Hey everyone,
We’re integrating OpenAI’s Deep Research API as a Dust App (via cURL + extract-code blocks). Each run can take >5min.
the App itself completes successfully and returns the Markdown (see Screenshot 1 of the run from Dust App Logs --> API)
However, when our Dust Agent calls that same App as a tool, the agent run times out and the tool’s output block is empty (Screenshot 2&3)
If we swap in a dummy App that just returns a hard-coded string, the Agent works fine → so the issue looks like an execution-time limit. (Screenshot 4)
Ask
- 1.
Is there a per-tool or per-agent execution timeout we can increase or bypass?
- 2.
If not, what’s the recommended pattern for long-running external calls inside an Agent?
Happy to provide link to chat, or ID of Dust app runs.
Many thanks in advance for your help!
