Theia API Documentation v1.74.0
    Preparing search index...

    Interface LanguageModelRequest

    interface LanguageModelRequest {
        clientSettings?: { keepThinking: boolean; keepToolCalls: boolean };
        compaction?: CompactionSettings;
        deferredToolIds?: string[];
        messages: LanguageModelMessage[];
        reasoning?: ReasoningSettings;
        response_format?:
            | { type: "text" }
            | { type: "json_object" }
            | ResponseFormatJsonSchema;
        serverTools?: string[];
        settings?: { [key: string]: unknown };
        tools?: ToolRequest<ToolInvocationContext>[];
    }

    Hierarchy (View Summary)

    Index

    Properties

    clientSettings? compaction? deferredToolIds? messages reasoning? response_format? serverTools? settings? tools?

    Properties

    clientSettings?: { keepThinking: boolean; keepToolCalls: boolean }
    compaction?: CompactionSettings

    Provider-agnostic server-side compaction settings, copied verbatim from the chat session. Resolved against the model's capability and default in the backend.

    deferredToolIds?: string[]

    Ids of tools whose definitions should be deferred and discovered on-demand via the provider's built-in tool search mechanism. Providers that do not support deferred loading should ignore this field.

    messages: LanguageModelMessage[]
    reasoning?: ReasoningSettings

    Provider-agnostic reasoning configuration; providers translate it to their native API.

    response_format?:
        | { type: "text" }
        | { type: "json_object" }
        | ResponseFormatJsonSchema
    serverTools?: string[]

    Ids of the provider's server tools (see ServerToolDescriptor) that are enabled for this request. Each provider translates the enabled ids into its native server tool configuration.

    settings?: { [key: string]: unknown }
    tools?: ToolRequest<ToolInvocationContext>[]

    Settings

    Member Visibility

    On This Page

    Properties