G-code injection¶

BamDude can inject operator-defined G-code snippets into a print's plate gcode at dispatch time — start snippet at the very beginning of the print, end snippet at the very end. The snippet survives even if BamDude crashes mid-print, runs at exactly the right point in the printer's own gcode sequence (before/after the firmware's reset routines), and is per-printer-model so the same project file can produce a different snippet on an X1C than on a P1S.

This is not the same as a macro. Macros are server-driven MQTT pushes at lifecycle events; G-code injection is bytes spliced into the actual file. See the comparison table on the macros page if you're picking between the two.

How it works¶

Configure — drop start / end snippets per printer model in Settings → Printing → G-code Injection.
Toggle on the print — when you start a print (PrintModal, queue item, dispatch), tick Inject G-code.
Dispatch path — the dispatcher opens the 3MF, locates the marker ; MACHINE_START_GCODE_END in each plate_*.gcode, splices the resolved start snippet immediately after it (and the end snippet immediately before ; EXECUTABLE_BLOCK_END), recomputes per-plate .gcode.md5 sidecars, repacks the 3MF, and uploads to the printer via FTP.
Audit trail — the resulting archive's applied_patches JSON records {name: "gcode_injection", model, plate_id} so you can see at a glance which prints had what injected.

graph LR
    A[Job dispatched] --> B[3MF opened]
    B --> C[Find ; MACHINE_START_GCODE_END]
    C --> D[Splice start + end snippets]
    D --> E[Recompute .gcode.md5]
    E --> F[Repack 3MF]
    F --> G[FTP upload to printer]

Configuring snippets¶

Settings → Printing → G-code Injection auto-discovers every distinct printer model from your linked printers and renders one collapsible <details> block per model.

Each model carries two textareas:

Snippet	Where it splices	Typical use
Start G-code	Right after `; MACHINE_START_GCODE_END` — i.e. once the firmware's heat-up + auto-home + bed-mesh phases finished, but before the slicer's own start sequence.	Park to a custom origin, set custom acceleration, log start-of-print to a serial probe.
End G-code	Immediately before `; EXECUTABLE_BLOCK_END` — after the printer's own machine-end sequence, but still inside the block the firmware executes.	Park to back-left for easy retrieval, purge tower clean-cut, post-cool-down nozzle wipe, pause for a manual swap.

Save on blur — the form auto-prunes empty entries. A "Configured" badge on the model summary row appears as soon as at least one of the two textareas has content.

Where the end snippet actually lands

The end snippet is spliced immediately before the ; EXECUTABLE_BLOCK_END marker, not at the physical end of the file. Bambu firmware (verified on a P1S) ignores any G-code that sits after that marker — appending to the file tail silently drops the snippet, so auto-eject / plate-clear moves never fire. Inserting before the marker keeps the snippet inside the block the firmware executes; slicers that don't emit the marker fall back to the old file-tail behaviour.

Two companion details keep the repacked 3MF valid on picky firmware: the per-plate .gcode.md5 sidecar is recomputed from the exact patched bytes (a stale hash makes the P1S reject the file at load), and every ZIP member keeps its original compression — STOREd members such as the embedded preview PNG aren't re-DEFLATE'd, which the P1S preview parser rejects.

Placeholder substitution¶

Snippets support {placeholder} substitution sourced from the slicer's gcode-header ; key = value lines (e.g. {first_layer_temperature}, {filament_type}, {nozzle_diameter}). Prusa-style aliases ({nozzle_temperature} → Bambu's {nozzle_temperature_initial_layer}) are mapped automatically so existing community snippets work without rewriting.

Example:

; Park to back-left, then drop bed for cool-down view
G1 X10 Y210 F6000
G1 Z{first_layer_print_height + 30}
M104 S0  ; nozzle off
M140 S{bed_temperature_other_layer / 2}  ; bed half-down — quick cool, gentle on the part

Per-job toggle¶

The PrintModal's Inject G-code checkbox is the per-job control. It defaults off — turn it on for a specific dispatch when the configured snippet matches what you want for that file. Stays per-PrintQueueItem so a queue full of mixed jobs can pick-and-choose.

Off + no configured snippet = the dispatcher takes a fast path that doesn't even open/repack the 3MF. Repacking is only done when there's something to inject — preserves the existing mesh_mode_fast_check no-op fast path.

When both Mesh-mode fast check and Inject G-code are toggled on the same job, the dispatcher does one open / patch / repack cycle covering both transforms — large multi-plate 3MFs aren't unzipped twice.

Virtual printers can opt in automatically

Prints that arrive through a virtual printer don't pass through PrintModal, so each Queue / Auto-queue VP card carries its own G-code injection toggle. Turn it on and every job that VP enqueues gets the per-model start/end snippets applied — off by default, a no-op unless snippets exist for the target model, and flipping it restarts the VP.

Permissions¶

Permission	Grants
`settings:update`	Edit the per-model snippet library.
`queue:create` / `printers:control`	Toggle Inject G-code on a job (whichever you'd already need to dispatch).

There's no separate printers:edit_gcode_snippets permission — admin gate via settings:update is sufficient given the operational nature of this feature (it's writing gcode, not changing auth).

Caveats¶

Test on a calibration print first

Bad start-gcode at the beginning of every print is a recipe for ruined builds. Validate a new snippet on a small calibration cube before letting it apply to your production farm.

Marker missing — if a 3MF doesn't carry ; MACHINE_START_GCODE_END (rare; some old slicers, custom files), the start snippet falls through to the file head as a best-effort. The end snippet likewise needs ; EXECUTABLE_BLOCK_END; when that marker is absent it falls back to the file tail.
Multi-plate — every plate gets the same snippets applied independently. Per-plate snippets aren't supported (would need a UI per plate × model and clutters the editor for negligible gain).
Reprint of an injected archive — the archive stores the injected output, so reprinting from Archives → Reprint reuses the injected file as-is (no re-injection). To re-inject with a fresh snippet, dispatch from the source library file instead.

Macros — server-side MQTT counterpart with a comparison table.
Virtual printer — per-VP toggle that auto-injects these snippets on slicer uploads.
Settings reference — gcode_snippets setting JSON shape.