistant

A comprehensive demonstration of advanced RAG concepts combining Graph RAG, PageIndex, Local SLMs, and Voice I/O.

Open About Demo Read README View Course Path
Ready
Mode: Hybrid RAG

📑 Ingested Documents

Upload PDFs or enter URLs to begin.

🔍 Knowledge Graph

Graph will appear after ingestion.

🤖 Answer

Ask a question to see the answer with citations...

Citations will appear here...

📖 System Architecture

  1. Ingestion: PDFs → Text → Chunks
  2. Indexing: PageIndex for page-level retrieval
  3. Graph Building: FalkorDB for entity relationships
  4. Query: STT → Hybrid RAG → TTS
Technical Stack
  • 📄 PDF.js - PDF text extraction
  • 🔢 PageIndex - Document indexing with citations
  • 🔗 FalkorDB - Knowledge graph (local)
  • 🧠 Local SLM - Phi-3.5/Mistral via llama.cpp
  • 🎤 Web Speech API - STT/TTS

Standard demo guide

Use this demo in a logical learning sequence

Starts immediately in browser with no installs, no API keys, and classroom-safe defaults.

What this demo is about

A comprehensive demonstration of advanced RAG concepts combining Graph RAG, PageIndex, Local SLMs, and Voice I/O.

Learning objectives

  • Explain the main rag/nlp decision that University Knowledge Assistant is designed to support.
  • Change input assumptions and predict how the output should respond before running the demo.
  • Interpret the result in plain language, not just as a number, chart, or AI recommendation.

Run mode and expectations

  • Supported modes: Browser
  • Starts immediately in browser with no installs, no API keys, and classroom-safe defaults.

Step 1: Inputs

  • Start with the default assumptions, then change one variable at a time so students can isolate cause and effect.
  • Treat each input as a lever that changes the scenario, baseline, or business context behind the result.

Step 2: Decision buttons

  • Use the main run or simulate action to compute the scenario after inputs are set.
  • Use export or reset actions, when present, to compare runs or return to a classroom-safe baseline.

Step 3: Outputs and what to notice

  • Read the top-line result first, then look for supporting metrics, tables, or narratives that explain why it changed.
  • Students should explain whether the output is descriptive, predictive, simulated, or recommended.
  • Look for student query, knowledge source, retrieved answer, and citation
  • Observe how answer quality changes when the knowledge base is specific versus generic

Available run modes

  • Browser: available for this demo.

How to proceed

  1. Choose the run mode that fits the class: Browser.
  2. Review the default assumptions before changing anything.
  3. Change one or two inputs, then use `Run the main action`.
  4. Read the output first, then compare any supporting metrics, charts, or AI text.
  5. Capture one insight, one limitation, and one action recommendation.