Two related queue issues surfaced when scheduling an ASAP print with
quantity > 1 on an H2D:
1. Double-dispatch — both items in the batch ended up in 'printing'
status on the same printer, logged as "BUG: Multiple queue items in
'printing' status for printer N". The scheduler seeded its busy
set empty each tick and relied on _is_printer_idle() reading live
MQTT state, but H2D / P1 series lag several seconds between the
print command and IDLE → RUNNING, so the next check_queue() tick
saw IDLE and dispatched the second batch item onto the already-
running printer. check_queue() now seeds busy_printers with every
printer_id that has a row in 'printing' status before iterating,
so any printer with an outstanding dispatched job is excluded
regardless of what MQTT currently reports.
2. Progress bar flashed 100% — immediately after dispatch the queue
item's per-row progress bar showed the prior print's final mc_percent
for a few seconds, then snapped back to 0% when the new print
started ticking. QueuePage.tsx now gates progress / remaining_time /
layer fields on status.state being RUNNING or PAUSE; in any other
state (FINISH from the prior print, IDLE, PREPARE while heating)
the bar renders at 0% with no stale ETA or layer count.
Regression coverage added in test_phantom_print_hardening.py
(TestBusyPrinterSeedingFromPrintingItems, 3 tests): seeding query
returns only printers with 'printing' rows, empty when none exist,
and end-to-end check_queue() does not call _start_print for a pending
item whose printer already has a 'printing' row even when
_is_printer_idle() is forced True.
_watchdog_print_start reverted queue items to "pending" at 45 s if
gcode_state hadn't changed, assuming the MQTT project_file was swallowed
by a half-broken session (#887/#967). H2D Pro firmware (01.01.00.00)
routinely keeps state=FINISH for 48-55 s after actually accepting the
command before transitioning to PREPARE. The watchdog reverted items
the printer had already started physically printing; the archive updated
normally via _active_prints, but the queue item was now "pending" again,
and the next scheduler tick after plate-clear re-dispatched the same
item as if it had never run. With one item left in the queue that looked
like a reprint of the just-finished job; with multiple items the
symptom was masked by item N+1 getting dispatched during the race.
Add a second "command landed" signal: subtask_id advancing past the
pre-dispatch value. Bambuddy already mints a unique submission_id per
project_file publish (#1042) and the printer echoes it back on the next
push_status as soon as it starts processing the command - well before
gcode_state transitions on slow-transition models. _start_print now
captures pre_subtask_id alongside pre_state and passes both to the
watchdog, which exits early on either a state change or a subtask_id
advance.
Raise default timeout 45 s → 90 s as belt-and-braces for printers that
neither flip state nor echo subtask_id inside the polling window.
Genuinely half-broken sessions (both signals unchanged across the full
90 s) still revert + force-reconnect exactly as before.
Transient subtask_id=None during reconnect is not mis-detected as a
change. pre_subtask_id=None falls back to state-only checking so the
fix is safe for printers that haven't reported a subtask_id yet.
New test_scheduler_watchdog.py pins the eight behaviours that matter:
pickup via state change; pickup via subtask_id change with state still
FINISH (the exact #1078 case); revert when neither signal changes;
default timeout is 90 s; pre_subtask_id=None state-only fallback;
current subtask_id=None not treated as change; printer disconnect
mid-watchdog leaves DB untouched; item that already moved on is not
clobbered.
PR #939 added the awaiting_plate_clear gate but stored it on
PrinterManager, not on PrinterState. printer_state_to_dict() — which
builds every WebSocket printer_status payload — never emitted the flag,
so the frontend's WS merge preserved the stale false value. The only
path that surfaced true was the 30s HTTP fallback poll, and incoming WS
ticks kept bumping React Query's dataUpdatedAt, pushing the refetch out
further on chatty printers.
Emit awaiting_plate_clear from printer_state_to_dict by reading
printer_manager.is_awaiting_plate_clear(printer_id) directly; returns
False when no id is passed. No frontend change needed — the existing WS
merge carries the flag end-to-end and the button now appears the instant
the printer transitions to FINISH.
Regression tests assert the WS dict always contains the key and surfaces
True when the manager has the flag set for that printer_id.
Affects every printer (A1/H2D/X1C) equally — transport-agnostic path.
A single legacy spool with a 7-char rgba ('FFFFFFF', missing one F)
caused GET /api/v1/inventory/spools to 500 with a pydantic
ResponseValidationError, leaving the reporter with a blank Filaments
page and "Add Spool" silently failing. Root cause spans three layers:
1. Write path: SpoolUpdate.rgba had no pattern constraint (only
SpoolCreate did), so PATCH could plant malformed values in the DB.
2. Frontend: ColorSection hex input's `val.length <= 6 ? 'FF' : ''`
emitted 7-char rgba for 5-char input (XXXXX + FF = 7) and for
7-char typed input (no alpha appended).
3. Read path: SpoolResponse inherited the write-side pattern, so a
single bad row 500'd the entire list endpoint instead of being
tolerated through serialize.
SpoolUpdate.rgba now carries the same ^[0-9A-Fa-f]{8}$ pattern as
SpoolCreate. The hex input emits a fully-formed 8-char RRGGBBAA on
every keystroke — 8-char paste passes through, 7-char drops the
stray, shorter input pads RGB with '0' and appends FF alpha.
SpoolResponse.rgba is now Optional[str] with no pattern — write-side
validation is the right place for format rules; responses must
tolerate historical rows.
Tests: 16 schema tests (SpoolCreate/Update reject, SpoolResponse
tolerate), 7 frontend tests covering every input length 0–8 plus
non-hex strip. A user who already has a bad row in their DB now sees
it render with a default color instead of having to hand-edit SQLite.
