orion

Orion turns NER's raw car telemetry into clean, analysis-ready data and shows it on a web dashboard.

The car's sensor readings are published to Penelope, NER's telemetry database. That data is raw: sensor names are inconsistent, timestamps from some sources are offset, and nothing is derived yet. Orion reads Penelope (it never writes to it), runs each run through a processing pipeline, stores the results in its own database (OrionDB), and serves them to a React dashboard.

What it does

• Browse runs. See every run recorded in Penelope and which ones Orion has already processed.
• Process a run. Trigger the pipeline for a run from the dashboard and watch its status. Jobs are queued and handled in the background, and re-running a run replaces its old output.
• Explore sensor data. View a run's sensor readings under consistent, clean sensor names and units.

The pipeline

Each run goes through a series of steps. Each step reads the previous step's output and saves its own, so any step can be re-run on its own.

Step	What it produces	Status
Clean	Raw readings mapped to clean sensor names and units, GPS time corrected	Done
State estimation	Estimated car state over the run	Not started
Laps	Lap detection and per-lap data	Not started

How it fits together

browser ──► dashboard ──► API ──► OrionDB      (Orion's own data)
                           │  └─► Penelope     (raw telemetry, read-only)
                           ▼
                     pipeline API ──► job queue ──► pipeline worker
                                                     reads Penelope,
                                                     writes OrionDB

• Dashboard (frontend/): React app for browsing runs, triggering the pipeline, and viewing data.
• API (services/backend/): serves runs, sensors, and pipeline status to the dashboard.
• Pipeline (services/pipeline/): an API that queues processing jobs, and a worker that runs them.
• Shared core (packages/orion_core/): settings, database models, and queries used by both services.

Getting started

You need Docker and a connection string for Penelope (the real database over the NER VPN, or penelope2, a copy hosted on Neon).

cp .env.example .env        # set PENELOPE_DATABASE_URL
docker compose up --build
docker compose exec orion-backend python -m orion_core.db.scripts.seed_sensors   # once

Then open the dashboard at http://localhost:5173.

Documentation

• DevREADME.md: architecture, services, configuration, local development, migrations, and CI.
• docs/ARCHITECTURE.md: design notes.