- AMS page: external spool slots (Ext/Ext-L/Ext-R), click-to-configure
  modal on all slots, temperature/humidity threshold-colored indicators,
  nozzle L/R badges for dual-nozzle printers, compact AMS-HT layout
- Dashboard: two-column layout with device status + printers list (left)
  and current spool card (right), state-colored scale/NFC icons, dashed
  border card styling
- Daemon: suppress redundant scale reports (±2g threshold + stability
  state change detection) to prevent weight display bouncing
- TopBar: auto-select online printers only, SpoolBuddy logo

NFCReader.__init__ imported read_tag and instantiated PN5180() outside
the try/except block, so a missing module crashed the entire daemon.
Moved the import inside the existing try/except so the daemon gracefully
skips NFC polling — matching the scale reader's existing behavior.

The daemon imports read_tag and scale_diag as bare modules, but they
live in spoolbuddy/scripts/ which isn't on sys.path when systemd runs
the daemon. Added scripts/ to sys.path at startup, resolved relative
to the module file. Also moved the read_tag import inside NFCReader's
try/except (was crashing the daemon instead of skipping gracefully)
and demoted hardware-not-available messages from ERROR to INFO.

5 samples at 100ms (500ms window) wasn't enough to smooth NAU7802 ADC
noise — the averaged value still varied by >2g between 1s report
intervals, and the stability state kept flipping, triggering a report
every cycle. Increased to 20 samples (2s window) so noise is smoothed
before reaching the reporting layer.

When ADC noise kept the spread hovering around the 2g stability
threshold, the stable flag toggled every cycle, forcing a report with
a slightly different weight each time. Now only actual weight changes
of >=2g trigger reports. The stable flag is still included in each
report for consumers that need it.

When Bambuddy auth is enabled, the SpoolBuddy kiosk gets redirected to
the login page because ProtectedRoute requires a user from GET /auth/me,
which only handled JWT tokens. The kiosk daemon already has an API key
but couldn't use it to satisfy the frontend auth check.

- Backend: /auth/me now accepts API keys (Bearer bb_xxx or X-API-Key)
  and returns a synthetic admin UserResponse with all permissions
- Frontend: AuthContext reads ?token= from URL on first load, stores in
  localStorage, and strips from URL (prevents history/referrer leakage)
- Install script: kiosk URL now includes ?token=${API_KEY}
- Tests: 3 new integration tests (Bearer API key, X-API-Key header,
  invalid key rejection)

Enlarge all interactive elements across 9 SpoolBuddy components to meet
44px minimum tap targets on the RPi touchscreen. Increase nav icons
(20→24px), labels (10→12px), bar heights, section headers, printer
buttons, spool visualizations, fill bars, and status indicators.
Compact the dashboard stats bar and remove the printers card. Add
fullScreen prop to ConfigureAmsSlotModal with two-column layout
(filament list left, K-profile + color right) to eliminate scrolling.

The timezone fix (ed36eaf) replaced datetime.utcnow() with
datetime.now(timezone.utc) across ~80 call sites, but SQLAlchemy's
SQLite DateTime columns strip tzinfo on read, returning naive datetimes.
Any Python-side comparison (subtraction, <, >) between an aware "now"
and a naive DB value raises TypeError.

Add `if value.tzinfo is None: value = value.replace(tzinfo=timezone.utc)`
guards at all 8 affected comparison sites:
- maintenance.py: last_performed_at subtraction (500 on /maintenance/overview)
- auth.py: API key expires_at check (2 locations)
- print_scheduler.py: scheduled_time comparison
- smart_plug_manager.py: auto_off_pending_since elapsed calc
- smart_plugs.py: power_alert_last_triggered cooldown
- main.py: last_runtime_update elapsed calc
- archives.py: timelapse completed_at fallback

The timezone fix (ed36eaf) replaced datetime.utcnow() with
datetime.now(timezone.utc) across ~80 call sites, but SQLAlchemy's
SQLite DateTime columns strip tzinfo on read, returning naive datetimes.
Any Python-side comparison (subtraction, <, >) between an aware "now"
and a naive DB value raises TypeError.