The "Print" button on a printer card (and drag-drop-onto-card) used
FileUploadModal to persist the file as a LibraryFile, then dispatched
through POST /library/files/{id}/print. The LibraryFile row + disk file
were left behind after every one-off print, polluting File Manager with
entries the user never asked to save.
FilePrintRequest.cleanup_library_after_dispatch (default False) opts
into post-dispatch cleanup. When set, _run_print_library_file stages
db.delete(lib_file) in the same transaction as archive_print so a
mid-flight FTP / start_print failure rolls both back cleanly, commits
together, then unlinks the library disk file + thumbnail after commit
succeeds. External library files (is_external=True) are never touched.
Only the Printers-page Direct-Print PrintModal sets the flag. Every
other api.printLibraryFile caller (File Manager Print, Project Detail
Print) leaves it unset — their entries are there by user intent.
Also moves formatPrintName out of PrintersPage.tsx into a new
utils/printName.ts module — fa1c46d9 (#881) exported it inline so its
test could import it, tripping react-refresh/only-export-components.
When two printers were running different plates of the same multi-plate
3MF, the Printers page cards displayed the same file name on both and
there was no way to tell them apart. The Queue view already had this
information by cross-referencing the archive's plate list; the card
didn't have the linkage.
Expose `current_archive_id` (resolved by matching the MQTT `subtask_id`
against `PrintArchive.subtask_id` — the bridge introduced in #972 for
restart-resume) and `current_plate_id` (parsed from `gcode_file` by a
new shared `parse_plate_id` helper) on the status endpoint. The helper
is also called from the WebSocket push path so plate transitions
reflect within 100 ms instead of waiting 30 s for the next REST poll;
the archive id itself stays REST-only since it's stable for the life
of a print and shouldn't make the push path touch the DB.
The card fetches plate metadata via the same `api.getArchivePlates()`
call QueuePage uses — shared React Query cache keeps it cheap across
polls — and renders the actual plate name (or a "Plate N" fallback)
only when `is_multi_plate` is true. Single-plate prints stay clean.
Falls back to the previous `plate_N.gcode` regex path when there's no
archive linkage (e.g. prints started directly from the printer LCD).
Tests cover the plate-id extraction across Bambu Studio path shapes
(backend parse_plate_id, printer_state_to_dict wiring) and the label
override precedence in formatPrintName (frontend).
After configuring an AMS-HT slot with a custom cloud preset, the slot
card and Configure modal kept showing "Generic PLA" even though the
printer and slicer had the correct preset. The `/slot-presets` response
keyed HT entries at `ams_id * 4 + tray_id = 512`, but frontend lookups
used `ams_id` directly (128 on PrintersPage via getGlobalTrayId, 64 on
SpoolBuddy via a one-off formula). All three agreed for regular AMS, so
the mismatch only surfaced on HT — the saved preset never reached the
UI and the render fell through to `tray.tray_type`.
Backend now keys via a helper that mirrors frontend `getGlobalTrayId`.
SpoolBuddy's AMS page switches to the shared helper. Regression test
covers regular, HT, and external slot keys.
Critical safety fix. The bed-jog dialog's "Home Z" button sent a bare
`G28 Z` over gcode_line. On Bambu printers where the Z endstop is at
the top (bed moves UP into it — H2C, H2D, H2S, X1 family), `G28 Z`
skips the toolhead-park step that a full `G28` runs first, so the bed
rises at full speed with nothing getting out of the way. The reporter
only escaped damage because the toolhead happened to be parked on the
purge chute.
The /printers/{id}/home-axes endpoint and BambuClient.home_axes() now
always send bare `G28` regardless of the axes argument, triggering the
firmware's safe multi-step routine (park toolhead → home XY → home Z).
The axes argument is kept for API compat but ignored; invalid values
still return 400.
Frontend retitles the button "Auto Home" and updates the dialog copy
in all 7 locales so users aren't surprised when X/Y motion happens
before Z. Parameterized regression test asserts z/xy/all all produce
bare G28.
Users integrating a Shelly plug through an external MQTT broker
(ioBroker, Zigbee2MQTT, HA's MQTT broker, etc.) lost the plug's
power/state/energy readings after every Bambuddy restart. The only
fix was opening Settings → Smart Plugs, renaming the topic to a dummy
value, saving, renaming back, and saving again.
Root cause: three code paths configure an MQTT smart plug's
subscriptions — the startup restore in main.py, the create route,
and the update route — and they had drifted. The create/update
routes used the newer per-type model (mqtt_power_topic /
mqtt_energy_topic / mqtt_state_topic with per-type paths,
multipliers and mqtt_state_on_value) while the startup restore was
still on the legacy single-topic model. Worse, the restore loop
short-circuited on `if plug.mqtt_topic:`, skipping any plug whose
topics were only set in the new per-type fields — exactly the shape
of a Shelly-via-ioBroker config, which publishes power and state on
separate topics. The "rename, save, rename back" workaround routed
through the update endpoint and re-established the subscription the
correct way.
Extracted the topic-resolution + service.subscribe() call into
subscribe_plug_to_mqtt() in mqtt_smart_plug.py and routed all three
paths through it so the schema can't drift again. The helper keeps
the legacy `mqtt_topic` field working as a fallback for all three
data types — matching the behaviour the startup restore used to
have via subscribe()'s internal `effective_*_topic or topic`
collapsing, and matching the change-detection dict already used
during updates.
Regression tests cover: per-type topics restored without a legacy
topic, legacy single-topic backward compat, per-type multipliers
overriding legacy, per-type winning when both are set, the
empty-config skip case, and topic-list de-duplication.
