kotlin-vscode/vscode/lm/invokeToolAsync

invokeToolAsync

fun invokeToolAsync(name: String, options: LanguageModelToolInvocationOptions<*>, token: CancellationToken = definedExternally): PromiseLike<LanguageModelToolResult>(source)

Invoke a tool listed in lm.tools by name with the given input. The input will be validated against the schema declared by the tool

A tool can be invoked by a chat participant, in the context of handling a chat request, or globally by any extension in any custom flow.

In the former case, the caller shall pass the toolInvocationToken, which comes from a chat request. This makes sure the chat UI shows the tool invocation for the correct conversation.

A tool result} is an array of {@link LanguageModelTextPart text- and prompt-tsx-parts. If the tool caller is using @vscode/prompt-tsx, it can incorporate the response parts into its prompt using a ToolResult. If not, the parts can be passed along to the LanguageModelChat via a user message with a LanguageModelToolResultPart.

If a chat participant wants to preserve tool results for requests across multiple turns, it can store tool results in the ChatResult.metadata returned from the handler and retrieve them on the next turn from ChatResponseTurn.result.

Parameters

name

The name of the tool to call.

options

The options to use when invoking the tool.

token

A cancellation token. See CancellationTokenSource for how to create one.