Merge branch 'main' into fix/lgraph-execution-loop-check

Both the do_not_catch_errors and catch-errors branches now call
doExecute (with onExecute guard) instead of onExecute directly.
This ensures execute_triggered visual feedback and onAfterExecuteNode
are triggered consistently regardless of error-catching mode.

src/lib/litegraph/src/LGraph.ts

@@ -579,9 +579,7 @@ export class LGraph
       for (let i = 0; i < num; i++) {
         for (let j = 0; j < limit; ++j) {
           const node = nodes[j]
-          // FIXME: Looks like copy/paste broken logic - checks for "on", executes "do"
           if (node.mode == LGraphEventMode.ALWAYS && node.onExecute) {
-            // wrap node.onExecute();
             node.doExecute?.()
           }
         }
@@ -597,8 +595,8 @@ export class LGraph
         for (let i = 0; i < num; i++) {
           for (let j = 0; j < limit; ++j) {
             const node = nodes[j]
-            if (node.mode == LGraphEventMode.ALWAYS) {
-              node.onExecute?.()
+            if (node.mode == LGraphEventMode.ALWAYS && node.onExecute) {
+              node.doExecute?.()
             }
           }