On bare-metal Raspberry Pi OS bookworm / armv7l / Python 3.11, 3MF
files larger than a few megabytes arrived complete via the
virtual-printer FTP server but the copy into data/archives/ was
silently truncated. The archive row was still written, the printer
card looked fine, and the problem only surfaced later when opening
the archive — the subsequent zipfile.ZipFile() in
GET /archives/{id}/plates raised BadZipFile and the UI came up blank
with no thumbnail, plate list, or filament data.
Two things conspired:
1. archive_print() used shutil.copy2, which takes Python's sendfile()
fast path on Linux. On the reporter's kernel/fs combination
sendfile returned a short count on the first call for the upload
sizes hit in practice and the destination ended up truncated.
Small files completed in one syscall and were fine.
2. ThreeMFParser.parse() caught the resulting BadZipFile in a bare
`except Exception: pass`, so the archive pipeline kept going with
empty metadata and left the bad file on disk — nothing in the
logs hinted anything had gone wrong until a support bundle came
in and the "Failed to parse plates" warning fired much later.
The archive copy is now an explicit chunked read/write with fsync —
sendfile is not in the path. After the copy, if the source was a
valid ZIP but the destination isn't, we refuse to create the archive
row, remove only the truncated file (and the archive directory if
empty — archive_dir is created with exist_ok=True so rmtree would be
unsafe if a same-second same-filename collision happened), and log
both sizes at ERROR so the condition is obvious in future support
bundles. The parser's silent catch now logs at WARNING for the same
reason.
All nine archive_print() call sites already check `if archive:` or
`if not archive:`, so returning None for corrupted ZIPs propagates
cleanly without behaviour changes elsewhere.
Regression tests cover single-chunk and multi-chunk copies, mtime
preservation via copystat, overwrite of an existing destination, a
ZIP roundtrip through a multi-megabyte 3MF, the new parser WARNING,
and a truncation sentinel verifying that zipfile.is_zipfile() flips
to False on a half-written ZIP — the exact post-condition
archive_print now trusts.
The wiki scraper silently returned no versions for P2S and X2D, causing
Bambuddy to fall back to the Bambu Lab download page, which still listed
01.01.01.00 as "latest" even though 01.02.00.00 shipped on 2026-04-09.
Two regex mismatches in _fetch_all_versions_from_wiki():
1. Heading anchor ids require an optional dash between version bytes and
date. H2D/X1/H2C/H2S use "h-01020000-20260409"; P2S and X2D publish
"h-0102000020260409" (no dash).
2. The text fallback only matched ASCII parens around release dates, but
P2S, X2D, A1 and A1-mini render dates in full-width parens （YYYYMMDD）
(U+FF08/U+FF09).
Anchor regex now accepts an optional dash; fallback accepts both paren
styles. Added regression tests for both shapes.
OrcaSlicer's Linux BBLNetworkPlugin publishes MQTT payloads with the
C-string null terminator included in the length, so decoded messages
arrived as `{…}\x00`. The strict json.loads() raised JSONDecodeError
and the publish handler silently returned — pushall, get_version, and
project_file were never answered, and the slicer hit its 60 s sync
timeout. Print_queue mode only (proxy mode tunnels MQTT). The b069b521
serial-adaptation fix was correct but ran past this earlier silent
failure.
_handle_publish now strips trailing \x00/whitespace before parsing and
logs the raw payload on any remaining decode failure so future silent
variants are visible in support bundles.
feat(cloud): support China region for token-based login
The /cloud/token endpoint always used the global Bambu API endpoint,
so users with China-region access tokens could not validate their
token. The password login flow already exposes a region selector; this
brings the token flow to parity.
Archive reprints and library-file prints built the MQTT project_file
command with hardcoded project_id="0", subtask_id="0", task_id="0".
Printers key per-job state (including gcode_start_time) on those IDs,
so reprints looked like continuations of the same job and third-party
MQTT observers (OctoEverywhere) reported compounding durations across
repeat replays — a 40 min job reprinted from archive showed ~1h40m,
and a second reprint of the same file showed ~4h. BambuStudio mints
fresh IDs per submission; bambu_mqtt.start_print() now does the same
using an epoch-millisecond timestamp for all three fields. md5 is
deliberately left empty to avoid activating firmware md5-validation
against a digest we can't compute without re-reading the upload.
Added 6 regression tests in TestStartPrintUniqueIdentityFields
covering non-zero IDs, md5 stays empty, uniqueness across successive
submissions, numeric-string format, and blast-radius guard on
unrelated payload fields. Updated CHANGELOG.
The 0.2.3b4 #1003 "fix" POSTed JPEG bytes as multipart form data,
but Obico's /p/ endpoint is declared methods=['GET'] upstream and
reads ?img=URL from the query string. Every POST was 405'd by
Flask's router before any handler ran, which is why the Obico
container logs were silent while Bambuddy kept reporting
"ML API call failed for printer N:" with a blank suffix —
raise_for_status() on the 405 produced an exception whose str()
rendered empty.
Restored the pre-#1003 nonce-URL approach (commit 3e434458):
capture locally with a 20s timeout we control, stash the JPEG
under a single-use 32-byte nonce, hand Obico a
GET /api/v1/obico/cached-frame/{nonce} URL that resolves in
<50ms so its hardcoded 5s read timeout never races RTSP.
Also guards against future silent exceptions: the error format
now falls back to type(exc).__name__ when str(exc) is empty.
Detection also early-returns with an explicit error if
external_url is unset instead of handing Obico a URL it can't
resolve.
The #1003 reverse-proxy scenario (Authelia/Authentik/CF Access
in front of Bambuddy) is addressed by documenting that the
/api/v1/obico/cached-frame/ path must be whitelisted from
external auth at the proxy layer — it is already public on
Bambuddy's side.
Backend: services/obico_detection.py, api/routes/obico.py,
main.py (PUBLIC_API_PATTERNS).
