Add FastAPI scanner service with 1Hz web dashboard

[devin-ai-integration]

Summary

The repo was previously empty. This PR introduces a small FastAPI service that polls a public crypto exchange once per second, detects market "situations", and serves a self-contained web dashboard that refreshes the data every second (per the original request).

• app/scanner.py — MarketScanner background task. Each tick it fetches the Binance 24h ticker, keeps a rolling in-memory history (HISTORY_LEN=120 samples) for the top TOP_N_BY_VOLUME=30 USDT pairs (selected on first tick), and produces a ScannerState snapshot.
• Situation detector emits four kinds:
  • price_spike_up_1m / price_spike_down_1m — ≥1% move in the last ~60s.
  • volume_surge — last-minute quote volume ≥2× the 10-minute baseline (estimated from the cumulative 24h quoteVolume delta).
  • new_24h_high / new_24h_low — last price touches the daily extreme.
• app/main.py — FastAPI app wiring the scanner into a lifespan, plus three routes: / (HTML), /api/state (JSON snapshot), /healthz.
• app/static/{index.html,styles.css,app.js} — vanilla HTML/JS dashboard. app.js polls /api/state in a loop using the server-provided poll_interval_seconds (defaults to 1.0s), flashes price cells green/red on change, and renders a "situations" panel and a tickers table.
• pyproject.toml / poetry.lock / README.md / .gitignore — project bootstrap (Poetry, Python ≥3.10).

Data source: I'm using https://data-api.binance.vision/api/v3/ticker/24hr (Binance's read-only data mirror) instead of api.binance.com because the main API returned HTTP 451 (geo-blocked) from the dev VM. The mirror exposes the same response shape.

Stack choice rationale: the user picked "На ваше усмотрение" — I went with FastAPI + plain HTML/JS so everything runs from a single uvicorn process with no build step.

Review & Testing Checklist for Human

• Confirm the stack/UX is what you wanted. I picked FastAPI + vanilla HTML/JS. If you wanted React, a fancier charting UI, WebSocket push instead of HTTP polling, a different exchange (Bybit/OKX/KuCoin/Coinbase all returned 200 from the VM as alternatives), or different/more situation types, this is the moment to redirect.
• Sanity-check the data source. data-api.binance.vision is Binance's read-only mirror; if you'll deploy this somewhere where api.binance.com is reachable, you may want to point BINANCE_TICKER_URL (or the MarketScanner(ticker_url=...) arg) back at the main API. Also confirm the once-per-second poll of the full 24h ticker payload (~600 KB, ~3000 symbols) is acceptable for your rate-limit budget — fetching only the tracked symbols would be cheaper.
• Review the situation heuristics. A few are approximations worth a second look:
  • volume_surge uses deltas of the cumulative 24h quoteVolume as a per-minute proxy; that's only a rough estimate and the baseline window only fills once history has ≥10 minutes of samples.
  • new_24h_high / new_24h_low fire whenever price >= high_24h / price <= low_24h. Because the latest price is itself included in the exchange's 24h extreme, this can trigger every second when the market is at the rolling boundary. You may want a debounce or a "strictly new" check.
  • Spike detection uses the oldest history sample within the window, so before 60s of history accumulates the "1m" change is computed against fewer than 60s of data.
• Run it end-to-end. poetry install && poetry run uvicorn app.main:app --port 8000, then open http://localhost:8000. Check that tickers populate, prices flash on change, the "Last update" timestamp ticks once per second, and curl http://localhost:8000/api/state returns a JSON snapshot with tickers and situations.

Notes

• I smoke-tested locally: /healthz returns 200, /api/state returns 30 tickers with error: null after the mirror swap, and one new_24h_high situation fired immediately on the boundary check (see the second bullet above).
• Existing .github/workflows/blank.yml is unchanged — CI is still just the scaffold "Hello world" job, so a green CI here doesn't validate the new code.
• No automated tests were added in this PR; happy to follow up with pytest coverage of SymbolHistory.price_change_pct, recent_volume_delta, and _detect_situations if you want.

Link to Devin session: https://app.devin.ai/sessions/2eb7aa7344844b0a841c827cba7ab0fa
Requested by: @evgetos

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring