DiSH Meeting Room Booking Agent

An AI-powered assistant that seamlessly manages your meeting room bookings and calendar in one conversation. Simply ask it to book a room, check availability, or schedule meetings. It handles the coordination between DiSH room reservations and Google Calendar automatically.

The meeting room booking website for my office (DiSH) was pretty terrible and made booking recurring meetings difficult and time consuming. I built this agent to save on all the manual time spent block booking meeting rooms. It has so far saved me so much time and I use it every week!

Key Features:

• 🗓️ Natural language scheduling — "Book me a room tomorrow at 2pm for an hour"
• 🔍 Smart availability checking across rooms and calendars
• 🔄 Automatic coordination between room bookings and calendar events
• 💬 Conversational interface with streaming responses

Built with Pydantic AI and the Model Context Protocol (MCP) for extensible tool integration.

Screenshots

Prerequisites

• Python 3.12+ with uv installed
• Bun for the frontend
• Docker and Docker Compose
• Access to the DiSH MCP repository (set DISH_MCP_PATH to its root; it must have a .venv with fastmcp installed via uv sync)

Quick Start

# 1. Clone and enter the repository
cd dish-booking-agent

# 2. Copy and configure environment variables
cp example.env .env
# Edit .env with your values (API keys, passwords, etc.)

# 3. Install dependencies
cd backend && uv sync && cd ..
cd frontend && bun install && cd ..

# 4. Start all services
make dev

This will start:

• Keycloak (authentication): http://localhost:8080
• Backend API (FastAPI): http://localhost:8000
• Frontend (Next.js): http://localhost:3000

Development Commands

Command	Description
make dev	Start all services (infrastructure + backend + frontend)
make infra	Start only Docker services (databases + Keycloak)
make backend	Start only the FastAPI backend
make frontend	Start only the Next.js frontend
make stop	Stop all running services
make logs	Tail Docker container logs
make clean	Stop services and remove Docker volumes
make help	Show available commands

Configuration

Environment Variables

Copy example.env to .env and configure:

cp example.env .env

Required variables:

Variable	Description
KEYCLOAK_POSTGRES_PASSWORD	Password for Keycloak's database
KEYCLOAK_BOOTSTRAP_ADMIN_PASSWORD	Keycloak admin console password
POSTGRES_PASSWORD	Password for the user database
OPENAI_API_KEY	Your OpenAI API key
DISH_MCP_PATH	Absolute path to the DiSH MCP repository
DISH_COOKIE	Your DiSH authentication cookie
TEAM_ID	Your DiSH team ID
MEMBER_ID	Your DiSH member ID
GOOGLE_OAUTH_CREDENTIALS	Path to Google OAuth JSON for Calendar MCP

Backend Configuration

The backend also reads from backend/.env for additional settings. See backend/example.env for reference.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         Frontend (Next.js)                      │
│                        http://localhost:3000                    │
└─────────────────────────────────┬───────────────────────────────┘
                                  │
                                  ▼
┌─────────────────────────────────────────────────────────────────┐
│                      Backend API (FastAPI)                      │
│                        http://localhost:8000                    │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────────┐  │
│  │ Pydantic AI │  │ DiSH MCP    │  │ Google Calendar MCP     │  │
│  │   Agent     │──│  (rooms)    │  │  (calendar events)      │  │
│  └─────────────┘  └─────────────┘  └─────────────────────────┘  │
└─────────────────────────────────┬───────────────────────────────┘
                                  │
          ┌───────────────────────┼───────────────────────┐
          ▼                       ▼                       ▼
┌─────────────────┐   ┌─────────────────┐   ┌─────────────────┐
│    Keycloak     │   │   Keycloak DB   │   │    User DB      │
│  (port 8080)    │   │   (PostgreSQL)  │   │  (port 5433)    │
└─────────────────┘   └─────────────────┘   └─────────────────┘

API Endpoints

• GET /health - Health check
• GET /send-message?message=<msg>&session=<id> - Send a message to the AI agent

The session parameter maintains separate conversational histories.

Running Individual Components

CLI Agent (without API)

cd backend
uv run src/agent.py

Backend Only

make infra  # Start databases first
make backend

Frontend Only

make frontend

Example Prompts

• "List my calendar events tomorrow morning."
• "Book a room from 14:00–15:00 tomorrow and create a calendar invite for it."
• "Reschedule the 13:00 sync to 15:00 and keep the same attendees."
• "What rooms are available next Monday afternoon?"

Troubleshooting

Services won't start

1. Ensure Docker is running
2. Check that ports 3000, 5433, 8000, and 8080 are not in use
3. Verify your .env file has all required variables set

Database connection issues

# Check if databases are healthy
docker compose ps

# View database logs
docker compose logs user-db
docker compose logs keycloak-db

Reset everything

make clean  # Stops services and removes volumes
make dev    # Fresh start

Testing

Unit Tests

The backend includes unit tests located in backend/tests/. Run them with:

cd backend
uv run pytest

Note: The unit tests were generated using Claude Opus 4.5 in the absence of time to write them manually. Whilst they provide some coverage and are better than no tests at all, they should not be fully relied upon and may require review or refinement for production use.

Manual Test Flow

1. Start all services with make dev
2. Navigate to http://localhost:3000
3. Authenticate via Keycloak
4. Ask for upcoming events on a given day
5. Book a DiSH room and request a matching Google Calendar invite
6. Reschedule or cancel events to verify the integration