Frontend: FailureDetectionSettings banner + client.ts type +
all 7 locales restored.
Tests: 15 unit + 5 integration tests pass.
Full-mode install booted into an unusable kiosk:
- Chromium opened before uvicorn → "can't connect to localhost"
- After reload, requires_setup=true hijacked /spoolbuddy → /setup
- Touch-only Pi has no keyboard to complete the setup wizard
- Declining auth left the user at / instead of the kiosk
Fixes, bundled:
1. backend/app/cli.py kiosk-bootstrap now, in one DB transaction:
- creates a scoped API key (can_read_status=True, rest false)
- upserts setup_completed=true
so AuthContext never redirects and the kiosk URL loads directly. Users
who want auth can still enable it from the admin UI; the provisioned
key keeps working.
2. install.sh full-mode runs the CLI as the bambuddy service user after
create_bambuddy_service and sed-replaces the CHANGE_ME_AFTER_SETUP
placeholder in spoolbuddy/.env.
3. The generated spoolbuddy-kiosk-launch polls ${backend_url}/health for
up to 60s before exec'ing chromium, so cold boots wait for uvicorn
instead of flashing ERR_CONNECTION_REFUSED.
Standalone mode was unaffected — users supply a real key from their
existing Bambuddy before install.
Full-mode install wrote CHANGE_ME_AFTER_SETUP as SPOOLBUDDY_API_KEY because
no admin exists yet to create a real one. On reboot the kiosk launched with
that placeholder, AuthContext rejected it, and the user hit the Bambuddy
login page instead of the kiosk. Standalone mode was unaffected — users
paste a real key from their existing Bambuddy before install.
Adds backend/app/cli.py with a kiosk-bootstrap subcommand that creates a
scoped APIKey row directly in the DB (can_read_status=True, everything else
false) and prints the full key to stdout. install.sh full-mode runs it as
the bambuddy service user after create_bambuddy_service, captures the key,
and sed-replaces the placeholder in spoolbuddy/.env. Idempotent with
--force for re-installs.
Drops the outdated "create an API key and edit .env" next-step block since
the kiosk is now provisioned automatically.
After hours idle the MQTT connection can degrade so telemetry still
flows but published commands never reach the printer. The existing
dev-mode probe only ran on first connect; this adds tracking for
user-initiated ams_filament_setting commands — two consecutive
unanswered commands (10 s timeout each) trigger force_reconnect.
The ML API previously called back into Bambuddy to fetch snapshots,
which failed behind reverse proxies with external auth (Authelia, etc.).
Now the detection loop captures the JPEG locally and POSTs it directly
as multipart form data — no callback URL, no nonce cache, no
external_url dependency.
The periodic camera cleanup task scans /proc for ffmpeg processes and
kills any not in the active-streams registry. The Obico detection
service's capture_camera_frame_bytes() spawns short-lived ffmpeg for
snapshots but never registered the PID — so cleanup killed it as
"orphaned" mid-capture (SIGKILL, exit -9), producing false errors and
missed detection frames.
Track capture PIDs in _active_capture_pids and exclude them from the
cleanup kill list.
Obico's ML API has a hardcoded 5s read timeout on the URL it fetches, which
our /camera/snapshot regularly exceeds on cold calls (TLS proxy + ffmpeg +
RTSP keyframe wait). The detection loop now captures the JPEG locally with
a 20s timeout we control, stashes the bytes under a single-use 32-byte
nonce, and hands Obico a new /api/v1/obico/cached-frame/{nonce} URL that
returns the cached bytes instantly. The 5s ceiling is no longer a factor.
The nonce is the credential (URL-safe, 256 bits of entropy, single-use,
30s TTL) so the endpoint can be unauthenticated without widening the
camera access surface. Replaces the previous camera-stream-token snapshot
URL approach, which remained vulnerable to the upstream 5s timeout even
when auth was disabled.
Thanks to @fblix for the detailed reproducer with timeout numbers.
The Bambu Lab X2D (launched April 2026, dual-nozzle, enclosed, hardened
steel rod gantry, AMS 2 Pro compatible) identifies itself as internal
model code N6 via SSDP/MQTT, and real serials begin with 20P9. None of
these identifiers existed in Bambuddy's registries, so the camera
service fell back to the chamber-image protocol on port 6000 (X2D
doesn't speak it), firmware-check logged "Unknown printer model: N6",
and the dual-nozzle K-profile paths — gated on the H2D serial prefix
"094" — would have treated X2D as single-nozzle.
Backend:
- Register N6 → X2D across every registry (PRINTER_MODEL_ID_MAP,
PRINTER_MODEL_MAP, STEEL_ROD_MODELS, ETHERNET_MODELS,
CHAMBER_TEMP_SUPPORTED_MODELS, firmware-check API keys + wiki path,
virtual-printer SSDP/product/serial tables, DB vp_model_fixes).
- supports_rtsp(): match the X2 display-name prefix and the N6 internal
code; camera now routes to RTSP on port 322.
- Dual-nozzle serial prefix check in bambu_mqtt.delete_kprofile and
kprofiles.set_kprofile broadened to ("094", "20P9") — X2D now takes
the H2D-style cali_idx in-place edit path.
- is_h2d model gate in bambu_mqtt.start_print extended with "X2D" so
timelapse / bed_leveling / flow_cali / vibration_cali / layer_inspect
are sent as integers and external-spool ams_id 254/255 routing is
preserved (H2D-style deputy-nozzle addressing).
X2D uses hardened steel rods like P2S — it is intentionally placed in
STEEL_ROD_MODELS, not CARBON_ROD_MODELS. A regression-guard test pins
the classification.
Frontend:
- mapModelCode in PrintersPage and SpoolBuddyAmsPage handle N6 and X2D.
