MeetingSecretaryAI

Overview

MeetingSecretaryAI is a comprehensive automated pipeline designed to generate structured meeting minutes from raw transcripts using Azure OpenAI models. It transforms unstructured meeting transcripts into well-organized outputs, including JSON, DOCX, and Markdown formats, following a standardized schema.

This tool is especially tailored for academic and scientific meetings, such as NIH-style study sections, providing high-quality, reproducible documentation of discussions, decisions, and actions.

🚀 NEW: Enhanced GUI with Smart Features

The latest version includes a powerful GUI with automatic meeting detection, audio processing, and intelligent workflow management.

✨ Key Features

Core Functionality

🤖 AI-powered transcript summarization using Azure OpenAI
📜 JSON output conforming to a strict meeting minutes schema
📝 DOCX and Markdown minutes generation with professional formatting
🛠️ Configurable via .env and config.ini
🔄 Cross-platform (Linux, MacOS) compatible via conda/mamba

🆕 Enhanced GUI Features

� Automatic Zoom Meeting Detection - Scans and lists recent meetings
🎵 Audio Processing - Direct WhisperX integration for transcription
👥 Participant Extraction - Automatically identifies speakers
📅 Smart Time Rounding - Rounds meeting times to 15-minute intervals
🎯 Context Management - Easy selection from preset context files
📋 Auto-Agenda Generation - Creates formatted agendas automatically

Multiple Input Methods

Zoom Meetings: Browse and select from recent meetings
Audio Files: Process with WhisperX for transcription
Transcript Files: Upload existing transcripts
Folder Input: Process entire folders containing transcripts

🎯 Quick Start

Need a guided, beginner-friendly install (Windows/macOS/Linux)? Follow the full tutorial in doc/INSTALLATION.md. It covers prerequisites, Conda/Mamba setup, ffmpeg installation, Hugging Face tokens, and the first GUI launch with checkpoints after each step.

Snapshot of the install steps

Install Git, Miniforge/Mambaforge, and ffmpeg using the commands for your OS (PowerShell, zsh, or bash) in the tutorial.
Clone this repository and create the meetingsecretaryai_env Conda environment with Python 3.9.
Run pip install -r requirements.txt, add your Hugging Face and Azure credentials, and verify with the provided smoke tests.
Launch the GUI with python meeting_secretary_gui.py (or double-click launch_with_diagnostics.command on macOS).

Fast commands (once prerequisites are in place)

git clone https://github.com/FritscheLab/MeetingSecretaryAI.git
cd MeetingSecretaryAI
conda create -n meetingsecretaryai_env python=3.9 -y
conda activate meetingsecretaryai_env
pip install --upgrade pip
pip install -r requirements.txt
python meeting_secretary_gui.py

Automated helper (macOS/Linux/WSL)

chmod +x setup_enhanced.sh
./setup_enhanced.sh

The script installs ffmpeg, creates the Conda env, and runs dependency checks. Windows users can follow the PowerShell instructions in doc/INSTALLATION.md to achieve the same result manually.

🔧 Main Scripts

Script	Description
`meeting_secretary_gui.py`	Graphical interface for processing transcripts and generating minutes.
`meeting_utils.py`	Helper library for tasks such as Zoom meeting detection and token management.
`scripts/transcript2json.py`	Converts raw transcripts into structured JSON minutes.
`scripts/json_refine.py`	Refines JSON minutes to reduce redundancy and improve readability.
`scripts/json2word.py`	Creates DOCX or Markdown minutes from JSON.
`scripts/generate_minutes.sh`	Convenience shell script that chains the conversion steps.

The typical workflow is to run scripts/transcript2json.py on a transcript file, pass the results through scripts/json_refine.py for redundancy and verbosity cleanup, and then use scripts/json2word.py to produce documents. For a graphical experience, launch meeting_secretary_gui.py, which orchestrates the full process. The shell script scripts/generate_minutes.sh demonstrates how these pieces fit together.

