Remi

Unfortunately, I don't think there is a workaround in this case today

Thanks for flagging ! I shared the thread with the team.

Awesome to hear ! There is an API available that allows you to embed Dust in you product/tools. You can reach out to support@dust.tt discuss moving to an enterprise plan with a pay-per-use pricing 😊

Splitting in two looks like the best approach indeed ! Also, docs uploaded in Dust from the Gdrive connection can be a bit larger too. You could push the file to Gdrive instead of a Dust folder.

Tomas can you send a message to support@dust.tt?

Tomas You can add the sub domain as starting point to the "website" connection and it will use extract only sub urls

Tomas Are you sure the info you are looking for is correctly extracted in the html? If yes, converting to markdown could help indeed !

Thanks for flagging Hugo Caisse! "Query table" should work just fine in this case. Do you confirm that the table was properly connected in the agent's settings? I'll flag this to the team in case it is a wider issue.

Etienne Dubois indeed, lags are often caused by delay in Claude's servers. Switching to the more recent gpt5 model should help. It's fast and smart ! Making sure the agent doesn't call tools when it doesn't need to can help too. You can do that through the instructions (eg: "never use tools unless asked")

Thank you all, the team is on it !