SQL & Analytics

SQL Execution & Warehouses

Skills: databricks-dbsql MCP Tools: execute_sql, execute_sql_multi, list_warehouses, get_best_warehouse

List all SQL warehouses in my workspace and show which ones are running.

Run this SQL query: SELECT * FROM main.default.my_table LIMIT 10

Execute these SQL statements in order:
1. CREATE SCHEMA IF NOT EXISTS main.analytics
2. CREATE TABLE main.analytics.daily_summary AS
   SELECT date, COUNT(*) as cnt FROM main.raw.events GROUP BY date

Find the best available SQL warehouse and run a query to show the top 10 largest
tables in the main catalog.

Run a SQL query to get the row count and column count for every table in the
main.sales schema.

Execute a DESCRIBE EXTENDED on main.production.customers and explain what each
field means.

---

AI Functions (SQL)

Skills: databricks-ai-functions MCP Tools: execute_sql

Classification & Extraction

Run a SQL query that uses ai_classify to categorize customer feedback in
main.support.tickets into ['bug', 'feature_request', 'question', 'complaint'].

Use ai_extract to pull out product names, prices, and quantities from
unstructured order description text in main.sales.raw_orders.

Classify the sentiment of customer reviews in main.reviews.comments using
ai_analyze_sentiment and store results in a new column.

Summarization & Generation

Summarize the top 50 longest support tickets in main.support.cases using
ai_summarize, keeping summaries under 100 words each.

Use ai_gen to generate product descriptions for items in main.catalog.products
that are missing descriptions, based on the product name and category.

Use ai_fix_grammar to clean up user-submitted text in main.content.posts.

Advanced AI Functions

Use ai_query to call my fine-tuned model endpoint with a custom prompt template
that includes data from multiple columns.

Run ai_forecast on main.sales.daily_revenue to predict the next 30 days of
revenue with 90% prediction intervals.

Build a document parsing pipeline:
1. Use ai_parse_document to extract text from PDFs stored in a volume
2. Chunk the extracted text into 500-token segments
3. Store the chunks in a table ready for Vector Search indexing

Use ai_mask to redact PII (names, emails, phone numbers) from customer
communications before sharing with the analytics team.

Use ai_translate to translate product descriptions in main.catalog.products from
English to Spanish, French, and German.

Use ai_similarity to find duplicate customer records by comparing name + address
fields between two tables.

---

AI/BI Dashboards

Skills: databricks-aibi-dashboards MCP Tools: create_or_update_dashboard, get_dashboard, delete_dashboard, publish_dashboard

Create an AI/BI dashboard with a dataset querying main.sales.transactions and add:
- A counter widget showing total revenue
- A bar chart showing revenue by product category
- A line chart showing daily revenue trends over the last 90 days

Build a dashboard with two pages:
- Page 1 "Overview": KPI counters for total users, active users, and churn rate
- Page 2 "Details": a table widget with user-level data and a filter for date range

Create a dashboard with a global date filter that applies to all datasets,
showing customer acquisition metrics.

Get the JSON definition of my existing dashboard and help me add a new chart
widget to it.

Publish my dashboard so it's accessible to all workspace users.

List all dashboards in my workspace and show which ones are published vs draft.

Create a dashboard that uses pre-aggregated datasets for performance — the
source table has 500M rows so I need the SQL to aggregate before visualization.

---

Genie Spaces

Skills: databricks-genie MCP Tools: create_or_update_genie, get_genie, delete_genie, ask_genie, migrate_genie

Creating Genie Spaces

Create a Genie Space called "Sales Explorer" that connects to
main.sales.transactions and main.sales.customers with sample questions:
"What was total revenue last month?", "Who are our top 10 customers?",
"Show monthly revenue trends".

Create a Genie Space for our finance team that includes tables
main.finance.gl_entries, main.finance.budgets, and main.finance.forecasts
with instructions to always filter by the current fiscal year.

Querying Genie Spaces

Ask my Genie Space "Sales Explorer": What were the top 5 product categories by
revenue in Q4?

Ask the Genie Space: Show me customer churn rate by region for the last 6 months
as a trend.

Have a multi-turn conversation with my Genie Space — first ask about total
revenue, then drill down into the top-performing region.

Managing Genie Spaces

List all Genie Spaces in my workspace.

Export the Genie Space "Sales Explorer" so I can import it into another workspace.

Import a Genie Space configuration from a JSON file into my current workspace.

Migrate my Genie Space from the dev workspace to production.

---

Metric Views

Skills: databricks-metric-views MCP Tools: manage_metric_views, execute_sql

Create a metric view for "total_revenue" defined as SUM(amount) from
main.sales.transactions, with dimensions for product_category, region,
and order_date.

Define a metric view for "monthly_active_users" that counts distinct user_ids
with activity in a calendar month, with dimensions for platform and country.

Create a set of related metric views for e-commerce KPIs:
- total_orders (COUNT)
- average_order_value (AVG)
- total_revenue (SUM)
- customer_count (COUNT DISTINCT)
All sourced from main.sales.orders with shared dimensions.

Query my metric view "total_revenue" grouped by region and month for the
last year.

List all metric views in the main.metrics schema.

---

Databricks SQL Advanced Features

Skills: databricks-dbsql MCP Tools: execute_sql, execute_sql_multi

SQL Scripting & Stored Procedures

Write a SQL stored procedure that accepts a date parameter, runs a data quality
check on main.production.orders for that date, and returns a summary of any
violations found.

Create a SQL script with variables, loops, and conditional logic that backfills
daily aggregation tables for a specified date range.

Pipe Syntax

Rewrite this SQL query using Databricks pipe syntax (|>):
SELECT department, AVG(salary) as avg_salary
FROM main.hr.employees
WHERE status = 'active'
GROUP BY department
HAVING AVG(salary) > 100000
ORDER BY avg_salary DESC

Geospatial

Write a geospatial query using H3 functions to aggregate delivery data by
hexagonal regions at resolution 7 and find the top 10 busiest zones.

Use ST_ spatial functions to find all stores within a 10km radius of a given
latitude/longitude coordinate.

Materialized Views & Recursive CTEs

Create a materialized view main.gold.daily_revenue_summary that pre-aggregates
revenue by day, product category, and region from main.silver.transactions.

Write a recursive CTE to traverse an employee-manager hierarchy in
main.hr.employees and show the full reporting chain for each employee.