maziggy/bambuddy v1.3.0 on GitHub

Bambuddy 1.3.0

What this is

The largest Bambuddy release so far, on top of 1.2.5.7: 40 new features, 11 changes, 320 fixes and three dependency advisories closed. 24 community PRs went in, and 83 people are credited.

The headline is archives that hold every print of a file. An archive used to show only its latest print; it now opens with each plate, every print of it, and each print's own photos, timelapse, note and outcome, and the calendar, Print Log and filters work per print. Reprinting a file from the File Manager adds a run to its archive instead of a duplicate. Groups can be limited to some printers, and jobs can wait for staff review before they print, for shared and school farms. Failure detection gets a second provider, OctoEverywhere, next to Obico. Notifications gain Notify! with iOS Live Activities, and Gotify. The File Manager can move folders, and users who see only their own files now see only their own folders. Manyfold joins MakerWorld under Model Sources. Printer cameras open from any page. The interface is available in Polish, and downloads about 70% less on each new version.

Most of the 320 fixes come from a review of the whole code base: queue and dispatch, inventory and Spoolman charges, archives, costs, permissions and the database layer on both SQLite and PostgreSQL. Read Before you update below: this release changes more defaults and API answers than usual.

New tables and columns are added automatically on both SQLite and PostgreSQL. Some API routes and settings fields were removed; see below.

If you are coming from 1.2.5.6 or earlier, read the 1.2.5.7 release notes too. If you are coming from 1.2.5 or earlier, read the 1.2.5 notes first; all of their upgrade callouts apply to you as well.

Before you update

These change behaviour you may rely on. Most need no action unless they apply to your setup.

The plate now moves after a print for the finish photo (#3240). Restore plate for finish photo is on by default and now really moves the bed. Turn it off under Settings > General to keep the old behaviour.
Queued jobs wait instead of printing on the wrong filament (#2799, #3280). A job whose filament can't be matched to a loaded tray is held for a manual start; Start offers Print Anyway for the old behaviour. With Force color match, a job waits for a spool of the exact colour and variant, and two slots asking for the same filament need two spools of it. A stored AMS mapping that doesn't fit the printer is recomputed.
A printer that refuses the upload no longer fails the job (#3409). SD card, USB and access-code refusals keep the job queued and retry the printer with a growing wait. They send Job Waiting instead of Job Failed, and "Any " jobs go back to the pool.
Groups can be limited to some printers (#1727). API keys, camera stream links, Cam Wall and overlay tokens now reach only the printers of whoever created them, and a key limited to some printers can't queue to "Any ". Nothing changes until an admin limits a group.
New permissions are granted once on upgrade, so nothing changes until you remove them. Print Without Review (#1620) goes to every group that could queue, start or run jobs. The Manyfold permissions (#1471) go to groups with the matching MakerWorld permissions. Permissions taken away from a group no longer come back on every restart (#3238).
Users who see only their own library files now see only their own folders (#3201). On upgrade, a folder whose files all belong to one user becomes theirs; every other existing folder is shared with everyone, so it stays visible.
Archives count prints, not files. Reprinting a File Manager or project file from the queue adds a run to its archive instead of a new archive, so scripts that count archives per print will see fewer. Aborted prints now count as failed. The CSV/Excel export has one row per print with new columns; spreadsheets built on the old layout need checking.
Costs can change for past prints. Energy is now costed per hour at that hour's price, and a price change no longer re-costs history, so Total consumption figures can differ (#1251). Recalculate Costs now reprices past usage from current spool prices (#3414). Cancelled prints are charged each filament's slicer weight times the share fed (#3317). Running prints are priced from their spools at start and marked ~ until they finish (#3261).
Obico scores like Obico (#3459). Pause print and Pause and cut power need a clearly stronger signal than before, Notify only fires at Obico's alert level, and a resumed print is not paused again.
Inventory slot assignments. Spools on an AMS that stays unplugged for more than two minutes are unassigned; RFID spools come back on their own (#3356). Spoolman mode no longer shows built-in inventory slot assignments (#3430). Reading a spool's tag in one slot removes it from any other slot (#3427). Archiving a spool unassigns it.
The sidebar's MakerWorld entry is now Model Sources (#1471). Old /makerworld links still work. Camera tokens and the Streaming Overlay builder moved from API Keys to the new Camera settings tab; old links redirect.
Streaming overlay (#3185). The Classic design no longer adds its own "Printer is idle" / "Printer offline" line. A Classic URL whose show= list leaves out status shows no idle or offline text; add status to get it back.
Password Reset email template. {password} and {login_url} are gone and the reset link is appended automatically. An untouched template is updated on upgrade; an edited one needs rewording for a link.
Built-in calibration on an A1, A1 Mini, P1 or A2L is no longer archived as a print. The printer kill switch still stops it on these models; switch the kill switch off to calibrate.
API, webhook and MQTT changes.
- Removed: POST /printers/test, POST /printers/{id}/debug/simulate-print-complete, GET /printers/{id}/runtime-debug, POST /spoolbuddy/devices/{id}/update-status and the global virtual printer fields of GET /settings and /settings/virtual-printer. Use PUT /virtual-printers/{id}.
- null for a non-nullable field on PATCH/PUT leaves it unchanged instead of answering 500.
- Bambu Cloud failures answer 502, and an API key that can't be checked answers 503 instead of "Invalid API key". A K-profile read the printer doesn't answer returns 504 instead of an empty list (#3404).
- Queue target_model is matched case-insensitively and stored in the printers' spelling, and batches can have status failed.
- Archive and statistics date filters take a tz parameter; without it, dates are days in Bambuddy's own TZ instead of UTC.
- MQTT relay: smart_plugs/energy is published once a minute, maintenance/alert carries real values and a unit field, and there are new offline, online and filament/low topics.
Databases. PostgreSQL 13 or newer is required (Bambuddy never started on 12 or older). On SQLite, the database is converted once at the first start so ids are never reused, deletes cascade like PostgreSQL, and old leftovers are repaired. The SQLite pool shrinks from 20 + 200 to 10 + 90 connections (#2883); a large SQLite farm may need DB_MAX_OVERFLOW or PostgreSQL.
Native installs need Python 3.11 or newer, and Python dependencies are now pinned in requirements.lock, which Docker, the installers and the update scripts install. The update scripts stop on an older Python and link to the migration steps. Docker and Windows ship Python 3.13.
Update the slicer sidecar image for the new filament lists, preset fixes and renamed printer presets (#3234, #3250). Some plates get a different automatic filament pick, for example Generic PA instead of Bambu Support For PA/PET.
Update the SpoolBuddy daemon on the kiosk (Settings > Updates) for the spool ID fix (#984). In Spoolman mode, the spool ID replaces the tag UID as the spool's tag.

How to update

Docker

docker compose pull
docker compose up -d

Native install - recommended path

sudo BRANCH=main /opt/bambuddy/install/update.sh

Native install - manual path

sudo systemctl stop bambuddy
cd /opt/bambuddy
sudo -u bambuddy git fetch --prune --tags --force origin
sudo -u bambuddy git checkout main
sudo -u bambuddy git reset --hard origin/main
sudo /opt/bambuddy/venv/bin/pip install -r requirements.lock
cd frontend && sudo npm i
sudo systemctl start bambuddy

Windows install

Download bambuddy-1.3.0-windows-x64-setup.exe from this release page (or the unversioned bambuddy-windows-x64-setup.exe alias). Existing Windows installs upgrade in place via the in-app Install Update flow.

New

Print outcome and archives:

Each archive opens with all its prints - a click on an archive card or its name opens the archive at /archives/<id>: the file, each plate with its estimate, photos and timelapse, and a table of every print of it. Each print has its own photos, video, note and outcome, with Reprint this print. The calendar, Print Log and the printer and Failed Prints filters now work per print.
Wear cost per printing hour (#694, requested by @wreuel) - an optional rate under Edit printer adds the printer's own wear to each print's cost, failed and cancelled prints included. It shows on the archive card, in a Wear Cost column in the Print Log, in Quick Stats, projects and batch orders; billing stays filament only.

