Preface
Introduction to Multimodal Chat Completion with Google Vertex AI recipe
- Multimodal Chat Completion with Google Vertex AI recipe contents
  - Multimodal Chat Completion with Google Vertex AI recipe assets
Using the Multimodal Chat Completion with Google Vertex AI recipe

Multimodal Chat Completion with Google Vertex AI

Back Next

Step 3. Configure and publish the process

Configure the Google Vertex AI LLM model version and grounding to connect the model output to verifiable sources of information and publish the process.

Open the

Chat with Google Vertex AI

process.

Optionally, on the Assignments
tab of the Set Flow Configuration
step, enter values for the following fields:

In the Model_LLM
field, enter the model ID of the LLM model. The default ID is

gemini-1.5-flash-002

Clear the check box corresponding to the Grounding
field. It is enabled by default. Grounding is the ability to connect the model output to verifiable sources of information.

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

<generationConfig>
      <maxOutputTokens>8192</maxOutputTokens>
      <temperature>1</temperature>
      <topP>0.95</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 close to 0 makes the output more deterministic, while a higher value close to 1 increases randomness and creativity. For example, if temperature is set to 0.5, the model balances between deterministic and creative outputs.
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.
maxOutputTokens	Defines the maximum number of tokens that the model can generate in its response. The value can't exceed the model's context length. Most of the models have a context length of 2048 tokens.