Vertex AI Search and Conversation has been renamed to Vertex AI Agent Builder. We are in the process of updating content to reflect the new branding.

Create a healthcare search data store

Caution:

Restrictions for healthcare: As a customer, you will not, and will not allow End Users to, use the Generative AI Services for clinical purposes (for clarity, non-clinical research, scheduling, or other administrative tasks is not restricted), as a substitute for professional medical advice, or in any manner that is overseen by or requires clearance or approval from any applicable regulatory authority. For more information, see Service Specific Terms.
For clarity, with respect to the use of Vertex AI Search to retrieve and summarize existing medical information, the restriction on the use for clinical purposes means the restriction on the use for direct diagnosis or treatment purposes without review by a licensed professional in compliance with applicable laws and regulations.
The generated output may not always be completely reliable. Due to the nature of LLMs and Generative AI, outputs may have incorrect or biased (for example, stereotypes or other harmful content) information and should be reviewed. All summaries or answers should be considered draft and not final.
This product's intended usage is not to provide information pertaining to the prevention, diagnosis or treatment of illness or disease. Questions regarding diagnosis or treatment recommendations are not intended to be addressed by the product. This product's intended use is to retrieve and summarize existing medical information provided by users.
Due to limited test data, this product may or may not be applicable to age group 0-18 and to age group 85 and above. Therefore, when reviewing the generated output, customers must consider the representativeness of subpopulations within their source data.

To search clinical data in Vertex AI Search, you can follow one of these workflows:

Create a healthcare data store, import FHIR R4 data into the data store, connect it to a healthcare search app, and query the clinical data.
Create a healthcare search app, create a healthcare data store and import FHIR R4 data into the data store during the app creation process, and query the clinical data. For more information, see Create a healthcare search app.

This page describes the first method.

About data import frequency

You can import FHIR R4 data into a data store in the following ways:

Batch import: a one-time import. Data is imported into a data store. For further incremental imports, see Refresh healthcare data.
Streaming import (Preview): a near real-time streaming data import. Any incremental changes in the source FHIR store are synchronized in the Vertex AI Search data store. Streaming requires a data connector, which is a type of a data store that contains an entity. An entity is also a data store instance.

The data streaming rate for a given Google Cloud project is dependent on the following quotas. If you exceed the quota you might experience streaming delays.
- The number of pending FHIR or BigQuery streaming writes per minute. For more information, see Quotas and limits.
- The number of FHIR read operations per minute per region. For more information, see Cloud Healthcare API FHIR quotas.

You can select the data import frequency at the time of data store creation and you cannot change this configuration later.

Streaming import is available for all the resources that Vertex AI Search supports. For more information, see Healthcare FHIR R4 data schema reference.

Before you begin

Before you create the healthcare data store and import data into it, complete these requirements:

Understand the relationship between apps and data stores for healthcare search. For more information, see About apps and data stores.
Prepare your FHIR data for ingestion.
Vertex AI Search for healthcare provides search services only in the US multi-region (us). Therefore, your healthcare search app and data stores must reside in the us multi-region.

Create a data store

You can create a data store either in the Google Cloud console or using the API. The two approaches differ in the following way:

In the Google Cloud console: Select the source FHIR store in the Cloud Healthcare API and import FHIR data as part of the healthcare search data store creation process. To stream FHIR data (Preview), your source FHIR store must be in the same Google Cloud project as the destination Vertex AI Search data store.
Through the REST API: You can import FHIR data from a Cloud Healthcare API FHIR store that's in the same Google Cloud project or a different one.
1. Use the dataStores.create method to create a healthcare data store.
2. Use the documents.import method to specify the FHIR store in Cloud Healthcare API and import FHIR R4 data.

To create a healthcare data store, complete the following steps.

Permissions required for this task

Grant the following permissions or the following Identity and Access Management (IAM) roles to the service-PROJECT_NUMBER@gcp-sa-discoveryengine.iam.gserviceaccount.com service account.