Add `if value.tzinfo is None: value = value.replace(tzinfo=timezone.utc)`
guards at all 9 affected comparison sites:
- maintenance.py: last_performed_at subtraction (2 code paths — days-based
  and hours-based intervals both crashed on /maintenance/overview)
- auth.py: API key expires_at check (2 locations)
- print_scheduler.py: scheduled_time comparison
- smart_plug_manager.py: auto_off_pending_since elapsed calc
- smart_plugs.py: power_alert_last_triggered cooldown
- main.py: last_runtime_update elapsed calc
- archives.py: timelapse completed_at fallback

The purple queue counter badge in the printer card header used the raw
unfiltered queue item count, so it appeared on ALL printers of the same
model when a job was scheduled for "any [model]" — even printers without
the matching filament color. The PrinterQueueWidget below it (which
shows "Clear Plate & Start") already filtered by filament type + color.

Apply the same filament compatibility filter (type check + color override
check) to the badge count so it only shows on printers that can actually
run the queued jobs.

The start_bambuddy.bat launcher had Unix (LF) line endings. When a
user's git is configured with core.autocrlf=false or input, the file
is checked out with LF endings and cmd.exe fails with "the syntax of
the command is incorrect" before reaching the hash check.

Add .gitattributes forcing CRLF for *.bat files so they always get
correct line endings on checkout regardless of git config.

  Frontend: Replace inline SpoolInfoCard/UnknownTagCard with full-screen
  TagDetectedModal that auto-opens on NFC tag detection. Known spools show
  remaining weight, fill bar, and offer "Assign to AMS" (new sub-modal with
  printer selector + AMS slot grid) and "Sync Weight". Unknown tags offer
  "Add to Inventory" and "Link to Spool". Modal stays open on tag removal,
  won't re-open for dismissed tags, reopens on re-place after removal.

  Daemon: Fix PN5180 NFC reader silently stopping tag detection when the
  reader drifts into a stuck state. Add auto-recovery (full hardware reset
  after 10 consecutive errors), preventive RF cycling every 60s when idle,
  and periodic status logging. Heartbeat now reports actual nfc_ok/scale_ok
  from reader instances instead of hardcoded True.

  Frontend: Replace inline SpoolInfoCard/UnknownTagCard with full-screen
  TagDetectedModal that auto-opens on NFC tag detection. Known spools show
  remaining weight, fill bar, and offer "Assign to AMS" (new sub-modal with
  printer selector + AMS slot grid) and "Sync Weight". Unknown tags offer
  "Add to Inventory" and "Link to Spool". Modal stays open on tag removal,
  won't re-open for dismissed tags, reopens on re-place after removal.

  Daemon: Fix PN5180 NFC reader silently stopping tag detection. The reader
  drifts into a stuck state where activate_type_a() returns None without
  raising exceptions, making error-based recovery useless. Changed the
  preventive idle maintenance from a lightweight RF off/on cycle to a full
  hardware reset (RST pin toggle + RF re-init) every 60s — the same reset
  that a service restart performs. Also added error-based auto-recovery
  after 10 consecutive exceptions, promoted poll errors to WARNING, added
  periodic status logging, and fixed heartbeat to report actual nfc_ok/
  scale_ok instead of hardcoded True.

  Frontend: Replace inline SpoolInfoCard/UnknownTagCard with full-screen
  TagDetectedModal that auto-opens on NFC tag detection. Known spools show
  remaining weight, fill bar, and offer "Assign to AMS" (new sub-modal with
  printer selector + AMS slot grid) and "Sync Weight". Unknown tags offer
  "Add to Inventory" and "Link to Spool". Modal stays open on tag removal,
  won't re-open for dismissed tags, reopens on re-place after removal.

  Daemon: Fix PN5180 NFC reader failing to maintain tag detection. After a
  successful SELECT the card stays in ACTIVE state and ignores subsequent
  WUPA/REQA, causing immediate false "tag removed" events. Added a brief
  RF off/on cycle (13ms) before each poll to force cards back to IDLE
  state. Also added a preventive full hardware reset every 60s when idle
  to recover from deeper stuck states where activate_type_a() silently
  returns None without exceptions. Heartbeat now reports actual nfc_ok/
  scale_ok instead of hardcoded True.

  Frontend: Replace inline SpoolInfoCard/UnknownTagCard with full-screen
  TagDetectedModal that auto-opens on NFC tag detection. Known spools show
  remaining weight, fill bar, and offer "Assign to AMS" (new sub-modal with
  printer selector + AMS slot grid) and "Sync Weight". Unknown tags offer
  "Add to Inventory" and "Link to Spool". Modal stays open on tag removal,
  won't re-open for dismissed tags, reopens on re-place after removal.

