14
Views
3
Comments
Call Agent in ODC Trial provides empty reponse
Question

I have tried to build a basic agent in agent builder in ODC Trial version. 

When testing the agent in the builder it works perfectly. 
If I try to call it from an application it keeps giving an empty response (+ TokenCount 0). 

I have tried increasing the timeout and checked the request by copying it to the agent builder interface and it works. 

I have checked the agent builder analytics and the network in dev tools and the request always returns 200 status code but no response. 

What could be other reasons for this behaviour?

2023-09-06 07-26-35
Sudip Pal

Hi @Nico Thumm 

If possible, can you please share a sample OML? There could be multiple reason. 

Thanks,

Sudip

2024-11-06 14-58-26
OJ JALLOW

HI @Nico Thumm,

How it works: You use the callAgent service action within ODC Studio. You will need to copy the Agent ID from the Agent Builder and provide it to your app.Remember! the trial version still have some constraints like no deployment to production....etc,only development. Please correct me if I'm wrong.

Best regards,

OJJallow



2025-12-22 13-50-43
Sherif El-Habibi
Champion

Hello @Nico Thumm,

I would approach this step by step. As you mentioned, you already investigated and confirmed that the response returns 200, which means the AI agent itself is functional.

Now, I would suggest stepping back a little and reviewing the system prompt carefully to see how it is written.

Then, compare it with the user prompt (the one you are passing dynamically) and check whether it properly aligns with the system prompt or if there are any conflicts or formatting issues.

If all of that looks correct, then inspect the response structure itself and verify how you are mapping or assigning the response in your logic.

If you are also confident that the mapping is implemented correctly, then I would recommend adding breakpoints and debugging step by step starting from the exact point where the agent is triggered.

At this stage, the issue might not actually be in the agent itself, but rather in how the prompt is being passed or how the response is being handled afterward.


Community GuidelinesBe kind and respectful, give credit to the original source of content, and search for duplicates before posting.