- Enclosure-door badge and airduct-mode whitelists include X2D.
- MaintenancePage.getMaintenanceWikiUrl routes X2D to P2S wiki URLs for
steel-rod lubrication, belt tension, cold-pull, and PTFE tube
(exported to enable direct unit testing).
Tests:
- test_printer_models.py: TestX2DModel (10 assertions).
- test_bambu_mqtt.py: X2D in start_print ams_mapping and is_h2d gate;
TestDeleteKProfileDualNozzleDetection across H2D, X2D, P2S, X1C.
- MaintenancePageWikiUrls.test.tsx: 15 assertions covering X2D, P2S
regression, X1C/H2D/A1Mini regression, and model-name normalisation.
Docs:
- README: added X2 series to the supported printers table.
- CHANGELOG: new entry under 0.2.3b4 Fixed.
Credit to @krautech for the report and debug bundle, and to @legend813
for PR #989 which seeded most of the registry changes — rod-type
classification was corrected (steel, not carbon) and the dual-nozzle /
K-profile / is_h2d gaps were added on top.
Second wave of #972 — reproducer on a 37.5 MB BambuStudio print to an A1
showed three stacking root causes when Bambuddy restarts mid-print.
1. Archive start_time lost on container restart. The name-based dedup
cancelled any "printing" archive older than 4h and recreated it with
started_at=now(), so a 13h print that saw a restart 10h in ended up
showing ~1.5h duration. Persist MQTT subtask_id on every archive and
match on that first, regardless of age — same id means same print,
resume in place. Also revives Stale-cancelled rows for users
upgrading mid-print.
2. 3MF FTP search tried non-existent paths for ~48 min. Order was
/cache → /model → /data → /data/Metadata → / with 11×30s retries
each; BambuStudio actually pushes to / on A1, so the real path was
tested last. Reorder to / first, and raise a new FileNotOnPrinterError
sentinel from download_to_file on 550 so with_ftp_retry short-circuits
via non_retry_exceptions. 425 / SSL EOF / connection resets still
retry as before.
3. Cover endpoint and archive flow downloaded the same 36 MB twice and
competed for the printer's single FTP socket, producing 425 errors
that fed cause-2's retry storm. Add an in-memory _threemf_path_cache
keyed on (printer_id, normalized filename); whichever flow fetches
first populates it, the other reuses the file read-only. Eviction
runs on on_print_complete and deletes the temp file.
Backend: 14 new tests across test_bambu_ftp.py and a new
test_subtask_archive_resume.py. Existing suite: 2737 pass. ruff clean,
frontend build clean.
Two-part root cause for missing photos/filament/cost on large prints
(#972). The configured ftp_timeout was only plumbed through as the FTP
socket timeout; the asyncio.wait_for wrapping run_in_executor stayed on
its 60s hardcoded default, so the user's 300s setting never applied.
Worse, asyncio.wait_for cannot cancel run_in_executor threads — after
the 60s outer timeout fired, the executor thread kept running
ftplib.retrbinary and frequently completed the download ~30–60s later,
but by then the async wrapper had returned False. with_ftp_retry kept
re-attempting the same path, each retry truncating the file the zombie
thread had just written, and the archive was ultimately persisted as a
fallback with no 3MF.
download_file_async now accepts timeout at each call site (plumbed from
ftp_timeout) and salvages post-timeout success via an explicit
completion flag the executor thread sets only after download_to_file
returns True. Per-attempt completion dict so a prot_p zombie can't
flip the flag for a later prot_c attempt. A cosmetic // prefix in the
directory-search download path is also fixed by replacing string
concatenation with posixpath.join.
Four attempts at making the printer-card SD badge stable on H2D all failed:
the final straw was powering on an A1 causing every connected H2D to flip to
red simultaneously. Bambu firmware SD signaling is not reliably derivable
from MQTT — the legacy `sdcard` field is sporadic and inconsistently typed,
and home_flag bits 8-9 are cleared on heartbeat pushes regardless of card
state with no clean way to distinguish heartbeats from full status reports.
Remove the badge from the Printers page card and the Printer Info modal,
drop `sdcard` from the frontend PrinterStatus type, and strip all home_flag
derivation and heartbeat-handling code from the MQTT parser.
`state.sdcard` is retained on the backend and populated only from a plain
truthy read of the `sdcard` field, because firmware_update.py uses it as a
precondition before starting firmware installs.
Third follow-up on the H2D SD card badge. The prior 3-strike downgrade still
lost the race: on idle printers, a nearby printer coming online (e.g. an A1
reconnecting) triggered an MQTT activity burst that let idle H2Ds accumulate
≥3 heartbeat home_flag pushes before the next full push_status, flipping every
H2D badge to red at once.
Reworked the derivation:
- the top-level `sdcard` field is authoritative when present (truthy check
handles bool / int / "HAS_SDCARD_NORMAL" string variants)
- home_flag bits 8-9 are only consulted on full push_status payloads
(detected via ≥2 of gcode_state, mc_percent, nozzle_temper, print_type,
stg_cur, ams)
- bare heartbeat pushes carrying home_flag alone no longer affect SD state
Removed the now-dead `_home_flag_seen` latch and 3-strike counter. Tests in
TestSdCardParsing rewritten to cover the new semantics.
H2D sends heartbeat-style home_flag pushes where bits 8-9 are clear
even when a card is inserted, so a single heartbeat flipped the badge
to red until the next full push. Downgrades true->false now require
three consecutive clear reads; upgrades apply immediately.
Settings dump now retains every key from the Settings table and replaces
sensitive values with [REDACTED] instead of dropping the row. New config
flags automatically surface in future bundles without a code change.
