Interface ChatRequest

A request to a chat participant.

Stubbed

interface ChatRequest {
    command: undefined | string;
    model: LanguageModelChat;
    prompt: string;
    references: readonly ChatPromptReference[];
    toolInvocationToken: never;
    toolReferences: readonly ChatLanguageModelToolReference[];
}

Properties

command model prompt references toolInvocationToken toolReferences

Properties

Readonly command

command: undefined | string

The name of the ChatCommand command that was selected for this request.

Readonly model

model: LanguageModelChat

This is the model that is currently selected in the UI. Extensions can use this or use chat.selectChatModels to pick another model. Don't hold onto this past the lifetime of the request.

Readonly prompt

prompt: string

The prompt as entered by the user.

Information about references used in this request is stored in ChatRequest.references.

Note that the ChatParticipant.name name of the participant and the ChatCommand.name command are not part of the prompt.

Readonly references

references: readonly ChatPromptReference[]

The list of references and their values that are referenced in the prompt.

Note that the prompt contains references as authored and that it is up to the participant to further modify the prompt, for instance by inlining reference values or creating links to headings which contain the resolved values. References are sorted in reverse by their range in the prompt. That means the last reference in the prompt is the first in this list. This simplifies string-manipulation of the prompt.

Readonly toolInvocationToken

toolInvocationToken: never

A token that can be passed to lm.invokeTool when invoking a tool inside the context of handling a chat request. This associates the tool invocation to a chat session.

Readonly toolReferences

toolReferences: readonly ChatLanguageModelToolReference[]

The list of tools that the user attached to their request.

When a tool reference is present, the chat participant should make a chat request using LanguageModelChatToolMode.Required to force the language model to generate input for the tool. Then, the participant can use lm.invokeTool to use the tool attach the result to its request for the user's prompt. The tool may contribute useful extra context for the user's request.

Settings

Member Visibility

Theme

On This Page

commandmodelpromptreferencestoolInvocationTokentoolReferences