LLM not responding with the tool response
# support
v
HI I am building a tool that the AI calls to get a list of appointments. The Assistant is doing the following request to the tool and getting the response below...everthing seems correct but the AI still doesnt talk back the response and says there are no upcoming appointments. Do I have to change anythiing on the prompt? Thanks. { "message": { "timestamp": 1750525825056, "type": "tool-calls", "toolCalls": [ { "id": "call_j92STzv6K7gIueLGSVa9kWce", "type": "function", "function": { "name": "get_appointments_aa1d8991", "arguments": {} } } ], "toolCallList": [ { "id": "call_j92STzv6K7gIueLGSVa9kWce", "type": "function", "function": { "name": "get_appointments_aa1d8991", "arguments": {} } } ], "toolWithToolCallList": [ { "type": "function", "function": { "name": "get_appointments_aa1d8991", "parameters": { "type": "object", "required": [], "properties": {} }, "description": "Get upcoming appointments for the patient calling. The tool will return a list of appointments, or any errors if there are no appointments." }, "async": true, ... And it is getting the following output. { "results": [ { "toolCallId": "call_j92STzv6K7gIueLGSVa9kWce", "result": "The tool returned the following output: Your upcoming appointments are 2024-06-20T10:00:00.000-07:00 - Unknown type, 2025-06-20T10:00:00.000-07:00 - Unknown type, 2025-06-20T10:00:00.000-07:00 - Unknown type, 2025-06-20T12:00:00.000-07:00 - Unknown type" } ] }
Also the messages on the call log was saying tool call was a success.
k
Update the system prompt to explain how to interpret the string, or return structured JSON so the AI can automatically parse and respond properly..
h
I've got this same issue, the result was only saying the tool call was a success but doesnt actually show the actual result with all the available time slots
did you managed to solve this?
v
I tried returning JSON response but changed it to string because the docs say a tool requires a string response. I tried all different variations without success. Also updating the descritption of the tool and also changing the prompt didnt have any effects.
Actually i found that if I turn async off in the tool then it reads the response correctly.
k
checking if this is resolved/solved for you?
v
Yes
p
Glad the issue was resolved. Please reach out to us if you have any other questions regarding this issue