Meeting Miner

Inspiration

Distributed teams lose critical meeting context. Action items get forgotten, decisions aren't tracked, and people who miss meetings are left in the dark. I wanted to fix this with intelligent orchestration, not just summarisation.

What it does

MeetingMiner extracts actionable intelligence from any meeting content: Action Items - Tasks, owners, deadlines, priorities automatically identified Key Decisions - What was decided and why Conflict Detection - Identifies blockers and risks Cross-Meeting Intelligence - Tracks recurring topics across multiple meetings Visual Dashboard - Priority-based overview of all action items Professional Exports - PDF reports and clipboard sharing

Upload screenshots, transcripts, or PDFs → Get structured insights in seconds.

How it's built

Gemini 3 Flash Preview - Multimodal analysis (images, PDFs, text) Node.js + Express - Backend orchestration Structured Prompting - Carefully designed prompts for consistent extraction In-memory History - Tracks up to 10 meetings for context PDFKit - Professional document generation

The key innovation: multi-step reasoning that compares meetings, identifies patterns, and maintains context over time.

Challenges

Markdown Parsing - Gemini's output format varied; had to build a robust parser Context Window Management - Balancing historical context vs token limits Action Item Extraction - Getting consistent structured data from unstructured content Deployment - Managing environment variables across platforms securely

Accomplishments

True Orchestration - Not just RAG or summarization, but multi-step reasoning across meetings Cross-Meeting Intelligence - The system actually "remembers" and connects discussions Production Ready - Clean architecture, error handling, professional UX Free Tier Friendly - Optimized for efficient token usage

What I learned

Prompt engineering is critical - small changes = massive output improvements Multimodal AI opens incredible possibilities for real-world workflows Good architecture matters - separating concerns made iteration fast Users value structure over summaries

What's next for MeetingMiner

Email Integration - Analyse email threads automatically Calendar Sync - Auto-fetch meeting recordings/transcripts Team Analytics - Identify patterns in team communication health Slack/Teams Bots - Real-time analysis in chat channels Smart Notifications - Alert owners when deadlines approach

Built With

• express.js
• gemini3
• node.js
• pdfkit