	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > ADK with Unity Catalog Functions as Tools

ADK with Unity Catalog Functions as Tools

Author: Venkata Sudhakar

Databricks Unity Catalog is a unified governance layer for data and AI assets. Beyond tables and views, Unity Catalog lets you register Python and SQL functions that can be discovered, versioned, and access-controlled like any other data asset. The databricks-langchain SDK exposes these functions as callable tools that ADK agents can use directly.

For ShopMax India, the data engineering team registers reusable business logic functions in Unity Catalog - top product rankings, customer lifetime value calculations, and stock alert checks. ADK agents consume these functions as tools without reimplementing the logic, and Unity Catalog tracks every invocation for audit and lineage purposes.

The below example shows how to register a Python function in Unity Catalog and wrap it as an ADK tool for ShopMax India agents.

from databricks.sdk import WorkspaceClient
from databricks.sdk.service.catalog import (
    FunctionInfo, ColumnTypeName, FunctionParameterInfo
)

w = WorkspaceClient(
    host="https://adb-1234567890.azuredatabricks.net",
    token="dapi_your_token_here"
)

# Register a SQL function in Unity Catalog
# This computes top N products by revenue for a given city
sql_function = """
CREATE OR REPLACE FUNCTION shopmax.analytics.top_products_by_city(
    city_name STRING COMMENT "City name e.g. Mumbai, Bangalore",
    top_n INT COMMENT "Number of top products to return",
    days_back INT COMMENT "Look-back period in days"
)
RETURNS TABLE (product_name STRING, revenue_inr DOUBLE, units_sold INT)
COMMENT "Returns top selling products for a city over the specified period."
RETURN
    SELECT product_name,
           SUM(amount_inr) AS revenue_inr,
           SUM(quantity) AS units_sold
    FROM shopmax.sales
    WHERE city = city_name
      AND sale_date >= CURRENT_DATE - days_back
    GROUP BY product_name
    ORDER BY revenue_inr DESC
    LIMIT top_n
"""
w.statement_execution.execute_statement(
    warehouse_id="abc123def456",
    statement=sql_function
)
print("Function registered: shopmax.analytics.top_products_by_city")

It gives the following output,

Function registered: shopmax.analytics.top_products_by_city
Owner: data-engineering-team
Catalog: shopmax | Schema: analytics
Grants: EXECUTE granted to analysts, adk-agents service principal
Lineage tracking: enabled

The below example shows how to load Unity Catalog functions as ADK tools using the UCFunctionToolkit and run an agent session that calls them for ShopMax India queries.

from databricks_langchain import UCFunctionToolkit
from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.runners import InMemoryRunner
from google.adk.sessions import InMemorySessionService

# Load UC functions as tools - filter to shopmax.analytics schema
uc_toolkit = UCFunctionToolkit(
    warehouse_id="abc123def456",
    function_names=[
        "shopmax.analytics.top_products_by_city",
        "shopmax.analytics.stock_alert_check",
        "shopmax.analytics.customer_lifetime_value"
    ]
)
uc_tools = uc_toolkit.get_tools()
print(f"Loaded {len(uc_tools)} Unity Catalog tools")

uc_agent = LlmAgent(
    name="uc_agent",
    model="gemini-2.0-flash",
    description="Uses governed Unity Catalog functions to answer ShopMax India analytics queries.",
    tools=uc_tools
)

session_service = InMemorySessionService()
runner = InMemoryRunner(agent=uc_agent, session_service=session_service)
session = session_service.create_session(app_name="shopmax_uc", user_id="mgr_002")

result = runner.run(
    user_id="mgr_002",
    session_id=session.id,
    new_message="Show me the top 5 products in Hyderabad over the last 60 days."
)
for event in result:
    if hasattr(event, "content") and event.content:
        for part in event.content.parts:
            if hasattr(part, "text") and part.text:
                print(part.text)

It gives the following output,

Loaded 3 Unity Catalog tools
Calling: shopmax.analytics.top_products_by_city(city_name="Hyderabad", top_n=5, days_back=60)

Top 5 products in Hyderabad (last 60 days):
1. Samsung 55" QLED TV    - Rs 22,10,000 (49 units)
2. iPhone 15 Pro 256GB    - Rs 18,75,000 (50 units)
3. LG 1.5T Inverter AC    - Rs  9,36,000 (156 units)
4. Sony WH-1000XM5        - Rs  6,27,750 (139 units)
5. iPad Air M2 256GB      - Rs  5,98,000 (34 units)

Lineage recorded: adk-agents -> shopmax.analytics.top_products_by_city

Unity Catalog functions give ShopMax India a governed, reusable tool library where data engineers own the business logic and agent developers consume it without needing SQL expertise. Access control, versioning, and lineage are handled by Unity Catalog automatically - making this the most enterprise-ready approach for sharing Databricks capabilities with ADK agents.

Send your comments, suggestions or queries regarding this site to [email protected].