# Auditor response — 2026-04-25 — dashboard wiring? deliberately not, here's a peek tool instead

## Answer

No, the watcher is not wired into the dashboard. That was deliberate scope — the original prompt listed "Dashboard UI for results" as out of scope to keep delivery fast.

## Trade-off

Wiring it in = several hours of dev (new HTTP endpoint that reads JSONL or streams SSE, plus a React panel). That delays the viability answer for ~zero analytical benefit — by the time the panel is useful you'd be 3 hours into a 6-hour collection the auditor will summarize anyway.

## Lightweight live peek (zero dev)

To eyeball progress, paste this into a terminal — it tails the JSONL and prints a fresh cross-venue spread snapshot every 5 seconds:

```bash
sudo -n -u arb bash -c 'cd /srv/projects/arb/tools/spread-watcher && python3 -c "
import json, time, collections, os
PATH = \"data/watcher-2026-04-25.jsonl\"
while True:
    rows = collections.defaultdict(list)
    with open(PATH) as f:
        f.seek(0, 2); end = f.tell(); f.seek(max(0, end - 65536))
        f.readline()
        for line in f:
            try: o = json.loads(line)
            except: continue
            rows[(o[\"exchange\"], o[\"symbol_canonical\"])].append(o)
    latest = {k: v[-1] for k,v in rows.items() if v}
    for sym in [\"BTC\",\"ETH\"]:
        present = [(ex, latest[(ex,sym)]) for ex in [\"kraken\",\"coinbase\",\"binance\",\"okx\"] if (ex,sym) in latest]
        if len(present) < 2: continue
        bids = [(ex, float(o[\"bid\"]), o[\"quote_ccy\"]) for ex,o in present]
        asks = [(ex, float(o[\"ask\"]), o[\"quote_ccy\"]) for ex,o in present]
        best_bid = max(bids, key=lambda x: x[1])
        best_ask = min(asks, key=lambda x: x[1])
        mid = (best_bid[1] + best_ask[1]) / 2
        bps = (best_bid[1] - best_ask[1]) / mid * 10000
        print(f\"{sym}  buy@{best_ask[0]:8s} {best_ask[1]:>10.2f}({best_ask[2]})  sell@{best_bid[0]:8s} {best_bid[1]:>10.2f}({best_bid[2]})  raw_spread={bps:+.2f}bps\")
    print(\"--\", time.strftime(\"%H:%M:%S\"))
    time.sleep(5)
"'
```

`Ctrl-C` to exit. **Peek tool only** — `raw_spread` ignores USDT/USD basis, so any USDT↔USD direction will look inflated. The real analysis at the end applies basis adjustment.

## Auditor's call

Don't wire to dashboard yet. Use the one-liner for occasional peeking. Wait for the 6-hour analysis — that's where the decision gets made.

If post-analysis the data flags an interesting venue-pair worth watching live ongoing, we can scope a focused dashboard panel then. Doing it before we know what to display is premature.