@x12i/ai-gateway 9.0.7 → 9.0.9

package/dist/gateway-utils.d.ts

@@ -19,3 +19,26 @@ export declare function ensureTaskTypeId(request: ChatRequest, logger: Logxer):
 export declare function mergeConfig(request: ChatRequest & {
     useInternalDefaults?: 'skill' | 'audit';
 }, config: GatewayConfig, logger: Logxer): Promise<ChatRequest['config']>;
+/**
+ * Maps provider/router usage objects to gateway token counts (`metadata.tokens`, Activix, trace attempts).
+ * Handles promptTokens/inputTokens, OpenAI-style snake_case, and missing total (sum prompt+completion).
+ */
+export declare function normalizeRouterUsageTokens(usage: unknown): {
+    prompt: number;
+    completion: number;
+    total: number;
+} | undefined;
+/**
+ * Reads token usage from every stable location the router may populate (see docs/PROVIDERS_ROUTER_DIAGNOSTICS_TRACE_REQUIREMENTS.md).
+ */
+export declare function extractTokenUsageFromRouterResponse(routerResponse: unknown): {
+    prompt: number;
+    completion: number;
+    total: number;
+};
+/**
+ * Best-effort USD cost from router/sync AIResponse shape: metadata.costUsd (preferred),
+ * metadata.attempts[].costUsd, response root, then common raw payload locations.
+ * Does not compute cost from tokens — adapters must populate normalized fields or raw usage.cost-style keys.
+ */
+export declare function extractCostUsdFromRouterResponse(routerResponse: unknown): number | undefined;

package/dist/gateway-utils.js

@@ -179,3 +179,102 @@ export async function mergeConfig(request, config, logger) {
     });
     return merged;
 }
+function firstFiniteNumber(...vals) {
+    for (const v of vals) {
+        if (typeof v === 'number' && Number.isFinite(v))
+            return v;
+    }
+    return undefined;
+}
+/**
+ * Maps provider/router usage objects to gateway token counts (`metadata.tokens`, Activix, trace attempts).
+ * Handles promptTokens/inputTokens, OpenAI-style snake_case, and missing total (sum prompt+completion).
+ */
+export function normalizeRouterUsageTokens(usage) {
+    if (usage == null || typeof usage !== 'object')
+        return undefined;
+    const u = usage;
+    const prompt = firstFiniteNumber(u.promptTokens, u.inputTokens, u.prompt, u.prompt_tokens) ?? 0;
+    const completion = firstFiniteNumber(u.completionTokens, u.outputTokens, u.completion, u.completion_tokens) ?? 0;
+    let total = firstFiniteNumber(u.totalTokens, u.total_tokens) ?? 0;
+    if (!total && (prompt || completion))
+        total = prompt + completion;
+    return { prompt, completion, total };
+}
+/**
+ * Reads token usage from every stable location the router may populate (see docs/PROVIDERS_ROUTER_DIAGNOSTICS_TRACE_REQUIREMENTS.md).
+ */
+export function extractTokenUsageFromRouterResponse(routerResponse) {
+    if (routerResponse == null || typeof routerResponse !== 'object') {
+        return { prompt: 0, completion: 0, total: 0 };
+    }
+    const r = routerResponse;
+    const meta = r.metadata != null && typeof r.metadata === 'object'
+        ? r.metadata
+        : undefined;
+    const buckets = [r.usage];
+    if (meta) {
+        buckets.push(meta.usage);
+        const nested = meta['ai-activities-response'];
+        if (nested != null && typeof nested === 'object') {
+            buckets.push(nested.usage);
+        }
+    }
+    const raw = r.rawResponse ?? r.raw;
+    if (raw != null && typeof raw === 'object') {
+        buckets.push(raw.usage);
+    }
+    for (const b of buckets) {
+        const n = normalizeRouterUsageTokens(b);
+        if (n && (n.prompt || n.completion || n.total))
+            return n;
+    }
+    return { prompt: 0, completion: 0, total: 0 };
+}
+/**
+ * Best-effort USD cost from router/sync AIResponse shape: metadata.costUsd (preferred),
+ * metadata.attempts[].costUsd, response root, then common raw payload locations.
+ * Does not compute cost from tokens — adapters must populate normalized fields or raw usage.cost-style keys.
+ */
+export function extractCostUsdFromRouterResponse(routerResponse) {
+    if (routerResponse == null || typeof routerResponse !== 'object')
+        return undefined;
+    const r = routerResponse;
+    const meta = r.metadata != null && typeof r.metadata === 'object'
+        ? r.metadata
+        : undefined;
+    const pick = (...vals) => firstFiniteNumber(...vals);
+    const fromMeta = pick(meta?.costUsd, meta?.cost);
+    if (fromMeta !== undefined)
+        return fromMeta;
+    const fromRoot = pick(r.costUsd, r.cost);
+    if (fromRoot !== undefined)
+        return fromRoot;
+    const attempts = meta?.attempts;
+    if (Array.isArray(attempts)) {
+        for (let i = attempts.length - 1; i >= 0; i--) {
+            const a = attempts[i];
+            if (a != null && typeof a === 'object') {
+                const o = a;
+                const c = pick(o.costUsd, o.cost);
+                if (c !== undefined)
+                    return c;
+            }
+        }
+    }
+    const raw = r.rawResponse ?? r.raw;
+    if (raw != null && typeof raw === 'object') {
+        const rawObj = raw;
+        const usage = rawObj.usage;
+        if (usage != null && typeof usage === 'object') {
+            const u = usage;
+            const fromUsage = pick(u.cost, u.costUsd, u.total_cost, u.totalCost);
+            if (fromUsage !== undefined)
+                return fromUsage;
+        }
+        const fromRawTop = pick(rawObj.cost, rawObj.costUsd);
+        if (fromRawTop !== undefined)
+            return fromRawTop;
+    }
+    return undefined;
+}

