Validation and Repair#
This page explains how SWC-Studio turns validation output into repair-oriented workflows.
What validation produces#
A validation run produces grouped pass, warning, and fail results together with detailed findings such as:
failing node IDs
failing section IDs
thresholds and metrics
blocked prerequisite summaries when required
The canonical rule and issue details live in:
What the issue system adds on top#
The issue list used by the GUI and swcstudio check is broader than the raw validation report.
The shared issue builder can combine:
direct validation findings
blocked checks
suspicious radii detected from the radii-cleaning backend
likely wrong labels inferred by comparing current labels against the auto-labeling backend
a simplification suggestion
This is why the issue list is useful as a repair queue rather than only as a check summary.
Repair paths#
SWC-Studio uses several repair paths depending on the issue type.
Validation and index repair#
For structural and ordering issues:
run validation
inspect the grouped findings
use
Index Cleanwhen the parent-before-child ordering or ID continuity needs correction
Label repair#
For annotation problems:
use
Manual Label Editingfor targeted node changesuse
Auto Label Editingto run the auto-typing engine for type reassignmentuse dendrogram subtree reassignment when a whole branch should inherit a new label
Radii repair#
For suspicious radii:
use
Manual Radii Editingfor direct node-level fixesuse
Auto Radii Editingorradii-cleanfor the shared multi-pass cleaner
Geometry repair#
For topology and geometry problems:
move nodes or subtrees
reconnect or disconnect branches
delete nodes or subtrees
insert new nodes
simplify geometry while preserving the main tree structure
Recommended repair loop#
The intended loop is:
validate or check the file
inspect one issue at a time
let the issue route you to the matching repair tool
apply a fix
rerun validation
review the new per-file history operation and any requested report
Type-suspicion analysis normally runs in a background worker so the rest of the issue list can appear first. If validation immediately follows an applied auto-label result, the GUI reuses that result instead of running the same inference again.
This is the same logic described in the GUI workflow and reflected by the CLI command set.