You can try out this Celebrity Matcher app at: https://mongodb-celeb-search.com/
This is a small FARM (FastAPI, React, MongoDB) project that illustrates integration with vector search, and AWS's Bedrock model hosting service.
You can run the development version with Docker Compose.
First, you must configure the application, using a .env file,
in the root directory of the project. (The one that contains compose.yml).
You need to set the following variables:
| Variable | Description |
|---|---|
| MONGODB_URI | The connection string (URI) for a MongoDB Atlas cluster. The connection string must contain the name of the database you want to connect to! |
| AWS_ACCESS_KEY | An AWS access key with permission to use the Bedrock service. |
| AWS_SECRET_KEY | The secret key associated with AWS_ACCESS_KEY. |
| DEBUG | Set this to "true" to enable stack traces and reload. Do not enable in production. |
Once you've set those, you can spin up the application with:
docker compose upThis will spin up three containers:
| Service | Port | Description |
|---|---|---|
| Frontend | 8002 | A React application, created with create-react-app, and running in development mode. |
| Backend | 8003 | A FastAPI application that routes requests from frontend to MongoDB & Amazon Bedrock. |
| Nginx | 8001 | A simple reverse-proxy that sits in front of frontend & backend, allowing them to be served from the same host & port. |
Both frontend and backend are configured to load any code changes made on-disk.
This is a small, stupidly-fast-moving project. Currently there are no tests (although there should be!), and no scripts for deployment. We'll get there!
All endpoints in the backend must be beneath the /api/ path prefix.
That's how Nginx knows to route to the backend, and not the frontend.