package/dist/gateway.js

@@ -8,7 +8,7 @@ import { ensureGatewayRequestIdentity } from './activity-manager.js';
 import { initializeGatewayComponents } from './gateway-config.js';
 import { buildMessages } from './message-builder.js';
 import { extractJsonFromFlexMd } from './flex-md-loader.js';
-import { mergeConfig } from './gateway-utils.js';
+import { extractCostUsdFromRouterResponse, extractTokenUsageFromRouterResponse, mergeConfig } from './gateway-utils.js';
 import { autoRegisterProviders } from './gateway-provider-auto-register.js';
 import { setGatewayLastJobId, setGatewayRuntimeClients } from './runtime-objects.js';
 import { gatewayLogDebug, withActivityIdentity } from './gateway-log-meta.js';
@@ -16,6 +16,25 @@ import { invokeWithRetry } from './gateway-retry.js';
 /** Error message thrown by the router when no provider is registered or specified */
 const NO_PROVIDER_ERROR = 'No provider specified and no providers registered';
 const NO_PROVIDER_HINT = ' Set OPEN_ROUTER_KEY (or OPENROUTER_API_KEY) in the environment to use OpenRouter, or register a provider with the router (e.g. via autoRegisterProviders or gateway config).';
+/** Warn when a successful call reports no tokens and/or explicit zero cost (often missing adapter metadata). */
+function warnIfSuccessfulInvokeReportsZeroUsageOrCost(logger, identity, meta, invokeKind) {
+    const { tokens, costUsd, cost } = meta;
+    const zeroTokens = tokens.prompt === 0 && tokens.completion === 0 && tokens.total === 0;
+    const zeroCostUsd = typeof costUsd === 'number' && costUsd === 0;
+    const zeroCost = typeof cost === 'number' && cost === 0;
+    if (!zeroTokens && !zeroCostUsd && !zeroCost)
+        return;
+    logger.warn('Successful provider response reported zero token usage and/or zero cost; verify router adapter usage and billing metadata', withActivityIdentity(identity, {
+        invokeKind,
+        zeroTokens,
+        zeroCostUsd,
+        zeroCostField: zeroCost,
+        tokens,
+        costUsd,
+        cost,
+        debugKind: gatewayLogDebug.anomaly
+    }));
+}
 /**
  * Simplified AI Gateway - Clean proxy implementation
  */
@@ -87,6 +106,8 @@ export class AIGateway {
                 },
                 mode: 'sync'
             });
+            const costUsdChat = extractCostUsdFromRouterResponse(response);
+            const metaChat = response?.metadata || {};
             // Create enhanced response
             const enhancedResponse = {
                 content: response.content || '',
@@ -94,15 +115,22 @@ export class AIGateway {
                     aiRequestId: request.aiRequestId,
                     identity: request.identity,
                     latencyMs: Date.now() - startTime,
-                    tokens: response.usage || { prompt: 0, completion: 0, total: 0 },
+                    tokens: extractTokenUsageFromRouterResponse(response),
                     taskTypeId,
-                    agentType: 'chat'
+                    agentType: 'chat',
+                    ...(typeof costUsdChat === 'number'
+                        ? {
+                            costUsd: costUsdChat,
+                            ...(typeof metaChat.cost === 'number' ? { cost: metaChat.cost } : { cost: costUsdChat })
+                        }
+                        : {})
                 }
             };
             // Track activity success if activity was started
             if (activity) {
                 try {
                     await this.activityManager.logSuccess(activity, {
+                        ...(typeof costUsdChat === 'number' ? { cost: costUsdChat } : {}),
                         response: enhancedResponse,
                         endTime: Date.now(),
                         duration: Date.now() - startTime
@@ -116,6 +144,11 @@ export class AIGateway {
                     });
                 }
             }
+            warnIfSuccessfulInvokeReportsZeroUsageOrCost(this.logger, request.identity, {
+                tokens: enhancedResponse.metadata.tokens,
+                costUsd: enhancedResponse.metadata.costUsd,
+                cost: enhancedResponse.metadata.cost
+            }, 'invokeChat');
             return enhancedResponse;
         }
         catch (error) {
@@ -340,12 +373,9 @@ export class AIGateway {
                                 const respAny = tryResp;