📁 Repository Structure

MeetingSecretaryAI_1.0/
├── meeting_secretary_gui.py          # Main GUI application
├── meeting_utils.py                   # Utility classes
├── setup_enhanced.sh                  # Automated setup script
├── requirements.txt                   # Python dependencies
├── config.ini                         # Configuration file
├── CONTRIBUTING.md                    # How to contribute
├── CHANGELOG.md                       # Release history
├── .env                              # Environment variables
├── scripts/                          # Core processing scripts
│   ├── transcript2json.py            # Transcript to JSON conversion
│   ├── json_refine.py                # JSON refinement for readability
│   ├── json2word.py                  # JSON to document conversion
│   ├── generate_minutes.sh           # Shell script pipeline
│   ├── prompt_*.md                   # Various detail level prompts
│   ├── prompt_refine.md              # Refinement prompt
│   └── minutes_schema.JSON           # Output schema definition
├── doc/                              # Documentation
│   ├── templates/                    # Template files
│   ├── transcript2json.md            # Script documentation
│   ├── json2word.md                  # Script documentation
│   ├── FAQ.md                        # Frequently asked questions
│   └── architecture.md              # Workflow diagram
├── tests/                            # Test scripts
│   ├── test_enhanced_features.py     # Feature testing
│   └── test_time_rounding.py         # Time rounding tests
└── archive/                          # Archived files
    ├── old_gui/                      # Previous GUI versions
    └── development_docs/             # Development documentation

External Data Structure:
../MeetingSecretaryAI_Data/
├── .hf_token.txt                     # HuggingFace token (for WhisperX)
├── context/                          # Context files for different meeting types
│   ├── COMPASS_PRS.md
│   ├── COMPASS_Genetics.md
│   └── ...
├── data/                             # Meeting data
└── output/                           # Generated minutes (default)

🖥️ Using the GUI

1. Input Methods Tab

Choose your input method:

Latest Zoom Meetings: Automatically detects recent meetings
Audio Files: Select .m4a, .mp3, .wav files for processing
Transcript Files: Upload existing transcript files
Input Folders: Process folders containing transcripts

2. Meeting Details Tab

Edit meeting information (auto-populated from Zoom)
Auto-generate formatted agendas
Manage participant lists

3. Context & Agenda Tab

Select from preset context files
Configure agenda input method
Paste or upload custom content

4. Settings Tab

Choose detail level (Concise, Action-Focused, Moderate, High Detail, In-Person)
Set output format (DOCX, Markdown, Both)
Configure HuggingFace token for audio processing
Set output directory

🎵 Audio Processing

WhisperX Integration

The system includes built-in WhisperX support for high-quality transcription:

Prerequisites:

ffmpeg must be installed on your system (handled automatically by setup script)
HuggingFace token for speaker diarization models

Usage:

Set HuggingFace Token: Get token from HuggingFace
Select Audio File: Choose your audio file
Process Audio: Click "Process Audio" button
Generate Minutes: Transcript automatically loads

Supported Audio Formats

.m4a (Zoom recordings)
.mp3
.wav
.flac

📝 Detail Levels

Choose from multiple detail levels:

Level	Description	Use Case
Concise	Brief summaries	Quick overviews
Action-Focused	Outcome- and action-oriented, de-duplicated	Browseable weekly minutes
Moderate	Balanced detail	Standard meetings
High Detail	Comprehensive coverage	Important decisions
In-Person	No speaker identification	Unreliable audio

🔧 Configuration

Environment Variables (`.env`)

MODEL=gpt-5.1
REFINEMENT_MODEL=o3-mini
OPENAI_API_BASE=https://api.umgpt.umich.edu/azure-openai-api
AZURE_OPENAI_API_KEY=your_api_key
OPENAI_ORGANIZATION=your_org_id
API_VERSION=2025-04-01-preview

Response Settings (`config.ini`)

