fix(34): resolve tool calling errors and stale workflow state

- Add execute functions to all three chat tools (pass-through pattern)
  so AI SDK produces proper tool-call/tool-result pairs in conversation
  history, fixing AI_MissingToolResultsError on subsequent messages
- Fix stale closure in ChatPanel customFetch by using a ref for
  workflowState, ensuring current canvas state is sent with each request
- Add EDITABLE NODE PROPERTIES section to system prompt so LLM uses
  correct data property names (e.g. "prompt" not "text" for prompt nodes)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

src/components/ChatPanel.tsx

@@ -22,7 +22,12 @@ export function ChatPanel({ isOpen, onClose, onBuildWorkflow, isBuildingWorkflow
   const messagesEndRef = useRef<HTMLDivElement>(null);
   const [input, setInput] = useState("");
 
-  // Create a custom fetch function that includes workflowState
+  // Use a ref so the fetch closure always reads the latest workflowState
+  // without needing to re-create the transport
+  const workflowStateRef = useRef(workflowState);
+  workflowStateRef.current = workflowState;
+
+  // Stable fetch function that reads workflowState from ref
   const customFetch = useCallback(async (input: RequestInfo | URL, init?: RequestInit) => {
     if (!init?.body) {
       return fetch(input, init);
@@ -31,17 +36,19 @@ export function ChatPanel({ isOpen, onClose, onBuildWorkflow, isBuildingWorkflow
     const body = JSON.parse(init.body as string);
     const bodyWithWorkflow = {
       ...body,
-      workflowState,
+      workflowState: workflowStateRef.current,
     };
 
     return fetch(input, {
       ...init,
       body: JSON.stringify(bodyWithWorkflow),
     });
-  }, [workflowState]);
+  }, []);
+
+  const [transport] = useState(() => new DefaultChatTransport({ api: "/api/chat", fetch: customFetch }));
 
   const { messages, sendMessage, setMessages, status } = useChat({
-    transport: new DefaultChatTransport({ api: "/api/chat", fetch: customFetch }),
+    transport,
   });
 
   const isLoading = status === "streaming" || status === "submitted";
@@ -57,22 +64,20 @@ export function ChatPanel({ isOpen, onClose, onBuildWorkflow, isBuildingWorkflow
         if (!("state" in part) || !("toolCallId" in part) || !("input" in part)) return;
         if (part.state !== "output-available") return;
 
-        const callId = part.toolCallId;
+        const callId = (part as Record<string, unknown>).toolCallId as string;
         if (processedToolCalls.current.has(callId)) return;
         processedToolCalls.current.add(callId);
 
         const toolName = part.type.replace("tool-", "");
         if (toolName === "createWorkflow" && onBuildWorkflow) {
-          // Extract description from tool args (input is the args)
-          const description = (part.input as { description?: string }).description;
+          const description = ((part as Record<string, unknown>).input as { description?: string }).description;
           if (description) {
             onBuildWorkflow(description);
           }
         }
 
         if (toolName === "editWorkflow" && onApplyEdits) {
-          // Extract operations from tool args (input is the args)
-          const operations = (part.input as { operations?: EditOperation[] }).operations;
+          const operations = ((part as Record<string, unknown>).input as { operations?: EditOperation[] }).operations;
           if (operations) {
             onApplyEdits(operations);
           }

src/lib/chat/tools.ts

@@ -108,7 +108,18 @@ ${formatContextForPrompt(workflowContext)}`;
 - Use **editWorkflow** when the user wants to ADD, REMOVE, CHANGE, or MODIFY nodes/connections in the CURRENT workflow.
 - Always explain what you're about to do BEFORE calling a tool.
 - When editing, reference nodes by their ID from the current workflow state.
-- After editing, summarize what changed.`;
+- After editing, summarize what changed.
+
+## EDITABLE NODE PROPERTIES
+
+When using editWorkflow with updateNode, you MUST use these exact property names in the data object:
+
+- **prompt** node: \`{ "prompt": "the text" }\`
+- **nanoBanana** (Generate Image) node: \`{ "resolution": "1K"|"2K"|"4K", "aspectRatio": "1:1"|"2:3"|"3:2"|"3:4"|"4:3"|"4:5"|"5:4"|"9:16"|"16:9"|"21:9", "useGoogleSearch": true|false }\`
+- **llmGenerate** node: \`{ "temperature": 0-2, "maxTokens": 256-16384 }\`
+- **Any node** title: \`{ "customTitle": "New Name" }\`
+
+Do NOT use "text", "content", or other guessed property names. Use ONLY the exact names listed above.`;
 
   return baseDomainExpertise + contextSection + toolUsageRules;
 }
@@ -130,8 +141,7 @@ export function createChatTools(nodeIds: string[]) {
           .string()
           .describe("The helpful answer to the user question"),
       }),
-      // No execute function - this is a "generate" tool pattern
-      // The LLM provides the answer directly in the tool call result
+      execute: async ({ answer }) => ({ answer }),
     }),
 
     createWorkflow: tool({
@@ -142,7 +152,7 @@ export function createChatTools(nodeIds: string[]) {
           .string()
           .describe("Description of what the workflow should do"),
       }),
-      // No execute function - the client handles calling quickstart API
+      execute: async ({ description }) => ({ description }),
     }),
 
     editWorkflow: tool({
@@ -199,7 +209,7 @@ export function createChatTools(nodeIds: string[]) {
             "Brief explanation of what changes are being made and why"
           ),
       }),
-      // No execute function - results are returned to client for application
+      execute: async ({ operations, explanation }) => ({ operations, explanation }),
     }),
   };
 }