MULTIBRAIN-AI

Turning messy meetings into organized tasks, timelines, and ownership.

Built with the tools and technologies:

---

Overview

MULTIBRAIN-AI is an AI-powered meeting and workload dashboard.
It turns meeting recordings or transcripts into structured tasks, assigns them to employees, syncs events to a calendar, and notifies the right people by email.

You can:

• Add events and see them in a calendar-style view.
• Upload audio recordings or transcripts from meetings.
• Let Whisper transcribe recordings.
• Use Ollama (LLaMA 2) to extract actionable tasks.
• Assign tasks to employees from the UI.
• Track each employee’s pending vs completed tasks.
• Filter and update task status.
• Automatically send:
  • notification emails to the whole team when a new event is added,
  • reminder emails to the specific assignee when tasks are created.

---

System Architecture

---

Features

Event & Calendar Management

• Create events with title, date, time, category and description.
• Events are stored in the backend and rendered in the calendar view (events.html).
• Every new event can trigger an email notification so the full team knows what’s coming up.

AI Task Extraction (Whisper + Ollama)

• Upload either:
  • a recording (.mp3, .wav) or
  • a transcript (.txt, .pdf, .docx).
• Whisper is used to transcribe audio files.
• The transcript is sent to Ollama with a strict JSON prompt.
• Ollama returns a list of tasks with:
  • description
  • suggested assignee + confidence
  • deadline
  • source quotes
• Tasks are stored in SQLite (Task + Meeting models).

Employee & Workload View

• Employee list with name, role, position, email, avatar.
• For each employee you can see:
  • total pending tasks
  • total completed tasks
• You can view all tasks for a specific employee.

Task Management & Filters

• See all tasks in the Tasks page with:
  • description
  • status (pending / complete)
  • AI-assignee + confidence
  • deadline
  • linked meeting file
• Filter tasks by status (pending / complete).
• Change status from the UI (calls PUT /tasks/<id>/status).

Email Notifications

• For AI-generated tasks, the app groups tasks by assignee and sends them a reminder email with:
  • task list
  • deadlines
  • short message.
• When you create new events, you can notify the whole team so they don’t miss any important meeting.
• Uses Gmail SMTP over SSL.

---

Project Structure

.
├── app.py
├── requirements.txt
├── .env                # local environment variables (not committed)
├── .gitignore
├── static/
│   ├── dash.css
│   ├── dash.js
│   ├── employee.css
│   ├── employee.js
│   ├── events.css
│   ├── events.js
│   ├── script.js
│   ├── style.css
│   ├── tasks.css
│   └── tasks.js
├── templates/
│   ├── dash.html       # main dashboard
│   ├── employee.html   # employees view
│   ├── events.html     # calendar / events page
│   ├── index.html      # auto-assign / upload page
│   └── tasks.html      # tasks list + filters
└── uploads/
    └── sample-0.mp3    # sample meeting recording

---

Setup & Installation

Clone the Repo

git clone https://github.com/hannahjan06/MULTIBRAIN.git
cd MULTIBRAIN

---

Create & Activate Virtual Env (Optional but recommended)

python -m venv venv
# macOS / Linux
source venv/bin/activate
# Windows
venv\Scripts\activate

---

Install Dependencies

requirements.txt should look roughly like:

Flask
Flask_SQLAlchemy
Werkzeug
requests
openai-whisper
python-dotenv

Then install:

pip install -r requirements.txt

⚠️ Whisper may require ffmpeg.
Example (macOS): brew install ffmpeg
Linux: use your package manager (apt, dnf, etc).

---

Install & Run Ollama

1. Download Ollama from its official site and install it.
2. Start Ollama (it will listen on http://localhost:11434).
3. Pull the LLaMA 2 model:

ollama pull llama2

4. (Optional) Test it:

ollama run llama2

The app expects:

"model": "llama2:latest"

and uses POST http://localhost:11434/api/generate.

---

Configure Environment Variables

Create a .env file in the repo root:

UNIQUE_KEY=some_random_secret_key

EMAIL_ADDRESS=[email protected]
EMAIL_PASSWORD=your_app_specific_password

• UNIQUE_KEY → Flask SECRET_KEY
• EMAIL_ADDRESS / EMAIL_PASSWORD → used for SMTP to send notifications. For Gmail, enable 2FA and generate an App Password (don’t use your real login password).

---

Run the App

python app.py

On first run, the app will:

• create multibrain.db
• add dummy employees
• add some dummy tasks for workload visualisation

Then open:

http://127.0.0.1:5000/

---

Typical Flow

1. Create events in the Events page → events stored in DB and your whole team gets notified.

2. Upload a recording or transcript in the Auto Assign page → Whisper transcribes (if audio), Ollama extracts JSON tasks.

3. Review AI tasks → check descriptions, deadlines, suggested assignees.

4. Assign tasks to employees → app links tasks to Employee records and emails each person their list.

5. Track progress

   • Dashboard shows workload.
   • Tasks page lets you filter pending/complete.
   • Change statuses as work is done.

---

Future Ideas

• Speaker diarization (who said what).
• Integration with Google Calendar / Outlook.
• Per-project boards instead of only per-employee view.
• Role-based access (manager vs employee dashboards).

---

Credits

Built with way too much coffee, Python, and curiosity.
AI stack: Whisper + Ollama (LLaMA 2)
Backend: Flask + SQLite + SQLAlchemy