[response_settings]
temperature = 0
max_tokens = 30384
max_completion_tokens = 80000
top_p = 1.0
frequency_penalty = 0.0
presence_penalty = 0.0
reasoning_effort_transcript = low
reasoning_effort_refine = medium

🚀 Advanced Usage

Command Line Interface

For automation and batch processing:

# Generate minutes from transcript
python scripts/transcript2json.py \
  --input_file transcript.txt \
  --context_file context.md \
  --agenda_file agenda.md \
  --output_file minutes.json

# Refine JSON for readability
python scripts/json_refine.py \
  --input_json minutes.json \
  --output_json minutes_refined.json

# Convert to documents
python scripts/json2word.py \
  --input_json minutes_refined.json \
  --output_dir output/ \
  --output_prefix meeting_minutes \
  --output_format both

Shell Script Pipeline

bash scripts/generate_minutes.sh

🧪 Testing

Test Enhanced Features

python tests/test_enhanced_features.py

Test Time Rounding

python tests/test_time_rounding.py

📊 Smart Features

Automatic Meeting Detection

Scans ~/Documents/Zoom for recent meetings
Extracts meeting names, dates, and times
Identifies participants from transcripts

Intelligent Time Handling

Rounds meeting times to 15-minute intervals
Example: 13:07:22 → 13:00:00

Context Management

Preset context files for different meeting types
Easy switching between contexts
Custom context support

🔍 Troubleshooting

Common Issues

Permission denied / Operation not permitted:
- Problem: Cannot access ~/Documents/Zoom folder
- Solution: System Preferences > Security & Privacy > Privacy > Files and Folders
- Find "Terminal" and check "Documents Folder"
- Restart the application
- Alternative: Use "Select Transcript File" or "Select Audio File" instead
No Zoom meetings found: Check ~/Documents/Zoom directory exists
Audio processing fails:
- Problem: "No such file or directory 'whisperx'" or "ffmpeg not found"
- Solution:
  - Verify HuggingFace token in Settings tab
  - Install ffmpeg: brew install ffmpeg (macOS) or sudo apt install ffmpeg (Ubuntu)
  - Ensure using correct Python environment: conda activate meetingsecretaryai_env
API errors: Check .env configuration file
Import errors: Ensure conda activate meetingsecretaryai_env

Quick Launch

Double-click: launch_with_diagnostics.command for guided startup with permission checking

Getting Help

Check the archive/development_docs/ for detailed documentation
Run test scripts to verify functionality
Review configuration files

📈 Output Examples

JSON Structure

{
  "meeting_info": {
    "title": "COMPASS PRS Meeting",
    "date": "2025-03-07",
    "time": "13:00:00",
    "participants": ["Dr. Smith", "Dr. Johnson"]
  },
  "agenda_items": [...],
  "decisions": [...],
  "action_items": [...]
}

Generated Documents

DOCX: Professional Word document with formatting
Markdown: Clean text format for web publishing
JSON: Structured data for further processing

🤝 Contributing

Fork the repository
Create a feature branch
Test your changes
Submit a pull request

📄 License

This project is licensed under the GNU General Public License v3.0.

👥 Author

Fritsche Lab
https://github.com/FritscheLab

🔄 Version History

v2.0: Enhanced GUI with Zoom integration, audio processing, and smart features
v1.0: Original command-line interface and basic functionality

Name		Name	Last commit message	Last commit date
Latest commit History 20 Commits
doc		doc
scripts		scripts
tests		tests
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
config.ini		config.ini
launch_with_diagnostics.command		launch_with_diagnostics.command
meeting_secretary_gui.py		meeting_secretary_gui.py
meeting_utils.py		meeting_utils.py
requirements.txt		requirements.txt
setup_enhanced.sh		setup_enhanced.sh
sitecustomize.py		sitecustomize.py

License

FritscheLab/MeetingSecretaryAI

Folders and files

Latest commit

History

Repository files navigation