Databricks
SQL-Abfragen ausführen und Jobs auf Databricks verwalten
Databricks is a unified data analytics platform built on Apache Spark, providing a collaborative environment for data engineering, data science, and machine learning. Databricks combines data warehousing, ETL, and AI workloads into a single lakehouse architecture, with support for SQL analytics, job orchestration, and cluster management across major cloud providers.
With the Databricks integration in AACFlow, you can:
- Execute SQL queries: Run SQL statements against Databricks SQL warehouses with support for parameterized queries and Unity Catalog
- Manage jobs: List, trigger, and monitor Databricks job runs programmatically
- Track run status: Get detailed run information including timing, state, and output results
- Control clusters: List and inspect cluster configurations, states, and resource details
- Retrieve run outputs: Access notebook results, error messages, and logs from completed job runs
In AACFlow, the Databricks integration enables your agents to interact with your data lakehouse as part of automated workflows. Agents can query large-scale datasets, orchestrate ETL pipelines by triggering jobs, monitor job execution, and retrieve results—all without leaving the workflow canvas. This is ideal for automated reporting, data pipeline management, scheduled analytics, and building AI-driven data workflows that react to query results or job outcomes.
Connect to Databricks to execute SQL queries against SQL warehouses, trigger and monitor job runs, manage clusters, and retrieve run outputs. Requires a Personal Access Token and workspace host URL.
Execute a SQL statement against a Databricks SQL warehouse and return results inline. Supports parameterized queries and Unity Catalog.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
warehouseId | string | Ja | The ID of the SQL warehouse to execute against |
statement | string | Ja | The SQL statement to execute (max 16 MiB) |
catalog | string | Nein | Unity Catalog name (equivalent to USE CATALOG) |
schema | string | Nein | Schema name (equivalent to USE SCHEMA) |
rowLimit | number | Nein | Maximum number of rows to return |
waitTimeout | string | Nein | How long to wait for results (e.g., "50s"). Range: "0s" or "5s" to "50s". Default: "50s" |
| Parameter | Typ | Beschreibung |
|---|
statementId | string | Unique identifier for the executed statement |
status | string | Execution status (SUCCEEDED, PENDING, RUNNING, FAILED, CANCELED, CLOSED) |
columns | array | Column schema of the result set |
↳ name | string | Column name |
↳ position | number | Column position (0-based) |
↳ typeName | string | Column type (STRING, INT, LONG, DOUBLE, BOOLEAN, TIMESTAMP, DATE, DECIMAL, etc.) |
data | array | Result rows as a 2D array of strings where each inner array is a row of column values |
totalRows | number | Total number of rows in the result |
truncated | boolean | Whether the result set was truncated due to row_limit or byte_limit |
List all jobs in a Databricks workspace with optional filtering by name.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
limit | number | Nein | Maximum number of jobs to return (range 1-100, default 20) |
offset | number | Nein | Offset for pagination |
name | string | Nein | Filter jobs by exact name (case-insensitive) |
expandTasks | boolean | Nein | Include task and cluster details in the response (max 100 elements) |
| Parameter | Typ | Beschreibung |
|---|
jobs | array | List of jobs in the workspace |
↳ jobId | number | Unique job identifier |
↳ name | string | Job name |
↳ createdTime | number | Job creation timestamp (epoch ms) |
↳ creatorUserName | string | Email of the job creator |
↳ maxConcurrentRuns | number | Maximum number of concurrent runs |
↳ format | string | Job format (SINGLE_TASK or MULTI_TASK) |
hasMore | boolean | Whether more jobs are available for pagination |
nextPageToken | string | Token for fetching the next page of results |
Trigger an existing Databricks job to run immediately with optional job-level or notebook parameters.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
jobId | number | Ja | The ID of the job to trigger |
jobParameters | string | Nein | Job-level parameter overrides as a JSON object (e.g., {"key": "value"}) |
notebookParams | string | Nein | Notebook task parameters as a JSON object (e.g., {"param1": "value1"}) |
idempotencyToken | string | Nein | Idempotency token to prevent duplicate runs (max 64 characters) |
| Parameter | Typ | Beschreibung |
|---|
runId | number | The globally unique ID of the triggered run |
numberInJob | number | The sequence number of this run among all runs of the job |
Get the status, timing, and details of a Databricks job run by its run ID.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
runId | number | Ja | The canonical identifier of the run |
includeHistory | boolean | Nein | Include repair history in the response |
includeResolvedValues | boolean | Nein | Include resolved parameter values in the response |
| Parameter | Typ | Beschreibung |
|---|
runId | number | The run ID |
jobId | number | The job ID this run belongs to |
runName | string | Name of the run |
runType | string | Type of run (JOB_RUN, WORKFLOW_RUN, SUBMIT_RUN) |
attemptNumber | number | Retry attempt number (0 for initial attempt) |
state | object | Run state information |
