Sync Databricks data to Enterpret

Sync user profiles, account attributes, and customer feedback from Databricks tables or views into Enterpret. Use user and account data to segment feedback by attributes such as user role or account plan.

Beta availability. If Databricks is missing from your integrations catalog, contact your Enterpret team to ask about enabling it for your workspace.

Before you connect

• Use an Enterpret Owner or Admin account with permission to manage integrations.
• Have your Databricks workspace URL, SQL warehouse ID, and the catalog, schema, and name of the table or view you want to sync.
• Get a Databricks personal access token for a user or service principal in that workspace. For a scoped token, include the Databricks SQL (sql) scope. The user or service principal needs CAN USE on the SQL warehouse, USE CATALOG on the catalog, USE SCHEMA on the schema, and SELECT on the table or view. See how to create a Databricks personal access token.

Use a view if you want to limit which fields Enterpret imports. Enterpret imports the exposed columns as metadata, except for columns mapped to feedback content. Include any identifiers you want to use to link feedback to users or accounts.

Connect in Enterpret

1. Go to Settings → Integrations → New Integration. Search for Databricks and select Continue.
2. Choose User Integration for individual profiles, Account Integration for organizations, or Feedback Integration for reviews, conversations, or survey responses. Create a separate connection for each source you want to sync.
3. Enter an Integration name, the Workspace URL, and the SQL warehouse ID. Use the HTTPS workspace URL without a path or query string. Copy the warehouse ID from Databricks.
4. Fill in Catalog, Schema, and Table or view. Paste your token into Personal access token.

Example connection details for a Databricks Feedback integration.

Map your columns

Under Column mapping, enter the source column name for each required field.

• ID column. Use a stable, nonempty identifier. The table or view should return one current row per ID.
• Created-at column. Use a non-null Databricks TIMESTAMP that records when the row was created.
• Updated-at column. Use a non-null Databricks TIMESTAMP that advances whenever the row changes. Enterpret uses it to find new and changed rows.

Example mapping for review feedback. User and Account connections use the same ID and timestamp fields.

Feedback content

For a Feedback connection, map at least one content column and provide content in each row. The JSON examples below show the contents of one cell.

Review column. Map a plain-text column containing the review.

Conversation JSON column. Map a JSON text column with a msgs array of messages.

{
  "msgs": [
    {
      "actor": "user",
      "text": "Export is slow."
    }
  ]
}

Survey JSON column. Map a JSON text column with a responses array. Give each response a nonempty question and include its answer or selectedOptions.

{
  "responses": [
    {
      "question": "How was setup?",
      "answer": "Easy"
    }
  ]
}

When a row has multiple mapped content columns, Enterpret checks conversation, survey, then review, and uses the first populated column. Malformed JSON in that column can interrupt ingestion.

Save and verify

For a Feedback connection, fill in the required description under Describe the type of feedback being integrated. Explain what feedback you’re importing, who submitted it, and what you want to learn from it.

Example description for feedback from a post-support survey.

1. Select Validate and Integrate to save the connection.
2. Open the saved integration. Go to Integration Details → Connection and select Validate Connection. Check Sample Data Preview for the expected IDs, timestamps, and data.

Sample Data Preview with five synthetic feedback rows.

After previewing the sample rows, check Status and Last Ingested Record in Integration Details to monitor the sync. If your workspace shows separate timestamps, use Latest Ingestion Timestamp to see when Enterpret last ingested a record.

Integration Details during the first sync, before any records have arrived.

After the first sync, check that the expected data appears in Enterpret. To enrich feedback with user or account attributes, configure links from the integration page using fields with matching values, such as a user ID or account ID. See how to link users and accounts to feedback.

By default, the first sync looks back one year from the date you create the connection, using the updated-at column. Later syncs recheck a 24-hour overlap to pick up late updates.

Edit the description

Open the saved Feedback integration and select Edit. Update the feedback description, leave Replacement personal access token blank to keep your existing token, and select Validate and Save.

Edit the description with the replacement token field empty.

Troubleshooting

• Validation fails. Confirm the token is active in the specified workspace, its identity has the required grants, and the warehouse ID and table or view names are correct.
• The token expired or was revoked. Edit the integration, enter a Replacement personal access token, then select Validate and Save.
• Rows are missing or stale. Check that mapped timestamps are populated, the updated-at value advances on every change, and feedback JSON is valid.

User and account records remain in Enterpret when their source rows are removed from Databricks.

To use Databricks tools in Agent sessions, see Connect Databricks to Enterpret Agent.