The filament list is positional from the modal down to the CLI's
filament_N.json parts, but for a source that already carries slice_info the
requirements endpoint returns only the slots the plate consumes. A
MakerWorld model declaring four filaments and painting with slot 4 alone
therefore showed one dropdown, whose PETG pick the CLI bound to slot 1 —
slot 4 sliced with the profile baked into the source, and the print came out
PLA.
The endpoint now takes full_slots, which widens that answer to every
project slot with used_in_plate flags, and only the slice modal passes it.
Print-time AMS matching shares the endpoint and keeps the used-only list, so
it still asks for exactly the spools the job needs.
setInterval does not await an async callback. Slicing a large project
blocks the backend for seconds, so poll ticks piled up behind one stalled
request, each holding a snapshot taken while the job was still active.
They resolved together, and every one of them ran the completion path —
one toast and two query invalidations each. A 20s stall against the 1.5s
interval produced 13 "Sliced X" toasts from a single slice.
Only one poll round is now in flight at a time, which also stops queueing
requests against a backend that is already saturated. Completion is
recorded once per job id, and a round still awaiting a response when the
effect tears down now returns instead of acting.
On a printer with an external camera, watching the live view while a print
ran meant the layer timelapse recorded almost nothing and the finish photo
went out with no image. The reporter measured 0 of 87 layer captures on one
print and 0 of 105 on another, both watched throughout. A USB camera allows
one V4L2 handle, so a capture during a live view fails outright.
The built-in camera has had this rule since #1348 and #1271: reuse the
viewer's buffered frame rather than opening a second connection. It was
never extended to the external paths, and it could not have been -- the
buffer it depends on was only ever populated by the built-in paths.
generate_mjpeg_stream yields multipart-wrapped chunks, so the route layer
could not recover the JPEG, and a guarded caller would have found an empty
buffer and skipped every time.
So the stream now publishes each raw frame through a new on_frame callback
(parallel to on_process from #2675), and the six one-shot consumers reuse
it: layer timelapse, the finish-photo moment and its background fallback,
the notification snapshot, Obico polling, and the plate check. A viewer
attached with nothing buffered yet skips that one attempt rather than
competing -- kicking the viewer off is worse than missing a frame.
on_frame exceptions are logged and swallowed, like iter_subscriber's
on_unsubscribe: buffering is a side effect and must never be able to take
the live stream down with it. The external stream's teardown now releases
the buffered frame too, ownership-checked so a concurrent viewer of the
same printer keeps its own.
Two side effects on paths not touched here, both improvements: the snapshot
endpoint and the finish-photo fallback chain consult get_buffered_frame and
can now serve an external camera's live frame. plate_detection's docstring
already claimed this behaviour while implementing it only for the built-in
fallback; that drift is resolved.
ffmpeg is spawned with stderr=PIPE and it was only read on the error
paths, so for the life of a working stream nobody read that pipe. ffmpeg
writes its banner, the input analysis, then a progress line at a steady
rate; a 64 KiB pipe fills eventually, ffmpeg blocks writing to it, frames
stop, and the stream's own 30s timeout fires -- logged as "RTSP read
timeout" with no hint that we starved it ourselves.
How long that takes is unmeasured and evidently long: one H2D upstream
ran 21m36s without stalling, and an earlier 512 B/s extrapolation of mine
was mostly the one-off startup banner. So this is a bounded resource being
treated as unbounded, not a fault anyone has reported.
_FfmpegStderrTail drains the pipe continuously and keeps a 16 KiB rolling
tail. That tail is what the error paths now report, which is better
material than before: it holds what ffmpeg said as things went wrong,
where the on-demand read returned whatever was printed first -- usually
the banner, which the summariser strips anyway.
Three readers wanted this one pipe, and asyncio rejects concurrent reads
on a StreamReader, so the collector is authoritative: it registers by pid,
_read_ffmpeg_stderr returns its tail when present and otherwise reads the
pipe unchanged, and _terminate_ffmpeg skips its own stderr drain when the
collector owns it (the collector keeps draining through teardown, which is
all wait() needs). The generator starts it after the immediate-failure
check, which reads the pipe directly because the process is already dead,
and releases it after _terminate_ffmpeg.
text() goes through _summarize_ffmpeg_stderr like every other stderr log
here, so the access code ffmpeg echoes in its input URL stays masked.
aclose() awaits the cancelled pump rather than firing and forgetting, so
no pending task survives into loop teardown.
Closing a camera view and reopening it immediately could leave the new
stream unregistered while it was running and delivering frames. The
damage was all indirect: is_stream_active() reported no viewer, so Obico
polling and snapshots opened a second camera connection against the live
view (the thing #1348 and #1271 exist to prevent); the janitor's /proc
scan found an ffmpeg missing from _active_streams and killed the live
stream as an orphan; and /camera/stop reported "Stopped 0" with a stream
running.
The fan-out stream id was f"{printer_id}-fanout" -- constant per printer,
so every successive stream shared one registry key, and the departing
generator's finally popped whatever was under it, including its
successor's entry. The same finally also cleared the per-printer frame
buffer unconditionally, discarding the new stream's frame. It needed the
two streams to overlap, which the 4s teardown made easy.
Each stream now gets its own key via _new_fanout_stream_id(), so a
generator can only clean up after itself -- the external-camera path
already does this (#2675) and this brings the fan-out path in line. The
per-printer dicts are released through _release_printer_frame_state(),
which checks that no other stream for the printer is still running; both
the RTSP and chamber-image cleanups had the same unconditional pop.
Also hoisted time and uuid to module level and dropped four
function-local `import time` statements. A local import shadows the name
for the whole function, so any use on a branch that doesn't reach the
import raises UnboundLocalError -- a real hazard in camera_stream, whose
external-camera branch imported both while the RTSP path needs them too.
A test pins camera_stream as free of function-local imports.
