Have Questions?  Need Expert Answers?  JOIN LUNCH N' LEARNS!

Chat Completion

Prev Next
Step Details
Introduced in Version9.25
Last Modified in Version--
LocationAI > AWS Bedrock

The Chat Completion Step is used to generate a Chat Response and is available for use across several AI Modules. The step enables Users to input Model and AI Prompt information to output a Chat Response. In v9.25, this functionality has been added to the AWS Bedrock Module. The step functions similarly to other AI Modules with additional settings, such as Override AWS Credentials.

Prerequisites 

This step requires installing the AI.Common, and AI.AWS.Bedrock Modules and adding their dependencies to the Project before they are available in the toolbox. 


Properties

AWS Settings

PropertyDescriptionData Type
AWS SettingsAllows Users to specify AWS Credentials.Boolean

Bedrock Settings

PropertyDescriptionData Type
Max TokensMaximum number of tokens allowed in the input stringInt32

Settings

PropertyDescriptionData Type
Select ModelCheckbox to enable the Model selection in settings. Boolean
Select PromptEnables access to prompts stored in the AI Prompt Manager through the "Select Prompt" property. When checked, this disables the direct Prompt input option.Boolean
PromptInput prompt for the LLM.String
Select System PromptEnables Users to select a system prompt. Selecting this option displays the following list of options:
  • Get System Prompt from Flow
  • Use TextMerge System Prompt
Boolean
Use AgentAllows Users to select and utilize an Agent. ---
Tool FlowsIf a Tool Flow exists in the platform. This setting enables Users to select and utilize it within the step. Boolean

Inputs

PropertyDescriptionData Type
Message HistoryProvides prior conversation context to help the LLM generate accurate and coherent responses.String
Override VariablesAllows Users to specify whether the step should include Override Variables. --
Response SchemaDefines the structured JSON format that the LLM must follow when generating its output. By specifying the expected fields and data types, the Response Schema ensures the model returns consistent, machine-readable data that can be accurately mapped into the workflow.--

Outputs

PropertyDescriptionData Type
Chat ResponseResponse from the specified model based on the input prompt.String


Chat Response Structure

The Chat Response returned from the Chat completion step consists of four components: ID, prompt, model, and completion. The completion component holds the model’s output and follows the structure defined by the Response Schema. When a schema, such as AI_StructuredOutput, is provided, the completion field returns a structured JSON object that matches the object defined by that schema. 

For example:
{ "StringValue": "Sky color", "StringValue2": "Ocean hue", "IntegerValue": 0, "IntegerValue2": 255, "BooleanValue": false, "BooleanValue2": true, "DecimalValue": 0.0, "DecimalValue2": 1.0 }

This ensures the output is predictable, machine-readable, and can be reliably passed to subsequent steps within the Flow.