  Daemon: Fix PN5180 NFC reader failing to maintain tag detection. After a
  successful SELECT the card stays in ACTIVE state and ignores subsequent
  WUPA/REQA. Added a conditional RF off/on cycle (13ms) before each poll,
  but only when a tag is present — resets card from ACTIVE to IDLE for
  re-selection. The cycle is skipped when idle to avoid degrading reader
  state with continuous RF toggling, which prevented new tags from being
  detected after removal. Also added a preventive full hardware reset every
  60s when idle to recover from deeper stuck states, error-based recovery
  after 10 consecutive exceptions, and accurate heartbeat reporting of
  NFC/scale health.

  Frontend: Replace inline SpoolInfoCard/UnknownTagCard with full-screen
  TagDetectedModal that auto-opens on NFC tag detection. Known spools show
  remaining weight, fill bar, and offer "Assign to AMS" (new sub-modal with
  printer selector + AMS slot grid) and "Sync Weight". Unknown tags offer
  "Add to Inventory" and "Link to Spool". Modal stays open on tag removal,
  won't re-open for dismissed tags, reopens on re-place after removal.

  Daemon: Fix PN5180 NFC reader failing to detect tags. Each
  activate_type_a() call returning None corrupts the PN5180 transceive
  state, silently preventing all subsequent tag detection. Fixed by
  performing a full hardware reset (RST pin toggle + RF re-init) before
  every idle poll (~240ms, ~1.8 Hz rate). When a tag is present, a light
  RF off/on cycle (13ms) resets the card from ACTIVE to IDLE state for
  continuous re-selection. Also added error-based recovery, periodic
  status logging, and accurate heartbeat NFC/scale health reporting.

  The tare and calibrate buttons on the Settings page queued commands
  but never executed them due to three broken links:

  1. Daemon received tare command via heartbeat but never called
     scale.tare() — the ScaleReader was available in shared dict
     but unused
  2. No API endpoint for the daemon to report the new tare offset
     back to the backend DB, so tare results were lost
  3. Heartbeat updated config but never called
     scale.update_calibration(), so ScaleReader kept initial values

  Added set-tare endpoint + API client method, and fixed heartbeat
  loop to execute tare, persist the result, and propagate calibration
  changes to the ScaleReader instance.

  The tare and calibrate buttons on the Settings page queued commands
  but never executed them due to three broken links:

  1. Daemon received tare command via heartbeat but never called
     scale.tare() — the ScaleReader was available in shared dict
     but unused
  2. No API endpoint for the daemon to report the new tare offset
     back to the backend DB, so tare results were lost
  3. Heartbeat updated config but never called
     scale.update_calibration(), so ScaleReader kept initial values

  Added set-tare endpoint + API client method, and fixed heartbeat
  loop to execute tare, persist the result, and propagate calibration
  changes to the ScaleReader instance. Skip calibration sync on the
  heartbeat cycle that delivers a tare command, since the response
  predates the tare and would overwrite it with stale values.

  The tare and calibrate buttons on the Settings page queued commands
  but never executed them due to four broken links:

  1. Daemon received tare command via heartbeat but never called
     scale.tare() — the ScaleReader was available in shared dict
     but unused
  2. No API endpoint for the daemon to report the new tare offset
     back to the backend DB, so tare results were lost
  3. Heartbeat updated config but never called
     scale.update_calibration(), so ScaleReader kept initial values
  4. The heartbeat response delivering the tare command still had
     pre-tare values, immediately overwriting the new offset to zero

  Added set-tare endpoint + API client method, and fixed heartbeat
  loop to execute tare, persist the result, propagate calibration
  changes to the ScaleReader, and skip calibration sync on the
  heartbeat cycle that delivers a tare command.

  Also replaced the calibration weight input with a touch-friendly
  numpad since the RPi kiosk has no physical keyboard.

