Hi there,
I’m using Make AI Agents via the “Run AI Agent in SSE Mode” API call. When rendering the output I ignore text in the ‘reasoning’ attribute. However recently (I think) reasoning has started to appear into the ‘content’ attribute which is disrupted the user experience. Is this an error in the API server or intentional? Ideally I’d like such messages to be confined only to the ‘reasoning’ attribute in the returns. See example below of JSON payload returned by Make:
PAYLOAD WITH REASONING IN THE CONTENT ATTRIBUTE
{
“role”: “assistant”,
“content”: “The user is asking me to retrieve projects from the Project Pipeline in ClickUp and filter for:\n1. Projects lacking data for “Signed SoW” or “Purchase Order” (so these fields are empty/missing)\n2. At a status of “In progress” or “Complete”\n3. Grouped by Oversight\n\nI need to retrieve the Project Pipeline list (ID: XXXXXXX) first to see what data we have.\n\nLet me call the tool to retrieve this list.”,
“toolCalls”: [
{
“id”: “XXXXXXXXXXXXXXX”,
“name”: “tool_retrieve_a_list_from_clickup”,
“toolType”: “scenario”,
“arguments”: {
“ListID”: “XXXXXXXX”
}
}
],
“tokenUsage”: {
“promptTokens”: 1411,
“completionTokens”: 181,
“totalTokens”: 1592
},
“executionTimeMs”: 2540,
“id”: “XXXXXXXXXXXXXXXXXXXXXXXXX”,
“agentIterationId”: “XXXXXXXXXXXXXXXXXXXXXXXXXX”
}