Cognitive Companion is an Agentic AI application designed to assist early-stage Alzheimer's patients with memory support, routine reminders, and emotionally comforting conversations. It can also notify loved ones via email when needed.
- Onboards the user and stores their routine/memory profile
- Uses Google Gemini (via Gemini API) to generate memory-aware, emotionally supportive replies
- Sets local reminders for daily routines or medications
- Sends email alerts to loved ones
- Saves user memory locally for context-aware interaction
- Secured using environment variables
| File | Purpose |
|---|---|
planner.py |
(To be implemented) Breaks down user goals into sub-tasks |
executor.py |
Calls Gemini API, generates responses using prompt + memory |
memory.py |
Saves and retrieves user memory (e.g., routines, meds, contacts) |
tools.py |
Contains helper functions like set_reminder() and send_notification() |
notifier.py |
Sends email to emergency contacts using Gmail SMTP |
python -m venv .venv
.venv\Scripts\activate # On Windows
source .venv/bin/activate # On macOS/Linuxpip install -r requirements.txt- Get a free Gemini API key via Google AI Studio
- Create a
.envfile in the project root and add:
# Gemini API Key (from Google AI Studio)
GEMINI_API_KEY=your_gemini_api_key_here
# Gmail SMTP credentials (App Password required)
EMAIL_SENDER=your_email@gmail.com
EMAIL_PASSWORD=your_gmail_app_password-
Go to: https://myaccount.google.com/security
→ Under "Signing in to Google", enable 2-Step Verification -
After enabling 2FA, go to: https://myaccount.google.com/apppasswords
→ Choose Mail as the app and Other (Custom name) as the device (e.g.,Promptonauts)
→ Click Generate -
Copy the 16-character App Password (e.g.,
abcd efgh ijkl mnop)
→ Paste it into.envlike this (no spaces):
EMAIL_PASSWORD=abcdefghijklmnop📌 Never commit .env to Git. It's already in .gitignore.
python src/main.pyWe have recorded a 5-minute demo video showcasing the full working of the Cognitive Companion Agent — from onboarding, planning, tool usage, to output generation and edge case handling.
📎 Watch the video demo: Demo Video Link
This is a hackathon prototype and not intended for medical use without proper safety, testing, and regulatory approval.