  The tare and calibrate buttons on the Settings page queued commands
  but never executed them due to five broken links:

  1. Daemon received tare command but never called scale.tare()
  2. No endpoint to persist tare offset back to backend DB
  3. Heartbeat never called scale.update_calibration()
  4. Heartbeat response with stale values overwrote new tare to zero
  5. set-factor used DB tare_offset (stale/zero), producing wrong
     calibration factor — empty scale showed ~5000g

  Fixed daemon to execute tare, persist result, and propagate
  calibration. Calibration step now captures raw ADC at tare time
  and sends it with step 2, making factor computation self-contained.

  Replaced calibration weight input with compact touch numpad for
  the RPi kiosk's 1024x600 touchscreen (no physical keyboard).

  Some firmware versions send MQTT payloads with non-UTF-8 bytes.
  UnicodeDecodeError was uncaught, silently dropping the entire message
  and leaving printer status as "unknown" with 0°C temps and no AMS.

  Fall back to decode(errors="replace") to keep JSON parseable.

  The H2C dual nozzle variant reports model code O1C2 via MQTT, but only
  O1C was recognized. This caused the camera to use the wrong protocol
  (chamber image on port 6000 instead of RTSP on port 322), producing a
  reconnect loop. Added O1C2 to all model ID maps across 8 files.

  Fix camera button permissions & ffmpeg process leak (#550)

  Camera button on printer card was clickable without camera:view
  permission. ffmpeg processes (~240MB each) accumulated after closing
  camera streams because: (1) stop endpoint called terminate() without
  wait()/kill(), (2) HTTP disconnect detection only ran between frames
  so was blocked when the generator was stuck on stdout read, and
  (3) no mechanism caught processes orphaned by generator abandonment
  or app restarts.

  - Add camera:view permission check + tooltip to camera button
  - Fix stop endpoint: terminate() → wait(2s) → kill() → wait()
  - Add background disconnect monitor (polls every 2s, kills ffmpeg
    directly on disconnect)
  - Add periodic /proc scan (every 60s) that SIGKILLs any ffmpeg
    with rtsps://bblp: not in an active stream
  - Add noCamera i18n key to all 6 locales
  - Fix camera API test mocks for async wait() and pid attribute

Inline callback refs on preset buttons called scrollIntoView on every
re-render, creating a scroll → re-render → scrollIntoView loop on
Windows. Replace with a useEffect + data attribute approach that only
scrolls when the selected preset actually changes.

* Add energy cost to archive card

* Replace Disc3 icon with Coins in ArchivesPage

---------

Co-authored-by: MartinNYHC <mz@v8w.de>

[Fix]: AMS slot modal infinite scroll loop on Windows

This reverts commit d4aa77f, reversing
changes made to 82178f8.

  The slicer_binary_path key was left in the settings table from earlier
  slicer integration research but is no longer referenced anywhere in the
  codebase. Add a startup migration step that deletes orphaned settings
  keys so they don't persist in backups.

  Closes #579 — adds HUF with symbol "Ft" to the supported currencies
  list so Hungarian users can track filament costs in their local currency.

  Profiles like "PLA Support for PETG PETG Basic @bambu Lab H2D" have
  filament_type PETG, but both the frontend and backend name parsers
  found "PLA" first (iterating material types in order). The MQTT command
  sent tray_type=PLA and tray_info_idx=GFL99 (PLA generic), so the
  slicer displayed PLA instead of PETG.

  - Frontend parsePresetName(): detect "X Support for Y" pattern, extract
    material after "Support for"
  - Frontend ConfigureAmsSlotModal: prefer corrected parsed material over
    stored localPreset.filament_type for tray_type, tray_info_idx, and
    temperature fallback
  - Backend _parse_material_from_name(): same "Support for" handling for
    future profile imports
  - Backend assign_spool: prioritize spool.material over lp.filament_type
    for generic filament ID lookup

  FTP upload: reduce chunk size from 1MB to 64KB for smooth progress bar
  updates (~1s intervals instead of 20+ second gaps). Skip voidresp() for
  all printer models — H2D delays the 226 response by 30+ seconds after
  data transfer, causing a hang at 100%. Add transfer speed and TLS
  handshake timing to logs for diagnosing slow connections.

  Print/Schedule modal: widen from max-w-lg (512px) to max-w-2xl (672px)
  to accommodate long filament profile names like "PLA Support for PETG
  PETG Basic @bambu Lab H2D 0.4 nozzle".