File Manager:

Move a folder (#3305, requested and contributed by @rojosinalma in #3306) - Move folder... in a folder's menu, or drag and drop, moves a folder to the root or into another folder. External folders can't be moved, and nothing can be moved into one.
Users who see only their own files now see only their own folders (#3201, requested by @Minebuddy, reported by @lonix) - folders now have an owner, and an admin can mark a folder Share with everyone. Uploads and imports go only into your own folders, shared ones and the top level, and folder counts no longer reveal other users' files. See Before you update.
Browse and import from your Manyfold library (#1471, requested by @hibikipr) - the MakerWorld sidebar entry is now Model Sources, with a new Manyfold tab to search models and import their 3MF, STL or STEP files, one at a time or in bulk. Bambuddy only reads from Manyfold, and two new permissions control the tab.

Inventory and spools:

Custom label sheets and saved label templates (#3296, requested and contributed by @koshilki in #3385) - Print spool labels gets a Custom sheet... entry with page, label size, columns, rows, margins and gaps, drawn to scale. Custom sheets can be saved under a name and shared by everyone on the install, in both inventory modes.
Add a scanned Bambu spool in SpoolBuddy in one tap (#2685, requested by @BenBBSLA, contributed by @hibikipr in #3389 and #3390) - an unknown Bambu Lab spool on the reader shows the material, colour and weight from its RFID tag, and Quick Add creates the spool with them, tray UUID and tare included.
Assign a spool by scanning its QR label (#1574, requested and contributed by @samedyuksel in #1606) - Assign by QR on the Inventory page and in SpoolBuddy puts the scanned spool into an AMS slot or a storage location, and the Assign Spool dialog gets Scan a spool QR. The live camera view needs HTTPS; over HTTP, Take photo uses the phone's camera.
New spools start with their product's presets and K profiles (#1326, requested by @mithkr) - tick Save as default for new ... spools on a spool's Printers tab, and every new spool of that product starts with its presets and K profiles, however it is added. Settings > Filament > Profile defaults lists them and can apply them to existing spools.
Spoolman mode uses Spoolman 0.27's native tags (#3168, requested and contributed by @Niko11111 in #3252) - on Spoolman 0.27 or later a spool can hold several tags, scans look the tag up directly and are faster, and Settings > Filament can copy existing tags over. Older Spoolman versions behave as before.
Each spool remembers when it was last dried (#2863, requested by @USER19464727) - a finished AMS drying run marks every spool in that AMS with the date, temperature and duration, shown in a Last dried column and on the spool card. It can also be set by hand for an external dryer.

Queue and printing:

Time to the next pause, and a heads-up before it (#2984, requested by @apizz) - the printer card shows the time until the next pause added in the slicer, e.g. "20m to pause - 20:27". A new Pause Coming Up notification is sent a set number of minutes before, off until turned on.
Larger queue thumbnails on hover (#3313, requested by @humplemann) - hovering a job's thumbnail on a Queue card shows it at 256x256, as in History (now also 256x256).
The queue says when a job printed in a colour the file didn't ask for (#3279, requested by @Lioncat55) - an amber line per slot on the queue row and in History, e.g. "Slot 1 printed in Black, the file asks for Silver". Turn on Force color match to have the queue wait instead.
Jobs can wait for staff review before they print (#1620, requested by @ImNanou) - users without the new Print Without Review permission can still queue, but their jobs wait with a Waiting for review badge until a reviewer starts them. It covers every way a job is queued, including API keys and the webhook.

Slicing and profiles:

Slice dialog: sorted presets, search, stock presets listed once (#3386, contributed by @pleasantone in #3393) - preset groups are sorted, every preset dropdown has a search, and a Bambu stock preset that both Bambu Cloud and the bundled Standard profiles list appears once, as the Bambu Cloud copy.
OrcaSlicer's third-party filaments and every brand-folder filament in the Slice dialog (#3234, requested by @nvx) - Elegoo, PolyLite, COEX, SUNLU and many more are now listed, the automatic pick prefers Bambu's own presets and real filament over support material, and custom presets based on brand or P1P filaments now slice with all their inherited settings. Needs the updated slicer sidecar image.
Show only printers that are online and spools that are loaded (#3172, requested by @wolfrage76) - two checkboxes in the Slice dialog narrow the printer and filament lists to what is connected and loaded right now, and a Pick button per filament row opens the loaded spools laid out like the AMS. Off until you turn them on.

Printers, AMS and calibration:

P2S printers show a P2S picture (#3366, requested and contributed by @andy-sorge in #3369) - instead of the P1S picture.
A page for managing printer locations (#2962, requested and contributed by @SVAN19 in #2995) - create, rename, colour and delete locations and move printers between them, one or many at a time. Renaming a location also updates queued jobs that target it.

Notifications and camera:

Notify! notification provider with iOS Live Activities and Lock Screen widgets (#2919, requested and contributed by @simplytoast1 in #3265) - Bambuddy talks to Notify! directly, without the companion container: alerts with photos, a Lock Screen tile per running print with a countdown and problems, and Lock Screen widgets per printer. See Before you update.
Gotify notification provider (#2743, requested by @MrCrunshy) - notifications to your own Gotify server, with a priority per event, and the camera photo when External URL is set.
Printer cameras from any page (#2737, requested by @apizz) - camera overlays stay open as you move around Bambuddy. A camera button in the sidebar, or the C key, opens one, and an overlay can pop out into its own window and back.
USB cameras are picked from a list - with USB Camera (V4L2), Settings > Camera lists the video devices the server sees by name and path, with a refresh button. Enter path manually still works.
Resize Cam Wall tiles with S / M / L / XL (#2735, requested by @cmerkle) - from one tile across the full width to five per row, remembered separately from the card size. A kiosk /camwall URL can set it with size=.
Chamber light on whenever the camera is used (#1655, requested by @dadino) - Settings → Camera can turn the light on for the live view, camera wall, overlay and notification photos, for all or selected printers, and off again 15 seconds after the last use. Off by default; a light you switched on yourself is left alone.

Failure detection:

OctoEverywhere AI failure detection (#3169, requested and contributed by @QuinnDamerell in #3428) - a second provider next to self-hosted Obico, using OctoEverywhere's cloud AI with a free Gadget API key and no ML container to run. Pick it under Settings > Failure Detection > Provider; it notifies, pauses or powers off as Obico does.
Failure detection can send Obico a different Bambuddy address (#3235, requested and contributed by @chrisbennight in #3255) - an optional Bambuddy Internal URL under Settings > Failure Detection, used only for the ML server's snapshots, for when External URL sits behind a login. Empty keeps using External URL.

Streaming overlay:

Your own logo and progress colours, and a see-through Version 2 background (#3104, requested and contributed by @adamspicedev in #3208) - upload a PNG or WebP logo, set a from/to colour for the progress bar, and fade Version 2's backgrounds so another OBS source shows through.
Create an overlay token in the URL builder, or import an existing overlay URL (#3193, requested and contributed by @adamspicedev in #3207) - Create overlay token puts a new token straight into the URL, and Existing overlay URL fills in every setting from a URL already in OBS.
Portrait overlay URLs for Classic and Version 2 (#3185, requested and contributed by @adamspicedev in #3209) - a new Layout choice of Landscape, Portrait or Both, each with its own URL, OBS size and preview. See Before you update.

Accounts, permissions and integrations:

A count on an app's sidebar link - an app in the sidebar can show a number on its link, as Bambuddy Orders does for its To do, using an API key with Send notifications. A count not updated for ten minutes disappears.
The MQTT relay publishes printer offline/online and low filament - new topics printers/{serial}/offline, printers/{serial}/online and printers/{serial}/filament/low for Home Assistant, Node-RED and other subscribers. A short disconnect sends neither offline nor online.
Groups can limit their members to some printers (#1727, requested by @McGagnor) - under Settings → Authentication → Printer access you limit a group to single printers or whole locations, and its members see and control only those printers everywhere in Bambuddy. Nothing changes until a group is limited; admins always see every printer. See Before you update.
Electricity price from a Home Assistant sensor, costed hour by hour (#1251, requested by @Roving-Ronin) - for tariffs that change during the day, each hour's energy is now costed at that hour's price, and a new price no longer re-costs past prints or months. See Before you update.

