Fix toolTokens over-counting when Anthropic Tool Search is enabled

When TST (Tool Search Tool) is enabled, most tools are sent with
defer_loading: true and don't count against the context window until the
model loads them via tool_search. However, the toolTokens calculation in
agentIntent counted ALL available tools, over-estimating by ~25-30K
tokens and causing premature compaction.

Fix: filter availableTools to only non-deferred tools (via
IToolDeferralService) before calling countToolTokens() when TST is
enabled. This gives an accurate budget that reflects what the API
actually charges against the context window.

Author: bhavyaus

src/extension/intents/node/agentIntent.ts

@@ -16,7 +16,8 @@ import { IAutomodeService } from '../../../platform/endpoint/node/automodeServic
 import { IEnvService } from '../../../platform/env/common/envService';
 import { ILogService } from '../../../platform/log/common/logService';
 import { IEditLogService } from '../../../platform/multiFileEdit/common/editLogService';
-import { CUSTOM_TOOL_SEARCH_NAME, isAnthropicCustomToolSearchEnabled } from '../../../platform/networking/common/anthropic';
+import { CUSTOM_TOOL_SEARCH_NAME, isAnthropicCustomToolSearchEnabled, isAnthropicToolSearchEnabled } from '../../../platform/networking/common/anthropic';
+import { IToolDeferralService } from '../../../platform/networking/common/toolDeferralService';
 import { IChatEndpoint } from '../../../platform/networking/common/networking';
 import { modelsWithoutResponsesContextManagement } from '../../../platform/networking/common/openai';
 import { INotebookService } from '../../../platform/notebook/common/notebookService';
@@ -363,6 +364,7 @@ export class AgentIntentInvocation extends EditCodeIntentInvocation implements I
 		@IExperimentationService private readonly expService: IExperimentationService,
 		@IAutomodeService private readonly automodeService: IAutomodeService,
 		@IOTelService override readonly otelService: IOTelService,
+		@IToolDeferralService private readonly toolDeferralService: IToolDeferralService,
 	) {
 		super(intent, location, endpoint, request, intentOptions, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, otelService);
 	}
@@ -388,7 +390,15 @@ export class AgentIntentInvocation extends EditCodeIntentInvocation implements I
 		}
 
 		const tools = promptContext.tools?.availableTools;
-		const toolTokens = tools?.length ? await this.endpoint.acquireTokenizer().countToolTokens(tools) : 0;
+		// When Anthropic tool search is enabled, deferred tools are sent with
+		// defer_loading: true and don't count against the context window until
+		// the model loads them via tool_search. Only count non-deferred tools
+		// so the budget isn't artificially reduced.
+		const toolSearchEnabled = isAnthropicToolSearchEnabled(this.endpoint, this.configurationService);
+		const effectiveTools = tools && toolSearchEnabled
+			? tools.filter(t => this.toolDeferralService.isNonDeferredTool(t.name))
+			: tools;
+		const toolTokens = effectiveTools?.length ? await this.endpoint.acquireTokenizer().countToolTokens(effectiveTools) : 0;
 
 		const summarizeThresholdOverride = this.configurationService.getConfig<number | undefined>(ConfigKey.Advanced.SummarizeAgentConversationHistoryThreshold);
 		if (typeof summarizeThresholdOverride === 'number' && summarizeThresholdOverride < 100 && summarizeThresholdOverride > 0) {
@@ -414,7 +424,7 @@ export class AgentIntentInvocation extends EditCodeIntentInvocation implements I
 		const safeBudget = useTruncation ? Number.MAX_SAFE_INTEGER : messageBudget;
 		const endpoint = toolTokens > 0 ? this.endpoint.cloneWithTokenOverride(safeBudget) : this.endpoint;
 
