K2 Intelligence service

The K2 Intelligence service type exposes the K2 Intelligence service instance and then create SmartObjects that utilize the service objects exposed by the service instance. It lets you create and run AI-powered tasks within your K2 environment by defining the task and its expected response. You can create more service instances which will include the AI tasks. The K2 Intelligence Service runs llama-server (from llama.cpp) as a managed Windows service. See the K2 Intelligence Service Windows service for more information, and best practices for creating and managing AI tasks.

You need to install K2 Intelligence to enable this service type. Please see the installation guide for more information.
The following authentication modes may be used with the service type:
  • Impersonate (default)
  • OAuth
  • ServiceAccount
  • Single Single-on
  • Static

Key Can be modified Data type Sample value Notes
ApiBaseURL Yes (required) Text

https://localhost:62137/v1

Your OpenAI provider endpoint.
ApiKey Yes (required) Text 98nklsdflkw3op4u!

API key for authentication.
DefaultModel Yes Text K2IntelligenceModel The name of the default AI model.
DefaultTemperature Yes number 0.7 Controls randomness, 0.7 is normally a good middle ground.
DefaultMaxTokens Yes number 512 Controls total token allowance, reasoning included.
OpenAIMaxRequests Yes number 60 The maximum requests returned
OpenAITimeWindowMinutes Yes number 1 The time window (in minutes) over which maximum request limits are enforced.
The K2 Intelligence Storage SmartObject is used by K2 Intelligence to store the base tasks and any custom tasks that you create as JSON. This SmartObject is managed by the service instance and should not be changed.

The service includes seven SmartObjects. Three of these, AI Task, AI Task Additional Property, and AI Task Output Field, are used to create and manage custom AI tasks. The remaining four are out-of-the-box objects that define the expected response format and handle execution of results. These are system objects and cannot be deleted.
These SmartObjects are:

SmartObject Description
AI Task Allows you to create and manage custom AI tasks.
AI Task Additional Property Lets you pass provider-specific parameters or settings that are not covered by the standard task configuration.
AI Task Output Field Defines what structured information you want the AI to return. This is how you get typed data back from the AI instead of just plain text.
Sentiment Analysis Analyzes the sentiment of the provided text by identifying emotionally significant words or phrases. It considers the overall tone, summarizes the content, and classifies the sentiment as positive, negative, or neutral.
Severity Analysis Uses scope, consequence, urgency, and reversibility to assess how critical or severe the situation, complaint, or issue described in the text is, regardless of domain or context. It is recommended that you modify this to meet your needs.
Summarization Summarizes the given text into concise key points.
Language Detection Identifies the language of the provided text. Only the language name is returned in the response if it is a supported language. If it is not supported, it is shown as “Other”.

See Create and manage AI tasks for more information about creating an AI task.

Properties


Properties Description
TaskId A unique ID for the task
TaskName A unique name to identify your task (used internally)
DisplayName A friendly name for the task
Description A description for the task
SystemPrompt Instructions given to the AI about its behavior
UserMessagePrefix Text that is automatically added before the user's message
ModelOverride Type a new default AI model for the specific task. This overrides the default AI model
TemperatureOverride Control randomness (0.0 = precise, 2.0 = creative)
MaxTokensOverride Sets a maximum total token limit. If the limit is reached, the process will return an error, providing a safeguard against excessive usage
SchemaEnforcementMode Controls how strictly the response format is enforced (ResponseFormat, PromptInjection, Both, or blank)
AllowPerCallOverrides Enables the caller to override settings
CreatedDate Date the task was created
IsPublished Indicates whether the task is published or not. If the task is not in a published state then you won't be able to generate the SmartObject for it
Revision A version number or label used to identify and track changes to a task over time, making it easy to determine the most current version
Success Indicates whether the execution was successful
ErrorMessage A system generated message informing you that a task failed or could not be completed
SanitizeUserMessage Clean user input before sending to AI (removes potential injection attempts, normalizes text)

Methods


Methods Description
Create AI Task Creates a new AI task
Delete AI Task Permanently deletes an AI task. This action cannot be undone
Get AI Task Get a single task by its ID
Get Published Task List List of published tasks
Get Task List List all tasks (including drafts)
Publish Task Publish a task to make it available as a dynamic SmartObject
Update AI Task Modify an existing task

See Create and manage AI tasks for more information about adding additional properties to an AI task.

Properties


Properties Description
PropertyId A unique ID for the property
TaskID A unique ID for the task
Key The parameter name
Value The parameter value
LogicalType The type of value (text, number, json)
Success Indicates whether the execution was successful
ErrorMessage A system generated message informing you that the execution failed or could not be completed

Methods


Methods Description
Add Additional property Use this method for provider-specific features (top_p, presence_penalty, frequency_penalty), custom sampling parameters, advanced model settings, feature flags or experimental options, and nested fields. Ensure logical type of JSON is set for nested fields
Delete Additional Property Permanently deletes a property
Get Additional Property Get a single property by its ID
List Additional Properties Lists all properties of a task
Update Additional Property Change a property's value or type

See Create and manage AI tasks for more information about AI task output fields.