Interface and languages:

Polish (#3303, requested by @lukg20) - the interface is available in Polish, under Settings > General > Language.
Remaining time instead of percentage in the browser tab (#3266, requested and contributed by @bgrr74 in #3283) - Tab display under Settings > Appearance switches the tab title to the time remaining, e.g. 1h 26m - Bambuddy. Percentage stays the default.
Dominican Peso (DOP) as a currency (#3388, requested by @rogerabm) - under Settings > General > Cost Tracking.
Egyptian Pound (EGP) as a currency (#3334, requested by @aammash) - under Settings > General > Cost Tracking.

Install, platform and database:

DB_POOL_PRE_PING to turn off the PostgreSQL connection check (#3396, requested and contributed by @oh-forf in #3413) - DB_POOL_PRE_PING=false skips the check before each pooled session and roughly halves the rolled-back transactions monitoring alerts on. The default stays on; turn it off only for a database on the same host or LAN.

Changed

A file printed again adds a run to its archive instead of a duplicate - printing a File Manager or project file from the queue again now joins its existing archive as a new run in the Print Log, with a Plate column for multi-plate files. Prints from the slicer, the printer or uploads are unchanged.
The archive export has one row per print - the CSV and Excel export now lists each print as a row, with new Print ID, Printer, Plate, Outcome and Duration (s) columns. See Before you update.
Unused API routes and settings removed - POST /printers/test, two printer debug routes, POST /spoolbuddy/devices/{id}/update-status and the old global virtual printer settings are gone. See Before you update.
A grouping that ended in failures is marked failed - a batch whose jobs all failed or were cancelled, with nothing printed, now shows as Failed with its own filter instead of staying active forever. Orders are unchanged.
Deleting a custom maintenance type says it applies to every printer - the confirmation now warns that the type and its history are removed from all printers.
The web interface downloads about 70% less on each new version (#3175, requested by @mason9468) - scripts and styles are now sent compressed and cached until an update brings new ones; the main script drops from 11 MB to 3 MB.
The web interface opens faster: pages and languages load when needed (#3175, requested by @mason9468) - each page and each language now loads on first use, so the login page needs 1.2 MB instead of 11 MB. A tab left open across an update reloads once onto the new version.
Notification providers are listed alphabetically - in the Add Notification Provider dialog, in the app's language.
The Streaming Overlay builder remembers your choices - fields, printer, layout, artwork, colours and the rest are kept in the browser across reloads, with a Reset choices link. The token is still never stored.
The camera settings stay short on large farms - External Cameras now lists only printers that have one, with Add external camera to pick a printer and a bin icon to switch back to the built-in camera.
Camera settings have their own tab, and the settings menu is sorted alphabetically - a new Camera tab holds External Cameras, Camera API Tokens and the Streaming Overlay builder; old links open it. General stays first, the other tabs follow alphabetically.

Fixes

AMS slots, K profiles and drying:

A spool loaded into a slot after a runout and assigned mid-print stayed invisible on the printer card until the print ended (#3434, reported by @0xCyberFortress). The card now shows it with an amber marker until the printer takes it.
AMS Filament Backup did not pool two rolls of the same filament when one was added by RFID and the other from the preset picker, so Start refused prints the printer could finish with Auto Refill (#1824, reported by @lightmaster). Pooling now follows what the printer reports for each tray.
A printer that did not answer the K-profile read looked like a printer with no K profiles, and a Git restore then sent every backed-up profile again as a duplicate (#3404, reported by @Spegeli). The page now says the profiles could not be read, and a restore skips that nozzle.
SUNLU filaments and about thirty P1/X1 presets put the wrong filament into an AMS slot, a Bambu PLA Silk spool went in as PLA Matte when signed in to Bambu Cloud, and Orca Cloud profiles based on SUNLU went in as Generic PLA (#3387, reported by @Spegeli). SUNLU and seven newer Bambu filaments, such as PLA Tough+ and PETG Matte, are now in the catalogue.
Configuring a slot through the API with a setting id such as GFSL99 left the slot without colour or type, and auto-unlink dropped its spool (#3444, reported by @Novastorm4332).
A colour set on the printer's touchscreen got a general name such as "Blue" when its hex was a step off Bambu's, as with Blue Grey (#3279, reported by @Lioncat55). It is now named after the closest Bambu Lab colour in the slot's own range.
Reset Slot cleared an AMS slot on a single click, even mid-print. It now asks first and is unavailable while a print runs.
On an H2D, a manual filament load the printer refused could later show the wrong active tray. The refused target is now forgotten after two minutes.
A printer refusing the K profile re-applied after an RFID read was not logged, which made wrong pressure-advance prints hard to diagnose.
Opening several pages at once could fetch the Bambu Cloud filament names for K profiles many times over. They now share one fetch.
Configure Slot reported a typed K value as applied when it could not be sent. It now warns that the filament was set but the K value was not.
An H2D could keep pointing at the old slot after a filament change, on the printer card and in usage tracking.
With "Wait for drying to complete" on, queue drying never dried an AMS ahead of a scheduled print. A printer whose job is scheduled for later is now dried ahead of it.
A Silk spool without a slicer preset could take the Matte preset already in the slot. The finish must now match too, otherwise the slot gets the generic preset for that material and finish.
Some Bambu Cloud presets showed their raw id instead of their name in the K profile list.
A K profile picked in Bambu Studio or on the printer's screen was set back to the stored one at the next AMS report. See Before you update.
Saving, deleting or selecting a K profile could change the nozzle size shown for the printer.
A K profile save, batch save or delete the printer never acknowledged was reported as done. It now says the profile was sent but not confirmed.
Stopping drying said "Drying stopped" as soon as the command went out. It now says "Stop command sent".
The "P1 / P2" print labels on AMS slots rarely appeared during a print. They now show, also after a Bambuddy restart.
"Rotate spool during drying" stayed available on some firmware with filament in the feed tube, and the printer refused the drying start.
On two-nozzle printers with several AMS, the run-out hint often could not name the slot.
A regular AMS reporting a single tray was shown as an AMS-HT.
Load and Unload on an AMS-HT slot were refused or aimed at a unit that does not exist. Loading from the A2L's AMS Lite now also sends the target Bambu Studio sends.
The K profile picker offered and pre-selected profiles measured on the other nozzle flow type on H2-series printers. They now sit under "Other", and every profile shows [HF] or [S].
With "Wait for drying to complete" on, an "Any " job could start on a printer whose AMS was drying, and queue auto-drying never dried for such a job.
The AMS drying button showed as available on a printer that cannot dry until the next live update after a page load.
A PP, PE, PCTG, PLA Silk or PLA High Speed spool without a slicer preset was sent to its AMS slot with no filament ID, and PVA, HIPS and Bambu support filaments got the wrong setting ID (#3273, reported by @sgiffhorn).
AMS Filament Backup showed "?" on the P1S, P1P, A1 and A1 Mini (#3259, reported by @HopNotch). The badge now shows the printer's real setting and follows changes made in the slicer.
H2D, H2D Pro and H2C with several AMS on one extruder briefly marked the wrong tray as loaded on every filament change (#3242, reported by @Thomansky). No wrong spool charge was found, but the wrong tray no longer enters the usage split.

