Streamlit Uber NYC Pickup Demo but it's in Sveltekit + FastAPI

Streamlit inspiration: https://demo-uber-nyc-pickups.streamlit.app/

Prerequisites

• just for commands
• Node.js for Javascript frontend
• uv for Python backend

Run (Windows)

One-time setup:

just install

Run both backend + frontend in parallel:

just dev

Or run them individually (in separate terminals):

just dev-backend   # FastAPI on http://localhost:8000
just dev-frontend  # SvelteKit on http://localhost:5173

Fullstack build (SPA served by FastAPI)

The SvelteKit app is configured as a SPA (@sveltejs/adapter-static with fallback: 'index.html'). FastAPI serves the built assets at / and exposes the API under /api/*, so there's no Vite proxy in production mode:

just build   # outputs frontend/build/
just run     # FastAPI in production mode on http://localhost:8000

The static mount in backend/app/main.py resolves frontend/build/ relative to the source tree and is skipped (with a warning) if the build is missing — so just dev keeps working without a build.

Regenerating frontend types from the backend

The frontend imports request/response types from frontend/src/lib/client/ — generated from FastAPI's OpenAPI schema by @hey-api/openapi-ts. Run this whenever you change a Pydantic model in backend/app/models.py or a route signature in backend/app/api/routes/:

just gen-client

Under the hood this dumps app.main.app.openapi() to frontend/openapi.json (gitignored), then runs npm run generate-client to emit types.gen.ts + index.ts. The generated client is committed to the repo, so npm install alone is enough on a fresh clone.

For a model to appear in the generated types, the route must expose it to OpenAPI — either via response_model=... on the decorator or a typed return annotation. Import the resulting type with import type { HistogramBin } from '$lib/client'.