Hugging Face's logo

Spaces:
aamanlamba
/
Lineage-graph-accelerator
Sleeping

App Files Community
Lineage-graph-accelerator / DEMO_VIDEO_SCRIPT.md
aamanlamba's picture
aamanlamba
Add demo video links to README and BUILD_PLAN
418445b
preview code
|
raw
history blame contribute delete
7.44 kB

A newer version of the Gradio SDK is available: 6.1.0

Upgrade

Demo Video Script - Lineage Graph Accelerator

Target: 2-3 minutes | Competition: Gradio Agents & MCP Hackathon Winter 2025

🎬 SETUP BEFORE RECORDING

Pre-recording Checklist:

Recording Tools (Choose One):

  1. macOS: QuickTime Player (File → New Screen Recording)
  2. Free: OBS Studio (https://obsproject.com)
  3. Browser: Loom (https://loom.com) - easiest, records directly

📝 SCRIPT

[0:00-0:15] Introduction (15 seconds)

[Show the HuggingFace Space with the purple gradient banner visible]

"Hi, I'm Aaman Lamba, and this is the Lineage Graph Accelerator - an AI-powered tool that automatically extracts and visualizes data lineage from your data pipelines, built with Gradio 6 and MCP integration."

ACTION: Scroll slowly down the page to show the "What You Can Do" section

[0:15-0:35] Problem Statement (20 seconds)

[Pause on the features table]

"Data engineers and analysts struggle to understand how data flows through their systems. Manual lineage documentation is tedious and quickly becomes outdated. This tool solves that by automatically parsing your dbt manifests, Airflow DAGs, SQL files, and even natural language descriptions."

[0:35-1:05] Feature #1: AI Assistant with Gemini (30 seconds)

[Click on "AI Assistant" tab]

"Let me show you the AI-powered assistant. I'll paste my Gemini API key and ask it to generate lineage for an e-commerce pipeline."

ACTION:

  1. Paste API key into the field
  2. Type in chatbot: "Create a lineage graph for an e-commerce platform with orders, customers, and products tables feeding into a sales mart"
  3. Click Send
  4. Wait for response (should generate JSON)
  5. Click "Use Generated JSON" button

SAY WHILE WAITING:

"The assistant uses Google Gemini to understand natural language and generate valid lineage JSON automatically."

AFTER CLICKING "Use Generated JSON": 6. Click "Extract Lineage" button 7. Show the generated Mermaid graph

"And there's our interactive lineage graph with color-coded nodes. You can click to zoom in."

[1:05-1:30] Feature #2: MCP Integration (25 seconds)

[Click back to "Extract Lineage" tab]

"Now let's try the MCP integration - this connects to Model Context Protocol servers for metadata."

ACTION:

  1. Check the "Use MCP Server for Metadata" checkbox
  2. Select "Local Demo MCP (Built-in)" from the dropdown
  3. In MCP Query field, type: "ecommerce"
  4. Click "Extract Lineage"
  5. Show the resulting graph

SAY:

"The MCP server provides pre-configured metadata. This works with any MCP-compatible server on HuggingFace."

[1:30-1:55] Feature #3: Demo Gallery & Export (25 seconds)

[Click on "Demo Gallery" tab]

"Need inspiration? The Demo Gallery has real-world examples."

ACTION:

  1. Click "Load Sample" button for "Complex E-commerce Platform (50+ nodes)"
  2. Show it loading in the text area
  3. Go back to "Extract Lineage" tab
  4. Click "Extract Lineage"
  5. Briefly show the large complex graph

[Scroll down to Export section]

"And you can export to any major data catalog - Collibra, Microsoft Purview, Alation, Apache Atlas, or OpenLineage standard."

ACTION:

  1. Select "Collibra" from dropdown
  2. Click "Generate Export"
  3. Show the JSON output briefly
  4. Click "Copy to Clipboard"

[1:55-2:15] Feature #4: Additional Capabilities (20 seconds)

[Scroll to show the graph visualization features]

"The graphs are fully interactive - click to zoom, download as PNG or SVG, or edit in Mermaid Live."

ACTION:

  1. Click on the graph to zoom
  2. Point to the "Download PNG" and "Download SVG" buttons
  3. Point to "Edit in Mermaid Live" link

SAY:

"All of this supports multiple input formats: dbt, Airflow, SQL DDL, BigQuery, and custom JSON."

[2:15-2:30] Call to Action (15 seconds)

[Scroll back up to show the banner with your website link]

"This project demonstrates the power of combining Gradio 6, MCP, and AI agents for productivity. Try it yourself on HuggingFace Spaces, check out the full documentation, and visit aamanlamba.com for more projects. Thanks for watching!"

ACTION:

  1. Hover over "By Aaman Lamba" button to show it's clickable
  2. Show the HuggingFace Space URL in the browser bar
  3. End recording

🎯 KEY MESSAGES TO EMPHASIZE

  1. AI-Powered: Google Gemini integration for natural language lineage generation
  2. MCP Integration: Connects to Model Context Protocol servers (competition requirement)
  3. Multiple Formats: dbt, Airflow, SQL, BigQuery, JSON
  4. Enterprise Ready: Exports to 5 major data catalogs
  5. Gradio 6: Modern, professional UI (competition requirement)
  6. Real-World Impact: Solves actual data governance problems

✅ POST-RECORDING

  1. Review the video:

    • Check audio quality
    • Ensure all features are clearly visible
    • Verify timing is under 5 minutes (ideally 2-3 min)

  2. Edit if needed:

    • Trim any long pauses
    • Add title card (optional): "Lineage Graph Accelerator | Gradio + MCP Hackathon 2025"
    • Add your website URL as overlay at the end (optional)

  3. Export settings:

    • Format: MP4
    • Resolution: 1080p (1920x1080) or 720p (1280x720)
    • Frame rate: 30fps
    • Bitrate: At least 5 Mbps for good quality

  4. Upload:

    • YouTube (recommended - can embed in README)
    • Loom (easiest, shareable link)
    • Google Drive (make sure it's publicly accessible)

  5. Update README.md:

    • Add video link under "Demo Video" section
    • Update BUILD_PLAN.md to mark video as complete

💡 RECORDING TIPS

DO:

  • Speak clearly and at moderate pace
  • Show each feature for at least 5 seconds
  • Let visual changes complete before moving on
  • Smile while talking (it affects your voice tone!)

DON'T:

  • Rush through features
  • Use filler words ("um", "uh", "like")
  • Apologize for anything
  • Go over 5 minutes (judges won't watch it all)

IF SOMETHING GOES WRONG:

  • Pause, take a breath, and continue from that section
  • You can edit out mistakes later
  • Or just re-record that section

🎥 EASIEST RECORDING METHOD (Loom)

  1. Install Loom browser extension: https://loom.com
  2. Click Loom icon → "Start Recording"
  3. Select "Screen + Camera" (or just "Screen Only")
  4. Choose "Current Tab"
  5. Click "Start Recording"
  6. Follow the script above
  7. Click "Stop" when done
  8. Loom auto-uploads and gives you a shareable link
  9. Download the MP4 if needed

Total time: 5-10 minutes including setup!

📊 SUGGESTED TIMELINE

Time Section Key Action
0:00 Intro Show banner, introduce yourself
0:15 Problem Explain data lineage challenge
0:35 AI Assistant Demo Gemini chatbot generating lineage
1:05 MCP Show MCP server integration
1:30 Demo Gallery Load complex sample + export
1:55 Features Interactive graph, downloads
2:15 CTA Website link, HF Space, thank you

Total: ~2.5 minutes ✅

Good luck with the recording! 🎬