Adds integrations.spoolbuddy with per-device firmware, NFC/scale hardware,
calibration, online state and uptime — anonymized (no hostnames, IPs or
device IDs). Both /support/bundle and the bug-report bubble benefit, since
they share _collect_support_info().
Adds a compact "Bed" badge in the printer-card controls row
between print speed and Stop/Pause. Opens a popover with up/down
arrows and a 1 / 10 / 50 mm step selector.
When the Z axis has not been homed since the last print, the
first jog per session opens a Bambu Studio-style modal with
Home Z / Move anyway / Cancel. "Move anyway" bypasses soft
endstops (M211 S0 ... M211 S1) for a single move and is
remembered for the browser session.
Backend:
- POST /printers/{id}/bed-jog?distance=N[&force=bool]
Emits G91 / G1 ZN F600 / G90 (with optional M211 wrap).
Distance validated server-side (non-zero, |N| <= 200 mm).
- POST /printers/{id}/home-axes?axes=z|xy|all
Emits G28 variants.
Both gated behind Permission.PRINTERS_CONTROL.
Frontend:
- New indigo-themed badge + popover in PrintersPage.
- Not-homed confirmation modal with sessionStorage "warned" flag.
- i18n keys under printers.bedJog.* in all 7 locales.
Tests:
- backend/tests/unit/test_bed_jog.py — 13 tests covering
404 / 400 / 500 / success paths for both endpoints, plus
gcode-payload assertions for force on/off.
Docs:
- README feature list, CHANGELOG (0.2.3b4 Unreleased),
printer-control wiki page, website features.html.
surfaced it as "Failed to get image", which Bambuddy reported back as a 400.
Fix: the snapshot endpoint already accepts a reusable camera-stream token (the same mechanism used by <img>-based camera consumers, since browsers can't send auth headers
on image loads). The detection service now appends that token to the URL it gives the ML API. The token is cached on the service, refreshed 5 min before its 60-min expiry,
and is simply ignored when Bambuddy auth is disabled — so no behavior change for users without auth.
Refs #172
Firmware update modal now shows every version from Bambu's wiki release
history, each badged Usable/Unavailable/Installed. Selecting a usable row
— newer or older than current — swaps the release notes and enables
install for that version, so rollback no longer requires hand-flashing.
Wiki scraper tightened to only read heading-anchor ids (h-XXXXXXXX-YYYYMMDD)
instead of any XX.XX.XX.XX substring, eliminating false positives like an
AMS firmware version mentioned in an H2D changelog being listed as H2D
firmware.
Refs #568
Settings → Support → Debug Logging elevated httpx/httpcore to DEBUG,
which makes httpx log every outbound request URL. For Discord and
generic webhook notifications the bearer token is embedded in the URL
path, so users who turned on debug logging to capture a support bundle
were writing their webhook tokens straight into bambuddy.log.
Pin httpx/httpcore to WARNING regardless of the debug toggle. paho.mqtt
still honours debug. Users who enabled debug logging while notifications
were sending must rotate any exposed Discord/webhook URLs — the token
is the path, so the whole URL has to be regenerated in the provider UI.
Parse `sdcard` from `home_flag` bits 8-9 (HAS_SDCARD_NORMAL /
HAS_SDCARD_ABNORMAL) when available and fall back to a type-tolerant
truthy check on the top-level `sdcard` field. Firmware ships that
field inconsistently (bool, int `1`, or string `"HAS_SDCARD_NORMAL"`),
so the previous `is True` identity check flipped the badge to red on
every report that carried a non-bool value.
Surface four Home Assistant-style controls on the Printers page card:
- SD Card badge in the top status row (green / red, icon-only).
- Enclosure Door badge in the top status row (green / yellow, icon-only).
Detection per printer family — X1/X1C/X1E read home_flag bit 23, all
others read top-level `stat` (hex string) bit 23 — so X1 firmware that
does not flip stat bit 23 stops false-triggering "open". WebSocket
status-change dedup key now includes door_open so toggling the door
alone publishes a push, no 30s REST-poll wait.
- Airduct Mode badge beside the speed control (cooling / heating)
for P2S/H2D/H2C/H2S; one-click dropdown calls the existing
set_airduct MQTT command via a new POST /printers/{id}/airduct-mode
route.
- Force Refresh entry in the kebab menu — calls the existing
/printers/{id}/refresh-status endpoint to request a pushall snapshot
without forcing a reconnect.
Tests: door-open parsing (X1 home_flag, non-X1 stat, ignore mismatched
source, invalid hex) and airduct route (validation, not-connected,
success, failure).
Adds a Failure Detection tab under Settings that wires Bambuddy to a
self-hosted Obico ml_api container — no cloud, no account, no WebSocket.
While a print is running, the detection service periodically hands the
printer's camera snapshot URL to the ML API and smooths scores over
time (30-frame warmup + EWM, alpha=2/13, short/long rolling means) so
one noisy frame can't trigger an action. When the smoothed score
crosses HIGH, the configured action fires exactly once per print:
notify, pause, or pause-and-cut-power (via linked smart plugs).
- Backend: new obico_detection + obico_smoothing + obico_actions
services, /obico/status and /obico/test-connection routes
(SETTINGS_READ / SETTINGS_UPDATE), six obico_* AppSettings fields
with validators for sensitivity/action/enabled_printers.
- Frontend: FailureDetectionSettings component (enable, ML URL + test,
sensitivity, action, poll interval, per-printer monitor list, live
status + detection history), new sidebar tab with service-active
bullet, toast on save.
- Tests: 17 detection unit tests + 15 smoothing unit tests + 4
frontend component tests.
- Docs: README bullet, CHANGELOG entry, wiki page under Analytics,
website features.html entry.
With Auto Off enabled and another job queued, the smart plug cut power when a
print finished and immediately re-powered the printer because the scheduler
saw pending items. The printer booted fresh into IDLE and the next job
auto-dispatched, bypassing the "Clear Plate & Start Next" confirmation.
