Menu
Run Databricks SQL queries, saved reports, and AI/BI Genie prompts
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Generate Quarto and HTML reports of local Codex session cost, token usage by model, project, coding vs cowork cost, weekly trends, annual projections, and top expensive sessions.
Check that core Sidekick integrations can authenticate and read common restricted files
Generate summary of 1:1 and meeting notes organized by audience, wins, and kudos
Capture missing TODOs from recent work context. Use when Codex needs to scan the user's Slack DMs with their manager or direct reports, manager/direct-report 1:1 docs, and leadership meeting notes for direct asks or action items assigned to the user, deduplicate against the user's configured todo app, and create missing tasks with source links and searchable metadata.
Draft concise interview scorecards from pasted long-form interview notes, nearby Google Calendar interview context, and memory/miclog/ transcript excerpts. Use when writing hiring feedback, interview scorecards, spikes/troughs, or candidate evaluation summaries.
Calculate a team's average Jira hours from Rovo Jira completed work, with optional Workday Approved Time Off absence adjustments for a date window and team roster.
| name | databricks |
| description | Run Databricks SQL queries, saved reports, and AI/BI Genie prompts |
| argument-hint | <operation> [args] |
| allowed-tools | Bash, Read |
Use the local Databricks client for Databricks SQL and AI/BI Genie work: $ARGUMENTS
Required in .env:
DATABRICKS_HOST=https://dropbox-prod.cloud.databricks.com
DATABRICKS_TOKEN=...
Optional defaults:
DATABRICKS_WAREHOUSE_ID=...
DATABRICKS_CATALOG=...
DATABRICKS_SCHEMA=...
DATABRICKS_GENIE_SPACE_ID=...
python3 -m sidekick.clients.databricks warehouses
Use this first when DATABRICKS_WAREHOUSE_ID is not configured.
python3 -m sidekick.clients.databricks query "SELECT 1" --warehouse-id WAREHOUSE_ID
With named parameters:
python3 -m sidekick.clients.databricks query \
"SELECT * FROM catalog.schema.table WHERE id = :id" \
--warehouse-id WAREHOUSE_ID \
--param id:STRING=abc123
python3 -m sidekick.clients.databricks query-file path/to/query.sql --warehouse-id WAREHOUSE_ID
Saved Databricks SQL queries are treated as reports by this client.
python3 -m sidekick.clients.databricks list-reports
python3 -m sidekick.clients.databricks list-reports "weekly active"
python3 -m sidekick.clients.databricks get-report "Report Name"
python3 -m sidekick.clients.databricks run-report "Report Name"
Saved reports can be addressed by query ID, exact display name, or a unique name substring.
For AI/BI dashboards, the client fetches the dashboard definition, extracts SQL datasets from serialized_dashboard, and executes those datasets on the dashboard warehouse.
python3 -m sidekick.clients.databricks dashboard-results "https://workspace/dashboardsv3/DASHBOARD_ID/published"
Run one dataset only:
python3 -m sidekick.clients.databricks dashboard-results DASHBOARD_ID --dataset DATASET_NAME
Use --row-limit N to cap fetched rows per dataset. This re-executes dashboard dataset SQL with the configured Databricks token, so dashboards published with embedded credentials can still fail if the token lacks direct table or schema permissions.
python3 -m sidekick.clients.databricks genie-spaces
Use this first when DATABRICKS_GENIE_SPACE_ID is not configured.
python3 -m sidekick.clients.databricks ask "How many active teams did we have last week?" --space-id SPACE_ID
The ask command starts or continues a Genie conversation, waits for completion, prints generated SQL when Genie returns it, and fetches query attachment results.
Continue a conversation:
python3 -m sidekick.clients.databricks ask \
"Break that down by team" \
--space-id SPACE_ID \
--conversation-id CONVERSATION_ID
python3 -m sidekick.clients.databricks genie-result SPACE_ID CONVERSATION_ID MESSAGE_ID ATTACHMENT_ID
Use --execute to execute or re-execute an attachment query when the cached result has expired.
By default, commands print compact tables for humans. Add --json for raw API payloads when another tool needs to parse the response.