Name the node behind each firmware row, not just the hive it sits in

[MacNite]

Why

Every row under Firmware → HiveInside nodes / HiveTraffic counters is headed by the hive name (shire-01). That says where a node sits and nothing about which unit it is, so a beekeeper running several identical nodes had no way to tell them apart from the dashboard — least of all when two rows both read relay failed and the question is which device to walk out to.

What changed

Each row's label carries a ? that names the node itself: the local name it advertises, and the BLE address it is paired on.

• Named node → Advertises itself as "HiveInside-8A3F". BLE address C4:9D:ED:8A:3F:11.
• No name → the address plus a note that HiveInside 0.5.0 onwards advertises HiveInside-XXXX, so the "why is this blank" question answers itself rather than reading as a fault.
• Neither → no ? at all. A tooltip that opens onto nothing is worse than no tooltip.

Firmware

hives[].ble.device_name / .mac. The name is captured from the scan response alongside the identity record, on the same terms (active scan, may arrive on a report carrying no measurement) and for every beacon type — a beekeeper telling two HolyIots apart has the same problem.

HiveTraffic

Same plumbing, end to end. bee_counter_wire.h reads an optional "name" from the counter's measurement JSON → hives[].bee_counter.device_name. No counter firmware sends one yet — a document without it parses exactly as before — so until then the row falls back to hives[].bee_counter.mac, which HiveHub writes before the !present return and therefore reports even for a counter that never answered. That is precisely the "relay failed" row where the identity matters most.

The counter's paired MAC is passed to writeSnapshotToHive rather than carried on beecnt::Snapshot: it is configuration, the call site already has it, and that struct is a MAX_HIVES-sized stack array (carrying it there cost ~324 bytes of stack for something already in hand).

Server

Two schema declarations, no migration. Both fields ride in hive_readings.raw_json rather than earning a column, like the HiveTraffic image version — they change only when a node is re-paired and nothing charts them. Declared on HiveBleIn / HiveBeeCounterIn so they are typed at the boundary rather than left to extra="allow".

The ble pair sits after the !present return, unlike the counter's: a hive with no beacon heard emits no ble object at all, and creating one just to carry an address would change what the presence of that object means for every reader of hive_readings. The dashboard reads the newest reading that has the field, so a node that missed one scan window keeps its identity anyway.

Files

Area	Files
Firmware — BLE	ble_sensor.{h,cpp}
Firmware — HiveTraffic	bee_counter_wire.h, bee_counter_client.{h,cpp}, sensors.cpp
Server	schemas.py
Dashboard	server/dashboard/assets/views.js, website/dashboard-demo/assets/views.js
Docs	hiveinside-ble-sensor.md, hivetraffic-bee-counter.md
Tests	test_bee_counter_wire.cpp

Testing

Passing locally:

• Wire decoder — new name cases: read when present, empty when absent, truncated not overrun, unterminated refused. Plus the existing fw:2–fw:5 contract suite.
• Firmware host tests — 111 / 35 / 41 checks, 0 failures.
• Server suite — all standalone scripts and the pytest set (39 passed).
• Server round trip — identity fields verified through _hive_reading_row_params → raw_json → _hive_reading_row_to_dict.
• Dashboard — rendered in headless Chromium against fixture data with nested hives[]. Confirmed all four cases (named node, unnamed node, named counter, identity-less counter → no ?), no console errors.

Not built: the ESP32 firmware. PlatformIO's registry (*.platformio.org) is blocked by this environment's network policy, so pio run cannot fetch scons or the NimBLE / ArduinoJson deps. The changed C++ is unbuilt here — CI is the first real compile.

Not included

• No firmware version bump — this repo bumps FIRMWARE_VERSION in separate commits.
• The Status card's In-hive sensors list could take the same tooltip; left alone as out of scope.
• The dashboard demo's sample data emits no nested hives[], so the relay sections do not appear there. views.js is kept in sync per dashboard-demo/README.md, but wiring up demo fixtures would be a separate change.

Companion

Needs MacNite/HiveInside#70 for HiveInside nodes to advertise a name at all. This side is useful without it — the address fallback works today — and the tooltip tells you what to relay to get a name.

🤖 Generated with Claude Code

https://claude.ai/code/session_01T9ZUAmc45AB5Mf7ETM82Uv

---

Generated by Claude Code