Closing a camera view logged "ffmpeg didn't terminate gracefully,
killing" followed by "ffmpeg did not exit within 2.0s of SIGKILL;
abandoning wait", on every single close. Both waits expired every time,
so teardown took a fixed 4.00s -- and since the firmware allows one
camera connection, that was 4s in which nothing else could use it.
ffmpeg is spawned with stdout and stderr as pipes and the teardown paths
have stopped reading them, so it sits blocked in write() on a full 64 KiB
pipe. SIGTERM cannot be acted on there: the handler only sets a flag that
the main loop polls, and the loop never gets back to the check. SIGKILL
does kill it, but asyncio resolves Process.wait()'s waiter through
_try_finish(), which requires every pipe transport to report
disconnected; paused, unread pipes never reach EOF, so wait() blocks with
returncode already set. A negative-control test shows returncode=-9 at
the instant the abandon fires.
Draining both pipes while stopping the process fixes both halves: 4.00s
becomes ~0.15s. The signal ladder and its bounds stay as backstops, so a
genuinely wedged process still cannot hang a stream, a Stop request or
the janitor.
This corrects _FFMPEG_KILL_TIMEOUT's premise and #2580's conclusion. That
12-hour hang was the unbounded form of this same self-inflicted stall, not
an ffmpeg stuck in uninterruptible I/O -- the process observed doing it
was in state S, which cannot survive a delivered SIGKILL. Bounding the
wait capped the symptom without removing the cause.
mqtt_relay reads state.left_aux_fan_speed, but the SimpleNamespace fixture in
test_plate_clear_mqtt_notification enumerates its fields explicitly, so the two
status-payload tests raised AttributeError. I updated the equivalent fixture in
test_printer_manager_status_broadcast and missed this one — running only the
touched suites is what hid it.
Also addresses the round-2 review notes:
- exhaust_fan_present: documented that the H2 series reports part 3 too, so the
flag is not model-specific despite the name.
- Mask the part id after shifting, matching get_flag_bits(id, 4, 8), for
consistency with the state decode. No behaviour change for any observed id.
- Noted the unmapped H2 id 6 beside the id branches.
- Reject fan=aux2 when the printer reports no left_aux_fan_speed, so a POST
against an A1 no longer sends M106 P10 for absent hardware. The UI already
hid the badge; this closes the same hole on the API.
- Added a test asserting EXHAUST_FAN_LABEL_MODELS and the frontend's
MODELS_WITH_EXHAUST_LABEL cannot drift apart.
Subprocess output and user-supplied URLs are scrubbed of credentials
before they reach the application log. Adds a shared redaction helper in
core/logging_filters and routes the existing support-bundle sanitizer
through the same pattern.
Timelapse was on, the video never reached the archive, and Scan for Timelapse
found nothing afterwards. Across 247 support bundles this was the norm, not an
edge case: 457 automatic scans scheduled, 262 attached.
The scan looked four times over ~65s. The attempt that found the video was #1
272 times, then 17 / 13 / 13 — flat against the cutoff, not decaying, i.e.
files were still arriving when we stopped. What ran afterwards searched for the
print name inside the filename; Bambu only writes "video_<timestamp>", so it
fired 159 times and matched zero.
The manual Scan had no baseline at all and matched on filename timestamp, FTP
mtime, or "there is only one video" — all reading a clock a LAN-only printer
cannot sync. The reporter's P1S was six and a half days out.
- Poll for minutes instead of ~65s; drop the name-match fallback.
- Persist the print-start baseline on the archive, so the diff survives a
restart mid-print and the manual Scan runs the same comparison. With a
baseline present the clock-based strategies are skipped entirely — they can
only turn an honest "pick one" into a confident wrong answer.
- When several files are new (a previous print's video landing late), exclude
the ones already attached to another archive instead of ordering the
candidates. Ordering could only be done on the printer's clock.
- Delete the video from the printer once archived. Keeps /timelapse to
unclaimed files, which is what makes the diff unambiguous, and stops P1S
cards filling with AVIs.
- Gate that delete on a verified transfer: download_file now compares against
the size from the listing. An FTPS connection closing early does not always
raise, so a partial buffer was being attached as a complete video — which
would also have been the one case where deleting the source lost data.
Bounded twice on purpose: wall-clock deadline plus a derived round cap, since
the deadline stops bounding the loop as soon as the sleeps are shortened.
Per-round logging only speaks when the listing changed — 31 rounds of full
listings would bury the interesting line in the support bundle.
Migration adds print_archives.timelapse_baseline as JSON, spelled the same on
both dialects so a migrated database matches a fresh one.
-----------
fix(finish-photo): add the timelapse frame to the archive after the notification (#2704)
When a print records a timelapse, its last frame is the better finish photo:
the firmware stops recording with the toolhead parked and before the end
G-code drops the bed, where a live grab at that moment catches a lowered
plate. Bambuddy waited 60s for the video and then gave up, because the
print-complete notification blocks on that photo and holding a notification
for minutes is worse than sending it with the live grab.
P1-series printers write MJPEG AVI rather than H.264 MP4 and serve it slowly.
Measured over 261 attaches in the support bundles: P1S median 33s, p90 167s,
worst 546s, while every other model finished inside 26s. So the printers that
most needed the better framing were the ones that never got it.
Keep the notification on the same bound, and keep waiting off to the side.
_capture_finish_photo_from_timelapse now reports whether it ran out of time or
concluded — a video that landed and failed extraction is not worth retrying,
one that never arrived is. On the first, schedule a background task that waits
up to 15 minutes and inserts the extracted frame at the front of the archive's
photo list, where the gallery opens.
The live grab stays on disk: the notification already links to that exact
file, so removing it would leave a broken image in Discord or Telegram.
The length check proves we received what the listing said, not that the file
was finished. The first look happens ~5s after the print ends, while the
printer may still be writing, so a growing file can be listed short, served
short, and pass. Re-list after the download and only accept the video once its
size has stopped changing — a failed re-list counts as not settled, since
"could not check" must not mean "safe to delete".