Properties

Properties Description
FieldId A unique ID for the output field
TaskId A unique ID for the task
Key A unique, single-word name for the output field (this becomes the property name)
DisplayName A user-friendly label for the output field shown as the output name when mapping the SmartObject
Description A description visible to the AI model that helps guide what is required for the specific field. For example, in sentiment analysis, the key might be “reason” and the description “the reason for the sentiment classification.” The model uses this description as guidance
LogicalType The type of data (text, number, boolean, datetime, int, string, yesno etc.)
IsRequired Indicates whether this field must be present in the AI response. It is left blank if this is not set to true.
EnumAllowedValues Use this to restrict the AI model to a predefined set of options instead of allowing free form responses. For example, in sentiment analysis, you can specify “positive, negative, neutral,” which limits the model to selecting one of these comma separated values
Success Indicates whether the execution was successful
ErrorMessage A system generated message informing you that the execution failed or could not be completed

Methods

Methods Description
Add Output Field Extracts specific pieces of information from the AI response, typed data (numbers, dates, yes/no values), validated responses (required fields, specific options) or data that can be used directly in yourworkflow
Delete Output Field Permanently deletes a field. This cannot be undone
Get Output Field Get a single output field by itsID
List Output Fields List all output fields of a task
Update Output Field Modify an existing output field

Use these SmartObjects to define and execute AI prompts. See AI Prompts for an example of executing AI prompts.

Properties


Properties Description
UserMessage The input text or data to be processed by the AI
ResponseContent The primary content extracted from the AI’s response
RawRequest The request that was sent, with any sensitive information redacted
RawResponse The complete response from the AI provider
Success Indicates whether the execution was successful
ErrorCode The error code returned if an issue occurs
Summary Summary of the UserMessage

Methods

Methods Description
Execute Executes a published task.a published task

Use the Sentiment Analysis, Severity Analysis, Summarization, and Language Detection SmartObjects to define and execute AI prompts.

As an example, when you execute a sentiment analysis prompt, the K2 Intelligence service generates a JSON raw request that includes the prompt below:
You are a sentiment analysis assistant.

Sentiment Analysis prompt scenario

Analyze the sentiment of the provided text by following these steps:
  1. Identify the emotionally significant words or phrases in the text.
  2. Consider the overall tone - is it favorable, unfavorable, or neither?
  3. Classify the sentiment as positive, negative, or neutral.
  4. Summarize your reasoning in one sentence.
Definitions:
  • Positive: expresses approval, happiness, satisfaction, or optimism
  • Negative: expresses disapproval, sadness, frustration, or criticism
  • Neutral: expresses facts or balanced views with no strong emotional lean

Return the classified sentiment along with a single concise sentence explaining the reasoning behind the classification.

Example:

To run a sentiment analysis prompt:

  1. Expand System, then K2 Intelligence Service/K2 Intelligence Service.
  2. Select Sentiment Analysis Task, then select the Execute method, and click Execute.
  3. In the User Message property, type the message you want to analyze under Value, and click Execute.
    In this example, we use the following input:
    User Message:
    "I received my order but the item was broken already, the delivery took too long and the driver was so rude. I'm not happy at all and I want a refund."

 

Results:

In the results, we get the following output.

RAW response content:

Copy

JSON response content

{
  "Reason": "The text expresses dissatisfaction with the order, delivery, and service, indicating a negative sentiment."
  ,
  "Sentiment": "negative"
}

Raw request:

Copy

JSON raw request

{
  "method": "POST",
  "url": "http://127.0.0.1:62137/v1/chat/completions",
  "headers": {
    "Content-Type": "application/json",
    "Authorization": "Bearer [REDACTED]"
  },
  "body": {
    "model": "K2IntelligenceModel",
    "messages": [
      {
        "role": "system",
        "content": " You are a sentiment analysis assistant. Analyze the sentiment of the provided text by following these steps:\n\n  1. Identify the emotionally significant words or phrases in the text.\n  2. Consider the overall tone - is it favorable, unfavorable, or neither?\n  3. Classify the sentiment as positive, negative, or neutral.\n  4. Summarize your reasoning in one sentence.\n\n  Definitions:\n  - positive: expresses approval, happiness, satisfaction, or optimism\n  - negative: expresses disapproval, sadness, frustration, or criticism\n  - neutral: expresses facts or balanced views with no strong emotional lean"
      },
      {
        "role": "user",
        "content": "[REDACTED]"
      }
    ],
    "temperature": 0.2,
    "response_format": {
      "type": "json_schema",
      "json_schema": {
        "name": "SentimentTask",
        "strict": true,
        "schema": {
          "type": "object",
          "properties": {
            "Reason": {
              "type": "string",
              "description": "Concise single sentence reason for sentiment classification."
            },
            "Sentiment": {
              "type": "string",
              "description": "The identified sentiment classificaiton.",
              "enum": [
                "positive",
                "negative",
                "neutral"
              ]
            }
          },
          "additionalProperties": false,
          "required": [
            "Reason",
            "Sentiment"
          ]
        }
      }
    }
  }
}

Raw response:

Copy

JSON raw response

{
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "role": "assistant",
        "content": {
          "Reason": "The text expresses dissatisfaction with the order, delivery, and service, indicating a negative sentiment.",
          "Sentiment": "negative"
        }
      }
    }
  ],
  "created": 1776407055,
  "model": "K2IntelligenceModel",
  "system_fingerprint": "b8390-b6c83aad",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 43,
    "prompt_tokens": 188,
    "total_tokens": 231
  },
  "id": "chatcmpl-5cWNNx15U4l7Htq06wMw6MoME4dC0jmg",
  "timings": {
    "cache_n": 0,
    "prompt_n": 188,
    "prompt_ms": 2431.119,
    "prompt_per_token_ms": 12.931484042553192,
    "prompt_per_second": 77.33064485942481,
    "predicted_n": 43,
    "predicted_ms": 2825.318,
    "predicted_per_token_ms": 65.70506976744187,
    "predicted_per_second": 15.219525731262815
  }
}