Google Cloud Discovery Engine

Warnings

• gotcha Resource names often require the Google Cloud project *number* instead of the project *ID*. Ensure you retrieve and use the numeric project ID when constructing resource paths like `projects/{project_number}/locations/{location}/dataStores/{data_store_id}`.
  Fix: Use `gcloud projects describe <PROJECT_ID> --format='value(projectNumber)'` to get the project number, or ensure your environment variable provides the number.
• gotcha The `google-cloud-discoveryengine` library currently primarily exposes the `discoveryengine_v1beta` module. Direct imports from `google.cloud.discoveryengine` might lack access to the latest APIs or features, leading to `AttributeError` or missing functionalities.
  Fix: Always import specific client modules using the version suffix, e.g., `from google.cloud import discoveryengine_v1beta as discoveryengine`.
• gotcha Authentication is crucial. Without proper Google Cloud credentials (e.g., via `gcloud auth application-default login`, `GOOGLE_APPLICATION_CREDENTIALS` environment variable, or service account key), API calls will fail with permission errors.
  Fix: Ensure your environment is correctly authenticated for Google Cloud APIs. Grant the service account the 'Discovery Engine User' role for basic operations.
• gotcha Discovery Engine's search and recommendation features rely on 'Serving Configs'. If you don't specify a valid `serving_config` in your search or recommendation requests, operations will fail or return incomplete results.
  Fix: Verify your Data Store has a 'default_config' or provide the full resource name of a custom serving config, e.g., `projects/{project_number}/locations/{location}/dataStores/{data_store_id}/servingConfigs/default_config`.

Install

• pip install google-cloud-discoveryengine Install stable version

Imports

• SearchServiceClient

  from google.cloud import discoveryengine_v1beta as discoveryengine
  client = discoveryengine.SearchServiceClient()

  The API is currently in v1beta; ensure you import from discoveryengine_v1beta for the most current features and classes. Direct import from `discoveryengine` might point to an older or empty module.
• CompletionServiceClient

  from google.cloud import discoveryengine_v1beta as discoveryengine
  client = discoveryengine.CompletionServiceClient()

• RecommendationServiceClient

  from google.cloud import discoveryengine_v1beta as discoveryengine
  client = discoveryengine.RecommendationServiceClient()

• UserEventServiceClient

  from google.cloud import discoveryengine_v1beta as discoveryengine
  client = discoveryengine.UserEventServiceClient()

Quickstart

This quickstart demonstrates how to perform a basic search query using the Discovery Engine SearchServiceClient. It shows how to construct a request with a serving config and iterate through the search results. Remember to set up proper authentication and replace placeholder values with your specific Google Cloud project details. Discovery Engine often requires the project *number* instead of the project *ID* for resource paths.

import os
from google.cloud import discoveryengine_v1beta as discoveryengine

def search_discovery_engine(
    project_id: str,
    location_id: str,
    data_store_id: str,
    query: str,
):
    # Construct the parent resource name for the serving config.
    # Format: projects/{project_number}/locations/{location}/dataStores/{data_store_id}
    # Note: Discovery Engine often expects project *number*, not project *ID*.
    parent = (
        f"projects/{project_id}/locations/{location_id}/dataStores/{data_store_id}"
    )

    client = discoveryengine.SearchServiceClient()

    # The request body for a search operation.
    request = discoveryengine.SearchRequest(
        serving_config=(
            f"{parent}/servingConfigs/default_config" # Or a custom serving config
        ),
        query=query,
        query_expansion_spec=discoveryengine.QueryExpansionSpec(
            condition=discoveryengine.QueryExpansionSpec.Condition.AUTO
        ),
        page_size=5 # Limit to 5 results for brevity
    )

    response = client.search(request=request)

    print(f"\nSearch results for query: '{query}'")
    if not response.results:
        print("No results found.")
        return

    for result in response.results:
        print(f"-" * 20)
        print(f"Document ID: {result.document.id}")
        if result.document.derived_struct:
            # Derived_struct contains parsed content like title, link, snippets.
            title = result.document.derived_struct.get('title')
            link = result.document.derived_struct.get('link')
            snippet = result.document.derived_struct.get('snippets')
            print(f"Title: {title if title else 'N/A'}")
            print(f"Link: {link if link else 'N/A'}")
            if snippet:
                print(f"Snippet: {snippet[0].get('snippet')}")

# Example usage:
if __name__ == "__main__":
    # Replace with your actual Google Cloud Project Number, Location, and Data Store ID.
    # Ensure your service account has 'Discovery Engine User' role.
    project_number = os.environ.get("GCP_PROJECT_NUMBER", "YOUR_PROJECT_NUMBER")
    location = os.environ.get("GCP_LOCATION", "global") # e.g., 'global', 'us-central1'
    data_store_id = os.environ.get("DISCOVERY_ENGINE_DATA_STORE_ID", "YOUR_DATA_STORE_ID")
    search_query = "Google Cloud AI solutions"

    if "YOUR_PROJECT_NUMBER" in project_number or "YOUR_DATA_STORE_ID" in data_store_id:
        print("Please set GCP_PROJECT_NUMBER, GCP_LOCATION, and DISCOVERY_ENGINE_DATA_STORE_ID environment variables ")
        print("or replace the placeholder values in the quickstart code to run it.")
    else:
        search_discovery_engine(project_number, location, data_store_id, search_query)