Review feedback on #2691: `state` is bit-packed like its sibling `range`
(end << 16 | start), and Bambu Studio decodes it with
get_flag_bits(state, 0, 8). Masking with & 0xFF before clamping means a
packed value decodes to the real percentage instead of clamping to 100.
Also moves the uses_exhaust_fan_label import to the top of printers.py with
the other imports.
Tests: packed value (60 << 16 | 45) decodes to 45, and plain 0-100 values
round-trip unchanged.
The fan-speed endpoint always reported 'Chamber fan set to N%', so on
P2S/X2D — where the printer card labels that fan 'Exhaust' — clicking
Exhaust produced a toast saying Chamber fan.
Adds uses_exhaust_fan_label() to printer_models so the badge label and the
API response share one source of truth, and uses it to pick 'Exhaust fan'
vs 'Chamber fan' in the response message.
Tests: helper coverage for P2S/X2D (incl. internal codes N7/N6), other
enclosed models, and unknown/missing model; API test asserting the message
matches the badge label per model.
The P2S/X2D have two fans bambuddy didn't fully handle. On the P2S both are
add-on kits; on the X2D they ship from the factory.
1. Left auxiliary part cooling fan — not shown or controllable. It is reported
ONLY as device.airduct part id 10 (raw id 160 >> 4; FAN_REMOTE_COOLING_1 in
Bambu Studio's DevFan::ParseV3_0) and is never mirrored into a flat
big_fanX_speed field, which is why it was invisible. This is the gap
identified in #2576, where the single 'Auxiliary' fan (big_fan1 / M106 P2)
only reaches the right-hand aux fan.
2. Chamber exhaust fan — shown on every P2S labelled 'Chamber Fan'. On P2S/X2D
Bambu's firmware/UI (and Bambu Studio's FAN_CHAMBER_0_IDX) call it 'Exhaust',
and it is a kit on the P2S rather than built in.
Both are now detected from device.airduct.parts, which lists only the fans that
physically exist, so each tile appears only when the hardware is present.
- bambu_mqtt: parse airduct part 10 -> left_aux_fan_speed (None when absent) and
part 3 presence -> exhaust_fan_present; set_fan_speed() accepts index 10 plus a
set_left_aux_fan() helper
- schema / status route / printer_manager broadcast / mqtt_relay expose both fields
- POST /printers/{id}/fan-speed accepts fan=aux2 -> M106 P10, the command Bambu's
official P2S machine profiles use
- frontend: 'Left Auxiliary Fan' tile shown when reported; big_fan2 tile labelled
'Exhaust' and presence-gated on P2S/X2D, unchanged 'Chamber Fan' elsewhere
- i18n: leftAuxiliary + exhaust for all 12 locales
Verified fan -> field map on a live P2S (fw 01.02.00.00), stable across cooling
and heating airduct modes:
Part cooling -> cooling_fan_speed / airduct id 1 (built in)
Aux -> big_fan1_speed / airduct id 2 (built in)
Exhaust -> big_fan2_speed / airduct id 3 (kit)
Left aux -> airduct id 10 only (kit; forced off in heating by mode config)
Tests: airduct id-10 parsing (raw 160 -> id 10, not literal 160), id-3 presence,
base-P2S absence, diff-push survival, clamping, malformed entries, M106 P10
emission, invalid-index rejection; fan-speed API aux2->10 mapping; frontend tile
presence and labelling per model/kit.
Published models often deviate from the stock Bambu profile on purpose -
five walls, 100% infill, a 0.1mm first layer. Re-slicing one for a
different printer discarded all of it: the picked process preset
overrides the file's embedded settings, and that override is precisely
what makes cross-printer re-slicing work, so it cannot just be dropped.
"Slice as designed" (#2611) does not help - it is all-or-nothing and
only offered when the picked printer already matches the design's
target.
The deviation list does not have to be computed. Bambu Studio writes it
into the 3MF as different_settings_to_system, laid out as
[process, *filaments, printer] - verified against real files at 2, 3 and
4 filament slots. The parser refuses any file whose array length
contradicts its own filament count rather than guessing an index, since
reading the printer slot as the process slot would carry the designer's
machine_start_gcode onto a foreign printer.
The slice dialog now lists exactly which print settings the author
changed and what each was set to, with a checkbox per setting. Design
intent - wall count, infill, layer and first-layer height, supports,
seam, brim, ironing - is ticked by default. Printer-specific values -
speeds, accelerations, jerk, fans, temperatures, prime-tower geometry -
are listed with a badge but start unticked: tuned for the author's
machine, they can be merely wrong on the target or outside the range its
profile accepts, which fails the slice outright.
Only ticked keys are sent, and only keys the source actually flags as
changed are applied. Values are written into the outgoing process JSON,
the same mechanism the support carry-over has used since #1881: for a
Standard preset pick that JSON is an inherits stub, so the patch is the
child in the chain and wins over the flattened parent. Process slot
only - filament picks are honoured as chosen.
The wiki's "this is not a settings merge" note under Slice as designed
described the gap this closes; rewritten to point at the new panel.
Translated in all locales; wiki updated. Covered by backend and frontend
tests.
When a print reaches a terminal state Bambuddy holds the queue until
someone confirms the build plate is clear. That gate was visible only in
the Web UI: the printer's own MQTT push reports nothing beyond RUNNING,
PAUSE, FAILED, FINISH and IDLE, so an external automation could not tell
"finished" from "finished and still waiting for a human".
The per-printer status topic now carries an awaiting_plate_clear field,
and every transition is additionally published on a new retained topic,
bambuddy/printers/{serial}/plate_clear. Retained, and published from the
flag itself rather than from printer telemetry: a subscriber learns the
state of every printer the moment it connects, and the state stays
correct after Auto Off powers a printer down - telemetry stops there,
which would otherwise leave the status topic frozen at false.
