Six silent failures found by installing this on a live system

Migrated the reference site off the YAML packages and onto the add-on. Every
bug below presented identically: the add-on starts, logs "started", serves its
UI, and cannot do its job.

- run.sh needs #!/usr/bin/with-contenv sh. s6-overlay sanitises the environment
  for services, so a plain shebang means SUPERVISOR_TOKEN is absent and every
  Core API call is 401 - while homeassistant_api: true makes permissions look
  granted. Startup now prints the token length and probes the API.
- Supervisor keys the image by config.yaml `version`, so rebuilding without a
  bump reuses the old image. Two fixes appeared not to work because of it.
- Alpine is musl and has no aiohttp wheel on PyPI; deps now come from apk so
  nothing compiles on a client's Pi.
- Alpine ships paho-mqtt 1.x, which has no CallbackAPIVersion. That raised at
  construction and took the control loop down with it - so MQTT setup is now
  wrapped too. Observability must never be able to stop the controller.
- MQTT discovery is published from on_connect: paho silently drops QoS-0
  publishes issued before the CONNACK, so the previous code announced nothing
  while logging "MQTT connected".
- Repeated failures now log once a minute. Six warnings a second rolled the log
  buffer and destroyed the startup diagnostics needed to find the 401.
- auto_start could never fire, because the store's defaults always supplied
  auto: False for the fallback to find.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016NckgXecasQb2eSsPYNSW6

goodwe_controller/app/main.py

@@ -384,13 +384,45 @@ async def amain() -> None:
     async with aiohttp.ClientSession() as session:
         hass = HomeAssistant(session)
 
-        broker = await hass.mqtt_service()
-        pub = MqttPublisher(
-            broker.get("host") if broker else None,
-            broker.get("port", 1883) if broker else 1883,
-            broker.get("username") if broker else None,
-            broker.get("password") if broker else None,
-        )
+        # Startup self-check: prove we can actually reach the Core API before
+        # anything tries to control an inverter with it. A 401 here is a
+        # permissions/token problem, not a configuration mistake, and saying so
+        # explicitly saves an installer from re-checking entity ids for an hour.
+        import os as _os
+        _tok = _os.environ.get("SUPERVISOR_TOKEN", "")
+        _LOG.info("supervisor token: %s (%d chars); env has: %s",
+                  "present" if _tok else "MISSING", len(_tok),
+                  ",".join(sorted(k for k in _os.environ if "TOKEN" in k.upper())) or "none")
+        try:
+            async with session.get("http://supervisor/core/api/",
+                                   headers={"Authorization": f"Bearer {_tok}"},
+                                   timeout=10) as _r:
+                _LOG.info("core api probe: HTTP %s %s", _r.status, (await _r.text())[:80])
+        except Exception as _e:  # noqa: BLE001
+            _LOG.error("core api probe failed: %s", _e)
+
+        # ⚠️ Status publishing must NEVER be able to stop the controller. A
+        # broken broker, a missing library, an API change in paho - all of it is
+        # observability, and the battery does not care. Caught broadly and on
+        # purpose: this crashed the add-on once already (paho 1.x vs 2.x) and
+        # took the control loop down with it.
+        try:
+            broker = await hass.mqtt_service()
+            pub = MqttPublisher(
+                broker.get("host") if broker else None,
+                broker.get("port", 1883) if broker else 1883,
+                broker.get("username") if broker else None,
+                broker.get("password") if broker else None,
+            )
+        except Exception as err:  # noqa: BLE001
+            _LOG.warning("MQTT unavailable (%s) - continuing without status entities", err)
+
+            class _NoMqtt:
+                enabled = False
+                def publish(self, *a, **k): pass
+                def close(self): pass
+
+            pub = _NoMqtt()
 
         controller = Controller(opts, hass, store, pub)
         runner = await web.start(controller, port=8099)