Connections Management

How do I add a new Connection to Dust?

• As an Admin, go to ️Build> Connections > Select the desired Connection, click Connect > Authenticate your account, and select the data you wish to synchronize with Dust.
• Please ensure to read the guides related to the Connection you are willing to set before setting it up → all the guides are here.
• Avoid having multiple admins, 2 or 3 is ideal. Ensure you edit Connection cautiously.
• How to update Connections: as an admin, ️ ️Build > Connections > Select the desired Connection, click Manage > Add/Remove data > Explore, and either select or deselect the data you want to synchronize with Dust.

What kind of data cannot be read by my assistants?

• Slack - Dust doesn't take into account:
  • private channels
  • group direct messages
  • external files
  • content behind URLs
  • Messages generated by a workflow.
  • Forwarded and “Also send in channel” messages are also not included.
• Notion: Dust doesn't take into account external files or content behind URLs.
• Google Drive:
  • Dust doesn't take into account files with more than 1Mb of extracted text
  • Supported files include GDocs, GSlides, pptx, docx, CSVs and .txt files;
    • Not PDFs (unless the feature was activated for your workspace. Contact us to know more)
    • Google sheets work with the Table Query and Extract data tools only, for structured data analysis
  • Shortcuts to documents in folders are not taken into account
  • Assistants can read document titles, but not paths / folder names
  • Images in documents are not read.
• Github: Dust only gathers data from issues, discussions, and top-level pull request comments, but not in-code comments in pull requests, nor the actual source code or other GitHub data
• Public Websites: Up to 500 web pages from a public website can be synchronized, but websites behind paywalls or that restrict webcrawling are not seen.
• Confluence: Dust does not synchronize private spaces. Dust does not access pages with view limitations and will not capture any content from a restricted page. This restriction also applies to all the child pages of such pages.
• Intercom:
  • Dust will index only the conversations from the selected Teams that were initiated within the past 90 days and concluded (marked as closed).
  • For the Help Center data, Dust will index every Article published within a selected Collection.

How long does it take to synchronize new data in one of my Connections?

Depending on the size of the data to synchronize, Dust syncs in minutes to several hours (up to c. 1/2 days). For larger synchronization, we recommend doing it later in the day to let the syncing happen overnight.

To check the last sync as an admin:

• Go to Build > Connections.
• Look for "last sync ~ x s ago."

For very large Google Drives, the synchronisation can sometimes take more time. If yours contains over c.50k files, do not hesitate to contact us so we check along the way.

How do I add data sources that are not supported as a Connection by Dust?

As a member, you can add your data to a connected platform like Notion or Google Drive. Ask an admin to verify if your added data are synchronized with Dust.

Admins/builders can add a Folders by:

• Going to Build > Folders.
• Clicking Add a new Folder.
• Naming it and adding a description (optional).
• Clicking create.
• Then upload your documents.

What are document uploads’ current size limits?

Documents up to 2MB can be uploaded manually via Folders and 10MB directly in the conversation.

How do I configure which data sources @dust has access to?

To configure the @dust assistant, got to Manage Assistants > Default tab and click on the Manage button next to the @dust assistant. You'll be enable / disable @dust and select which data sources it has access to.

Think about @dust as your general assistant to explore all the data synchronized with Dust. Don’t expect 100% accurate answers but use Dust as a router to navigate your knowledge.

Does Dust use user / company data to train models?

No, Dust does not use user or company data to retrain models, nor do model providers. Any data sent is retained for a limited time and this is strictly for debugging purposes.

You can have more details by checking the Security page on our website, our sub-processors list and our trust center.