Meet Michelle, the appointment manager assistant

This repository contains a working prototype of Michelle, an appointment management agent for Shine & Style, a local hair salon. Built using LangGraph, Streamlit, and OpenAI models, this project showcases a conversational assistant capable of handling salon bookings.

Please note that this a work in progress (its Version0), and serves as an early-stage prototype. The goal is to eventually evolve it into a production-ready system. The current system can currently book, reschedule and cancel appointments.

In progress

• Appointment rescheduling or cancellation features.
• Decouple LangGraph agent (backend) from Streamlit (frontend); expose the backend as a containerized FastAPI endpoint.
• Answer FAQ questions about the salon timings, etc
• Allow to book with specific stylists at the salon (as is the case with certain well-known salons).
• Make SQL backend (checking availability) more efficient to handle lots of appointment requests.
• Make this a voice agent eventually for practical applications.

Simple demo

Simple System Design

Get started

1. Setting Up Backend Service

1. Install Requirements: Make sure you have Python 3.10+ installed. Then run:

   pip install -r backend_service_requirements.txt

2. Setting Up Environment Variables: Set up your OPENAI_API_KEY environment variable.
3. Run the service: Run the following command to start the backend service:

   python src/run_service.py

   Running the service would create a new sqlite database file bookings.sqlite in the current directory if it doesn't exist.

2. Setting up Front-End (Streamlit)

1. Install Requirements: Make sure you have Python 3.10+ installed. Then run:

   pip install -r frontend_requirements.txt

2. Run the service: Run the following command to start the Streamlit front-end:

   streamlit run src/streamlit_app.py

The app will open in your browser at http://localhost:8501 by default.

Disclaimer

This is a work-in-progress project and not suitable for production use yet. Features may be incomplete, and the codebase is subject to frequent changes.