Set verbosity to "low" for model (#1813)

* Set verbosity to "low" for model

* Keep createRequestBody sync for simplicity in endgame

Author: roblourens

src/extension/prompt/node/chatMLFetcher.ts

@@ -113,7 +113,7 @@ export class ChatMLFetcherImpl extends AbstractChatMLFetcher {
 		}
 
 		const postOptions = this.preparePostOptions(requestOptions);
-		const requestBody = chatEndpoint.createRequestBody({
+		const requestBody = await chatEndpoint.createRequestBody({
 			...opts,
 			requestId: ourRequestId,
 			postOptions

src/platform/endpoint/common/chatModelCapabilities.ts

@@ -27,6 +27,8 @@ const VSC_MODEL_HASHES_B = [
 	'df610ed210bb9266ff8ab812908d5837538cdb1d7436de907fb7e970dab5d289',
 ];
 
+let hiddenModelBFamily: string | undefined;
+const HIDDEN_MODEL_B_HASH = '8f398886c326b5f8f07b20ac250c87de6723e062474465273fe1524f2b9092fa';
 
 function getModelId(model: LanguageModelChat | IChatEndpoint): string {
 	return 'id' in model ? model.id : model.model;
@@ -39,7 +41,11 @@ export async function isHiddenModelA(model: LanguageModelChat | IChatEndpoint) {
 
 export async function isHiddenModelB(model: LanguageModelChat | IChatEndpoint): Promise<boolean> {
 	const h = await getCachedSha256Hash(model.family);
-	return h === '8f398886c326b5f8f07b20ac250c87de6723e062474465273fe1524f2b9092fa';
+	if (h === HIDDEN_MODEL_B_HASH) {
+		hiddenModelBFamily = model.family;
+		return true;
+	}
+	return false;
 }
 
 
@@ -150,3 +156,10 @@ export function modelNeedsStrongReplaceStringHint(model: LanguageModelChat | ICh
 export function modelSupportsSimplifiedApplyPatchInstructions(model: LanguageModelChat | IChatEndpoint): boolean {
 	return model.family.startsWith('gpt-5');
 }
+
+export function getVerbosityForModelSync(model: IChatEndpoint): 'low' | 'medium' | 'high' | undefined {
+	if (model.family === hiddenModelBFamily) {
+		return 'low';
+	}
+	return undefined;
+}

src/platform/endpoint/node/chatEndpoint.ts

@@ -246,7 +246,7 @@ export class ChatEndpoint implements IChatEndpoint {
 
 	createRequestBody(options: ICreateEndpointBodyOptions): IEndpointBody {
 		if (this.useResponsesApi) {
-			const body = this._instantiationService.invokeFunction(createResponsesRequestBody, options, this.model, this.modelMetadata);
+			const body = this._instantiationService.invokeFunction(createResponsesRequestBody, options, this.model, this);
 			return this.customizeResponsesBody(body);
 		} else {
 			const body = createCapiRequestBody(options, this.model, this.getCompletionsCallback());

src/platform/endpoint/node/responsesApi.ts

@@ -18,18 +18,19 @@ import { IInstantiationService, ServicesAccessor } from '../../../util/vs/platfo
 import { ConfigKey, IConfigurationService } from '../../configuration/common/configurationService';
 import { ILogService } from '../../log/common/logService';
 import { FinishedCallback, IResponseDelta, OpenAiResponsesFunctionTool } from '../../networking/common/fetch';
-import { ICreateEndpointBodyOptions, IEndpointBody } from '../../networking/common/networking';
+import { IChatEndpoint, ICreateEndpointBodyOptions, IEndpointBody } from '../../networking/common/networking';
 import { ChatCompletion, FinishedCompletionReason, TokenLogProb } from '../../networking/common/openai';
 import { IExperimentationService } from '../../telemetry/common/nullExperimentationService';
 import { ITelemetryService } from '../../telemetry/common/telemetry';
 import { TelemetryData } from '../../telemetry/common/telemetryData';
-import { IChatModelInformation } from '../common/endpointProvider';
+import { getVerbosityForModelSync } from '../common/chatModelCapabilities';
 import { getStatefulMarkerAndIndex } from '../common/statefulMarkerContainer';
 import { rawPartAsThinkingData } from '../common/thinkingDataContainer';
 
-export function createResponsesRequestBody(accessor: ServicesAccessor, options: ICreateEndpointBodyOptions, model: string, modelInfo: IChatModelInformation): IEndpointBody {
+export function createResponsesRequestBody(accessor: ServicesAccessor, options: ICreateEndpointBodyOptions, model: string, endpoint: IChatEndpoint): IEndpointBody {
 	const configService = accessor.get(IConfigurationService);
 	const expService = accessor.get(IExperimentationService);
+	const verbosity = getVerbosityForModelSync(endpoint);
 	const body: IEndpointBody = {
 		model,
 		...rawMessagesToResponseAPI(model, options.messages, !!options.ignoreStatefulMarker),
@@ -48,7 +49,8 @@ export function createResponsesRequestBody(accessor: ServicesAccessor, options:
 			? { type: 'function', name: options.postOptions.tool_choice.function.name }
 			: options.postOptions.tool_choice,
 		top_logprobs: options.postOptions.logprobs ? 3 : undefined,
-		store: false
+		store: false,
+		text: verbosity ? { verbosity } : undefined,
 	};
 
 	body.truncation = configService.getConfig(ConfigKey.Internal.UseResponsesApiTruncation) ?

src/platform/networking/common/networking.ts

@@ -101,6 +101,9 @@ export interface IEndpointBody {
 	truncation?: 'auto' | 'disabled';
 	include?: ['reasoning.encrypted_content'];
 	store?: boolean;
+	text?: {
+		verbosity?: 'low' | 'medium' | 'high';
+	};
 }
 
 export interface IEndpointFetchOptions {