Root cause: the plate-clear gate lived only in PrinterManager._plate_cleared
(in-memory set) and _is_printer_idle treated IDLE as unconditionally idle. On
power cycle the in-memory flag was lost and the IDLE-on-boot state skipped
the gate entirely.
Fix:
- Replace the in-memory flag with an awaiting_plate_clear column on the
printers table, rehydrated into the PrinterManager at startup.
- Set the flag in on_print_complete for completed/failed prints (not user
cancellations); clear it on ack and on scheduler dispatch.
- _is_printer_idle now short-circuits to not-idle whenever require_plate_clear
is on and the flag is set, regardless of the currently reported state —
so the gate holds through power cycles, Bambuddy restarts, and the printer
booting back into IDLE.
- /printers/{id}/clear-plate no longer requires the printer to report
FINISH/FAILED; it accepts the ack whenever the flag is raised.
- Frontend widgets (PrinterQueueWidget, Layout, BulkPrinterToolbar) gate on
the flag rather than reported state.
Tests: added regression tests for IDLE+awaiting=True (the #961 case) and
full DB round-trip tests for the persistence layer.
The Printer tab AMS popup and spool auto-provisioner resolved color
names from hardcoded tray_id_name tables with a suffix-code fallback —
and suffix codes like "R1" are not globally unique across material
families. A17-R1 (PLA Translucent Cherry Pink) fell through the
fallback and resolved to "Scarlet Red" (A01-R1, PLA Matte), baking
the wrong name into auto-created inventory spools.
The fix removes the hardcoded tables entirely. Backend resolves color
names via the existing color_catalog table by hex; frontend fetches a
compact {hex: name} map once per session via a new
GET /inventory/colors/map endpoint (auth-gated but not on
inventory:read — read-only views need it too) and stores it in a
ColorCatalogProvider context. A useSyncExternalStore hook cascades a
re-render into pages mounted before the fetch completes so they
refresh from HSL-fallback names once the catalog loads.
Existing auto-provisioned spools keep their stored names; only new
provisioning and live display benefit. Co-Authored-By is intentionally
omitted here per project convention — set it via git config if needed.
LDAP auto-provisioning hit a NOT NULL constraint error on upgraded SQLite
installs because the existing migration only ran on PostgreSQL. The SQLite
branch now patches sqlite_master via writable_schema and bumps schema_version
so the change takes effect without a restart. Fresh installs were unaffected.
The Statistics page reported "Gesamt" (All Time) kWh correctly but showed
zero for Today/Week/Month in total-consumption mode. Two bugs drove it:
1. The starting plug counter was kept in an in-memory dict
`_print_energy_start` that was lost on any backend restart mid-print, so
the per-print `energy_kwh` delta silently never got computed. The stats
endpoint's fallback path `SUM(PrintArchive.energy_kwh)` therefore summed
to zero for users running in total mode.
2. Total-consumption mode has no per-print delta by design — it includes
idle/preheat/standby — so the fallback to archive rows was the wrong
strategy even when the data existed.
Fix, in two parts:
- Persist `energy_start_kwh` on the archive row and read it back from a
fresh session at print end. Deletes `_print_energy_start` and its 5
call sites, replacing them with a single `_record_energy_start()` helper.
Per-print tracking is now restart-resilient regardless of tracking mode.
- Add hourly `smart_plug_energy_snapshots` table + `_snapshot_loop()` in
SmartPlugManager. Rewrote the `/archives/stats` energy branch as
`_sum_snapshot_deltas()` which computes per-plug
`max(0, last-in-range - baseline)` where baseline is the latest snapshot
at or before the range start, falling back to the earliest-ever snapshot
and signalling `energy_data_warming_up` when no pre-range baseline
exists (fresh upgrade). MQTT plugs are skipped from snapshots since they
only report "today" and have no lifetime counter.
Frontend: QuickStatsWidget renders an AlertTriangle next to Energy Used /
Energy Cost with a tooltip when `energy_data_warming_up` is true, so the
"low values right after upgrading" situation is explained in-product.
Fully localised across 7 UI languages.
Tests: new backend unit tests cover the snapshot delta arithmetic
(baseline/endpoint, counter reset clamp, multi-plug, warming-up fallback,
endpoint windowing), per-print restart resilience via expunge_all, and the
snapshot task lifecycle (start idempotent, stop cancels). Frontend tests
assert the warning icon appears only when the flag is set and only on the
energy tiles.
Docs: updated `CHANGELOG.md`, `README.md`, wiki `features/energy.md`,
wiki `features/statistics.md`, and website `features.html` with the new
behaviour and warming-up explanation.
OrcaSlicer's "Send job" flow sat on "Synchronizing device information…"
until it gave up, even though FTP upload worked when the user clicked
"Send job anyway". The virtual printer's MQTT server gated all incoming
command handling on `f"device/{self.serial}/request" in topic` — if the
slicer's cached serial for the VP didn't exactly equal the VP's computed
self.serial (model prefix + per-VP serial_suffix), every get_version,
pushall, and project_file publish was silently dropped. Nothing was
logged past the initial "MQTT publish to …" line, so the slicer never
received a push_status or get_version response on its subscribed
device/{serial}/report topic and hit its sync timeout. Responses were
also unconditionally published on device/{self.serial}/report, so even
when the inbound check happened to pass, replies targeted a topic the
slicer wasn't listening on if its serial had drifted.
Both directions are now serial-adaptive:
- `_handle_publish` accepts any authenticated publish on a
`device/*/request` topic and extracts the serial from the topic itself
rather than comparing against self.serial.