Publishing is edge-triggered. The queue clears the gate on every
dispatch whether or not it was up, and no subscriber should see a
"plate cleared" for a plate that was never dirty. Persistence and the
WebSocket broadcast stay unconditional; they are idempotent and predate
this.
A matching Plate Clear Required notification event was added, off by
default on every provider because it fires after every print at the
same moment as the print-complete alert. Only the rising edge notifies.
Acknowledging still goes through POST /printers/{id}/clear-plate.
Two tests in test_printer_manager_status_broadcast.py asserted
_schedule_async.call_count == 2 for the setter. The new emission makes
it three on a transition, so they now assert that the persist and
broadcast coroutines are actually scheduled - which is the contract
Translated in all locales; wiki updated. Covered by backend and
frontend tests.
Projects made of many distinct files that each need N prints (e.g. 13
plates x 10 sets = 130 prints) only had aggregate progress. Finding out
"how many times have I printed plate_7?" meant reading the Activity
Timeline line by line, unusable at 130 events.
Projects now take an optional Copies per File target. Every printable
file in the project's linked folders shows an X / N badge with a mini
progress bar (gray not started, amber in progress, green done), and the
progress card gains a Complete Sets bar - the minimum per-file count,
i.e. how many finished assemblies can be shipped right now. Without the
target, printable files show a plain printed-count badge.
Counting matches the aggregate project stats: completed runs only,
served by a new /projects/{id}/file-progress endpoint. Runs attribute
to a file via a new library_file_id stamp on queue-dispatched archives,
falling back to content hash and then filename for historical rows.
Also fixed: files queued from a project-linked File Manager folder now
inherit that project, so their prints count toward project statistics -
previously only prints started from the project page were attributed.
Test-harness fix along the way: the test suite's get_db override never
committed, unlike production get_db, so endpoints relying on the
request-scoped commit silently lost their writes in tests. The override
now mirrors production commit/rollback semantics.
Library folders have no ownership tracking, so folder deletion was
gated entirely behind library:delete_all - a user with
library:delete_own could create folders and delete their own files,
but the emptied folder sat there until an admin removed it.
Users with library:delete_own can now delete folders that are truly
empty: no subfolders and no files, including trashed ones - folder
deletion cascades, so removing a folder that holds another user's
trashed file would silently break trash restore. External folders
(operator-configured mounts) and folders linked to a project or
archive still require library:delete_all even when empty, since
deleting them affects more than the folder itself. The bulk-delete
endpoint applies the same rule instead of skipping all folders for
non-admin users.
The folder tree's Delete entry enables accordingly and shows a
"You can only delete empty folders" hint on non-empty folders. The
backend stays authoritative - a folder that only contains trashed
files is invisible in the tree but still refuses deletion.
The live Obico classification was only visible under Settings ->
Failure Detection, so tracking how detection matched an ongoing print
meant flipping between the Printers screen and Settings.
Each printer card's badge row now shows an AI badge whenever detection
is enabled for that printer, like the other health badges: gray Idle
while no print is being watched, then green Safe, amber Warning, or
red Failure while a print is actively monitored. The tooltip carries
the current smoothed score; clicking jumps to the full detection
status and history in Settings. Printers excluded from the monitored
subset show no badge.
Served by a new lightweight /obico/printer-status endpoint readable
with printer permissions alone - it exposes only the enabled flag, the
monitored-printer set, and per-printer classification, keeping ML URL
and other configuration behind the existing settings-gated endpoint.
The Most Expensive record on the Statistics page ranked prints by
filament cost alone, ignoring the per-print energy cost Bambuddy
already measures via an attached smart plug. It now ranks by
filament + measured energy cost; prints without smart-plug data
compete on filament cost alone, as before.
Filament Trends gains an Energy Over Time chart: kWh per day (per
hour for ranges of a week or less, per week for long ranges), with
the range's total kWh and energy cost in the header. The chart only
renders when the selected range contains measured energy data, so
setups without smart plugs see no change.
The /archives/slim stats feed now carries each run's energy_kwh /
energy_cost from print_log_entries. Translated in all locales.
Covered by backend and frontend tests.
Closing an external USB (V4L2) camera view abruptly could leave its ffmpeg
running and holding /dev/videoN open -- LED stuck on, and reopening the view
failed or took 10-30s fighting for exclusive device access. Same class of leak
as #776 (built-in RTSP path), but the external path was never wired into that
fix: external streams registered into none of the _active_streams /
_disconnect_events / spawned-PID registries, so /camera/stop returned
{"stopped": 0} for a live USB stream and the orphan janitor's /proc net matched
only rtsp(s)://bblp: cmdlines. Cleanup ran only via the stream generator's own
finally, which an abrupt disconnect can skip.
- Thread an on_process callback + stop_event through generate_mjpeg_stream into
_stream_usb / _stream_rtsp; register the process before the startup probe so a
process that hangs on a locked device (not just one that exits) is reapable.
- Register external streams into the shared registries under a unique
{printer_id}-ext-{token} id so /camera/stop and cleanup_orphaned_streams find
and kill them; stop_event prevents the reconnect loops from respawning.
- Extend the /proc safety-net scan to also match USB (-f v4l2) ffmpeg, excluding
still-active streams and unrelated ffmpeg.
The folder tree's "sort by recent activity" and the file pane's date sort
put external (mapped/NAS) files in a near-random order instead of ls -t's
newest-first. Nothing captured the files' on-disk mtime: the sort keyed off
the DB updated_at/created_at, which for a bulk external scan is the same
scan instant for every row, so a whole block tied and sorted arbitrarily;
only rows Bambuddy had later touched individually looked "partially right."
The tree also bubbled up only immediate child-file activity, so a file added
deep in a subtree never lifted its parent folders.
- Add nullable fs_modified_at to LibraryFile and LibraryFolder (dialect-
branched migration, mirroring the #2615 dispatching_at pattern).
