Skip to main content

Set up Genie Search

Genie Search builds and maintains a search index across the sources where your organization keeps its knowledge: Google Drive, Jira, Confluence, and Workday. Databricks ingests and indexes the content along with each document's permissions, and keeps both current as they change.

With Genie Search, Genie searches one index that covers all of your connected sources, instead of querying each source separately. Genie Search also extends Genie Ontology beyond Databricks, so Genie can draw on your documents and tickets as context when it answers questions. Users see only the content they can already open in the source.

Workday also powers org directories in Genie One. See Connect a people source for Genie One.

Requirements​

  • Databricks account admin privileges
  • An admin for each source you connect, such as a Google Workspace or Atlassian admin, to authorize access. This can be a different person from the Databricks account admin.

Supported sources​

Genie Search indexes the following sources.

Source

What's indexed

How you authenticate

Google Drive

Google Workspace files, Microsoft Office files, PDFs, and text files. See the supported file types in Limitations.

Google Cloud service account with domain-wide delegation

Jira Cloud

All projects that the authorizing user can access

Your own Atlassian OAuth 2.0 app

Confluence Cloud

All spaces that the authorizing user can view

Your own Atlassian OAuth 2.0 app

Workday

Worker profiles from a custom Workday Reports-as-a-Service (RaaS) report, for org directories in Genie One

See Connect a people source for Genie One.

Source

What's indexed

How you authenticate

Google Drive

Google Workspace files, Microsoft Office files, PDFs, and text files. See the supported file types in Limitations.

Google Cloud service account with domain-wide delegation

Jira Cloud

All projects that the authorizing user can access

Your own Atlassian OAuth 2.0 app

Confluence Cloud

All spaces that the authorizing user can view

Your own Atlassian OAuth 2.0 app

Workday

Worker profiles from a custom Workday Reports-as-a-Service (RaaS) report, for org directories in Genie One

See Connect a people source for Genie One.

Connect a source​

Start each connection from the account console.

  1. In the account console, click Genie Ontology.
  2. Click Connect source.
  3. Select the source.
  4. In Data residency Geo, select the geo where the source's index is stored. You can select only a geo where your account has a workspace. If the account already has a source, select the same geo. See Storage and location.

Then complete the steps for your source:

You need a Google Cloud project where you can create a service account, and a Google Workspace admin who can grant domain-wide delegation in the Google Admin console.

  1. Create a service account and a JSON key:

    1. In the Google Cloud console, go to IAM & Admin > Service Accounts.
    2. Click Create service account.
    3. Enter a name and description for the service account, then click Create and continue.
    4. Click Done. You don't need to grant project-level roles.
    5. On the Service Accounts page, click the service account that you created.
    6. On the Keys tab, click Add key > Create new key.
    7. Select JSON, then click Create. The JSON key file downloads to your machine.
  2. Enable the Google APIs that the service account uses:

    1. In the Google Cloud console, go to APIs & Services > Library.
    2. Search for Google Drive API, click it, and click Enable.
    3. Search for Admin SDK API, click it, and click Enable.
  3. Grant domain-wide delegation:

    1. In the Google Cloud console, copy the service account's Unique ID (client ID) from the Details tab.

    2. In the Google Admin console, go to Security > Access and data control > API controls.

    3. Under Domain-wide delegation, click Manage domain-wide delegation.

    4. Click Add new.

    5. In Client ID, paste the service account's unique ID.

    6. In OAuth scopes, enter the following read-only scopes, separated by commas:

      Text
      https://www.googleapis.com/auth/drive.readonly,https://www.googleapis.com/auth/admin.directory.user.readonly,https://www.googleapis.com/auth/admin.directory.group.readonly,https://www.googleapis.com/auth/admin.reports.audit.readonly
    7. Click Authorize.

  4. In the Connect Google Drive dialog in Databricks, paste the full contents of the JSON key, including the opening and closing braces, into Service account JSON.

  5. In Admin email, enter the email of a Google Workspace admin who can read users, groups, and shared drives. The service account acts as this admin.

  6. Click Start ingestion.

If you skip any of the delegation, API, or scope steps, ingestion can fail or miss content until you finish that setup.

Initial indexing can take hours or days, depending on how much content is in the source. The source status is Provisioning until indexing finishes, then Searchable. If setup fails, the status is Failed.

Use indexed content​

After a source is Searchable, Genie One chat uses Genie Search automatically when it answers questions. Users can also find indexed content in Genie One Discover, alongside Databricks assets. See Use Genie Search in Genie One.

Genie Search and MCP connectors​

Genie Search and Model Context Protocol (MCP) connectors work together, and Databricks recommends enabling both for the sources your teams rely on. Genie Search lets Genie search everything an account admin has indexed in one place. MCP connectors let each user connect applications such as Gmail and Slack with their own credentials, so Genie One can look up live information and take action, such as drafting an email, posting to a Slack channel, or updating a Jira issue.

When both are enabled, Genie chooses the approach that fits each request. For example, it can use Genie Search to find the design docs, wikis, and tickets about a launch, then use the Atlassian connector to update the related Jira issues.

Aspect

Genie Search

MCP connectors

Setup

An account admin connects each source one time for the whole account.