Spools, inventory, Spoolman and SpoolBuddy:

In Spoolman mode, slot assignments kept from the built-in inventory put printer slots on unrelated Spoolman spools that could not be cleared (#3430, reported by @sean3318). Spoolman mode now uses only Spoolman's assignments; the built-in ones are kept for when you switch back.
A spool whose tag was read in one printer's slot stayed assigned to another printer's slot too, and that printer's prints were charged to it (#3427, reported by @Spegeli).
In Spoolman mode, a Bambu spool created by AMS sync lost its filament preset, so on the external spool a Bambu PLA Basic became Generic PLA (#3433, reported by @ojimpo).
Spoolman mode had no per-print usage history, so the spool dialog listed no prints and the stock forecast never reached its Trend rate (#3407, reported by @ojimpo). The forecast's fallback rate in both modes also counted shelf time as use, and now counts only the time a spool was in use.
In Spoolman mode, the stock forecast used a built-in spool's history when the ids matched, so a spool using 2.5 g a day could show 46 g a day (#3406, reported by @ojimpo).
The stock forecast left out used-up spools once they were archived, so a SKU in weekly use could show 0.2 g a day and the Reorder and Stock Break alerts came late or never (#3399, reported by @ojimpo). Spools archived in the last 90 days now count towards the rate.
The stock forecast named a SKU's colour twice, as in "PLA Basic Black Black" (#3400, reported by @ojimpo).
In Spoolman mode, an external-spool print on an X2D or H2D was charged to the wrong holder, or not at all from the right holder (#3405, reported by @juandelarossa).
A Spoolman slot change made from another tab or client, such as a filament scale using the API, did not reach open pages without a reload (#3402, reported by @Eifel-Joe).
SpoolBuddy sometimes read a Bambu tag on the first scan without its tray UUID (#3377, reported and contributed by @hibikipr in #3378).
SpoolBuddy with a PN5180 reader misread NTAG tags with 7-byte UIDs (#3258, contributed by @Thomansky in #3346). Spools linked with the short form are still found and upgraded on the next scan.
Bambu PLA Silk+ spools added by RFID arrived without a colour name, in both inventory modes. The update files the seeded Silk+ colours correctly and names the affected spools.
A spool replaced mid-print after a runout was charged for the whole print (#3325, reported by @HarryCoff2). The filament is now split between the old and new spool at the layer of the swap.
A print that failed or was cancelled could be charged for more filament than it used, sometimes more than the finished print (#3317, reported by @ojimpo; thanks to @Andrewdlec for the files).
With Spoolman, writing a SpoolBuddy tag that another spool already carried left it on both spools (#3312, reported by @sgiffhorn).
A SpoolBuddy kiosk got stuck on the login page until the next reboot when it started while the database was unavailable (#3360, reported by @sargonas). Pages now show "Can't reach Bambuddy. Retrying..." and continue once it answers.
Searching "#123" in Assign Spool found nothing when that spool's material did not match the slot (#3301). It now shows the spool, or says why it cannot be offered.
After a printer was deleted, spools still showed as loaded in it for up to 30 seconds.
In Spoolman mode, writing a SpoolBuddy sticker to a Bambu spool created from the AMS broke its AMS link, so the next sync created a duplicate spool. Such spools are no longer listed as untagged. See Before you update.
SpoolBuddy's Write Tag warned "color_name not set" for every Spoolman spool. It now warns only when the colour name is really missing.
A SpoolBuddy update reported success when installing its dependencies timed out. A timeout now stops the update with an error, and the time allowed went from 2 to 5 minutes.
A failed Unassign on the printer card showed no message, and the spool simply stayed assigned.
The Inventory "Full" column always read "-" and has been removed.
In Spoolman mode, a print whose 3MF arrived late was charged from the AMS remaining percentage instead of the 3MF.
The Spoolman AMS syncs could replace a spool assigned by hand, and the Sync buttons in Spoolman settings did not update other open pages or the slot's filament preset.
Several Spoolman requests misbehaved: a storage location could be deleted while Spoolman was unreachable, labels for an archived spool failed the batch, an archived spool could be assigned to a slot, and resetting "Total Consumed" reported success when it failed.
A finished print's filament could be charged twice when an AMS update arrived right as the print ended.
Spool weights and slot assignments written at the same moment could undo each other or fail with a server error on PostgreSQL.
Inventory views and printer cards did not show deleted spools and new weights until a refresh, and deleting a printer on SQLite left its spools assigned to it.
Saving the SpoolBuddy backend config said only "Config queued." although it needs a daemon restart. It now says so and offers "Restart Daemon".
A used weight entered for a new spool was not locked, so the first AMS sync could overwrite it. See Before you update.
Every AMS sync and scale weighing stamped "last used" on Spoolman spools. Only print usage sets it now.
On the A2L, filament used from the AMS Lite was never charged to its spool, or went to a regular AMS slot of the same number.
A spool Spoolman created from the AMS for a 250 g refill matched to a 1000 g filament was created as a 1000 g spool showing 75% used.
An unregistered SpoolBuddy that was still running did not register again on its next heartbeat.
In Spoolman mode, a failed print that switched to the backup spool charged everything to the spool that ran out, a second copy of the same archive printing at once was never deducted, and unused filament slots could be charged to the external spool.
Spoolman slot assignments were not checked against what the printer reported back, and Configure AMS Slot opened without the Spoolman spool's slicer preset.
Archiving or deleting a spool left it in its AMS slots, where it was charged the next print's usage, and "Restore" in an expanded group row did nothing. See Before you update.
"Sync Weights from AMS" counted a running print twice when used mid-print and skipped locked spools for good, and SpoolBuddy scale weights could be overwritten by the AMS sync. See Before you update.
SpoolBuddy's Write Tag page could stay on "Writing tag...", scanned tags failed silently when Spoolman was unavailable, and lookups fell back to the built-in inventory when the Spoolman URL was refused.
The filament of a print spanning a Bambuddy restart could be counted twice.
In Spoolman mode, SpoolBuddy could write a hidden built-in spool with the same id to the tag while reporting success, so the Spoolman spool never got its tag. Writing a tag now uses only the inventory in use.
Deleting a spool deleted its usage history on PostgreSQL, and on SQLite the next new spool could take over a deleted spool's number with its prints and consumption. Usage history now stays, prints keep their spool cost, and spool numbers are never reused.
On PostgreSQL databases created before v0.2.4.8, reporting usage to Spoolman failed and archives never got the colour, material and cost of their Spoolman spools. The affected columns are converted on the next start.
SpoolBuddy knew a Bambu Lab spool only by the tag that was scanned, so scanning its other side offered to add it again, and kiosk and AMS spools were not matched to each other (#984, block 9 read from @Keybored02's #1200, measured by @Thomansky and @Sawtaytoes). Update the SpoolBuddy daemon to get the fix; existing spools pick up their spool ID on the next scan.
In Spoolman mode, lifting a spool off SpoolBuddy while Assign Spool or Add to Inventory was open lost its tag.
Inventory deducted only the first filament when two filaments of a print came from the same spool (#3230, reported by @Andrewdlec).

Queue and dispatch:

