Using Databricks with SoterAI

Connect your Databricks workspace to SoterAI and query the data in your lakehouse directly inside your chats and workflows — no switching tabs, no exporting tables.

Before you connect

You'll need two details from your Databricks workspace: your Workspace URL and a personal access token. Have both ready before you start.

To create a personal access token:

Open your Databricks workspace and click your username in the top-right corner.
Go to Settings → Developer → Access tokens.
Click Generate new token, give it a name, and copy the token. It starts with dapi.

⚠️ Copy your token as soon as it's generated — Databricks only shows it once. If you lose it, generate a new one.

What SoterAI can access

Once connected, SoterAI has read-only access to:

Tables and schemas — the data structures in the workspace your token can reach
Query results — the rows returned when SoterAI runs a read query on your behalf
Metadata — table names, column names, and data types

Read-only access: SoterAI can read and analyse data in your Databricks workspace, but cannot create, edit, or delete tables, rows, or any other objects.

What SoterAI cannot do

Cannot create, edit, or delete tables, schemas, or rows
Cannot run write, update, or administrative operations
Cannot access data outside the workspace your token is scoped to
Cannot change workspace settings, permissions, or user access

How to connect

Click the puzzle piece icon in the chat input bar to open the Connectors panel. Find Databricks and click Connect.

A secure connection window opens. Enter the following:

Workspace URL — your Databricks workspace web address, for example https://dbc-abc123.cloud.databricks.com. Copy it from your browser address bar. It must match the workspace your token is from, with no trailing slash.
Access Token — the personal access token you created above (starts with dapi).

Click Connect Account and return to the chat — you're connected. Your connection is persistent, so you won't need to reconnect each session.

If the connection fails, double-check that the Workspace URL has no trailing slash and matches the workspace your token was generated in.

Example queries

Exploring your data

"What tables are available in my Databricks workspace?"
"Describe the columns in the incidents table"
"Show me a sample of rows from the safety_observations table"

Reading and analysing

"How many incidents were recorded last quarter in Databricks?"
"Summarise the injury records by site from my Databricks data"
"What are the top five hazard types in the observations table?"

Using Databricks data in workflows

"Pull last month's incident counts from Databricks and draft a summary report"
"Compare incident rates across sites using my Databricks tables"

Use cases

Querying enterprise safety data

Many enterprise customers store data from Applied Epic and other agency systems in Databricks. Connecting SoterAI lets you ask questions about that data in plain language, without writing SQL or exporting tables.

Cross-source reporting

Combine data from your Databricks lakehouse with SoterAI's safety analysis to build reports that draw on your organisation's full data picture.

Ad-hoc data exploration

Quickly check counts, trends, or specific records in your warehouse during a conversation, rather than opening the Databricks interface and building a query by hand.

Troubleshooting

Issue	Solution
Connection fails or returns an error	Check that the Workspace URL is correct, has no trailing slash, and matches the workspace your token is from. Confirm the access token is valid and starts with dapi.
Token no longer works	Personal access tokens can expire or be revoked. Generate a new token in Databricks (Settings → Developer → Access tokens) and reconnect via the Connectors panel.
SoterAI can't see a table	Check that the token's user has permission to view that table. SoterAI can only read data your token is authorised to access.
Slow responses on large tables	Large tables may take longer to query. Try narrowing your question — for example, limiting to a date range or specific columns.