  The isSlicedFile() check only matched .gcode or .gcode.3mf extensions,
  so archives from .3mf prints (the standard Bambu slicer output) showed
  a "SOURCE" badge instead of "GCODE". Since .3mf can be either sliced or
  a raw CAD export, now checks the archive's total_layers and
  print_time_seconds metadata to distinguish them. Also passes the
  original filename when creating archives from file manager prints.

- Extract shared detectSystemDateFormat() helper to eliminate
  duplicated system locale detection in date.ts
- Derive SUPPORTED_CURRENCIES from CURRENCY_SYMBOLS map instead
  of maintaining a duplicate hardcoded list
- Consolidate redundant time-parsing regex paths into one
- Use applyTimeFormat() consistently in formatETA()
- Add exhaustive switch default for type safety

Extract repeated helper functions from components into shared utils
(colors.ts, date.ts, file.ts) to reduce duplication:
- isLightColor, parseFilamentColor, hexToColorName → utils/colors.ts
- formatMediaTime, formatDurationFromHours → utils/date.ts
- formatFileSize guard hardened for NaN/negative → utils/file.ts
- formatBytes (ToastContext), formatTime (Timelapse*) deduplicated

Functions with subtly different behavior are intentionally kept local
(formatFilament, formatStorageSize, formatBytes, formatUptime,
formatDateTime, formatDate, formatTimeAgo) to preserve exact UI output.

  The previous fix only filtered status codes (< 0x4000) from the
  print_error field. Firmware can also send the same false positive
  through the hms array in MQTT, which had no such filter. Apply the
  same < 0x4000 check to the HMS parser so status/phase indicators
  are skipped regardless of which MQTT field carries them.

  The firmware check only mapped display names (e.g., "H2D Pro") to
  API keys but not raw SSDP device codes (e.g., "O1E", "O2D"). If a
  printer's model was stored as the raw SSDP code, the firmware check
  either failed or matched the wrong model — H2D Pro matched against
  H2D firmware, showing a false update-available badge.

  Add all known SSDP device codes to MODEL_TO_API_KEY so firmware
  versions resolve correctly regardless of how the model was stored.

)

  Parse home_flag bit 18 (0x00040000) from MQTT to detect ethernet
  connections. When set, the printer card shows a green "Ethernet"
  badge with a cable icon instead of WiFi signal strength in dBm.
  The printer info modal also displays "Ethernet" instead of WiFi
  signal details.

Refactor frontend utility functions to reduce duplication

…cy controls

  Floating bug report button submits issues via bambuddy.cool relay (no GitHub
  token needed locally). Collects 30s debug logs with printer push_all, sanitizes
  all sensitive data, uploads logs as files to GitHub. Screenshot upload/paste/drag
  with JPEG compression. Translated into all 7 languages. Includes 21 tests.

* Fix AMS slot modal not scrolling to preselected filament

The scroll useEffect fired before the loading spinner was replaced
by the preset list, so the DOM element didn't exist yet. The ref
guard was then set, preventing any retry once buttons appeared.

- Move isLoading computation before the scroll effect and add it
  as a dependency so the effect re-runs when loading completes
- Gate scroll on !isLoading so it only runs when buttons exist
- Only mark scrolledToRef after the element is actually found
- Use requestAnimationFrame to scroll after browser layout
- Use block:'center' so the selected item is clearly visible

Preselect filament for printer-configured AMS slots

When an AMS slot was configured directly on the printer (no
savedPresetId), the modal didn't preselect the filament even though
it was in the list. Add a third fallback that matches trayInfoIdx
against the full filteredPresets list (cloud and builtin sources).

* Restore scrolledToRef lost during merge

The useRef import, scrolledToRef declaration, and its reset were
dropped during the 0.2.2b1 merge, causing a ReferenceError in the
scroll-to-preselected-filament effect.

* Fix preset scroll loop and wire up useEffect scroll logic

- Remove inline ref callbacks with scrollIntoView on both preset
  button lists (fullscreen + standard) to stop the infinite scroll
  loop on Windows
- Add data-preset-id attribute so the existing useEffect/querySelector
  scroll logic can actually find the elements
- Replace filteredPresets dependency with builtinFilaments in the
  selection useEffect to prevent search keystrokes from overriding
  manual preset selection

* Adjust scroll behavior in ConfigureAmsSlotModal

---------

Co-authored-by: MartinNYHC <mz@v8w.de>