↳ lifeCycleState | string | Lifecycle state (QUEUED, PENDING, RUNNING, TERMINATING, TERMINATED, SKIPPED, INTERNAL_ERROR, BLOCKED, WAITING_FOR_RETRY) |
↳ resultState | string | Result state (SUCCESS, FAILED, TIMEDOUT, CANCELED, SUCCESS_WITH_FAILURES, UPSTREAM_FAILED, UPSTREAM_CANCELED, EXCLUDED) |
↳ stateMessage | string | Descriptive message for the current state |
↳ userCancelledOrTimedout | boolean | Whether the run was cancelled by user or timed out |
startTime | number | Run start timestamp (epoch ms) |
endTime | number | Run end timestamp (epoch ms, 0 if still running) |
setupDuration | number | Cluster setup duration (ms) |
executionDuration | number | Execution duration (ms) |
cleanupDuration | number | Cleanup duration (ms) |
queueDuration | number | Time spent in queue before execution (ms) |
runPageUrl | string | URL to the run detail page in Databricks UI |
creatorUserName | string | Email of the user who triggered the run |
List job runs in a Databricks workspace with optional filtering by job, status, and time range.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
jobId | number | Nein | Filter runs by job ID. Omit to list runs across all jobs |
activeOnly | boolean | Nein | Only include active runs (PENDING, RUNNING, or TERMINATING) |
completedOnly | boolean | Nein | Only include completed runs |
limit | number | Nein | Maximum number of runs to return (range 1-24, default 20) |
offset | number | Nein | Offset for pagination |
runType | string | Nein | Filter by run type (JOB_RUN, WORKFLOW_RUN, SUBMIT_RUN) |
startTimeFrom | number | Nein | Filter runs started at or after this timestamp (epoch ms) |
startTimeTo | number | Nein | Filter runs started at or before this timestamp (epoch ms) |
| Parameter | Typ | Beschreibung |
|---|
runs | array | List of job runs |
↳ runId | number | Unique run identifier |
↳ jobId | number | Job this run belongs to |
↳ runName | string | Run name |
↳ runType | string | Run type (JOB_RUN, WORKFLOW_RUN, SUBMIT_RUN) |
↳ state | object | Run state information |
↳ lifeCycleState | string | Lifecycle state (QUEUED, PENDING, RUNNING, TERMINATING, TERMINATED, SKIPPED, INTERNAL_ERROR, BLOCKED, WAITING_FOR_RETRY) |
↳ resultState | string | Result state (SUCCESS, FAILED, TIMEDOUT, CANCELED, SUCCESS_WITH_FAILURES, UPSTREAM_FAILED, UPSTREAM_CANCELED, EXCLUDED) |
↳ stateMessage | string | Descriptive state message |
↳ userCancelledOrTimedout | boolean | Whether the run was cancelled by user or timed out |
↳ startTime | number | Run start timestamp (epoch ms) |
↳ endTime | number | Run end timestamp (epoch ms) |
hasMore | boolean | Whether more runs are available for pagination |
nextPageToken | string | Token for fetching the next page of results |
Cancel a running or pending Databricks job run. Cancellation is asynchronous; poll the run status to confirm termination.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
runId | number | Ja | The canonical identifier of the run to cancel |
| Parameter | Typ | Beschreibung |
|---|
success | boolean | Whether the cancel request was accepted |
Get the output of a completed Databricks job run, including notebook results, error messages, and logs. For multi-task jobs, use the task run ID (not the parent run ID).
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
runId | number | Ja | The run ID to get output for. For multi-task jobs, use the task run ID |
| Parameter | Typ | Beschreibung |
|---|
notebookOutput | object | Notebook task output (from dbutils.notebook.exit()) |
↳ result | string | Value passed to dbutils.notebook.exit() (max 5 MB) |
↳ truncated | boolean | Whether the result was truncated |
error | string | Error message if the run failed or output is unavailable |
errorTrace | string | Error stack trace if available |
logs | string | Log output (last 5 MB) from spark_jar, spark_python, or python_wheel tasks |
logsTruncated | boolean | Whether the log output was truncated |
List all clusters in a Databricks workspace including their state, configuration, and resource details.
| Parameter | Typ | Erforderlich | Beschreibung |
|---|
host | string | Ja | Databricks workspace host (e.g., dbc-abc123.cloud.databricks.com) |
apiKey | string | Ja | Databricks Personal Access Token |
| Parameter | Typ | Beschreibung |
|---|
clusters | array | List of clusters in the workspace |
↳ clusterId | string | Unique cluster identifier |
↳ clusterName | string | Cluster display name |
↳ state | string | Current state (PENDING, RUNNING, RESTARTING, RESIZING, TERMINATING, TERMINATED, ERROR, UNKNOWN) |
↳ stateMessage | string | Human-readable state description |
↳ creatorUserName | string | Email of the cluster creator |
↳ sparkVersion | string | Spark runtime version (e.g., 13.3.x-scala2.12) |
↳ nodeTypeId | string | Worker node type identifier |
↳ driverNodeTypeId | string | Driver node type identifier |
↳ numWorkers | number | Number of worker nodes (for fixed-size clusters) |
↳ autoscale | object | Autoscaling configuration (null for fixed-size clusters) |
↳ minWorkers | number | Minimum number of workers |
↳ maxWorkers | number | Maximum number of workers |
↳ clusterSource | string | Origin (API, UI, JOB, MODELS, PIPELINE, PIPELINE_MAINTENANCE, SQL) |
↳ autoterminationMinutes | number | Minutes of inactivity before auto-termination (0 = disabled) |
↳ startTime | number | Cluster start timestamp (epoch ms) |
