Repository navigation
v1.1.93
Bambu printer MCP 1.1.93
Added
counterpart_self_check(counterpart/1) reports whether this install's own runtime dependencies and its bundled bambu-node patch loaded: one check per dependency inpackage.json, with the installed version (an exact pin,^or~range must be met; any other form fails as one the check cannot judge), and a read-only check that every hunk of the bundled patch is in the installed bambu-node, plus an in-memory H2D version report that only the patched bambu-node reads.okis false when any check fails, and that check'smessagesays why. It loads modules from the server's own install and sends nothing to the printer. Checked against the real tarball in a local install, before and after its bambu-node is broken (Counterpart #644).package.jsonhas acounterpartfield, readable withnpm view bambu-printer-mcp counterpartbefore downloading:contract(counterpart/1), one sentence ofnotesabout the release, andautoUpdate, which a release sets to false to opt out of the app's automatic update. The MCPinitializeversion stays the package version, now covered by a test (Counterpart #644).counterpart_printer_inforeturnsprinter.displayName, the model with its maker as Bambu names it ("Bambu Lab P1S", "Bambu Lab H2D Pro"), so a client shows it as given. It is absent when the model is unknown (Counterpart #698).
Fixed
-
Preserve the patched bambu-node declaration in the desktop extension, overriding the packer’s default type-file exclusion so counterpart_self_check verifies every bundled patch hunk. Exercise the self-check on the extracted archive.
-
Keep cancellation polling until the slicer output closes, so a cancel after the slicer exits still reaches the bounded pipe cleanup when a helper holds its output. Covered with an exit-before-abort subprocess fixture; no printer contact.
-
At the start of a print,
counterpart_print_progresscould read the previous job's last layer against the new job's layer count ("315/150") for one poll, because a P1/A1 printer's first pushes for the new job carry its name and layer count before its layer number. A report that names a different job (name, file or task id), prepares after any end, or runs after a finished or idle job now leaves out the previous job's layer, percentage and remaining time until the printer reports the new job's own; mcp-3D-printer-server reports the same. A job running again straight after FAILED is that job resuming, and a revised layer count is the same job, so neither loses its progress. Empty values and LAN's task id "0" name no job. Checked with bambu-node's own parser fed recorded-style pushes, no printer connection (Counterpart #702). -
A cancelled
slice_stlcall now stops the slicer and, on macOS and Linux, every helper process it started (the slicer leads its own process group), and settles once they have all gone. The slice's folder is removed only after that, so nothing writes into it afterwards. A slicer or helper that ignores the stop is killed after 5 s; if a helper outside that group (a daemonized one, or any helper on Windows) still holds the slicer's output a second later, the server lets go of it and the call settles anyway. The server logs the removal, and the cancellation test waits for that line instead of reading the folder as soon as the process is gone, which failed under load (Counterpart #699). -
Four changelog entries carried the day their version was bumped rather than the day it was published: 1.1.80 to 1.1.83 were published on 2026-11-10 (Counterpart #699).