-		this.logService.debug(`AgentIntent: rendering with budget=${safeBudget} (baseBudget: ${baseBudget}, toolTokens: ${toolTokens}), summarizationEnabled=${summarizationEnabled}`);
+		this.logService.debug(`AgentIntent: rendering with budget=${safeBudget} (baseBudget: ${baseBudget}, toolTokens: ${toolTokens}${toolSearchEnabled ? `, totalTools: ${tools?.length ?? 0}, nonDeferredTools: ${effectiveTools?.length ?? 0}` : ''}), summarizationEnabled=${summarizationEnabled}`);
 		let result: RenderPromptResult;
 		const props: AgentPromptProps = {
 			endpoint,

src/extension/intents/node/askAgentIntent.ts

@@ -12,6 +12,7 @@ import { IEnvService } from '../../../platform/env/common/envService';
 import { ILogService } from '../../../platform/log/common/logService';
 import { IEditLogService } from '../../../platform/multiFileEdit/common/editLogService';
 import { IChatEndpoint } from '../../../platform/networking/common/networking';
+import { IToolDeferralService } from '../../../platform/networking/common/toolDeferralService';
 import { INotebookService } from '../../../platform/notebook/common/notebookService';
 import { IOTelService } from '../../../platform/otel/common/otelService';
 import { IPromptPathRepresentationService } from '../../../platform/prompts/common/promptPathRepresentationService';
@@ -128,8 +129,9 @@ export class AskAgentIntentInvocation extends AgentIntentInvocation {
 		@IExperimentationService expService: IExperimentationService,
 		@IAutomodeService automodeService: IAutomodeService,
 		@IOTelService otelService: IOTelService,
+		@IToolDeferralService toolDeferralService: IToolDeferralService,
 	) {
-		super(intent, location, endpoint, request, { processCodeblocks: true }, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, logService, expService, automodeService, otelService);
+		super(intent, location, endpoint, request, { processCodeblocks: true }, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, logService, expService, automodeService, otelService, toolDeferralService);
 	}
 
 	public override async getAvailableTools(): Promise<vscode.LanguageModelToolInformation[]> {

src/extension/intents/node/editCodeIntent2.ts

@@ -13,6 +13,7 @@ import { IEnvService } from '../../../platform/env/common/envService';
 import { ILogService } from '../../../platform/log/common/logService';
 import { IEditLogService } from '../../../platform/multiFileEdit/common/editLogService';
 import { IChatEndpoint } from '../../../platform/networking/common/networking';
+import { IToolDeferralService } from '../../../platform/networking/common/toolDeferralService';
 import { requestHasNotebookRefs } from '../../../platform/notebook/common/helpers';
 import { INotebookService } from '../../../platform/notebook/common/notebookService';
 import { IOTelService } from '../../../platform/otel/common/otelService';
@@ -89,8 +90,9 @@ export class EditCode2IntentInvocation extends AgentIntentInvocation {
 		@IExperimentationService expService: IExperimentationService,
 		@IAutomodeService automodeService: IAutomodeService,
 		@IOTelService otelService: IOTelService,
+		@IToolDeferralService toolDeferralService: IToolDeferralService,
 	) {
-		super(intent, location, endpoint, request, intentOptions, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, logService, expService, automodeService, otelService);
+		super(intent, location, endpoint, request, intentOptions, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, logService, expService, automodeService, otelService, toolDeferralService);
 	}
 
 	public override async getAvailableTools(): Promise<vscode.LanguageModelToolInformation[]> {

src/extension/intents/node/notebookEditorIntent.ts

@@ -12,6 +12,7 @@ import { IEnvService } from '../../../platform/env/common/envService';
 import { ILogService } from '../../../platform/log/common/logService';
 import { IEditLogService } from '../../../platform/multiFileEdit/common/editLogService';
 import { IChatEndpoint } from '../../../platform/networking/common/networking';
+import { IToolDeferralService } from '../../../platform/networking/common/toolDeferralService';
 import { IAlternativeNotebookContentService } from '../../../platform/notebook/common/alternativeContent';
 import { getCellId } from '../../../platform/notebook/common/helpers';
 import { INotebookService } from '../../../platform/notebook/common/notebookService';
@@ -107,8 +108,9 @@ export class NotebookEditorIntentInvocation extends EditCode2IntentInvocation {
 		@IExperimentationService expService: IExperimentationService,
 		@IAutomodeService automodeService: IAutomodeService,
 		@IOTelService otelService: IOTelService,
+		@IToolDeferralService toolDeferralService: IToolDeferralService,
 	) {
-		super(intent, location, endpoint, request, intentOptions, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, logService, expService, automodeService, otelService);
+		super(intent, location, endpoint, request, intentOptions, instantiationService, codeMapperService, envService, promptPathRepresentationService, endpointProvider, workspaceService, toolsService, configurationService, editLogService, commandService, telemetryService, notebookService, logService, expService, automodeService, otelService, toolDeferralService);
 	}
 
 	protected override prompt = NotebookInlinePrompt;