--export-slicedata

After a successful slice, --export-slicedata <dir> serializes the computed slice geometry for every plate to a directory tree, so a subsequent invocation can load it with --load-slicedata and skip the geometry pass entirely. The flag accepts a path to an existing parent directory. Inside that directory it creates one subdirectory per plate, named by plate index (1, 2, and so on). Within each plate subdirectory it writes one JSON file per print object, with a filename of the form obj_<id>_<region_count>_<config_hash>.json. The identity portion comes from the model instance's recorded ID (falling back to an internal instance ID for older 3MF files); the region count and a combined hash of all print-region configs for the entire print are appended so that the cache filename changes automatically if the print settings change.

Each per-object JSON file captures the full geometric output of the slicing pass: every print layer with its Z position, slice height, and layer-height values; the sliced polygons and their bounding boxes; overhang polygons and their bounding box; the per-region extrusion data for each layer region; every support layer with its support islands and support fill extrusion paths; and the first-layer volume groupings used for first-layer checks. The JSON is written without indentation unless the log level is set to 4 (debug) or higher, in which case it is pretty-printed with four-space indentation.

Only objects that were actually sliced in the current run (those the engine marks as needing re-slicing, not shared copies reusing another object's geometry) are written. If a per-plate subdirectory cannot be created, or if writing any file fails, the CLI deletes the partial plate subdirectory and exits with a non-zero error code rather than leaving a partial cache behind. The export step runs only after slicing and G-code generation succeed for a plate; a slice error on any plate aborts before export occurs for that plate.