One-time import of FHIR data from FHIR stores in Cloud Healthcare API.
Permissions
- healthcare.datasets.get
- healthcare.datasets.list
- healthcare.fhirStores.get
- healthcare.fhirStores.list
- healthcare.fhirStores.export
Roles
- Healthcare FHIR Store Administrator (roles/healthcare.fhirStoreAdmin)
Streaming import of FHIR data from FHIR stores in Cloud Healthcare API (Preview).
Permissions
- healthcare.datasets.get
- healthcare.datasets.list
- healthcare.fhirStores.get
- healthcare.fhirStores.update
- healthcare.fhirStores.list
- healthcare.fhirStores.export
- healthcare.fhirResources.get
Roles
- Healthcare FHIR Store Administrator (roles/healthcare.fhirStoreAdmin)
- Healthcare FHIR Resource Reader (roles/healthcare.fhirResourceReader)
Import FHIR data that references files in Google Cloud. These are granted by default if the referenced files are in the same Google Cloud project as the Vertex AI Search app.
Permissions
- storage.objects.get
Roles
- Storage Object Viewer (roles/storage.objectViewer)

Console

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

Agent Builder
In the navigation menu, click Data stores.
Click Create data store.
In the Select a data source pane, select Healthcare API (FHIR) as your data source.
To import data from your FHIR store, do one of the following:
- Select the FHIR store from the list of available FHIR stores:
  1. Expand the FHIR store field.
  2. In this list, select a dataset that resides in a permitted location and then select a FHIR store that uses FHIR version R4.
    To stream FHIR data (Preview), your source FHIR store must be in the same Google Cloud project as the destination Vertex AI Search data store.
- Enter the FHIR store manually:
  1. Expand the FHIR store field.
  2. Click Enter FHIR store manually.
  3. In the FHIR store name dialog, enter the full name of the FHIR store in the following format:
    project/PROJECT_ID/locations/LOCATION/datasets/DATASET_ID/fhirStores/FHIR_STORE_ID
  4. Click Save.
In the Synchronization section, select one of the following options. This selection cannot be changed after the data store is created.
- One time: to perform a one-time batch data import. For further incremental imports, see Refresh healthcare data.
  1. Click Continue.
  2. In the Your data store name field, enter a name for your data store.
  3. Click Create.
- Streaming: to perform a near real-time streaming data import. To stream data, you must create a data connector, which is a type of a data store. This is a Preview feature. To set up a streaming data store using the REST API, contact your customer engineer.
  1. Click Continue.
  2. In the Your data connector name field, enter a name for your data connector.
  3. Click Create.

REST

Create a data store.

curl -X POST\
 -H "Authorization: Bearer $(gcloud auth print-access-token)" \
 -H "Content-Type: application/json"\
 -H "X-Goog-User-Project: PROJECT_ID" \
"https://us-discoveryengine.googleapis.com/v1alpha/projects/PROJECT_ID/locations/us/collections/default_collection/dataStores?dataStoreId=DATA_STORE_ID" \
 -d '{
    "displayName": "DATA_STORE_DISPLAY_NAME",
    "industryVertical": "HEALTHCARE_FHIR",
    "solutionTypes": ["SOLUTION_TYPE_SEARCH"],
    "searchTier": "STANDARD",
    "searchAddOns": ["LLM"]
}'

Replace the following:

PROJECT_ID: the ID of your Google Cloud project.
DATA_STORE_ID: the ID of the Vertex AI Search data store that you want to create. This ID can contain only lowercase letters, digits, underscores, and hyphens.
DATA_STORE_DISPLAY_NAME: the display name of the Vertex AI Search data store that you want to create.

Response

You should receive a JSON response similar to the following. If the value for the done key is true, it indicates that the operation to create the data store was completed. Otherwise, the data store creation operation was unsuccessful.

{
  "name": "projects/PROJECT_NUMBER/locations/us/collections/default_collection/operations/DATA_STORE_CREATION_OPERATION_ID",
  "done": true
}

If the source FHIR store and the target Vertex AI Search data store are in the same Google Cloud project, call the following method to perform a one-time batch import. If they're not in the same project, go to the next step.
```
curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-H "X-Goog-User-Project: PROJECT_ID" \
"https://us-discoveryengine.googleapis.com/v1alpha/projects/PROJECT_ID/locations/us/dataStores/DATA_STORE_ID/branches/0/documents:import" \
-d '{
   "reconciliation_mode": "FULL",
   "fhir_store_source": {"fhir_store": "projects/PROJECT_ID/locations/CLOUD_HEALTHCARE_DATASET_LOCATION/datasets/CLOUD_HEALTHCARE_DATASET_ID/fhirStores/FHIR_STORE_ID"}
}'
```
Replace the following:
- PROJECT_ID: the ID of your Google Cloud project.
- DATA_STORE_ID: the ID of the Vertex AI Search data store.
- CLOUD_HEALTHCARE_DATASET_ID: the ID of the Cloud Healthcare API dataset that contains the source FHIR store.
- CLOUD_HEALTHCARE_DATASET_LOCATION: the location of the Cloud Healthcare API dataset that contains the source FHIR store.
- FHIR_STORE_ID: the ID of the Cloud Healthcare API FHIR R4 store.
Response

