Skip to content

Latest commit

 

History

25 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

5-day-ai-agent

I am participating in a capstone project of 5 day ai agent.

🏥 CareFlow — AI Healthcare Appointment Assistant (Agents for Good)

CareFlow is a multi-agent AI assistant that helps patients find, schedule, reschedule, and cancel medical appointments using natural language conversations. It reduces scheduling friction, avoids calendar conflicts, and remembers patient preferences — supporting both patients and clinic staff.

Submission for Kaggle – Agents Intensive Capstone Project Track: 🩺 Agents for Good

Problem

In many clinics — especially in low-resource settings — appointments are still managed through:

  • Phone calls
  • Notebooks or spreadsheets
  • Manual calendar checking

This creates:

Long wait times

Missed or double bookings

Frustration for both patients and staff

Scheduling logistics consume time that could instead be used for care.

💡 Solution

CareFlow automates medical appointment management using AI agents. Patients can express their needs naturally, e.g.:

“I need a cardiology appointment next week after 5 pm.”

CareFlow will:

Understand the intent

Collect necessary information using questions only if needed

Check doctor/clinic availability

Suggest suitable appointment slots

Finalize booking, reschedule, or cancellation

Send a clear confirmation summary

The system remembers doctor preferences, time preferences, and past appointments for smoother repeat interactions.

🧠 Key Concepts from the Course

CareFlow demonstrates the following course requirements:

Concept Implementation Multi-Agent System Orchestrator, Triage, Scheduler, Notification Tools Custom tools for clinic scheduling & calendar conflict checks Memory Session memory + long-term patient preference memory Context Engineering Conversation compaction to prevent token overload Observability Structured logs for agent decisions & tool calls Evaluation Script-based test scenarios validating booking flows 🏗 Architecture & Flow User ↓ Orchestrator Agent ↓ Triage Agent → extracts structured request from natural language ↓ Scheduler Agent → calls tools (clinic scheduler + calendar) ↓ Orchestrator Agent → confirms selected slot ↓ Notification Agent → sends friendly summary/confirmation ↓ User

Agents share session memory and patient profile memory (doctor & time preferences).

📂 Project Structure agents/ orchestrator_agent.py triage_agent.py scheduler_agent.py notification_agent.py

tools/ clinic_scheduler_tool.py calendar_tool.py

memory/ session_service.py patient_memory.py

evaluation/ test_scenarios.py

main.py requirements.txt README.md

🛠 Installation & Setup 1️⃣ Clone the repository git clone https://github.com/Kulsoom0201/5-day-ai-agent.git cd 5-day-ai-agent

2️⃣ Create a virtual environment (optional but recommended) python -m venv .venv source .venv/bin/activate # Windows: .venv\Scripts\activate

3️⃣ Install dependencies pip install -r requirements.txt

⚠️ API keys (if any) must be added via environment variables — never committed to the repository.

💬 Usage Example Start the agent: python main.py

Then interact with messages like:

I need a dermatology appointment next Tuesday morning Reschedule my appointment with Dr. Mehta to Friday evening Cancel my appointment tomorrow at 3pm

The agent will:

Ask follow-up questions if needed

Check availability with scheduling tools

Suggest slots

Book/reschedule/cancel after confirmation

And return a summary such as:

✔ Your appointment with Dr. Mehta (Cardiology) is confirmed for Wednesday 6:30 PM at Sunrise Clinic. A reminder will be sent 24 hours prior.

🧪 Evaluation / Testing

To run scripted test cases:

python -m evaluation.test_scenarios

This verifies:

Booking flow

Rescheduling flow

Cancellation flow

Logs show each agent call + each tool call step-by-step.

⚠️ Limitations & Future Work Current Limitations Planned Improvements Uses mock scheduling APIs Integrate real clinic APIs with authentication No mobile/voice interface Add WhatsApp / SMS / voice support English-only Add multilingual support Console-only interaction Deploy as web app / chatbot front-end No caregiver mode Add “book for family member” feature 🛡 License & Disclaimer

License: MIT (or any preferred open-source license)

Disclaimer: CareFlow does not provide medical advice and does not replace professional healthcare judgment. It supports appointment logistics only.

🙌 Acknowledgements

Built as part of the Kaggle + Google Agents Intensive Program. Inspired by the need to make healthcare more accessible and reduce administrative burden.

About

I am participating in a capstone project of 5 day ai agent.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages