Hi everyone,
I’ve been working on InferenceFlow, a backend project for submitting telescope-observation features and processing prediction requests asynchronously. I wanted to get practice connecting an API, database, message broker, and Python worker in one system.
https://github.com/mysaas2024/inferenceflow-gamma-hadron
The current flow is:
An Express API validates a request containing ten numeric features.
It stores the request in PostgreSQL using Prisma.
It publishes a job to RabbitMQ.
A Python/FastAPI worker consumes the job and publishes a result message.
The project also includes a Docker Compose observability setup with Prometheus, Grafana, Loki, and an OpenTelemetry Collector.
Tech stack: Node.js, Express, Python, FastAPI, PostgreSQL, Prisma, RabbitMQ, Redis, Prometheus, Grafana, Loki, OpenTelemetry, Docker Compose.
A couple of important caveats: the classifier is currently a simple demo heuristic—not a trained or scientifically validated model.…