You should receive a JSON response similar to the following. The response contains an identifier for a long-running operation. Long-running operations are returned when method calls might take a substantial amount of time to complete. Note the value of IMPORT_OPERATION_ID. You need this value to verify the status of the import.
```
  {
    "name": "projects/PROJECT_ID/locations/us/collections/default_collection/dataStores/DATA_STORE_ID/branches/0/operations/IMPORT_OPERATION_ID",
    "metadata": {
      "@type": "type.googleapis.com/google.cloud.discoveryengine.v1alpha.ImportDocumentsMetadata"
    }
  }
```
If the source FHIR store and the target Vertex AI Search data store are in different Google Cloud projects, call the following method to perform a one-time batch import. If they're in the same project, go back to the previous step.
```
curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-H "X-Goog-User-Project: TARGET_PROJECT_ID" \
"https://us-discoveryengine.googleapis.com/v1alpha/projects/TARGET_PROJECT_ID/locations/us/dataStores/DATA_STORE_ID/branches/0/documents:import" \
-d '{
   "reconciliation_mode": "FULL",
   "fhir_store_source": {"fhir_store": "projects/SOURCE_PROJECT_ID/locations/CLOUD_HEALTHCARE_DATASET_LOCATION/datasets/CLOUD_HEALTHCARE_DATASET_ID/fhirStores/FHIR_STORE_ID"}
}'
```
Replace the following:
- TARGET_PROJECT_ID: the ID of the Google Cloud project that contains the Vertex AI Search data store.
- DATA_STORE_ID: the ID of the Vertex AI Search data store.
- SOURCE_PROJECT_ID: the ID of the Google Cloud project that contains the Cloud Healthcare API dataset and FHIR store.
- CLOUD_HEALTHCARE_DATASET_ID: the ID of the Cloud Healthcare API dataset that contains the source FHIR store.
- CLOUD_HEALTHCARE_DATASET_LOCATION: the location of the Cloud Healthcare API dataset that contains the source FHIR store.
- FHIR_STORE_ID: the ID of the Cloud Healthcare API FHIR R4 store.
Response

You should receive a JSON response similar to the following. The response contains an identifier for a long-running operation. Long-running operations are returned when method calls might take a substantial amount of time to complete. Note the value of IMPORT_OPERATION_ID. You need this value to verify the status of the import.
```
{
  "name": "projects/TARGET_PROJECT_ID/locations/us/collections/default_collection/dataStores/DATA_STORE_ID/branches/0/operations/IMPORT_OPERATION_ID",
  "metadata": {
    "@type": "type.googleapis.com/google.cloud.discoveryengine.v1alpha.ImportDocumentsMetadata"
  }
}
```
Optional: To set up a streaming data import (Preview) using the REST API, contact your customer engineer.

Verify data store creation and FHIR data import

This task shows you how to verify whether a data store was created successfully and whether FHIR data was imported into the data store successfully.

In the Google Cloud console: Select the data store and verify its details.
Through the REST API:
1. Use the dataStores.get method to get the healthcare data store details.
2. Use the operations.get method to get the details of the import operation.

To verify data store creation and data import, complete the following steps.

Console

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

Agent Builder
In the navigation menu, click Data stores.

The Data stores page displays a list of data stores in your Google Cloud project with their details.
Verify whether the data store or the data connector that you created is in the data stores list.
Select the data store or the data connector and verify its details.
- For a data store:
  - The summary table lists the following details:
    - The data store ID, type, and region.
    - The number of documents indicating the number of FHIR resources imported.
    - The timestamp when the last document was imported.
    - Optionally, click View details to see the document import details, such as the details about a successful, partial, or failed import.
  - The Documents tab lists the resource IDs of the imported FHIR resources and their resource types in a paginated table. You can filter this table to verify whether a particular resource was imported.
  - The Activity tab lists the document import details, such as the details about a successful, partial, or failed import.
- For a data connector:
  - The summary table lists the following details:
    - The collection ID, type, and region.
    - The name of the connected app.
    - The state of the connector, which is either active or paused.
  - The Entities table shows the entity within the data connector. The entity's name is the source FHIR store name. The entity's ID is the data connector's ID appended with the source FHIR store name.
    - Click the entity name to see its details. Because an entity is a data store instance within a data connector, the entity details are the same as a data store details.

