🎙️ AI Meet Agent: Intelligent Meeting Studio

Deniz Calik

Mohamed Shefeeque

Aadel Mohamed

Navneet Sinha

AI Meet Agent is a high-performance, end-to-end meeting automation ecosystem. It transforms raw audio recordings or transcripts into professional, boardroom-ready PDF reports featuring intelligent executive summaries, precise action items, and participant engagement analytics.

Designed for the modern edge, AI Meet Agent supports Cloud AI (GPT-4o) and Local Edge AI (Phi-3 Mini) running natively on Qualcomm Snapdragon NPU (Windows ARM64), ensuring maximum privacy and ultra-low latency.

---

⚡ Main Interfaces

The AI Meet Agent ecosystem provides three distinct ways to interact with the pipeline:

1. 📊 Interactive Dashboard (Primary): A modern, glassmorphic web UI (frontend/index.html) for drag-and-drop processing and live log monitoring.
2. 🚀 Streamlit Studio: A quick-launch Streamlit application (streamlit_app.py) for simplified deployment and visualization.
3. Terminal Pipeline: A robust command-line interface (main_pipeline.py) for batch processing and automated workflows.

---

🧪 Pipeline Architecture

AI Meet Agent operates a high-fidelity 3-phase pipeline:

• Phase 0: Transcription & Diarization (transcribe.py) Uses WhisperX to convert audio/video into speaker-labeled transcripts. Supports x64 emulation on Snapdragon for desktop-class accuracy.
• Phase 1: Intelligent Analysis (transcript_to_json.py) Orchestrates the LLM (Cloud GPT-4o or Local NPU Phi-3 mini) to parse the raw text into structured meeting data (JSON).
• Phase 2: Boardroom Rendering (gene rate_report.py) A professional LaTeX engine transforms data into a polished PDF using a custom boardroom template (meetreport.sty).

---

🚀 Installation & Setup

1. Prerequisites

• Python 3.11+ (ARM64 recommended for Snapdragon hardware).
• MiKTeX — Required for LaTeX PDF generation (miktex.org).
• FFmpeg — Required for audio pre-processing.

2. Rapid Installation

Run the one-click industrial setup script:

./QuickSetup.bat

3. WhisperX x64 Helper (Snapdragon/ARM64)

If you are on ARM64 and need transcription, run the dedicated helper to set up the x64-emulated environment:

./setup_whisperx.bat

---

🛠 Usage

Launching the Dashboard (Recommended)

1. Start the Flask Backend:

   python app.py

2. Open frontend/index.html in your browser.

Launching via Streamlit

streamlit run streamlit_app.py

Running from CLI (Automation)

python main_pipeline.py --input meeting_transcript.txt --output "Executive_Summary"

---

📝 Configuration (.env)

Configure your .env (refer to .env.sample):

• OPENAI_API_KEY: API Key for Cloud Analysis.
• HF_TOKEN: Required for speaker diarization models.
• WHISPER_PYTHON_PATH: Path to your transcription environment.
• PHI3_MODEL_PATH: Local path to the Phi-3 ONNX model folder.

---

📦 Project Structure

├── app.py                # Flask Backend & Dispatcher
├── streamlit_app.py      # Streamlit Web Wrapper
├── main_pipeline.py      # CLI Automation Pipeline
├── transcribe.py         # Transcription Logic
├── QuickSetup.bat        # Standard Installer
├── setup_whisperx.bat    # x64 Emulation Installer
├── frontend/             # Dashboard Source
└── backend/
    └── report_generator/ # Analysis & LaTeX Core

---

Created for the Snapdragon AI Hackathon. Built with passion for Advanced Agentic Coding.