Get started with media recommendations

Prepare the dataset

You use the Cloud Shell to import the Movielens dataset and restructure the dataset for Vertex AI Search for media.

Open the Cloud Shell

1. Open Google Cloud console.
2. Select your Google Cloud project.
3. Take note of the project ID in the Project info card on the dashboard page. You will need the project ID for the following procedures.

4. Click the Activate Cloud Shell button at the top of the console. A Cloud Shell session opens inside a new frame at the bottom of the Google Cloud console and displays a command-line prompt.

   

Import the dataset

The Movielens dataset is available in a public Cloud Storage bucket to make it easier to import.

1. Run the following using your project ID to set the default project for the command-line.

   gcloud config set project PROJECT_ID
   

2. Create a BigQuery dataset:

   bq mk movielens
   

3. Load movies.csv into a new movies BigQuery table:

   bq load --skip_leading_rows=1 movielens.movies \
     gs://cloud-samples-data/gen-app-builder/media-recommendations/movies.csv \
     movieId:integer,title,genres
   

4. Load ratings.csv into a new ratings BigQuery table:

   bq load --skip_leading_rows=1 movielens.ratings \
     gs://cloud-samples-data/gen-app-builder/media-recommendations/ratings.csv \
     userId:integer,movieId:integer,rating:float,time:timestamp
   

Create BigQuery views

In this step, you will restructure the Movielens dataset so it follows the expected format for media recommendations. Media recommendations require user events data in order to create a model. For this guide, you will create fake view-item events during the past 90 days from positive ratings (< 4).

1. Create a view that converts the movies table into the Google-defined Document schema:

   bq mk --project_id=PROJECT_ID \
    --use_legacy_sql=false \
    --view '
     WITH t AS (
       SELECT
         CAST(movieId AS string) AS id,
         SUBSTR(title, 0, 128) AS title,
         SPLIT(genres, "|") AS categories
         FROM `PROJECT_ID.movielens.movies`)
       SELECT
         id, "default_schema" as schemaId, null as parentDocumentId,
         TO_JSON_STRING(STRUCT(title as title, categories as categories,
         CONCAT("http://mytestdomain.movie/content/", id) as uri,
         "2023-01-01T00:00:00Z" as available_time,
         "2033-01-01T00:00:00Z" as expire_time,
         "movie" as media_type)) as jsonData
       FROM t;' \
   movielens.movies_view
   

   Now, the new view has the schema that the Vertex AI Agent Builder API expects.

2. Go to the BigQuery page in Google Cloud console.

   Go to BigQuery

3. In the Explorer pane, expand your project name, expand the movielens dataset and click movies_view to open the query page for this view.

   

4. Go to the Table explorer tab.

5. In the Generated query pane, click the Copy to query button. The query editor opens.

6. Click Run to see movie data in the view that you created.

7. Create fictitious user events from movie ratings by running the following Cloud Shell command:

   bq mk --project_id=PROJECT_ID \
    --use_legacy_sql=false \
    --view '
    WITH t AS (
     SELECT
       MIN(UNIX_SECONDS(time)) AS old_start,
       MAX(UNIX_SECONDS(time)) AS old_end,
       UNIX_SECONDS(TIMESTAMP_SUB(
       CURRENT_TIMESTAMP(), INTERVAL 90 DAY)) AS new_start,
       UNIX_SECONDS(CURRENT_TIMESTAMP()) AS new_end
     FROM `PROJECT_ID.movielens.ratings`)
     SELECT
       CAST(userId AS STRING) AS userPseudoId,
       "view-item" AS eventType,
       FORMAT_TIMESTAMP("%Y-%m-%dT%X%Ez",
       TIMESTAMP_SECONDS(CAST(
         (t.new_start + (UNIX_SECONDS(time) - t.old_start) *
         (t.new_end - t.new_start) / (t.old_end - t.old_start))
       AS int64))) AS eventTime,
       [STRUCT(movieId AS id, null AS name)] AS documents,
     FROM `PROJECT_ID.movielens.ratings`, t
     WHERE rating >= 4;' \
     movielens.user_events
   

Activate Vertex AI Agent Builder

1. In the Google Cloud console, go to the Agent Builder page.

   Agent Builder

2. Read and agree to the Terms of Service, then click Continue and activate the API.

Create an app for media recommendations

The procedures in this section guide you through creating and deploying a media recommendations app.

1. In the Google Cloud console, go to the Agent Builder page.

   Agent Builder

2. Click Create app.

3. On the Create app page, under Media recommendations, click Create.

4. In the App name field, enter a name for your app, such as quickstart-media-recommendations. Your app ID appears under the app name.

5. Under Recommendations type, make sure Others you may like is selected.

6. Under Business Objective, make sure Click-through rate (CTR) is selected.

7. Click Continue.

8. If you completed the Get started with media search tutorial and you still have the data store (suggested name quickstart-media-data-store), then select it, click Create, and skip to Train the recommendation model.

9. On the Data Stores page, click Create data store.

10. Enter a display name for your data store, such as quickstart-media-data-store, and then click Create.

11. Select the data store you just created, and then click Create to create your app.

Import data

Next, import the movies and user events data that were formatted earlier.

Import documents

1. Open the Requirements tab.

2. Click Import Documents.

3. Select BigQuery.

4. Enter the name of the movies BigQuery view you created and click Import.

   PROJECT_ID.movielens.movies_view
   

5. Wait until all documents have been imported, which should take about 15 minutes. There should be 86537 documents when complete.

   You can check the Activity tab for the import operation status. When the import is complete, the import operation status changes to Completed.

Import user events

1. Open the Requirements tab.

2. Click Import Events.

3. Select BigQuery.

4. Enter the name of the user_events BigQuery view you created and click Import.

   PROJECT_ID.movielens.user_events
   

5. Wait until at least a million events have been imported before proceeding to the next step, in order to meet the data requirements for training a new model.

   You can check the Activity tab for the operation status. The process takes about an hour to complete because you are importing millions of rows.

   It can take some time for the Requirements tab to update its status to Data requirements met.

Train the recommendation model

1. Go to the Configurations page.

2. Click the Serving tab. A serving config has already been created.

   If you want to adjust the Recommendation demotion or Result diversification settings, you can do so on this page.

3. Click the Training tab.

   After the data requirements have been met, the model begins training automatically. You can view the training and tuning status on this page.

   It might take a couple of days for the model to train and become ready to query. The Ready to query field indicates Yes when the process is complete. You need to refresh the page to see the change No to Yes.

Preview recommendations

After the model is ready to query:

1. In the navigation menu, click Preview.

2. Click the Document ID field. A list of document IDs appears.

3. Enter a seed document (movie) ID, such as 4993 for "The Lord of the Rings: The Fellowship of the Ring (2001)".

   

4. Select the Serving config name from the drop-down menu.

5. Click Get recommendations. A list of recommended documents appears.

Deploy your app for structured data

There is no recommendations widget for deploying your app. To test your app before deployment:

1. Go to the Data page, Documents tab, and copy a document ID.

2. Go to the Integration page. This page includes a sample command for the servingConfigs.recommend method in the REST API.

3. Paste the document ID you copied earlier into the Document ID field.

4. Leave the User Pseudo ID field as is.

5. Copy the example request and run it in Cloud Shell.

For help integrating the recommendations app into your web app, see the code samples at Get media recommendations.