Each user signs in to each connector with their own credentials.

Search

Searches all indexed content across connected sources at the same time.

Queries each source live. Genie chooses which connectors to call for each request.

Freshness

Shows changes after the next index refresh.

Returns the current state of the source.

Actions

Read only.

Creates and updates content in the application as the user. See Write actions.

Permissions

Enforces each document's source permissions. Permissions refresh with content, and non-public documents are rechecked at query time.

Uses the user's own permissions in the application.

Governance

Account admins manage each source on the Genie Ontology page in the account console.

Admins control connectors, tools, and policies in Unity Gateway. See Govern an MCP.

Sources

Google Drive, Jira, Confluence, and Workday.

Databricks-provided connectors and custom MCP services. See Connect Genie One to external applications with MCP.

Aspect

Genie Search

MCP connectors

Setup

An account admin connects each source one time for the whole account.

Each user signs in to each connector with their own credentials.

Search

Searches all indexed content across connected sources at the same time.

Queries each source live. Genie chooses which connectors to call for each request.

Freshness

Shows changes after the next index refresh.

Returns the current state of the source.

Actions

Read only.

Creates and updates content in the application as the user. See Write actions.

Permissions

Enforces each document's source permissions. Permissions refresh with content, and non-public documents are rechecked at query time.

Uses the user's own permissions in the application.

Governance

Account admins manage each source on the Genie Ontology page in the account console.

Admins control connectors, tools, and policies in Unity Gateway. See Govern an MCP.

Sources

Google Drive, Jira, Confluence, and Workday.

Databricks-provided connectors and custom MCP services. See Connect Genie One to external applications with MCP.

Manage sources​

Account admins manage connected sources on the Genie Ontology page in the account console.

To stop serving a Google Drive, Jira, or Confluence source without deleting its indexed content, turn off its Enabled switch. Genie stops using the source. You can turn it back on later.

To replace the credentials of a Google Drive, Jira, or Confluence source without deleting its indexed content, click the Kebab menu icon. kebab menu on the source row, then click Re-authorize. To revoke a credential, remove the source.

To remove a source:

  1. Click the Kebab menu icon. kebab menu on the source row.
  2. Click Remove source.
  3. Confirm. Ingestion stops, indexed content is deleted, and reconnecting requires a full ingestion.

How Genie Search handles your data​

Genie Search stores a copy of your indexed content and its permissions in Databricks. The following sections describe where that data lives, how source credentials and permissions are handled, and how long data is retained.

Storage and location​

Each account gets a dedicated catalog, and each source gets its own search index. Indexed data is stored in the data plane and isolated from other customers.

Indexed data stays in a single geo, which an account admin chooses when they connect a source. All sources in the account use the same geo. Users can query indexed data from any workspace in that geo, and queries from workspaces in other geos are blocked. Account-level Genie One chats can use indexed data when their region is in the same geo.

Credentials​

The credential you provide when you connect a source is stored in a Unity Catalog connection in a Databricks-managed account. Databricks uses it only to ingest the data you authorize. To replace or revoke a credential, see Manage sources.

Permissions​

Databricks ingests each document's permissions along with its content and enforces them when Genie retrieves results, so users see only the content they can already access in the source.

Permissions refresh at the same cadence as content, on a schedule that varies by source and depends on factors such as source API limits and data volume. For non-public documents, Genie Search also rechecks access against the source at query time, using a short cache.

Org directories are the exception. Everyone who uses Genie One in your account sees the same profile details. See What is ingested.

Retention and deletion​

Content that's deleted in the source is removed from the index during the next refresh. To delete all of a source's indexed data, remove the source. Reconnecting it later requires a full ingestion. To stop using a source without deleting its data, turn off the source instead. See Manage sources.

Limitations​

The following limitations apply to Genie Search.

  • Genie Search isn't available in AWS GovCloud regions.
  • Genie Search content is available in Genie One only. Genie Code doesn't use it.
  • You can't exclude content from a source, such as specific Google Drive folders, Jira projects, or Confluence spaces.
  • Google Drive indexes only the following file types. Other files, such as images and videos, are skipped.
    • Google Docs, Google Sheets, and Google Slides
    • Microsoft Word (.doc and .docx)
    • Microsoft Excel (.xls and .xlsx)
    • Microsoft PowerPoint (.ppt and .pptx)
    • PDF (.pdf)
    • Plain text (.txt)
    • Markdown (.md)
    • Rich Text Format (.rtf)
    • CSV (.csv)
  • Google Drive files larger than 100 MB are skipped, and files larger than 10 MB keep metadata only.
  • PDFs are indexed from extracted text, without optical character recognition (OCR).
  • Only Jira Cloud and Confluence Cloud are supported. On-premises Jira and Confluence are not.
  • Jira and Confluence results follow each item's permissions, but Genie Search doesn't check whether the user still has a Jira or Confluence license.
  • There is no API for managing sources. Setup is in the account console.
  • All sources in an account use the same geo. To change the geo, remove all sources, then connect them again.

Pricing​

At this time, customers do not incur costs for Genie Search. However, Databricks may update pricing for the feature at any time by notifying you and/or by updating our pricing page.