- External scan records each file's and directory's real os.stat().st_mtime
and refreshes it on every re-scan, so a file edited over the mount
re-sorts and existing installs backfill on the next scan.
- list_folders computes each folder's activity as a recursive newest-
descendant roll-up (post-order), so a fresh deep file lifts every ancestor.
- Folder tree sort and the file pane's date sort now use the real mtime,
falling back to created_at for managed uploads with none.
- New toolbar toggle shows/hides each item's last-modified date in the right
pane (grid + list), with strings in all locales.
Store the mtime as naive UTC to match the other timestamp columns so activity
comparisons never mix naive and aware values on either dialect. Covered by
integration tests (mtime capture, re-scan refresh, deep-file recursive bubble,
folder mtime) and a frontend test proving fs_modified_at is preferred over
created_at.
Pairs the top-down plate preview with the slicer's per-object pick mask
(Metadata/pick_N.png), whose pixel colours encode the same identify_id the
firmware's skip command takes, so a click resolves to a real object rather
than an inferred bounding box. Several objects can be selected before one
confirmation; selected and already-skipped items are highlighted on the
plate; the checklist stays available when no mask exists.
view=pick serves only the active plate's mask and 404s otherwise, unlike
every other view. A render returned in a mask's place would be decoded as
object IDs — dark pixels yield small integers that collide with real ones —
and a click would then skip an arbitrary object, mid-print, irreversibly.
The 404 is what tells the UI to fall back to the checklist.
Click mapping goes through the contained rect, since the canvas paints at
mask resolution under object-contain; clicks on a letterbox bar are rejected
rather than clamped onto whichever object touches the border. Confirming
names the object when one is selected and counts them when several are,
which is what plates of identically-named clones need.
No printer-control command path was added or changed; the layer, permission
and existing skip-command guards are untouched.
Slicing for a P2S failed with "filament preset Bambu PLA Basic @BBL X1C 0.2
nozzle (slot 1) is not compatible with printer Bambu Lab P2S 0.4 nozzle" —
naming a profile shown nowhere in the dialog. The picked profile was
"Overture PLA Matte @0.2", whose inheritance chain roots in that X1C profile.
The dialog classifies a profile by its compatible_printers list and falls back
to reading the printer out of its name. That name carries no model, and the
list — present on the imported copy — is not shipped by every source: Bambu
Cloud omits it deliberately (rate limits), and Orca Cloud shipped it but
Bambuddy only mined filament type and colour from the same content.
Orca Cloud entries now carry their own compatible_printers, and the existing
same-name enrichment bridge carries the list onto entries that lack one, in
both directions between the cloud tiers. A bare "@<size>" name tag is read as
a nozzle size as a last resort: it can rule a printer out but never rules one
in, and implausible values are ignored rather than guessed at.
An end-of-print auto-off on a plug that powers a filter fan marked the linked
printer offline and forced its state to "unknown". The mark was unrecoverable:
connected heals on the next MQTT message but state does not (only frames
carrying gcode_state rewrite it, and steady-state push_status frames are
partial), so the printer stayed "unknown" until a manual Force Refresh and the
queue never dispatched to it again.
The offline mark is now an explicit presumption: mark_power_off records the
state it overwrites and _on_message undoes it as soon as the printer sends
another report on its own topic, since inbound traffic proves the power was
never cut. A reconnect discards the saved state, so a genuine power cut is
unaffected. Each plug also gains a controls_printer_power flag (default true,
backfilled) that gates all five power-off paths, and the queue's power-on step
now picks the flagged plug instead of whichever linked plug came first.
The A2L reports its 4-slot AMS Lite as unit id 16, but its slot-presence
bitmasks sit at bit base 24 (id 6) and it reports tray_now as a local 0-3
slot. Fed the raw id 16, the ams_id*4+slot convention probed bits 64-67
(always zero) and marked loaded slots empty; the local tray_now was read as
global, so usage deducted from the wrong spool (or not at all); and the
ams_id<=7 DB constraint rejected id-16 Spoolman links.
Normalise the Lite 16->6 at the MQTT ingest boundary so global tray ids land
at 24-27 - matching the firmware's own bit base, working with every existing
ams_id*4+slot consumer, colliding with nothing, and passing the DB
constraint. Globalise tray_now to 24+slot, widen the valid-tray guards, label
the unit "AMS Lite", and build the confirmed ams_mapping2 {ams_id:16,
slot_id:0-3} / flat 0-3 for dispatch. Outbound slot commands translate 6->16
on the wire via a single helper. Self-scoping: only unit id 16 is touched, so
all other printers/AMS types are unaffected. One uncaptured wire field (the
physical global tray on load/cali) is extrapolated and isolated to the helper.
Assigning a spool to an AMS tray pushed ams_filament_setting +
extrusion_cali_sel and reported success immediately, whether or not the
tray accepted it. A silently-dropped assignment never surfaced, and since
a print only deducts from the spool on the exact tray it pulls from, it
also recorded no filament usage - which made the whole thing feel random.
Read the AMS telemetry back after every assign (inventory assign_spool and
the Configure Slot modal) and toast the outcome: loaded when the tray
echoes the pushed tray_info_idx, a warning when the filament loaded but the
K-profile (cali_idx) did not, or not-confirmed after ~30s. Verification
uses the periodic per-tray push (the command ack hardcodes sequence_id 0
and can't be correlated); an on-demand pushall is nudged so it lands
quickly. Covers regular AMS, AMS-HT and external slots; stays silent rather
than inventing a failure if the printer goes quiet. The read-back check
runs on every AMS push because the change-hash excludes tray_info_idx.
