AI Chat Task

Pricing Templates

The AI Chat task is a system task that integrates AI-powered chat completion capabilities. It uses pre-trained language models, such as GPT (Generative Pre-trained Transformer), to generate natural, human-like responses to user inputs.

There are several tabs available where you can configure a AI Chat task:

General tab
Input Parameters tab
Configuration tab
Retry tab
Routing tab

The configuration of the General, Input Parameters, and Routing tabs is similar to that of a user task. We discuss only the Configuration tab here.

Configuration tab

From the Configuration tab, set up the following fields:

Field	Required	Description
Credentials	Yes	The saved credentials to use for the chat model.
Chat model	Yes	A configuration for chat model by a JSON Object.

Chat model field top-level fields

Field	Type	Required	Description
`userMessages`	JSON Array	Yes	A list of messages that the model should follow. Please provide either the user messages, or system messages, or both. User messages are user-provided prompts or additional context information. System messages are developer-provided instructions that the model should follow, regardless of messages sent by the user.
`systemMessages`	JSON Array	Yes
`provider`	String	Yes	The Large language model (LLM) provider to use. Supported Providers: AmazonBedrockConverse AnthropicClaude AzureOpenAI Cerebras DeepSeek GoogleGemini Groq MistralAI NVIDIA Ollama OpenAI Perplexity
`model`	String	Yes	The model to use from the LLM provider.
`endpoint`	String	No	Required if the provider is AzureOpenAI. The endpoint from the Azure AI OpenAI `Keys and Endpoint` section under `Resource Management`.
`temperature`	Number	No	Scales randomness of token selection. Lower is more deterministic and higher is more varied.
`topP`	Number	No	Samples from the smallest set of tokens whose cumulative probability ≥ p, ignoring the long tail. We generally recommend altering this or temperature but not both.
`topK`	Integer	No	Samples only from the top K most probable tokens at each step, discarding the rest.
`responseJSONSchema`	JSON Object	No	Ensures the model will always generate responses that adhere to the supplied JSON Schema for the supported provider. SimWorkflow adopts the 2020-12 version.
`mcpServers`	JSON Array	No	List of MCP (Model Context Protocol) servers the LLM can use. See the top-level fields of the MCP Server JSON object for details.

MCP Server JSON object top-level fields

Field	Type	Required	Description
`credentials`	Credentials Id	No	The saved credentials to use for the MCP server.
`transportType`	String	Yes	The MCP server transport type. Supported Types: StreamableHTTP, SSE
`baseUrl`	String	Yes	The base URL of the MCP server.
`endpoint`	String	No	The endpoint to use for MCP server. Default is `/mcp` for `StreamableHTTP` transport type and `/sse` for `SSE` transport type.
`requestTimeoutSeconds`	Integer	No	The timeout for an MCP server request, in seconds. Default is 20 seconds.
`toolFilter`	JSON Array	No	Select a specific subset of tools from the MCP server, rather than exposing all of them to your agent.

Note:

The following fields in a Chat Model can reference task inputs using JSONPath expressions, the expression must be inside a placeholder ${ }:
- userMessages
- systemMessages
- model
To escape a JSONPath expression, prefix it with an extra $ character, e.g., $${}. See JSONPath for more information about JSONPath.
Alternatively, JSONata expressions can be used instead of JSONPath by prefixing the expression with jsonata: within ${} placeholders. See JSONata for more information about JSONata.

Here is an example AI Chat model:

{
  "userMessages": [
    "Extract price as number and availability as boolean from https://www.costco.com.au/Appliances/Kitchen-Appliances/Air-Fryers-Deep-Fryers/Cuisinart-Express-Air-Fry-Oven-TOA-65XA/p/88195 and return a JSON payload"
  ],
  "provider": "OpenAI",
  "model": "gpt-5-nano",
  "responseJSONSchema": {
    "type": "object",
    "properties": {
      "inStock": {
        "type": "boolean"
      },
      "price": {
        "type": "number"
      }
    },
    "required": [
      "inStock",
      "price"
    ],
    "additionalProperties": false
  }
}

The Test AI chat button allows you to test the AI Chat model configuration with the task input parameters JSON data.

Response for AI chat model

The response for AI hat model is a JSON Object with the following fields:

Field	Type	Description
`result`	JSON Object	The AI chat model result.
`usage`	JSON Object	The AI chat model usage JSON Object.

Usage JSON object top-level fields

Field	Type	Description
`promptTokens`	Number	The number of tokens used in the prompt of the AI request.
`completionTokens`	Number	The number of tokens returned in the generation of the AI's response.
`totalTokens`	Number	The total number of tokens from both the prompt of an AI request and generation of the AI's response.

Pricing Templates Integrations Blog Contact Us Help

Terms of Service Privacy Policy