400 Missing required parameter: 'input[N].name'

[ashil816]

TanStack AI version

0.14.0

Framework/Library version

Tanstack start rc and react 19.2

Describe the bug and the steps to reproduce it

When a tool with [needsApproval: true] completes execution during a continuation flow, [TextEngine.buildToolResultChunks()] emits a [TOOL_CALL_START] chunk with only [toolName] but omits the required [toolCallName]field:

// src/activities/chat/index.ts ~L1217
chunks.push({
  type: 'TOOL_CALL_START',
  timestamp: Date.now(),
  model: finishEvent.model,
  toolCallId: result.toolCallId,
  toolName: result.toolName,   // ← present
  // toolCallName is missing!  // ← @ag-ui/core requires this
} as StreamChunk)

Impact:

The client [StreamProcessor.handleToolCallStartEvent] reads [chunk.toolCallName] (which is undefined), creating a ghost [ToolCallPart] with [name: ""] in a new assistant message. On the next user message, [convertMessagesToInput] in @tanstack/ai-openai serializes this as a function_call input item without the required [name] field, causing:
400 Missing required parameter: 'input[N].name'

Reproduction:
Define a server tool with [needsApproval: true]
Have the LLM call the tool
Approve the tool via [addToolApprovalResponse]
Tool executes successfully (continuation flow triggers [buildToolResultChunks] with [argsMap]
Send any follow-up message → 400 from OpenAI

Expected fix:
Add [toolCallName: result.toolName] to the [TOOL_CALL_START] chunk in [buildToolResultChunks] same as the [TOOL_CALL_END] chunk already does:

chunks.push({
  type: 'TOOL_CALL_START',
  timestamp: Date.now(),
  model: finishEvent.model,
  toolCallId: result.toolCallId,
  toolCallName: result.toolName, // ← add this
  toolName: result.toolName,
} as StreamChunk)

Your Minimal, Reproducible Example - (Sandbox Highly Recommended)

Screenshots or Videos (Optional)

No response

Do you intend to try to help solve this bug with your own PR?

None

Terms & Code of Conduct

• I agree to follow this project's Code of Conduct
• I understand that if my bug cannot be reliable reproduced in a debuggable environment, it will probably not be fixed and this issue may even be closed.