A printer whose SD card or USB drive refused the file, or that rejected the access code, failed one queued job after another: an H2D with a bad USB stick failed seven in an hour (#3409, reported by @simonkreuchauff-oss). The jobs now stay queued, the printer is retried with a growing wait, "Any " jobs can move to another printer, and a file name no card accepts still fails with a request to rename it.
"Force color match" was never saved on a job for a specific printer, so a part set to black could print in green (#3445, reported by @frnzzle).
A job with "Force color match" printed on any spool of the same material when the colour was missing (#3280, reported by @Lioncat55). It now waits with "Needs PLA #A4AAAC" until that exact colour, and variant, is loaded.
A job queued for "Any" model kept returning to a dual-nozzle printer whose filament sat on the wrong nozzle, or to a printer with one spool for two slots of the same material. Such printers are now passed over, and the job waits with the filament it needs.
Queue prints to a printer without an AMS, such as a P1S on its spool holder, failed with "Failed to get AMS mapping table" (#3356, reported by @dogod3814t). An unplugged AMS also now disappears from the printer card without a restart.
Opening the Queue page on a large print farm could stall Bambuddy for over a minute. History now loads 50 runs at a time, and Clear History is a single request.
The queue's "View in File Manager" icon opened an empty page (#3336, reported by @Magical-Penguin). It now opens the file's folder with the file marked, and the archive icon marks its archive.
A late slicer update through a virtual printer could rewrite a queue job already being sent, so the job and the spool charged could differ from what was printed.
Saving an edited queue item showed "Queue item updated" twice and closed the dialog twice.
The queue timeline and history rows showed plate 1 for every plate of a multi-plate file.
An "Any model" job with a filament override could wait forever for an equivalent filament under another name, preheat heated the bed for the hottest filament of the whole project instead of the plate, and the queue page showed generic colour names.
Cancelling a pipeline run, or stopping a job on an offline printer, could leave a printer preheating or cost centre budget reserved, and a print start the printer refused was taken as accepted for three minutes. A refused start now goes back to the queue at once.
Queue jobs added through the API or webhook could miss their printers: a model in lower case was refused, a webhook reprint ignored the slicer's AMS slots, and a multi-plate file queued without a plate waited for every plate's filament. See Before you update.
A queue job edited while the scheduler was working through the queue could lose the edit or be sent to the wrong printer. It now waits for the next round.
An "Any model" job waiting for a drying printer read "Busy" as if that printer were printing. It now says "(drying)".
The queue bulk edit showed a neutral, English-only message even when nothing was updated. It now shows an error or a warning.
A plate printed only from the external spool turned the AMS back on when another plate of the same file used it, and the printer refused the job (also #3087).
A virtual printer for any printer of a model stored the slicer's AMS on/off with the job, although the slicer could not see the real printer's AMS.
A nozzle mapping on an "Any model" job was sent unchanged to whichever H2C picked it up instead of that printer's rack positions.
Jobs added from the File Manager, webhook, virtual printer, pipeline or an order sent no "Job added" notification or MQTT event, and some ignored the default print options. See Before you update.
The filament override picker merged identical left and right external spools on an H2D or X2D without an AMS.
Prints queued through a virtual printer lost the slicer's nozzle offset calibration, rack nozzles and AMS choices, or took the old printer's nozzle positions to another printer.
The print dialog had several gaps: Print Anyway in a multi-printer print waved through printers that were never checked, an untouched plate dropdown with printer alternatives queued the job with no plate, H2D and H2C matched nozzles from every plate instead of the chosen one, and a sliced 3MF named without .gcode could not be dropped on a printer card or picked in Upload for Print.
Orders and groupings whose last job was cancelled, stopped, cleaned up or failed before reaching the printer stayed active until the next restart.
Cancelling a queued job while it was being sent was overwritten when the upload then failed, sending a failure notification and skipping the next dependent job. A stopped job could also be put back in the queue or leave its file on the printer.
"Only if the previous print succeeded" checked the wrong printer for jobs queued to any printer of a model, so it nearly always let the job through. It now checks the printer it is about to use, takes another of the model if needed, and shows a Resume after failure banner when every free one failed last.
A queued job moved to another printer kept the old printer's AMS mapping, Print Anyway and nozzle choice, so it could print the wrong colour or skip filament checks, and Dispatch remaining of an order pinned every copy to one printer. These are now worked out again for the new printer.
A slicer pipeline run stayed dispatching forever once a copy's queue entry was deleted, and Retry failed could not tell what to retry. Each copy now keeps its final state.
A printing queue job showed the slot AMS Filament Backup had switched away from as Empty. The card now shows the backup spool in use.
A queue job whose file name started or ended with a space stayed "printing" for five minutes after it finished, holding up the next job (#3241, reported by @Thomansky).
An "Any model" job waited for filament on one printer while another idle printer of that model had enough (#3137, reported by @RambachTJ). The queue now picks a printer with enough filament.
A job moved to a location or model could print from an empty external spool on a printer that had the filament in its AMS, stopping with "External filament is missing" (#3239, reported by @hyyz17200).
A job moved to a location or model could print in the wrong colour from the tray number it had on its original printer (#3239, reported by @hyyz17200). Trays are now matched by material and colour on the printer that runs it, and the job is held when nothing fits.
A virtual printer with Save AMS mapping lost the slicer's external-spool pick, so the print could stop with 0700-8012 "Failed to get AMS mapping table" (#3237, reported by @erabti).
One model queued to several printers used the first printer's AMS mapping for all of them and could print in the wrong filament (#2799, reported and contributed by @grolmus in #2803). Each printer is now mapped against its own AMS.
A job whose filament was not loaded was sent anyway and refused by the printer after the whole upload (#2799, reported and contributed by @grolmus in #2803). It now waits for a manual start and the queue row says which filament it needs.

Archives, print log and statistics:

A print that finished while Bambuddy was restarting was archived as aborted and its queue job cancelled (#3310, reported by @Thomansky). It is now closed as completed, or as failed with the printer's error.
Printing one plate of a multi-plate file was measured against the whole file, so the card showed +200 % (#3375, reported by @sargonas). Each print is now compared with its own plate's estimate, on the card, in the archive, in statistics and in notifications.
Reprinting a plate of a multi-plate file from an archive, the outcome dialog or the queue history opened on plate 1 and could print the wrong part (#3395, reported by @Thomansky).
An old video still on the printer's card could be attached to a new print, deleted from the printer and used for its finish photo, when the card could not be read at print start (#3412, reported by @ojimpo).
The second plate or a reprint of an archive lost its finish photo. Photos lost this way are restored once at startup if the file is still on disk.
A late timelapse left two finish photos on an archive.
An archive printed more than once played only the latest print's video. The button now shows how many videos there are and opens the archive when there is more than one.
"Change outcome" in an archive's print list opened a dialog with nothing to choose.
A print started over LAN, from Handy or on the printer was not linked to its archive while it ran, and its print time came out short (#3327, reported by @Andrewdlec).
"Print again" on the printer's screen produced an empty archive when the project's title differed from its file name (#3307, reported by @elizabethburg).
A large 3MF that timed out at print start was downloaded a second time, holding the printer's file connection for up to 5 more minutes (#3308, reported by @elizabethburg).
Statistics' Today, This Week and the other ranges used UTC days, so near midnight they showed the wrong day's prints.
"Purge old archives" could delete by an age that had not been previewed when you clicked straight after changing it.
A stray Stop on an idle printer could relabel a later failed print as cancelled by the user.
A timelapse could go to the wrong print, or none, when the next plate or a reprint started soon after, or when Scan for Timelapse matched another archive's video.
The Archives page showed a timelapse picked, uploaded, converted or deleted by hand only after a reload.
Editing an archive while it was being reprinted rewrote the previous print's result in the print log.
Archives made without their 3MF showed blank colour swatches and got no filament from an AMS-HT. Existing archives are repaired on startup.
An archive showed the nozzle temperature of the project's first filament instead of the one its plate prints.
A print stopped while its printer was offline was missing from the print log, and a printer reporting the same end twice logged the run twice.
A print Bambuddy did not start could be matched to an archive whose name only contained its name, so finishing "Clip" could complete and bill "Cable Clip".
When the 3MF could not be downloaded directly, print start could archive another file whose name merely contained the print's name.
The built-in calibration on A1, A1 Mini, P1 and A2L was archived and announced as a print. See Before you update.
The recovery download for a print's 3MF gave up after 90 seconds, so a large file on a slow SD card never made it. The deadline now follows the file size, up to 15 minutes.
A print stopped by the printer kill switch was archived as failed with no reason. It is now cancelled, with "Stopped by the printer kill switch".
A queue job stopped while its printer was offline stored an untranslated sentence as its failure reason. It now stores "User cancelled", and existing archives are converted on startup.
A re-sliced archive showed as printed and counted as a finished print. It now shows as "Not printed" with an estimated cost.
A reprint of a single-plate archive left the old timelapse behind as an unlinked file.
Edit Archive saved "Cancelled" as "aborted", opened a cancelled archive as "Completed", and allowed setting "Printing" by hand.
An archive left on "printing" after a missed end took over the next print of the same file for two hours.
Print History, usage and thumbnails could describe the wrong plate of a multi-plate file, also after a restart or for prints started outside Bambuddy.
Reprints had several archive problems: a completed reprint still showed the previous run's failure reason, a queued job whose archive was deleted printed with no archive, notification or usage tracking, and a timelapse deleted during MP4 conversion came back afterwards.
Deleting an archive for good failed with a server error on PostgreSQL when an inventory spool had been used for it, and Recalculate costs could add the spool cost of other prints with the same name to an archive.
Archiving a virtual-printer upload into a project that does not exist failed after archiving on PostgreSQL or saved a broken link on SQLite. It is now refused up front with "Project not found".
Deleting a printer could delete the archive of a print starting on it at that moment on PostgreSQL, even with "keep archives" on.
A printer deleted while its print was starting still got an archive, which showed "printing" forever on SQLite, plus a timelapse and tracking.
Starting the next plate of a Send All deleted the previous plate's timelapse (#3275, reported by @dovmesiz). A queued print running over 2 hours also lost its plate, AMS mapping and cost centre data.
A large timelapse could not be attached when it took more than 5 minutes to download (#3272, reported by @dovmesiz).
Statistics counted a multi-material print priced from spools as one combined material, and the Archives material filter missed such prints (#3262, reported by @ojimpo).
A print started from the printer's screen took about six seconds longer to archive, and could not be reattached after a Bambuddy restart (#3009, reported by @bdwilson).

