Function Calling using Google Vertex AI

Back Next

Step 5. Configure and publish the processes

Configure the Google Vertex LLM model version and service URL, and publish the processes.

To publish the Generic_executor
process, click Actions
in the row that contains the process and select publish.

Open the

Google Vertex Function Calling

process.

On the Assignments
tab of the Set Flow Configuration
step, enter values in the following fields:

In the LLM_Model
field, enter the model ID of the model that you want to use. The model ID is set to gemini-1.5-pro-002, by default.

In the Generation_Config
field, enter the instructions using the Expression Editor, as shown in the following sample code:

<generationConfig>
    <temperature>0.2</temperature>
    <maxOutputTokens>1024</maxOutputTokens>
    <topP>0.8</topP>
</generationConfig>

For the Generation_Config
field, enter values for the following properties:

Property	Description
temperature	Controls the randomness of the model's output. A lower value makes the output more deterministic, while a higher value increases randomness and creativity. For example, a temperature of 0.5 balances between deterministic and creative outputs.
maxOutputTokens	Defines the maximum number of tokens the model can generate in its response. Setting a limit ensures that the response is concise and fits within the desired length constraints.
topP	Determines the cumulative probability threshold for token selection. The model considers the smallest set of tokens whose cumulative probability meets or exceeds topP. For example, if topP is set to 0.1, the model considers only the top 10% most probable tokens at each step.

To add additional generation config parameters, see the Generative AI on Vertex AI documentation.

On the Assignments
tab of the Prepare Default Function Declaration
step, enter the process URL of the recipe in the Function_Declaration
field using the Expression Editor, as shown in the following sample code:

<tools>
    <functionDeclarations>
        <name>Update_Amazon_Bedrock_Knowledge_Base_executor</name>
        <description>This function processes a user-provided prompt to identify and extract key information based on predefined parameters. For instance, it can detect specific identifiers such as Knowledge_Base_ID or Email_Address by scanning the text for patterns or matches associated with these keywords. The function returns the extracted data for further processing or validation, facilitating automated workflows where structured information is derived from unstructured input.</description>
        <parameters>
            <type>object</type>
            <properties>
                <Process_URL>
                    <type>string</type>
                    <description>Fixed URL to execute process</description>
                    <enum><Service_URL_of_the_Application_Integration_Process></enum>
                </Process_URL>
                <Payload>
                    <type>array</type>
                    <description>Structured payload for processes</description>
                    <items>
                        <description>Fields in payload for processes</description>
                        <type>object</type>
                        <properties>
                            <Bucket_Name>
                                <type>string</type>
                                <description>The bucket name in AWS S3 is a unique identifier for an S3 bucket.</description>
                            </Bucket_Name>
                            <Knowledge_Base_ID>
                                <type>string</type>
                                <description>The unique identifier of the knowledge base. Sometimes, the knowledge base is abbreviated as KB.</description>
                            </Knowledge_Base_ID>
                            <Email_Address>
                                <type>string</type>
                                <description>Email address where the process result will be sent.</description>
                            </Email_Address>
                            <Data_Source_ID>
                                <type>string</type>
                                <description>The unique identifier of the data source. Sometimes, the data source is abbreviated as DS.</description>
                            </Data_Source_ID>
                        </properties>
                        <required>Knowledge_Base_ID</required>
                        <required>Email_Address</required>
                    </items>
                </Payload>
            </properties>
            <required>Process_URL</required>
            <required>Payload</required>
        </parameters>
    </functionDeclarations>
    <functionDeclarations>
        <name>Synchronize_ServiceNow_Incidents_with_Jira_Issues_executor</name>
        <description>This function processes a user-provided prompt to identify and extract key information based on predefined parameters. For instance, it can detect specific identifiers such as Jira_Issue_Type_ID or Jira_Project_ID by scanning the text for patterns or matches associated with these keywords. The function returns the extracted data for further processing or validation, facilitating automated workflows where structured information is derived from unstructured input.</description>
        <parameters>
            <type>object</type>
            <properties>
                <Process_URL>
                    <type>string</type>
                    <description>Fixed URL to execute process</description>
                    <enum><Service_URL_of_the_Application_Integration_Process></enum>
                </Process_URL>
                <Payload>
                    <type>array</type>
                    <description>Structured payload for processes</description>
                    <items>
                        <description>Fields in payload for processes</description>
                        <type>object</type>
                        <properties>
                            <Date_For_Search>
                                <type>string</type>
                                <description>Search date in the YYYY-MM-DD format.</description>
                            </Date_For_Search>
                            <Filter>
                                <type>string</type>
                                <description>Name of the category to be filtered</description>
                            </Filter>
                            <Email>
                                <type>string</type>
                                <description>Email address where the process result will be sent.</description>
                            </Email>
                            <Jira_Issue_Type_ID>
                                <type>string</type>
                                <description>The unique identifier of Jira issue type</description>
                            </Jira_Issue_Type_ID>
                            <Jira_Project_ID>
                                <type>string</type>
                                <description>The unique identifier of Jira project</description>
                            </Jira_Project_ID>
                            <Sync_Assignee>
                                <type>boolean</type>
                                <description>Boolean variable whether to synchronize the Assignee</description>
                            </Sync_Assignee>
                        </properties>
                        <required>Email</required>
                        <required>Jira_Issue_Type_ID</required>
                        <required>Jira_Project_ID</required>
                    </items>
                </Payload>
            </properties>
            <required>Process_URL</required>
            <required>Payload</required>
        </parameters>
    </functionDeclarations>
</tools>

This example applies to and can execute the Synchronize_ServiceNow_Incidents_with_Jira_Issue
and Update_Amazon_Bedrock_Knowledge_Base
recipes.

Save and publish the process.

Using the Function Calling using Google Vertex AI recipe

Download Guide

Watch

Comments

Cloud Application Integration Homepage