[qoncept96]

Problem

getMetaDataFrom3mf() returns {} on every failure path (RequestException,
BadZipFile, ParseError, bare Exception — tools_3mf.py:293-323). Both callers
then immediately hard-index it:

# mqtt_bambulab.py:266  (cloud / project_file path)
print_id = insert_print(PRINTER_STATE["print"]["subtask_name"], "cloud", PENDING_PRINT_METADATA["image"])

# mqtt_bambulab.py:316  (local path)
print_id = insert_print(PENDING_PRINT_METADATA["file"], PRINTER_STATE["print"]["print_type"], PENDING_PRINT_METADATA["image"])

So any transient FTP/network hiccup, or a 3MF the parser doesn't like, turns into
KeyError: 'image' or KeyError: 'file' inside the MQTT callback.

This is the crash in #78 ("When print starts from printer - openspoolman throws
errors", KeyError: 'file'). It's also the root of the reports where print history
silently stops populating: the exception escapes mid-way through building
PENDING_PRINT_METADATA, leaving the module global half-initialised.

Note the if PENDING_PRINT_METADATA: guard added at :310 does not help — by the
time it runs, {} has already had print_type/task_id/subtask_id written onto
it, so it is truthy but still missing file and image.

The downstream ["complete"] cascade from this was already fixed in #76; this is
the same class of bug at the two sites that actually raise first.

Fix

.get() at both call sites. insert_print() already stores these straight into
nullable columns, so None is fine and the print is still recorded — just without
a thumbnail/filename — instead of taking down MQTT handling.

-      print_id = insert_print(PRINTER_STATE["print"]["subtask_name"], "cloud", PENDING_PRINT_METADATA["image"])
+      print_id = insert_print(PRINTER_STATE["print"]["subtask_name"], "cloud", PENDING_PRINT_METADATA.get("image"))

-            print_id = insert_print(PENDING_PRINT_METADATA["file"], PRINTER_STATE["print"]["print_type"], PENDING_PRINT_METADATA["image"])
+            print_id = insert_print(PENDING_PRINT_METADATA.get("file"), PRINTER_STATE["print"]["print_type"], PENDING_PRINT_METADATA.get("image"))

Testing

These guards have been running in production on a LAN-mode P1S since 2026-08 (on
0.3.0, rebased onto main and redeployed 2026-09-03). I hit the 'image'
variant originally as a repeating crash: the exception escaped mid-way through
building PENDING_PRINT_METADATA, and every subsequent MQTT message then died
until the container was restarted, so tracking went silently dead.

I have not written an automated repro for either call site — the change is a
narrow .get() on two values that insert_print() already accepts as NULL. The
'file' traceback in #78 is a reporter's, not mine.