Costs, energy and finance:

"Recalculate Costs" ignored changed spool prices and updated nothing (#3414, reported by @BurgerKerman). It now reprices past usage from the spools' current prices, and the Print Log, statistics, projects and orders follow.
A personal cost centre's budget could not be changed or set to 0 (#3358, reported by @airforceixi).
A failed or cancelled print's card showed the whole file's cost and grams with no hint (#3318, reported by @ojimpo). A tooltip now says it is one full print's cost and points to the Print Log.
A print that failed on its first run left the archive costing only the part it printed, so reprints were estimated and billed from that fraction.
In Spoolman mode, a very short print could end with its estimated cost instead of the real one.
Editing a transaction to "No cost center" kept the old cost centre.
In Spoolman mode, the print dialog's cost estimate, budget check and "g left" came from the built-in inventory instead of the Spoolman spool in the slot.
A Spoolman spool priced only through its filament showed no Cost/kg in the Inventory or on SpoolBuddy, although prints were charged that price. The column now shows it, marked "(filament price)".
A print's filament and cost could be charged with another plate's or another run's AMS mapping.
An archive cost set through the API or by Recalculate costs did not reach the statistics, project totals or batch sums. See Before you update.
An archive whose 3MF arrived during the print kept the default filament price instead of being priced from its spools.
In Spoolman mode, failed or stopped prints were logged with no cost, and queued library files were priced at the default filament cost.
Recalculate Costs and Rescan gave an archive printed three times the cost of all three, editing an archive's weight did not reprice it, and a spool without a price was charged nothing at completion.
A print whose dispatch failed left its cost centre behind, and the next print of that archive was charged to it.
A print without a user kept its budget reservation for good, so the cost centre showed less budget left. Leftover reservations are released at the next start; booked charges don't change.
Editing a smart plug could save values that broke the smart plug list and the plug with server errors until fixed in the database, and could give a printer a second Tasmota plug. Edits are now checked the same way as adding a plug.
A print whose user was deleted while it ran was never charged to its cost centre and its budget reservation stayed active. It is now charged and settled, and ledger entries without a user read "No user".
A running print's cost used the default rate instead of its spools, then changed without notice at completion (#3261, reported by @ojimpo). It is now priced from its spools at start, marked with ~ while it runs.
A smart plug that reports energy in watt-seconds, such as a myStrom Switch, could not be saved because the multiplier was too small (reported by @CLKRUN in #1251).
The Statistics energy total stayed at 0 for Shelly and other REST plugs with only a lifetime counter (#3232, reported and contributed by @rojosinalma in #3233).

Slicing, profiles and 3MF:

An unsliced project saved as .gcode.3mf was treated as a printable file, and printing it to a cost centre failed with an unrelated error (#3435, reported by @Mounting9612). It is now stored as a project and offered for slicing.
Orca Cloud processes were hidden for every printer, and a process made for a user printer preset was refused by the slicer (#3380; contributed by @pleasantone in #3381).
With OrcaSlicer preferred, the Slice dialog could preselect the wrong material for H2C filaments, such as Generic PETG for a PETG-CF slot (#3403, reported by @pleasantone).
Slicing for the Cool Plate SuperTack, or for the Smooth PEI Plate, failed with "Filaments are not compatible with the plate type" (#3359, reported by @Akaviiii). Saved pipelines with either plate now work without being saved again.
Slicing a model coloured with the paint tool could turn several filaments into one colour (#3355, reported by @cartman75).
A slow progress answer from the slicer sidecar was reported as "Slicer sidecar unreachable". It now says the sidecar did not answer in time.
The Slice dialog hid renamed printer profiles while "Only printers that are online" was on (#3172, reported by @gregspatrick).
Slicing failed with a renamed printer preset, saying the printer was not compatible with the process preset (#3250, reported by @bobobo1618). Needs the updated slicer sidecar image.

Printers, MQTT and connection:

When a dropped printer connection came back, commands still waiting for an acknowledgement were sent again, and a 4-hour print was cancelled seconds after the reconnect (#3454, reported and diagnosed by @SaadShaya7). A command sent in the seconds before a drop now has to be clicked again.
On the X2D, the left and right auxiliary fans were swapped on the printer card (#3373, reported by @Spegeli).
An H2D Pro reported no chamber temperature and refused chamber heating, and queued prints got no chamber preheat or airduct switch; the X1E's chamber heater was not offered either (#3335, reported by @Spegeli).
An AMS Lite showed no serial number or firmware version on the A1, A1 Mini and A2L (#3348, reported by @Spegeli).
Firmware could be uploaded to a printer in the middle of a print. Upload is now offered only while the printer is idle.
The MQTT Debug log showed messages already changed by Bambuddy instead of what the printer sent.
The cooling/heating mode switched on the printer showed the old mode on the printer card until the next refresh.
An H2C rack slot's material label disappeared from the printer card until the next refresh.
Time a printer spent offline mid-print was counted as print time, so hours-based maintenance came due early.
Two pages could create the same maintenance item twice, and marking it performed could update the hidden copy. Existing duplicates are merged on upgrade.
The maintenance API answered "performed" in hours for an item set to days.
Skip Objects and the printer connection diagnostic held a database connection while waiting on the printer.
An upload the printer never confirmed was not checked. Its size is now read back and a mismatch is logged.
A skip the printer refused stayed marked as skipped. It is now taken back off the list and logged with the printer's reason.
A nozzle change, firmware update or developer-mode switch reached the printer card only with the 30-second refresh.
A print resumed after a Bambuddy restart could lose the plate height the plate restore uses.
Saving any virtual printer restarted every proxy-mode virtual printer and dropped the slicer's connection. Changing a printer's IP and renaming a virtual printer now restart the right ones.
With the internet down, every firmware check logged a full error for every Bambu Lab page it tried. The outage is now logged once per site as a warning, with a line when it is reachable again.
When Bambu Cloud was unreachable, the cloud preset, device and firmware views answered with a server error, and a pasted token, password or email code was reported as wrong. These now say Bambu Cloud is unreachable.
A printer added after deleting another could take over the deleted printer's running print and its owner, filament tracking, plate-clear prompt, pending notifications, camera picture, Obico verdict and plate detection references, and skipped its own first-connect steps.
The SD-card cleanup could delete a file the printer was printing right then (#3009, reported by @bdwilson).

Notifications and HMS:

Mail from Bambuddy could be quarantined by strict mail servers because it had no Date or Message-ID header, and a sender name with accents or a comma broke the From address (#3376, reported by @BenjaminCs).
The same printer fault could be listed twice when the printer reported it both as a print error and in its HMS list.
The daily digest was lost when it could not be sent, or when its time fell in quiet hours. See Before you update.
A later print could be attributed to, and emailed to, the user of an earlier queued print of the same name.
"Print Stopped" after a printer came back online showed the time it was offline as the print's duration. It now shows the sliced estimate.
The template editor had no variables or preview for the print outcome confirmation, and seven other events had no names.
A notification channel limited to one printer also received other printers' queue and waiting events. See Before you update.
CallMeBot dropped the "+" from the phone number.
A print started outside Bambuddy on a printer with auto-archive off could be reported to the owner of the previous queue job. A queue job whose archive couldn't be found now still reaches its owner with the right duration.
Smart plug Power Alerts were never sent and broke the plug's status request on the printer card. They are now checked in the background, sent once per crossing as Printer Error events.
A printer fault that kept coming and going was notified every time it came back (#3226, reported by @sgiffhorn). It is now notified again only after it has been gone for 10 minutes.

