Maxence Jouve

Commented on Bypassing Gmail Limits for Dust Agent Email Integr...·Posted inHelp

Indeed, only a snippet is returned by the tool. The full email body is truncated.

Bypassing Gmail Limits for Dust Agent Email Integration

The gMail tool seems to limit what can be visible (email content) to the Dust agent. Has anyone managed to bypass this limitation? The Dust agent manages to retrieve the correct emails, but only sees a snippet of the email body. This limits the power of such an integration as a lot of the context is lost.

2 attachments

1Comment

Posted in Feedback·

Maxence Jouve

Building a Custom Attio Connector with Dust API Features

Hey, I am building a custom connector for Attio (CRM), using Dust API. It would be great to either have :

a way to retrieve documents based on their parent id
an option to deleted children automatically when a parent is deleted

0Comments

Commented on Integrating OpenAI's Deep Research API: Timeout Is...·Posted inHelp

Maxence Jouve

Thanks Alban for your answer

Posted in Help·

Maxence Jouve

Integrating OpenAI's Deep Research API: Timeout Issues and Solutions

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!

4 attachments

2Comments

Commented on Enabling Full_Text Param for Improved LLM Context...·Posted inHelp

Maxence Jouve

all good, found the answer myself. you can do it via dust apps!

Posted in Help·

Maxence Jouve

Enabling Full_Text Param for Improved LLM Context in API

I realized that through the API, you can load the entire document in the context of the LLM when using the Search tool with the full_text param. Is it possible to enable this param through the UI? Could be useful for queries requiring "reasoning" to give more context to the LLM. Many thanks in advance for your help!

1 attachment

1Comment