Witty Autocomplete Agent

Witty Autocomplete Agent is an AI agent that helps users to fill Data Contract schemas faster. It leverages the Witty Autocomplete Agent microservice which, by contacting an LLM provider of your choice, generates fields suggestions.

Prerequisites

• Witty Autocomplete Agent microservice must be up & running and reachable from Witboost Core.

Configuring the agent

In your app-config add these lines:

witty:
  scaffolder:
    smartSuggestions:
      baseUrl: <base URL where Witty Autocomplete Agent is running>
      apiKey: <(optional) api key>
      enabled: true

This will tell Witboost Core where to send autocomplete requests coming from enabled templates.

Setting up a Witty template

Witty Autocomplete is available both for Creation and Edit templates kinds. However, as of now, you can only enable Witty for tables and its inner rows.

Now, let's take this template as a reference:

apiVersion: witboost.com/v1
kind: EditTemplate
metadata:
  name: edit-template
  title: Edit Template
  description: Edit Template
spec:
  owner: agile_lab
  useCaseTemplateId:
    - urn:dmb:utm:my-template:0.0.0
  parameters:
    - title: Data Contract Schema
      properties:
        schema:
          title: Schema
          description: Data Contract Schema
          type: array
          ui:ArrayFieldTemplate: ArrayTableTemplate
          default: []
          items:
            type: object
            ui:ObjectFieldTemplate: TableRowTemplate
            required:
              - name
              - dataType
            properties:
              name:
                type: string
                title: Name
              surname:
                type: string
                title: Surname
  # ... other template stuff ...

To enable Witty on this template table, we will modify the schema property, which is an ArrayTableTemplate, and set it to be a WittyTableTemplate as follows:

apiVersion: witboost.com/v1
kind: EditTemplate
metadata:
  name: edit-template
  title: Edit Template
  description: Edit Template
spec:
  owner: agile_lab
  useCaseTemplateId:
    - urn:dmb:utm:my-template:0.0.0
  parameters:
    - title: Data Contract Schema
      properties:
        schema:
          title: Schema
          description: Data Contract Schema
          type: array
          ui:ArrayFieldTemplate: WittyTableTemplate
          default: []
          items:
            type: object
            ui:ObjectFieldTemplate: TableRowTemplate
            required:
              - name
              - dataType
            properties:
              name:
                type: string
                title: Name
              surname:
                type: string
                title: Surname
  # ... other template stuff ...

But we are not done yet! We have just enabled Witty on this table. Now, we need to decide which are the fields to be autocompletable by the agent.

To do so, let's define both name and surname as Witty-enabled fields:

schema:
  title: Schema
  description: Data Contract Schema
  type: array
  ui:ArrayFieldTemplate: WittyTableTemplate
  default: []
  items:
    type: object
    ui:ObjectFieldTemplate: TableRowTemplate
    required:
      - name
      - dataType
    properties:
      name:
        type: string
        title: Name
        ui:field: WittyTextPicker
      surname:
        type: string
        title: Surname
        ui:field: WittyTextPicker

That's it! Now both Name and Surname fields will be adorned by the Witty Autocomplete button. Moreover, the Witty Table autocomplete button will show up.

Witty Enabled Fields

To check the list of available witty-enabled fields, look into the Pickers section of the docs.

Advanced settings

You can supply some additional settings to a Witty table. To do so, you can add properties into the WittyTableTemplate. Here you have a list of available properties:

apiVersion: witboost.com/v1
kind: EditTemplate
metadata:
  name: edit-template
  title: Edit Template
  description: Edit Template
spec:
  owner: agile_lab
  useCaseTemplateId:
    - urn:dmb:utm:my-template:0.0.0
  parameters:
    - title: Data Contract Schema
      properties:
        schema:
          title: Schema
          description: Data Contract Schema
          type: array
          witty:
            maxRowsPerChunk: 30 # tells witty how many rows at most to attach into each request towards the Witty service
            context: # a list of strings that can be attached into the request to give additional context to the underlying LLM model
              - each row of this table is a set of metadata about columns of a dataset.
              - each metadata row follows the open metadata specification
          ui:ArrayFieldTemplate: WittyTableTemplate
          default: []
          items:
  # ... other template stuff ...

Supported models

The Witty Autocomplete Agent microservice has been tested against the following models/providers:

LLM Provider	Model
Azure OpenAI	gpt-4o
Google Gemini AI	gemini-2.0-flash

Below you can check how to configure your LLM provider on the Witty Autocomplete Agent side.

Azure OpenAI

generator:
  kind: AzureOpenAIChatGenerator
  parameters:
    azure_endpoint: https://<your azure endpoint>.cognitiveservices.azure.com
    azure_deployment: gpt-4o
    generation_kwargs:
      temperature: 0

You can find some more configuration properties here

Google GeminiAI

generator:
  kind: GoogleAIGeminiChatGenerator
  parameters:
    model: gemini-2.0-flash
    generation_config:
      response_mime_type: application/json
      temperature: 0

You can find some more configuration properties here