Camera, stream overlay and failure detection:

Obico failure detection added whatever the camera always sees, such as plate texture, to its score, so it acted too early, and it paused a print again after you resumed it (#3459, reported by @jasoncronje). Scoring now matches Obico's, and the pause actions wait for Obico's stronger pause level.
With "Restore plate for finish photo" on, a tall print could drive the plate past the bottom of its travel, as on a 238 mm print on an H2C. The plate now goes back to where the print's end G-code left it.
A camera opened from a printer card or the Cam Wall followed you onto every page (#2737). Only cameras opened from the sidebar's Cameras button now show everywhere.
In the iOS home-screen web app, a camera opened in window mode took over the app with no way back (#3419).
A camera rotated 90 or 270 degrees was cut off or letterboxed in the floating camera window, on Cam Wall tiles and in the stream overlay (#3362, reported by @BurgerKerman).
Stitching a layer timelapse could use gigabytes of memory, so a container with a 2 GiB limit was killed at the end of a long print and lost its timelapse (#3292, reported by @fma965). It now peaks around 300 MiB, and ffmpeg no longer keeps running after an error or cancel.
The timelapse player pushed its controls off screen for a portrait video, and some of its labels were always in English (#3281, reported by @BurgerKerman).
Saving in the timelapse editor could fail on long videos, and the saved music was always at full volume. Saving now runs in the background with progress and Cancel.
Escape could be ignored right as the Build Plate Check result appeared, leaving the chamber light on.
With the camera light setting on, the plate check could switch the chamber light off under a live view.
A single-client external MJPEG camera made timelapse frames, finish photos and Obico checks fail while someone watched. They now use the live view's frame.
The X1C live view could stop instead of reconnecting.
Camera test, diagnose, plate check and plate calibration held a database connection for the whole capture, which could exhaust the pool on a large farm.
The OBS stream overlay ignored live updates and moved only with its 2-second refresh.
A plate check that could not run reported an empty plate. It now warns with the reason, and the print still goes ahead.
A second Bambuddy on the same machine cut the other's live camera every minute.
A restart with a camera view open logged an ERROR with a full traceback for each stream.
A layer timelapse of a long print was stopped after 5 minutes and its frames deleted. Stitching now gets up to an hour, and frames are kept for about 24 hours if ffmpeg fails.
A frozen external RTSP camera kept showing the same picture. It is now reconnected, like the printer's own camera.
A USB camera could not be watched from several places at once; the second viewer got a blank picture. See Before you update.
Camera snapshots and the camera Test button failed for a USB camera while someone watched the live view. See Before you update.
USB camera finish photos, timelapse frames and snapshots were sometimes killed mid-capture, and other programs' ffmpeg on a USB camera were stopped too.
The camera Test button read "Camera connected640x480".
With an external RTSP or USB camera, closing one view of a printer stopped the stream for every other open view.
A minimized camera window kept its stream open, and the camera window and Camera page left the current stream running after a reconnect, keeping the camera busy and the chamber light on.
Cam Wall tiles froze on their last frame when the server ended a stream, and a long-running overlay browser source piled up camera connections. Live tiles now reconnect every minute and show "No signal" if the camera stays down.
The finish photo never raised the plate back into view, on any printer (#3240, reported by @Thomansky). The plate now moves when Restore plate for finish photo is on, which is the default.

Login, users and permissions:

API keys of Operators were refused with 403 on their owner's own queue jobs, archives and library files.
LDAP and SMTP "Test" checked the saved settings instead of what the form held.
A user's email address could not be removed.
Clear all on the shopping list, Clear in the Git backup history and removing a linked SSO account acted without asking. An SSO-only account can no longer remove its last SSO link.
After a 2FA or SSO login, LDAP and OIDC users were offered "Change Password" until the page was reloaded.
Admins could not remove a locked-out user's 2FA from Edit user. Settings -> Users -> Edit user now has "Disable 2FA".
Users without settings:read got default values for settings many pages need: with billing on, the print dialog hid the cost centre picker and the print was refused, and the queue order, Slice in the File Manager, filament warnings, Spoolman mode, theme and currency were wrong too. These settings now reach every signed-in user, without credentials.
A project's bill of materials could show the names of archives from printers outside the user's groups.
Adding archives or queue jobs to a project, or removing them, accepted archives and jobs from printers outside the user's groups and other users' prints. Assign to project now reports how many were actually added.
Editing an archive could move it to a printer outside the user's groups, or link it to a printer or project that does not exist.
The "Password Reset" email template was never used, and its default text still promised a new password. The reset email is now built from the template with the reset link added at the end.
Removing live-update or camera permission from a group, a user or an API key did not end open dashboards, camera views, camera links or Cam Wall and Home Assistant camera tokens.
An open camera view kept showing live video after a password change, deactivation, revoked token, group change or similar. Streams now re-check every 30 seconds that the viewer may still watch.
Signing a user out everywhere (password change or reset, 2FA removal, deactivation, backup restore) left their open dashboards receiving live updates. Turning authentication on and disabling or expiring an API key now close the dashboards they opened too.
A deleted API key's open dashboards and camera and WebSocket links passed to the next key created on SQLite.
With login off, any unexpected error was reported as "Authentication service temporarily unavailable" (503), and with login on a brief database hiccup signed you out. Errors are now reported as 500 and logged, and a hiccup is retried.
Permissions removed from a group, such as View and Import MakerWorld, came back after a restart (#3238, reported by @Minebuddy).

File Manager:

A large tag catalogue pushed the file list off the page (#3411, reported by @iammenasco). Tags now take up to two rows, with Show all and a tag search.
The File Manager's toolbar ran past the edge of a phone screen and pushed the page sideways (#3295, reported by @apizz).
Changing a library file's folder through the API could move it into or out of an external folder without moving the file. See Before you update.
A slow folder listing on the printer showed as "printer unavailable", and every refresh started another listing.
Printer files written in the evening could be dated a year back on hosts behind UTC, so their camera recordings were not matched to their prints.
Uploading two files with the same name showed the first file's result on both rows.
Opening a printer's 3MF in the file browser downloaded the whole file again for the plate list and every plate thumbnail.
Printer file downloads and previews reported every failure as "File not found", and large files hit a fixed 10-minute limit. They now get the time their size needs, up to 30 minutes.
A timed-out printer download kept running beside its retry.
Large STL files froze the web interface and printer updates while their thumbnail was made.
The File Manager and Archives page loaded every thumbnail at once, slowing down the server. Thumbnails now load as they scroll into view.
File names could vanish from the File Manager's list view when the folder sidebar left little room (#3105, reported and contributed by @akshayy29 in #3119).

Interface:

The full-page G-code preview stayed black and the scrollbar flickered on wide screens with classic scrollbars (#3371, reported by @Underscan007).
The print dialog's slot picker was too narrow to read on a phone held upright (#3294, reported by @apizz).
The Printers page sent hundreds of requests on a print farm, about 550 with 59 printers, so cards took up to 20 seconds to fill. Requests are now combined.
A bug report was filed as a public GitHub issue the moment debug logging stopped, also on the 5-minute auto-stop (#3363, reported by @choumarin). You now review it before submitting.
Check Belt Tension on a P1 printer linked to a general maintenance page instead of Bambu Lab's belt guide (#3302).
Removed the "Save thumbnails" and "Notification Language" settings, which did nothing. See Before you update.
The printer temperature history retention had no setting, and a retention of 0 set through the API wiped all history. See Before you update.
"Max copies per run" and the Docker compose directory were never saved, "Default printer" and "Auto-archive prints" were never used, and "Create Archive Entry When Printing" did nothing. See Before you update.
Importing or creating a project could give it a priority or status the project editor refuses, making it uneditable, and failed on PostgreSQL. Such projects now get normal and active, also when restoring an older SQLite backup onto PostgreSQL.
A tab whose live connection was closed by the server froze until reloaded. It now goes to the login page if the session ended, or reconnects.
A project's timeline stayed empty while one of its jobs was queued or printing.
Printer cards never showed the last print, nor the Good / Reject prompt for it. The cards now share one request instead of one per card.
The Printer Locations page had no way back to the Printers page. It now has a "Back to Printers" link.

