Add candidateModel and stickyOverride to automode.routerDecision telemetry

The automode.routerDecision event was only logging predictedLabel and
confidence, making it impossible to distinguish what model the router
recommended vs. what model was actually used after sticky-provider and
vision overrides.

New properties:
- candidateModel: the router's top pick (candidate_models[0]) before
  any same-provider or vision fallback overrides are applied
- stickyOverride: whether the router applied a sticky override (1/0)
- conversationId: enables joining with response.success events
- vscodeRequestId: enables joining with response.success events

This allows analysts to detect when the router recommends model A but
the client uses model B due to provider affinity or vision fallback.

Author: Aashna Garg

src/platform/endpoint/node/automodeService.ts

@@ -368,7 +368,7 @@ export class AutomodeService extends Disposable implements IAutomodeService {
 				previous_model: entry?.endpoint?.model,
 				turn_number: (entry?.turnCount ?? 0) + 1,
 			};
-			const result = await this._routerDecisionFetcher.getRouterDecision(prompt, token.session_token, token.available_models, undefined, contextSignals);
+const result = await this._routerDecisionFetcher.getRouterDecision(prompt, token.session_token, token.available_models, undefined, contextSignals, conversationId, chatRequest?.id);
 
 			if (!result.candidate_models.length) {
 				return { lastRoutedPrompt: prompt, fallbackReason: 'emptyCandidateList' };
@@ -465,7 +465,7 @@ export class AutomodeService extends Disposable implements IAutomodeService {
 			};
 
 			const result = await this._routerDecisionFetcher.getRouterDecision(
-				prompt, token.session_token, token.available_models, undefined, contextSignals);
+					prompt, token.session_token, token.available_models, undefined, contextSignals, conversationId);
 
 			if (!result.candidate_models.length) {
 				this._logService.trace('[AutomodeService] resolveEndpointForSummarization: empty candidates');

src/platform/endpoint/node/routerDecisionFetcher.ts

@@ -48,7 +48,7 @@ export class RouterDecisionFetcher {
 	) {
 	}
 
-	async getRouterDecision(query: string, autoModeToken: string, availableModels: string[], stickyThreshold?: number, contextSignals?: RoutingContextSignals): Promise<RouterDecisionResponse> {
+	async getRouterDecision(query: string, autoModeToken: string, availableModels: string[], stickyThreshold?: number, contextSignals?: RoutingContextSignals, conversationId?: string, vscodeRequestId?: string): Promise<RouterDecisionResponse> {
 		const startTime = Date.now();
 		const requestBody: Record<string, unknown> = { prompt: query, available_models: availableModels, ...contextSignals };
 		if (stickyThreshold !== undefined) {
@@ -100,20 +100,28 @@ export class RouterDecisionFetcher {
 			"automode.routerDecision" : {
 				"owner": "lramos15",
 				"comment": "Reports the routing decision made by the auto mode router API",
+				"conversationId": { "classification": "SystemMetaData", "purpose": "FeatureInsight", "comment": "The conversation ID in which the routing decision was made" },
+				"vscodeRequestId": { "classification": "SystemMetaData", "purpose": "FeatureInsight", "comment": "The VS Code chat request ID in which the routing decision was made" },
 				"predictedLabel": { "classification": "SystemMetaData", "purpose": "FeatureInsight", "comment": "The predicted classification label (needs_reasoning or no_reasoning)" },
+				"candidateModel": { "classification": "SystemMetaData", "purpose": "FeatureInsight", "comment": "The top candidate model recommended by the router before any overrides" },
 				"confidence": { "classification": "SystemMetaData", "purpose": "FeatureInsight", "isMeasurement": true, "comment": "The confidence score of the routing decision" },
 				"latencyMs": { "classification": "SystemMetaData", "purpose": "PerformanceAndHealth", "isMeasurement": true, "comment": "The latency of the router API call in milliseconds" },
-				"e2eLatencyMs": { "classification": "SystemMetaData", "purpose": "PerformanceAndHealth", "isMeasurement": true, "comment": "The end-to-end latency of the router request in milliseconds, including network overhead" }
+				"e2eLatencyMs": { "classification": "SystemMetaData", "purpose": "PerformanceAndHealth", "isMeasurement": true, "comment": "The end-to-end latency of the router request in milliseconds, including network overhead" },
+				"stickyOverride": { "classification": "SystemMetaData", "purpose": "FeatureInsight", "isMeasurement": true, "comment": "Whether the router applied a sticky override (1) or not (0)" }
 			}
 		*/
 		this._telemetryService.sendMSFTTelemetryEvent('automode.routerDecision',
 			{
+				conversationId: conversationId ?? '',
+				vscodeRequestId: vscodeRequestId ?? '',
 				predictedLabel: result.predicted_label,
+				candidateModel: result.candidate_models[0] ?? '',
 			},
 			{
 				confidence: result.confidence,
 				latencyMs: result.latency_ms,
 				e2eLatencyMs: e2eLatencyMs,
+				stickyOverride: result.sticky_override ? 1 : 0,
 			}
 		);
 		return result;