AI

AI plugin to use. Provide the class here. Available plugins are:

• OpenAIPlugin - use OpenAI API
• AnthropicPlugin - use Anthropic API
• GooglePlugin - use Google API

You can also provide a custom plugin by extending the AbstractApiPlugin class.

Shows an AI model selector in the SettingsPanel where the user can select between the available API:s

Set this to true or a config object to add a ChatButton

Set to true to automatically console.log things that happen internally. AI responses and audio processing for example.

An array of strings which will be circulated as waiting status texts when awaiting long AI responses Defaults to ['Waiting', 'Waiting ...', 'Still waiting', 'Still waiting ...']

The URL to send feedback (thumbs up/thumbs down) to

Additional options to pass to fetch requests

By default, data item which are targeted in an AI action will be highlighted in the UI. Set this to false to prevent this behavior.

Set to false to not manage chat history on the client

Max tokens setting (if applicable)

AI model name (if applicable)

Use this to provide more details of data models currently available for the AI agent to work with. Provide an object with model class names as keys and configuration objects as values.

A setup in a Bryntum Grid application which uses the class Project for its rows and also have a related class Member could look like this:

{
    Member  : { store : memberStore },
    Project : {
        relations : {
            members : {
                relatedModelName : 'Member'
            },
            owner : {
                relatedModelName : 'Member'
            }
        }
    }
}

The URL to send text prompts to

Enables support for redoing data modifying actions performed by the AI. Shows a redo icon under the last redoable message in the ChatPanel. Requires undo to be enabled.

Set to true to always prompt the user for confirmation before allowing the AI to add records

Set to true to always prompt the user for confirmation before allowing the AI to remove records

Set to true to always prompt the user for confirmation before allowing the AI to update records

Set to true to upon creation, send a test prompt to verify the connection to the AI backend. If the test fails, a warning icon will be shown on the chat panel.

Temperature setting (if applicable). Overrides the default temperature set by the apiPlugin.

The URL to send text for speech synthesizing

A number of milliseconds a prompt is allowed to take before being aborted. Defaults to 60000 (1 minute).

Use this to add custom tools to the AI model's tool list or to remove existing tools. Provide an object with the tool names as keys and configuration objects as values. Please note that modifying existing tools is not supported, only removing them.

Specific tools for Grid AI are:

• getRecords - Used by agent to retrieve records from the store
• updateRecords - Updates records in the store
• deleteRecords - Deletes records from the store
• addRecord - Add a new record in the store
• filterRecords - Filters the Grid's store
• unfilterRecords - Removes all filters from the Grid's store
• sortRecords - Sorts the Grid's store
• selectRecords - Selects records in the Grid
• lockRows - Locks rows to the top (LockRows feature required)
• reloadGrid - Reloads data. Only available if the Grid can be reloaded

Also, the tools from tools are available

Use this to provide example prompt scenarios to "teach" the AI of more complex tasks. Must be configured as an array containing arrays which contains the AI messages for the complete conversation.

Please be advised that the content here will be provided in each request sent to the AI API and will therefore consume tokens and take up space in the context window.

Adding an example to the "training data" will often solve one particular case, but may "confuse" the AI in ways not expected.

The URL to send audio for transcribing

Enables support for undoing data modifying actions performed by the AI. Shows an undo icon under the last undoable message in the ChatPanel.

A configuration object containing the user's preferred AI feature settings. If there is a matching AI feature class config, the user setting will take precedence JSON.

Experimental configuration of the AI verbosity. Valid values are "1", "2" and "3", where "1" is lowest.

Configuration settings for voice activation

Forces voice only mode, which hides the chat panel and only shows a record chat button.