Make service health-check self-diagnosing

When a service group's cog stays disabled the log now says WHY:
_service_health returns the actual connection error (ConnectionError
= refused/down, ConnectTimeout = firewall/slow, gaierror = DNS/wrong
host) instead of a bare 'unreachable', and on_ready logs the resolved
FILE/LIBRARIAN/RADIO addresses so an env-var that never reached the
process (address falls back to the 192.168.1.15:5000 default) is
obvious at a glance.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

bot.py

@@ -120,13 +120,15 @@ SERVICE_EXTENSION_GROUPS = {
 SERVICE_RECHECK_SECONDS = 300
 
 
-def _service_alive(url: str) -> bool:
-    """True when the service answers HTTP at all (any status code counts)."""
+def _service_health(url: str):
+    """Return None when the service answers HTTP at all (any status counts),
+    otherwise the connection error explaining WHY it's unreachable (refused vs
+    timeout vs DNS - the difference points straight at the cause)."""
     try:
         requests.get(url, timeout=3)
-        return True
-    except requests.exceptions.RequestException:
-        return False
+        return None
+    except requests.exceptions.RequestException as exc:
+        return exc
 
 
 async def _load_extension_safe(name: str) -> bool:
@@ -152,12 +154,14 @@ async def _load_service_groups() -> bool:
         missing = [e for e in group["extensions"] if e not in client.extensions]
         if not missing:
             continue
-        alive = await asyncio.to_thread(_service_alive, group["health_url"])
-        if not alive:
+        err = await asyncio.to_thread(_service_health, group["health_url"])
+        if err is not None:
             logger.warning(
-                "Service '%s' unreachable (%s) - cogs stay disabled: %s",
+                "Service '%s' unreachable (%s) [%s: %s] - cogs stay disabled: %s",
                 service,
                 group["health_url"],
+                type(err).__name__,
+                err,
                 ", ".join(missing),
             )
             continue
@@ -204,6 +208,16 @@ async def on_ready():
     for extension in CORE_EXTENSIONS:
         await _load_extension_safe(extension)
 
+    # Log the ACTUALLY-resolved service addresses. When one shows the built-in
+    # default (192.168.1.15:5000) it means the matching CONJURER_* env var never
+    # reached the process - the single most common cause of "service unreachable"
+    # confusion. Printing them makes env-vs-default obvious at a glance.
+    logger.info(
+        "Resolved service addresses -> musician(file): %s | librarian: %s | radio: %s",
+        FILE_SERVICE_ADDRESS,
+        LIBRARIAN_SERVICE_ADDRESS,
+        RADIO_SERVICE_ADDRESS,
+    )
     await _load_service_groups()
     logger.info("Sensors: online")