- A per-connection `_client_serials` dict tracks the serial the slicer
actually uses, populated from the first SUBSCRIBE or PUBLISH seen on
each connection and cleared on disconnect/stop.
- `_send_status_report`, `_send_version_response`, `_send_print_response`
now take an optional `serial` parameter (defaulting to self.serial)
so every outgoing publish — including the periodic 1-second status
push — targets the topic the slicer subscribed to.
- The version response's embedded `module[].sn` fields now also carry
the client's serial so the payload is internally consistent with the
topic.
- When the client's serial differs from self.serial an INFO log records
the adaptation so it's visible in future support bundles.
The working case (slicer's cached serial equals self.serial, as in my
own H2D-1 Proxy setup) is bit-for-bit identical to the old behavior —
the new check is strictly more permissive and only affects cases the
old code silently dropped.
Regression tests cover:
- `_extract_serial_from_topic` valid/invalid topic shapes
- mismatched-serial publish → handler runs, response topic and sn field
both use the client's serial
- non-`/request` topics → still rejected
- pushall → status_report routed to the client's subscribed topic
- `_client_serials` cleared on stop()
detect_current_branch() was reading .git/HEAD from settings.base_dir,
which points at the data volume (DATA_DIR=/app/data in Docker) and
never contains .git. The repo is at /app, so the lookup always failed
and the code fell through to the GIT_BRANCH env-var → "main" fallback.
The SpoolBuddy device was therefore checking out `main` regardless of
which branch Bambuddy itself was running.
The old subprocess-based implementation had the same bug but it was
masked: the stock Docker image has no `git` binary, so `git rev-parse`
raised FileNotFoundError, the except clause swallowed it, and the
fallback kicked in. Swapping to filesystem reads exposed the wrong
lookup path.
Add a module-level _APP_DIR constant (parents[3] of the module file,
same depth as config.py uses for its own _app_dir) and read `.git/HEAD`
from there. A regression test plants a decoy .git in the data dir and
asserts we still pick up the real one from the app root.
Per your NO GIT WRITES rule, nothing is staged or committed.
Follow-up to the asyncssh migration. asyncssh.connect() internally
calls getpass.getuser() for ~/.ssh/config host matching, regardless
of the explicit `username=` passed for the remote login. Under an
arbitrary Docker PUID with no /etc/passwd entry, getpass.getuser()
raises "No username set in the environment" (OSError in Python 3.13+,
previously a bare KeyError).
Fix: set LOGNAME=bambuddy, USER=bambuddy, HOME=/app in the Dockerfile.
getpass.getuser() tries env vars before pwd.getpwuid(), so the lookup
never touches the passwd database and works for any PUID the operator
picks — no helper code, no image rebuild for different UIDs.
Also pass config=[] to asyncssh.connect() so it does not try to load
~/.ssh/config (whose default path needs a resolvable home directory).
An earlier draft of this fix added a Python helper that caught the
KeyError and injected LOGNAME at module import. That was both more
code than needed and broken on Python 3.13, which wraps the KeyError
in an OSError the helper didn't catch — so the module import itself
crashed, producing a 500 on /spoolbuddy/devices/{id}/update. Reverted
in favour of the one-line ENV fix.
Follow-up to the previous commit that swapped the `ssh`/`ssh-keygen`
subprocesses for asyncssh. asyncssh.connect() internally calls
getpass.getuser() to resolve the *local* username for ~/.ssh/config
host matching, regardless of the explicit `username=` we pass for the
remote login. Under an arbitrary Docker PUID with no /etc/passwd
entry, getpass.getuser() tries LOGNAME/USER/LNAME/USERNAME (all unset
in python:3.13-slim) and falls back to pwd.getpwuid(), which raises
KeyError. asyncssh rewraps that as "Unknown local username: set one
of LOGNAME, USER, LNAME, or USERNAME in the environment" — which
surfaced in the UI as "ssh connection failed: no username set in the
environment".
Fix is two-part:
- _ensure_local_username_env() runs at module import. If getpass
.getuser() already works, or any of LOGNAME/USER/LNAME/USERNAME is
set, it is a no-op. Otherwise it sets LOGNAME=bambuddy so asyncssh
can proceed. Native installs are untouched.
- asyncssh.connect() is now called with config=[] to skip the
default ~/.ssh/config load, which relies on a resolvable home
directory that may not exist under arbitrary Docker PUIDs.
Three new unit tests cover the env-var fallback, including the case
where the operator has set USER but the passwd lookup still fails.
Commit 67749565 eliminated ssh-keygen from the SpoolBuddy remote-update
flow, but the update path still shelled out to the OpenSSH `ssh` client
for every command. Like ssh-keygen, the `ssh` binary calls
getpwuid(getuid()) during startup and aborts with "No user exists for
uid <N>" when the container runs under an arbitrary PUID that isn't in
/etc/passwd (python:3.13-slim only ships a root entry, so any
`user: "1000:1000"` compose setup trips the same error).
detect_current_branch() had a related problem: when the git repo is
bind-mounted into the container, .git exists inside Docker, so the code
tried to run `git rev-parse`. Git isn't in the image, so the subprocess
silently fell back to the GIT_BRANCH env var — and if git ever were
added, it could hit the same getpwuid trap.
The entire update path is now subprocess-free:
- _run_ssh_command uses asyncssh (pure-Python, built on the already
installed cryptography library). Connection errors map to rc=255 to
match `ssh`'s convention; asyncio.timeout handles the timeout path.
- detect_current_branch reads .git/HEAD directly (handling git-worktree
`gitdir:` pointer files too), keeping the same GIT_BRANCH → "main"
fallback chain.
- shutil and the inline `import subprocess` are gone from the module.
Regression tests assert that neither keypair creation, branch
detection, nor command execution spawns any subprocess. Native installs
are unaffected.
maziggy