Show raw input token counts in json output. (#15021)

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

packages/core/src/telemetry/uiTelemetry.test.ts

@@ -173,6 +173,7 @@ describe('UiTelemetryService', () => {
           totalLatencyMs: 500,
         },
         tokens: {
+          input: 5,
           prompt: 10,
           candidates: 20,
           total: 30,
@@ -227,6 +228,7 @@ describe('UiTelemetryService', () => {
           totalLatencyMs: 1100,
         },
         tokens: {
+          input: 10,
           prompt: 25,
           candidates: 45,
           total: 70,
@@ -301,6 +303,7 @@ describe('UiTelemetryService', () => {
           totalLatencyMs: 300,
         },
         tokens: {
+          input: 0,
           prompt: 0,
           candidates: 0,
           total: 0,
@@ -345,6 +348,7 @@ describe('UiTelemetryService', () => {
           totalLatencyMs: 800,
         },
         tokens: {
+          input: 5,
           prompt: 10,
           candidates: 20,
           total: 30,

packages/core/src/telemetry/uiTelemetry.ts

@@ -43,6 +43,7 @@ export interface ModelMetrics {
     totalLatencyMs: number;
   };
   tokens: {
+    input: number;
     prompt: number;
     candidates: number;
     total: number;
@@ -80,6 +81,7 @@ const createInitialModelMetrics = (): ModelMetrics => ({
     totalLatencyMs: 0,
   },
   tokens: {
+    input: 0,
     prompt: 0,
     candidates: 0,
     total: 0,
@@ -171,6 +173,10 @@ export class UiTelemetryService extends EventEmitter {
     modelMetrics.tokens.cached += event.usage.cached_content_token_count;
     modelMetrics.tokens.thoughts += event.usage.thoughts_token_count;
     modelMetrics.tokens.tool += event.usage.tool_token_count;
+    modelMetrics.tokens.input = Math.max(
+      0,
+      modelMetrics.tokens.prompt - modelMetrics.tokens.cached,
+    );
   }
 
   private processApiError(event: ApiErrorEvent) {