feat: add reasoning format processing to LLMNode for <think> tag handling (#23313)

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>

web/app/components/workflow/nodes/llm/use-config.ts

@@ -315,6 +315,14 @@ const useConfig = (id: string, payload: LLMNodeType) => {
     return [VarType.arrayObject, VarType.array, VarType.number, VarType.string, VarType.secret, VarType.arrayString, VarType.arrayNumber, VarType.file, VarType.arrayFile].includes(varPayload.type)
   }, [])
 
+  // reasoning format
+  const handleReasoningFormatChange = useCallback((reasoningFormat: 'tagged' | 'separated') => {
+    const newInputs = produce(inputs, (draft) => {
+      draft.reasoning_format = reasoningFormat
+    })
+    setInputs(newInputs)
+  }, [inputs, setInputs])
+
   const {
     availableVars,
     availableNodesWithParent,
@@ -355,6 +363,7 @@ const useConfig = (id: string, payload: LLMNodeType) => {
     setStructuredOutputCollapsed,
     handleStructureOutputEnableChange,
     filterJinja2InputVar,
+    handleReasoningFormatChange,
   }
 }