Bed levelling, flow calibration, and nozzle-offset calibration were on/off
only, so the sole way to run bed levelling was to force a full level before
every print. Bambu Studio has always offered a third "Auto" state that lets
the printer skip the calibration when it was done recently -- the state most
users actually want. Make these three options tri-state (off/on/auto),
defaulting to auto, and leave vibration/layer-inspect/timelapse as on/off
(Bambu Studio exposes no auto for those).
Wire encoding follows Bambu Studio's source exactly: each option sends a JSON
bool (true only for "on") plus a companion int -- off=0, on=1, auto=2. The
bool fields stay booleans (the #1478 H2S regression); only the companion int
widened from {0,1} to {0,1,2}. #1721's observation that stage 8/39 stays
queued when sending 2 is the auto contract (queued, skipped at runtime if
recent), not a broken "off".
- schemas: TriState = Literal[off/on/auto] with a BeforeValidator coercing
legacy bool / 0-1 / true-false so old clients and un-migrated rows validate
- model + migration: boolean columns -> String; SQLite via column affinity +
data backfill, PostgreSQL via ALTER COLUMN TYPE guarded on information_schema
(verified on both dialects); settings rows normalised true/false -> on/off
- MQTT: start_print takes the tri-state strings and emits the paired bool+int
- Virtual Printer: reconstructs the slicer's auto/on/off from the int companion
(auto_bed_leveling / extrude_cali_flag) in both capture paths
- frontend: CalibrationMode type; off/auto/on segmented controls in the print
dialog, queue bulk-edit, and Settings -> Workflow; calibrationMode_* strings
in all 11 locales
The /overlay/{id} route renders without a login, but everything it draws is
auth-gated: printer status and name (PRINTERS_READ), one setting (SETTINGS_READ),
and the camera stream (a camera-stream token). A signed-in browser rides its JWT
from local storage; OBS is a fresh browser with no session, so the overlay stayed
blank whenever authentication was enabled. Cloudflare/remote access was never the
cause -- an incognito window fails identically.
Give the overlay a self-contained kiosk-token mode, mirroring the Cam Wall:
- New `overlay` long-lived-token scope, kept separate from `camwall`: the overlay
names the printed file on screen, which a Cam Wall token is trusted never to
expose, so folding it in would silently widen every existing wall token.
- New token-authed GET /printers/{id}/overlay-status returning exactly the fields
the overlay draws and nothing else; added to the auth-middleware allowlist so it
reaches its own RequireOverlayTokenIfAuthEnabled gate.
- StreamOverlayPage reads ?token= and, in that mode, authenticates its status and
camera calls with the token and skips the WebSocket (the 2s poll is the feed).
The logged-in path is unchanged.
- Token-mint UI (Settings > API Keys) offers the scope with a ready-made
/overlay/{id}?token= URL copied once on creation.
A queue row stays status='pending' for the whole FTP upload; status only flips
to 'printing' at the end. The edit routes only blocked non-pending rows, so a
PATCH during the upload window was accepted while the in-flight dispatch kept
using its snapshotted printer -- splitting the queue row from the archive /
expected-print / physical command across two printers, and enabling a duplicate
dispatch on restart. The #1853 CAS guards cancellation, not reassignment.
Add a dispatching_at claim, stamped atomically (WHERE status='pending' AND
dispatching_at IS NULL) before any slow I/O and cleared on every exit. While
held, the single-item PATCH returns 409 (re-checked just before the write),
bulk edits skip the row, and the scheduler won't re-select it. Startup
reconciliation clears claims orphaned by a crash mid-dispatch. The row stays
pending throughout, so no status/UI/completion/reconciliation path changes.
New column print_queue.dispatching_at (nullable, dialect-safe DDL). Covered by
scheduler tests (claim exclusivity, non-pending rejection, release-on-exit,
skip-already-claimed, startup stale-clear) and API tests (reassign 409,
printer_id unchanged, bulk skip, unclaimed row still edits).
Server-side slicing always applied the picked printer/process/filament
triplet via --load-settings, which overrides the designer's embedded
project_settings.config — so a MakerWorld model set up for 5 walls came
out at the picked profile's default 2. That override is correct for
re-slicing a design onto your own printer/AMS, but there was no way to
slice a file the way its author configured it.
SliceModal now offers a "Use the file's built-in settings" checkbox when
the source 3MF carries embedded settings AND the picked printer matches
the design's target model. It routes to the existing embedded-settings
slice path (previously only a crash fallback), so walls/infill/filament
come from the file. Ticking it locks all four preset dropdowns — printer
included, since it's unused on this path and changing it would drop the
match and hide the toggle. The printer-match gate stops embedded settings
being honoured across models (wrong bed); there is no cross-printer
re-targeting on this path.
- schema: use_embedded_settings on SliceRequest
- route: embedded_mode branch; crash-fallback guarded against re-running
- frontend: gated checkbox locking all four dropdowns, resets on mismatch
- 2 i18n keys across all 11 locales
- tests: backend (flag skips triplet / ignored for STL) + frontend
(toggle offered on match, locks dropdowns + sends flag / hidden on mismatch)
A print queued from a specific plate of a multi-plate 3MF showed as Plate 1
in Print History after cancellation: the archive derives its plate from the
filename, but a whole multi-plate 3MF uploads under one name with no plate
suffix, so the parser defaulted to plate 1 and nothing copied the queue
item's plate_id onto the archive (which had no plate field).
Add a nullable print_archives.plate_id, copy it from the queue item at
dispatch (archive- and library-file paths), expose it in the archive API,
and render it in Print History. A startup backfill copies the plate onto
existing archives from their linked queue rows. Column add + backfill are
identical on SQLite and Postgres.
Also fix a related lifecycle bug: stopping a printing item while the printer
was offline left the linked archive stuck at "printing" (queue row
cancelled, but no MQTT completion ever arrives to reconcile the archive).
The offline-stop path now closes the archive out directly; the online path
still defers to the MQTT completion event.
