Fix todo update failures by removing ID from create_todos schema

This fixes two related issues with todo management:

1. "todo X not found" errors when updating todos
   - LLMs were providing IDs in create_todos (e.g., "id": "1") because
     the schema included an id field
   - Backend auto-generates proper IDs (e.g., "todo_1") but LLMs would
     later try to update using the wrong ID format
   - Solution: Created CreateTodoItem struct without id/status fields
     for the create_todos schema, preventing LLM confusion

2. Missing todo update events in TUI
   - TUI was updating todos on ToolCall event (before execution)
   - If a tool failed, TUI would show the update even though it never
     succeeded in the backend
   - Solution: Moved todo updates to ToolCallResponse event with error
     checking to ensure TUI only updates after successful execution

Changes:
- Add CreateTodoItem struct with only description field
- Update CreateTodosArgs to use CreateTodoItem instead of Todo
- Move TUI todo update logic from ToolCall to ToolCallResponse
- Update tests to match new schema

Author: trungutt

pkg/tools/builtin/todo.go

@@ -24,12 +24,16 @@ type Todo struct {
 	Status      string `json:"status" jsonschema:"New status (pending, in-progress,completed)"`
 }
 
+type CreateTodoItem struct {
+	Description string `json:"description" jsonschema:"Description of the todo item"`
+}
+
 type CreateTodoArgs struct {
 	Description string `json:"description" jsonschema:"Description of the todo item"`
 }
 
 type CreateTodosArgs struct {
-	Todos []Todo `json:"todos" jsonschema:"List of todo items"`
+	Todos []CreateTodoItem `json:"todos" jsonschema:"List of todo items"`
 }
 
 type UpdateTodoArgs struct {

pkg/tools/builtin/todo_test.go

@@ -77,22 +77,12 @@ func TestTodoTool_Tools(t *testing.T) {
 			"items": {
 				"type": "object",
 				"required": [
-					"id",
-					"description",
-					"status"
+					"description"
 				],
 				"properties": {
 					"description": {
 						"type": "string",
 						"description": "Description of the todo item"
-					},
-					"id": {
-						"type": "string",
-						"description": "ID of the todo item"
-					},
-					"status": {
-						"type": "string",
-						"description": "New status (pending, in-progress,completed)"
 					}
 				},
 				"additionalProperties": false
@@ -191,7 +181,7 @@ func TestTodoTool_CreateTodos(t *testing.T) {
 
 	// Create multiple todos
 	args := CreateTodosArgs{
-		Todos: []Todo{
+		Todos: []CreateTodoItem{
 			{Description: "First todo item"},
 			{Description: "Second todo item"},
 			{Description: "Third todo item"},
@@ -222,7 +212,7 @@ func TestTodoTool_CreateTodos(t *testing.T) {
 
 	// Create multiple todos
 	args = CreateTodosArgs{
-		Todos: []Todo{
+		Todos: []CreateTodoItem{
 			{Description: "Last todo item"},
 		},
 	}

pkg/tui/page/chat/chat.go

@@ -242,20 +242,23 @@ func (p *chatPage) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 	case *runtime.ToolCallEvent:
 		spinnerCmd := p.setWorking(true)
 		cmd := p.messages.AddOrUpdateToolCall(msg.AgentName, msg.ToolCall, msg.ToolDefinition, types.ToolStatusRunning)
+		return p, tea.Batch(cmd, p.messages.ScrollToBottom(), spinnerCmd)
+	case *runtime.ToolCallResponseEvent:
+		spinnerCmd := p.setWorking(true)
+		cmd := p.messages.AddToolResult(msg, types.ToolStatusCompleted)
 
 		// Check if this is a todo-related tool call and update sidebar
+		// Only update after successful execution (in response, not during call)
 		toolName := msg.ToolCall.Function.Name
 		if toolName == "create_todo" || toolName == "create_todos" ||
 			toolName == "update_todo" || toolName == "list_todos" {
-			// Log error but don't fail the tool call
-			// Could add logging here if needed
-			_ = p.sidebar.SetTodos(msg.ToolCall)
+			// Only update if the response doesn't contain an error
+			// Response starting with "Error calling tool:" indicates failure
+			if len(msg.Response) < 19 || msg.Response[:19] != "Error calling tool:" {
+				_ = p.sidebar.SetTodos(msg.ToolCall)
+			}
 		}
 
-		return p, tea.Batch(cmd, p.messages.ScrollToBottom(), spinnerCmd)
-	case *runtime.ToolCallResponseEvent:
-		spinnerCmd := p.setWorking(true)
-		cmd := p.messages.AddToolResult(msg, types.ToolStatusCompleted)
 		return p, tea.Batch(cmd, p.messages.ScrollToBottom(), spinnerCmd)
 	case *runtime.MaxIterationsReachedEvent:
 		spinnerCmd := p.setWorking(false)