Integrations (Home Assistant, MakerWorld, Spoolman, API):

Every MakerWorld import, Save and Save & Slice failed with "MakerWorld did not return a download URL" from 3 November on (#3447, reported by @waynem-bris; #3448, reported by @frederic3114).
Home Assistant's connection test said "Could not connect" when the problem was an untrusted https certificate (#3300). It now names the certificate problem and what to do.
An API client reading right after a successful write could miss it, and a save that failed at the end was hidden behind a success (#3320, reported by @MorganMLGman).
The legacy PUT /settings/virtual-printer route answered 200 for settings it no longer applies. See Before you update.
GET /queue?status=pending,printing with several statuses returned nothing.
A request with Infinity or NaN in a field that failed validation ended in a server error instead of 422.
Home Assistant "Run when printer turns off" scripts ran only when the plug was switched by hand, not on auto-off, schedules or failure detection. See Before you update.
A truncated or broken MakerWorld download was stored as the model's 3MF.
Unused API pieces were removed and the rest documented. See Before you update.
The archive export's date_to left out the whole end day, and an invalid print speed answered 500 instead of 422.
An id of 0 when queueing, creating a batch or creating a project skipped the existence check, causing a server error on PostgreSQL or a link to nothing on SQLite. 0 now means "none", and batches check their archive or library file.
A smart plug or virtual printer could be linked to printer 0. It is now refused with "Printer not found".
null for a field that can't be empty caused a server error on about a dozen PATCH/PUT routes. It now leaves the field unchanged, as on other routes.
The /webhook/... routes answered a database problem while checking the API key with a 500. They now answer 503, and 401 for an invalid key.
The MQTT relay never sent smart_plugs/energy, sent smart_plugs/on and /off only for manual switching, and maintenance/alert always reported 0. All three now match the wiki.
The webhook endpoints listed in Settings → API Keys didn't exist or had the wrong path.
A valid API key could be refused as invalid during a brief database problem. It now answers 503, which a client can retry.
Stopping or cancelling a print through the API failed with an error, and cancel left the print running. Both now work, count as stopped rather than failed, and stop also stops a paused print.

Install, platform and database:

Restoring a Git backup brought back archives without their prints, so statistics and the Print Log were empty. Backups now include prints, and older backups restore as before.
The connection diagnostic told containers on a Docker macvlan network to switch to host networking (#3417, reported by @ddavidebor). It now recognises macvlan and ipvlan, and warns about ipvlan L3.
Deleting a local backup failed silently when the server refused it.
A print running when a Git backup was made came back as a failed print after restore, with a failure notification. It is now restored as cancelled.
Spools came back from a Git backup restore as never used.
A Git backup restore brought every reset "Total Consumed" counter back to its full lifetime figure.
The virtual printers kept running during a backup restore.
Restoring an old SQLite backup onto PostgreSQL could put some timestamps hours off.
A GitHub backup restore copied damaged or hand-edited numbers into the totals. Invalid values are now left empty and listed in the restore summary.
Bug reports kept the reporter's email locally. Stored emails are cleared at startup.
An in-app update that hung blocked every further update until a restart. Each step now has a time limit.
The SpoolBuddy installer in full mode, installed under /home, wrote a Bambuddy service that failed with 203/EXEC (also #1685).
SQLite gave a deleted user's, API key's, printer's, archive's or camera token's id to the next one created. These tables now never reuse an id; existing databases are converted once at start.
On SQLite, deleting archives, library files, queue jobs, batches or projects left references behind, so a queue job could print a different file and a new project could show a deleted one's files. Deletes now clean up as on PostgreSQL, and each start repairs what earlier deletes left.
PostgreSQL databases from older versions were missing links a new install has, so deleting an archive, user or library file left their id behind, and 51 columns differed from a new install. They are fixed on the next start.
Each install or update took whatever Python dependency versions were newest that day. They are now pinned to tested versions.
Native installs on Python 3.10 failed to start while the installer and docs said 3.10 was enough. Python 3.11 is now required, and the installer picks up python3.11 on Ubuntu 22.04.
Every PostgreSQL start logged about 320 errors from migrations that had already run. PostgreSQL 13 or newer is required.
Shutting Bambuddy down logged an error with a traceback when nothing was wrong (#3243, reported by @Thomansky).
Running out of file descriptors could corrupt a SQLite database (#2883, reported and contributed by @M2ABRAMSTANK in #2884). Bambuddy now raises its open-file limit at start and uses a smaller SQLite pool.

Security (dependencies)

dompurify to 3.4.17, zod to 4.6.7 and Vitest to 4.1.12. Bambuddy did not use the affected DOMPurify options; zod and Vitest are development-only and not in the shipped image.
virtualenv to 21.14.9, filelock to 4.2.1 and setuptools to 84.1.0, all development-only and not in the shipped image.
pyasn1 to 0.6.5 and minimatch to 11.0.4. pyasn1 ships with LDAP sign-in, where crafted input could cost excessive time or memory; minimatch is development-only.

Merged community PRs in this release

#1606 by @samedyuksel - assign a spool by scanning its QR label.
#2803 by @grolmus - map each printer against its own AMS, hold jobs whose filament isn't loaded.
#2884 by @M2ABRAMSTANK - SQLite pool and open-file limit.
#2995 by @SVAN19 - printer locations page.
#3119 by @akshayy29 - File Manager list column widths.
#3207, #3208, #3209 by @adamspicedev - overlay token creation and URL import, logo and progress colours, portrait layouts.
#3233, #3306 by @rojosinalma - lifetime energy total for REST plugs, moving folders in the File Manager.
#3252 by @Niko11111 - Spoolman native tags.
#3255 by @chrisbennight - Bambuddy Internal URL for Obico.
#3265 by @simplytoast1 - Notify! with Live Activities and Lock Screen widgets.
#3283 by @bgrr74 - remaining time in the browser tab.
#3346 by @Thomansky - full tag UIDs on the SpoolBuddy PN5180 reader.
#3369 by @andy-sorge - P2S printer picture.
#3378, #3389, #3390 by @hibikipr - reliable Bambu tag reads on SpoolBuddy, Bambu tag data with each scan, Quick Add from a scanned tag.
#3381, #3393 by @pleasantone - Orca Cloud printer lists, sorted and searchable Slice dialog presets.
#3385 by @koshilki - custom label sheets and saved templates.
#3413 by @oh-forf - DB_POOL_PRE_PING.
#3428 by @QuinnDamerell - OctoEverywhere AI failure detection (brought in by hand).

Thanks

To everyone who requested, reported, captured logs, or contributed: @0xCyberFortress, @aammash, @adamspicedev, @airforceixi, @Akaviiii, @akshayy29, @Andrewdlec, @andy-sorge, @apizz, @bdwilson, @BenBBSLA, @BenjaminCs, @bgrr74, @bobobo1618, @BurgerKerman, @cartman75, @choumarin, @chrisbennight, @CLKRUN, @cmerkle, @dadino, @ddavidebor, @dogod3814t, @dovmesiz, @Eifel-Joe, @elizabethburg, @erabti, @fma965, @frederic3114, @frnzzle, @gregspatrick, @grolmus, @HarryCoff2, @hibikipr, @HopNotch, @humplemann, @hyyz17200, @iammenasco, @ImNanou, @jasoncronje, @juandelarossa, @Keybored02, @koshilki, @lightmaster, @Lioncat55, @lonix, @lukg20, @M2ABRAMSTANK, @Magical-Penguin, @mason9468, @McGagnor, @Minebuddy, @mithkr, @MorganMLGman, @Mounting9612, @MrCrunshy, @Niko11111, @Novastorm4332, @nvx, @oh-forf, @ojimpo, @pleasantone, @QuinnDamerell, @RambachTJ, @rogerabm, @rojosinalma, @Roving-Ronin, @SaadShaya7, @samedyuksel, @sargonas, @Sawtaytoes, @sean3318, @sgiffhorn, @simonkreuchauff-oss, @simplytoast1, @Spegeli, @SVAN19, @Thomansky, @Underscan007, @USER19464727, @waynem-bris, @wolfrage76, @wreuel.

Sponsors

Bambuddy is sustainable thanks to people who put their money where their use is. If this release saved you time or kept your farm running, the project runs on recurring contributions - there's no paid tier, no telemetry, no upsell, just sustainable maintenance.

GitHub Sponsors (recurring, 5 tiers from $5/mo to $300/mo) - https://github.com/sponsors/maziggy
Ko-fi (one-time or recurring) - https://ko-fi.com/maziggy