Snacker Tracker Interface

A read-only web interface showing which snacks are taken from the office pantry. A Python pipeline (a stepdag graph) turns raw scanner events into static JSON; a React app renders it.

• project.md: the original brief
• PRD.md: requirements and the static JSON contract
• GitHub issues: specs and tickets, labelled by feature

Layout

Path	What it is
backend/	The Python + Polars graph that builds the site's data, managed with uv
stepdag/	The framework that runs it on Step Functions, one Lambda per asset
frontend/	Vite + React + TypeScript app
data/	Raw scanner events and OpenFoodFacts, copied locally (git-ignored)
public-data/	Generated JSON, written by the graph (git-ignored)

Running locally

Requires uv and Node 22 (see frontend/.nvmrc).

1. Build the data:

   cd backend
   scripts/build-local-data.sh

   This runs the graph over the raw events in data/, and writes public-data/data/…. Rerun it whenever the raw data changes: it reads only raw files newer than the last run's. The first run copies the 8 GB OpenFoodFacts file into data/files/. See backend/README.md.

2. Run the frontend:

   cd frontend
   npm install
   npm run dev

   Vite serves public-data/ as its public directory, so the app loads the built JSON from /data/…. npm run build produces a static site in frontend/dist/, with the data copied in.

Checks

cd backend && uv run pytest && uv run mypy      # backend tests and typecheck
cd frontend && npx tsc -b && npm run lint       # frontend typecheck and lint