The Configure AMS Slot modal sends built-in / local / Orca-generic presets
with a GF* tray_info_idx but an empty setting_id, and configure_ams_slot
forwarded that empty value to ams_filament_setting. The firmware treats a
filament-id-without-setting-id slot as half configured: it shows the new
material briefly, then reverts to its previously stored profile.
Back-fill setting_id from the resolved tray_info_idx via
filament_id_to_setting_id when the client sent none (e.g. GFB99 -> GFSB99),
mirroring the derivation the inventory/assignment path already does. Doing
it server-side also protects API callers and future frontends. P* user
presets and already-GFS* values are left unchanged, and an explicit
setting_id still passes through untouched.
Three more idle-in-transaction / thundering-herd paths from farm testing:
- print_scheduler: _start_print commits before the FTP delete/upload and
_preheat_and_soak commits before the heat-soak wait, so the per-item
session no longer sits idle-in-transaction across preheat + upload.
- cloud/filament-info: rollback the request transaction after the token
read and before the sequential Bambu Cloud calls; single-flight
concurrent misses for the same setting_id through one shared call.
- printers/cover: coalesce identical in-flight cover requests so followers
serve from the cache the leader fills instead of duplicating the
multi-path FTP + 3MF extraction.
Also adds pool_use_lifo (PostgreSQL default on, DB_POOL_USE_LIFO override,
shown in /system/db-pool) so a bursty farm keeps a small hot connection set.
Two or three concurrent UI logins exhausted the PostgreSQL pool on the
reporter's 93-printer farm: QueuePool limit of size 10 overflow 20 reached,
with all 30 sessions idle in transaction on the auth_enabled SELECT. Three
regressions had landed on dev after an earlier configurable-pool change was
reverted and never re-applied (only the route-by-route session fixes were).
- Pool sizing is env-configurable again (DB_POOL_SIZE / DB_MAX_OVERFLOW /
DB_POOL_TIMEOUT / DB_POOL_RECYCLE); the PostgreSQL default returns to
20 + 80 with pool_pre_ping and pool_recycle=1800, and GET
/api/v1/system/db-pool reports resolved config + live gauges without
checking out a connection. SQLite unchanged (20 + 200).
- is_auth_enabled caches for 30s again. Only enabled=True is ever cached, so
a stale read can only fail closed (require auth), never open; set_auth_enabled
invalidates immediately. An autouse test fixture resets the module cache
between tests to keep ordering deterministic.
- Every authenticated request checked out two pooled connections: the
permission dependency held one and the revoked-jti check opened another.
is_jti_revoked now reuses the caller's session; the token dependencies and
the auth-middleware gateway were restructured to open one session and pass
it in, so each request makes a single checkout.
The firmware's runout HMS text says "insert into the same AMS slot", which is
wrong under AMS Filament Backup: the firmware won't re-accept the depleted slot
and advances to the next compatible one. Bambuddy parsed print.ams.tray_now only
and dropped tray_tar/tray_pre, so the expected slot never reached the UI.
Capture tray_tar/tray_pre on PrinterState and, while paused, resolve them to
global tray IDs (expected_tray/previous_tray) on both the REST and WebSocket
status payloads via a shared resolver: single-AMS passthrough, multi-AMS
snow-mapping resolution, AMS-HT/external passthrough, and an honest null when the
slot can't be placed. The AMS graphic highlights the expected slot (amber) and
the ran-out slot (red); the HMS modal re-describes runout codes to name both,
falling back to "check the printer" when unresolved. Runout copy translated in
all 11 locales.
Reporter @Jostxxl confirmed tray_pre=1/tray_tar=2 during the pause (ran out in
Slot 2, printer expected Slot 3).
reconcile_stale_active_prints closes out stale status="printing" archives by
synthesising an aborted on_print_complete, which logged a PrintLogEntry whose
duration was completed_at - started_at — the whole multi-day disconnect gap,
since a reconciled archive's real end time is unknown. Across a farm of stale
rows this inflated Total Print Time by hundreds of hours, and the Stats total
recomputed the same value from the timestamps even when duration was NULL/0.
Reconciled completions now log duration_seconds=0, the two Stats time paths
trust a stored 0 instead of recomputing, and reconciled aborts get an honest
"Stale - reconciled ..." failure_reason instead of "User cancelled". Genuine
long prints are untouched (no cap; still-running >24h prints aren't stale).
The remaining routes of the idle-in-transaction class: the file-manager,
storage, camera-snapshot and timelapse routes each took their printer row
via Depends(get_db) and then talked FTP/camera on the same held session, so
a farm dashboard polling cover/snapshot tiles (offline printers included)
crept the pool to exhaustion over ~23h. They now read in a short session and
release before the I/O; timelapse re-opens a fresh session only for the write.
Also caps the four bare-executor FTP helpers with asyncio.wait_for so a
saturated 48-worker pool can't pin a caller (and its DB connection)
indefinitely, and runs the synchronous smtplib send off the event loop with
an explicit timeout so a wedged relay can't freeze the loop.
GET /printers/{id}/cover took its printer row via Depends(get_db), whose
yield-dependency session stays open for the whole request — including the
3MF cover download (up to 8 remote paths x retries with backoff, minutes
under FTP contention). One pooled connection sat idle-in-transaction the
entire time; on a large farm a wall of dashboards drained the pool. The
route now fetches the printer in a short-lived async_session() and releases
the connection before the download (expire_on_commit=False keeps printer.*
readable). Pinned by a signature-inspection guard that fails if get_db is
ever re-added.
fix(print-start): release the DB connection across plate detection and 3MF download (#2572)
on_print_start held one session from top to bottom of the handler, across
two slow I/O blocks that need no database: the plate-detection camera grab
and, on the new-archive path, the multi-path 3MF FTP download (its own
comments cite worst cases of tens of minutes). The connection sat idle-in-
transaction for both, once per starting print. It now commits at each
boundary — only read SELECTs have run on those paths (every write branch
returns earlier), so the commit persists nothing and simply returns the
connection to the pool for the I/O; the next query re-acquires, and
expire_on_commit=False keeps printer.* readable.
fix(startup): connect to printers concurrently so the API serves within seconds (#2572)
init_printer_connections awaited each printer's connection serially, and
connect_printer ends in a fixed 1s settle wait. The MQTT connect is non-
blocking (connect_async + loop_start), so that 1s x fleet size was pure
serial dead air the FastAPI lifespan blocked on before uvicorn began
serving — ~100s before port 8000 responded on a 93-printer farm. The
connections are now started with asyncio.gather, so the step takes ~1s
regardless of fleet size. return_exceptions=True isolates each result: one
unreachable printer no longer aborts the rest, or startup itself.