REST

Verify the data store creation.

curl -X GET -H "Authorization: Bearer $(gcloud auth print-access-token)" \
 -H "Content-Type: application/json"\
 -H "X-Goog-User-Project: PROJECT_ID" \
 "https://us-discoveryengine.googleapis.com/v1alpha/projects/PROJECT_ID/locations/us/collections/default_collection/dataStores/DATA_STORE_ID"

Replace the following:

PROJECT_ID: the ID of your Google Cloud project.
DATA_STORE_ID: the ID of the Vertex AI Search data store.

Response

You should receive a JSON response similar to the following. The response contains details of the created data store.

{
  "name": "projects/PROJECT_ID/locations/us/collections/default_collection/dataStores/DATA_STORE_ID",
  "displayName": "DATA_STORE_DISPLAY_NAME",
  "industryVertical": "HEALTHCARE_FHIR",
  "createTime": "DATA_STORE_CREATION_TIMESTAMP",
  "solutionTypes": [
    "SOLUTION_TYPE_SEARCH"
  ],
  "defaultSchemaId": "default_schema",
  "documentProcessingConfig": {
    "ocrConfig": {
      "enabled": true
    },
    "defaultParsingConfig": {
      "ocrParsingConfig": {}
    }
  }
}

Verify whether the FHIR data import operation is complete.
```
curl -X GET \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
"https://us-discoveryengine.googleapis.com/v1alpha/projects/PROJECT_ID/locations/us/collections/default_collection/dataStores/DATA_STORE_ID/branches/0/operations/IMPORT_OPERATION_ID"
```
Replace the following:
- PROJECT_ID: the ID of your Google Cloud project.
- DATA_STORE_ID: the ID of the Vertex AI Search data store.
- IMPORT_OPERATION_ID: the operation ID of the long-running operation that's returned when you call the import method
Response

You should receive a JSON response similar to the following. The import operation is a long-running operation. While the operation is running, the response contains the following fields:
- successCount: Indicates the number of FHIR resources that were imported successfully so far.
- failureCount: Indicates the number of FHIR resources that failed to be imported so far. This field is displayed only if there are FHIR resources that failed to be imported.
When the operation is complete, the response contains the following fields:
- successCount: Indicates the number of FHIR resources that were imported successfully.
- failureCount: Indicates the number of FHIR resources that failed to be imported. This field is displayed only if there are any FHIR resources that failed to be imported.
- totalCount: Indicates the number of FHIR resources that are present in the source FHIR store. This field is displayed only if there are any FHIR resources that failed to be imported.
- done: Has the value true to indicate that the import operation is complete.
- errorSamples: Provides information about the resources that failed to be imported. This field is displayed only if there are any FHIR resources that failed to be imported.
- errorConfig: Provides a path to a Cloud Storage location that contains the error summary log file.
```
{
 "name": "projects/PROJECT_ID/locations/us/collections/default_collection/dataStores/DATA_STORE_ID/branches/0/operations/IMPORT_OPERATION_ID",
 "metadata": {
   "@type": "type.googleapis.com/google.cloud.discoveryengine.v1alpha.ImportDocumentsMetadata",
   "createTime": "START_TIMESTAMP",
   "updateTime": "END_TIMESTAMP",
   "successCount": "SUCCESS_COUNT",
   "failureCount": "FAILURE_COUNT",
   "totalCount": "TOTAL_COUNT",
 },
 "done": true,
 "response": {
   "@type": "type.googleapis.com/google.cloud.discoveryengine.v1alpha.ImportDocumentsResponse",
  "errorSamples": [ERROR_SAMPLE],
  "errorConfig": {
     "gcsPrefix": "LOG_FILE_LOCATION"
   }
 }
}
```

What's next

Create a healthcare search app and connect your data store to it.
Search for healthcare data in your healthcare search app.
Refresh healthcare data.