Google ADK retail location strategy

Compare cities with maps, durable context, and current reporting.

What you’ll build

A Google ADK agent that compares cities for the first location of a specialty coffee roaster. It evaluates place-level competition, durable city context, and recent local developments before recommending one market.

Default brief: Choose between Austin, Raleigh, and Denver. Compare each city consistently and return evidence, tradeoffs, a recommendation, and source URLs.

How the agent works

  1. Inspect place patterns. Maps search looks for competitors and neighborhood signals at the local level.
  2. Add context and recency. Web search supplies durable city facts while news search finds recent local changes.
  3. Compare in one session. An in-memory ADK session keeps the evidence together while the agent builds the final strategy.

Core agent setup

This excerpt shows the ADK agent’s three evidence surfaces. The full script also configures its session service, transient-error retries, credentials, and Markdown output.

agent = Agent(
    name="serpapi_retail_location_strategist",
    model=MODEL,
    instruction=(
        "Use maps for place-level evidence, web for durable context, "
        "and news for current signals."
    ),
    tools=[
        maps_search(provider="google-adk"),
        web_search(provider="google-adk", allowed_engines=["google_light", "bing"]),
        news_search(provider="google-adk"),
    ],
)
runner = Runner(agent=agent, app_name=app_name, session_service=session_service)

async for event in runner.run_async(
    user_id=user_id,
    session_id=session_id,
    new_message=message,
):
    ...

View the complete main.py →

Run the recipe

Set SERPAPI_API_KEY and GEMINI_API_KEY (or GOOGLE_API_KEY) in the repository-root .env, then run:

uv run --isolated --no-project --with 'serpapi-search-tools[google-adk]' --with python-dotenv cookbook/google-adk/main.py

Optional controls: GOOGLE_ADK_MODEL, COOKBOOK_PROMPT, and COOKBOOK_OUTPUT_DIR.

Resilient live run: The recipe retries transient Gemini availability failures up to three times while preserving the same ADK session and task.

Inspect the result

The recipe writes cookbook-output/google-adk-location-strategy.md. Expect a consistent city comparison, place-level evidence, current local signals, tradeoffs, one recommendation, and direct source URLs.

Use COOKBOOK_PROMPT to replace the cities, retail concept, or decision criteria without changing the agent setup.

SerpApi tools used

maps_search reveals competitor and neighborhood patterns, web_search supplies durable city context, and news_search tracks recent local developments. Using all three helps the agent compare each city with the same place-level, contextual, and current evidence.

Framework pattern: The city-comparison workflow follows Google’s retail AI location strategy sample and uses SerpApi maps, web, and news tools.