About data connections

AI Hub supports setting up connections to external systems, such as storage drives or mailboxes, which can function in some or all of the following ways:

• Input — Sources from which AI Hub can read files for processing in conversations, chatbots, automation apps, and deployments. These connected sources appear in the AI Hub file explorer or can be referenced by file path when using the AI Hub API.

• Output — Destinations where AI Hub can write processed run results from automation apps and deployments.

• Storage — Certain drive types can be designated as the default drive, making them the default storage location for all AI Hub resources, including automation project files, conversation files, temporary processing files, and configuration files.

  By default, the Instabase Drive (1 TB of included storage) serves as the default output and storage drive. However, organizations can use their own connected drive as a default drive to ensure all content remains within their own storage system.

  The default drive is also used as the default location for all output when no other destination is specified.

Additionally, some data connections can be used for deployment-specific input and output integrations:

• Upstream integration — A deployment configuration that automatically pulls files or folder contents from connected sources for processing.

• Downstream integration — A deployment configuration that automatically sends processed results to external systems.

  Downstream integrations aren’t limited to connected data sources. Other destinations, such as email addresses, are supported.

Review the following table for an overview of which data connections support each capability:

Adding data connections

You can add data connections from the Data tab of any workspace. See the following configuration guides for details.

• Connecting drives

• Connecting mailboxes

Connection guidelines

When adding data connections, admins have the option to select an audience: organization members or workspace members. Selecting Organization members establishes the connection at the organization level, making the data source available to every workspace. Data connections limited to workspace members are available only in the selected workspace.

In general, when adding a data connection, select Workspace members as the audience. AI Hub uses workspaces to maintain separation of access and resources, and workspace-level connections respect this separation. Additional guidelines include:

• Add data connections only in the workspaces where they’re needed. You can connect the same data source to multiple workspaces if you need access across workspaces.

  When the same drive is connected to multiple workspaces, any output files are saved to workspace-specific folders in the drive. Those folders and their contents are visible in all workspaces that have access to the drive.

• Connect a drive at the organization level only if intending to use it as the organization default drive.

  Adding drives at the organization level isn’t recommended as a shortcut to supporting input and output across all workspaces for the following reasons:

  • Drives configured at the organization level can’t be used for input.

  • While organization drives can be referenced as an output destination by using the output_dir parameter when running an automation app or deployment by API, this isn’t the preferred approach. Instead, accept the deployment configuration for deployment runs or use the output_workspace parameter for app runs.

Additional input and output options

Beyond connected data sources, some AI Hub features support alternative input and output options. Support varies between features and can differ between the user interface and API.

Additional input options include:

• Local upload — You can upload files from your local machine to conversations and to automation app or deployment runs. Local upload is also supported when adding files to a conversation or app or deployment run by API.

• Upload from URL — Conversations support uploading content from a URL. Content at the destination is captured as a PDF.

Additional output options include:

• Manual export — Automation app or deployment results can be manually exported to your local machine—or a connected drive—in .csv or .xlsx format. This option is available in the user interface only. Open an app or deployment run, click the dropdown indicator next to the app name, then select Export results.