The Queue listing serialized each item by opening its 3MF and re-parsing
slice_info.config three times (print time, filament usage, bed type) on
every poll, per connected client, even for unchanged files. Add a single
combined extract_plate_metadata_from_3mf() cached by (path, plate_id,
mtime_ns, size); the three legacy helpers delegate to it. An unchanged
queue now does no repeat 3MF parsing.
OrcaSlicer shipped a first-class external-app pairing API (OAuth 2.0 Device
Authorization Grant), so the Supabase-PKCE copy-paste flow is replaced end to
end. Connecting is now: click Connect, approve a short code on the Orca Cloud
settings page, done — no redirect, no callback paste, no client secret, works
from a LAN IP / localhost / behind a proxy.
Backend: services/orca_cloud.py rewritten to device-code request + poll (the
four RFC outcomes) + refresh_token grant + introspection + external sync pull;
routes expose /device/start and /device/poll (device_code kept server-side in
the reused orca_cloud_pending_* columns, no migration). Requests sync:read
(read-only feature). Prod endpoint by default, ORCA_CLOUD_API_BASE overrides
to staging. Wired the shared httpx client (fixes a per-request socket leak).
Frontend: device-code connect UI + api client methods; all 11 locales updated.
After an RTSP read timeout the stream cleanup killed the stalled ffmpeg
and then awaited process.wait() unbounded. A SIGKILLed ffmpeg stuck in
uninterruptible I/O on a dead RTSP socket can take arbitrarily long to
be reaped, so the fan-out stream coroutine sat parked in that wait (12
hours in the reported case) while every new viewer attached to the
stalled broadcaster and received no frames.
Bound the post-kill wait to 2s in all three places it existed: the
stream generator's _terminate_ffmpeg (the reported hang), the camera
stop endpoint (which would hang the recovery request itself; now uses
the shared helper instead of an inline copy), and the orphan-cleanup
janitor (whose hang would disable the safety net). On timeout the
zombie is abandoned; the janitor's /proc scan reaps it next pass and
the stream proceeds to its normal reconnect.
A queue item's "Any <model>" button labeled itself from the file's slice
metadata while the scheduler used the row's target_model, so an X1C-sliced
item targeting H2D showed "Any X1C" above "assign to first idle H2D". The
mismatch itself was created silently: sliced-for metadata loads async, and
switching to model mode before it arrived pre-selected the alphabetically
first model (H2D on a mixed farm), after which the model dropdown hid
itself. Nothing validated compatibility, so the scheduler would hand X1C
G-code to an H2D.
Frontend: never default the target silently, keep the dropdown visible in
model mode (incompatible models disabled), label from the actual target,
warn on mismatch, block submit when incompatible.
Backend: new GCODE_COMPAT_FAMILIES table (X1/X1C/X1E/P1P/P1S interchange;
everything else exact-match; missing metadata never blocks). Queue create
and update reject incompatible targets with 400; the scheduler holds back
pre-existing mismatched rows with an actionable waiting_reason instead of
dispatching them.
Manual jog could drive an axis past its travel limit into a collision.
Instrumenting the exact G-code to an H2D showed Bambuddy sending a clean
move at the limit (G91 / G1 Z-1.00 F600 / G90, no M211) that the printer
ran straight past, while its own touchscreen refuses the identical move.
This is a Bambu firmware bug: soft endstops are not enforced on G-code
received over MQTT, and no axis position is reported, so the move cannot
be clamped firmware- or client-side from position.
Two changes: (1) jogs no longer wrap moves in M211 S0/S1 — that disabled
the firmware's soft endstops globally, breaking even the touchscreen's
limits until a power cycle; a bare move keeps the touchscreen protected.
(2) The jog panel shows a prominent warning that travel limits are not
enforced during manual moves due to the firmware bug. Client-side
dead-reckoning enforcement is tracked separately.
/camera/stream took its printer row via Depends(get_db). get_db is a
yield dependency, so its session stayed open until the response body
finished streaming — for a live MJPEG stream, as long as the browser
tab is open (hours). Every open camera tile pinned one pooled DB
connection idle-in-transaction, draining the pool on large farms.
Fetch the printer in a short-lived async_session() and release the
connection before returning the StreamingResponse. expire_on_commit=
False keeps the already-loaded columns readable during the stream.
Large PostgreSQL farms exhausted the fixed pool (pool_size=10 +
max_overflow=20): with ~93 printers every connection sat idle in
transaction and unrelated requests waited out the 30s pool timeout or
failed in the auth middleware.
- Make pool sizing env-configurable (DB_POOL_SIZE / DB_MAX_OVERFLOW /
DB_POOL_TIMEOUT / DB_POOL_RECYCLE); raise the Postgres default to
20 + 80 with pool_pre_ping + pool_recycle=1800.
- Cache the auth_enabled probe (30s) to drop a per-request DB round-trip.
Only enabled=True is cached, so staleness fails closed; set_auth_enabled
invalidates immediately.
- Add GET /api/v1/system/db-pool exposing resolved config + live
checked_out/checked_in/overflow gauges without consuming a connection.
Session-hygiene (connections held across MQTT/FTP/camera/3MF I/O) is a
separate follow-up.
MartinNYHC