                                 if (ok && respAny) {
                                     const meta = respAny.metadata || {};
-                                    const usage = respAny.usage || meta?.['ai-activities-response']?.usage;
-                                    const prompt = usage?.promptTokens ?? usage?.inputTokens ?? 0;
-                                    const completion = usage?.completionTokens ?? usage?.outputTokens ?? 0;
-                                    const total = usage?.totalTokens ?? 0;
+                                    const tokenCounts = extractTokenUsageFromRouterResponse(respAny);
                                     a.usage = {
-                                        tokens: { prompt, completion, total },
+                                        tokens: tokenCounts,
                                         maxTokensRequested: typeof meta?.maxTokensRequested === 'number'
                                             ? meta.maxTokensRequested
                                             : typeof mergedConfig?.maxTokens === 'number'
@@ -372,17 +402,9 @@ export class AIGateway {
                                     a.routing.requestIds = requestIds;
                                     a.modelUsed =
                                         meta?.modelUsed || meta?.model || respAny.model || candidate.model;
-                                    const costUsd = typeof meta?.costUsd === 'number'
-                                        ? meta.costUsd
-                                        : typeof meta?.cost === 'number'
-                                            ? meta.cost
-                                            : typeof respAny?.costUsd === 'number'
-                                                ? respAny.costUsd
-                                                : typeof respAny?.cost === 'number'
-                                                    ? respAny.cost
-                                                    : undefined;
-                                    if (typeof costUsd === 'number')
-                                        a.costUsd = costUsd;
+                                    const attemptCostUsd = extractCostUsdFromRouterResponse(respAny);
+                                    if (typeof attemptCostUsd === 'number')
+                                        a.costUsd = attemptCostUsd;
                                     if (includeRawProviderPayload) {
                                         // Size-capped preview only.
                                         const raw = respAny.rawResponse ?? respAny.raw ?? respAny;
@@ -491,23 +513,9 @@ export class AIGateway {
             }
             contentType = 'structured';
             parsingMethod = 'flex-md';
-            // Extract token usage properly
-            let tokens = { prompt: 0, completion: 0, total: 0 };
-            if (routerResponse.usage) {
-                tokens = {
-                    prompt: routerResponse.usage.promptTokens || routerResponse.usage.inputTokens || 0,
-                    completion: routerResponse.usage.completionTokens || routerResponse.usage.outputTokens || 0,
-                    total: routerResponse.usage.totalTokens || 0
-                };
-            }
-            else if (routerResponse.metadata?.['ai-activities-response']?.usage) {
-                const usage = routerResponse.metadata['ai-activities-response'].usage;
-                tokens = {
-                    prompt: usage.promptTokens || usage.inputTokens || 0,
-                    completion: usage.completionTokens || usage.outputTokens || 0,
-                    total: usage.totalTokens || 0
-                };
-            }
+            const tokens = extractTokenUsageFromRouterResponse(routerResponse);
+            const resolvedCostUsd = extractCostUsdFromRouterResponse(routerResponse);
+            const routerMetaForCost = routerResponse?.metadata || {};
             const enhancedResponse = {
                 content: content,
                 parsedContent: parsedContent,
@@ -520,6 +528,14 @@ export class AIGateway {
                     agentType: 'ai',
                     contentType,
                     parsingMethod,
+                    ...(typeof resolvedCostUsd === 'number'
+                        ? {
+                            costUsd: resolvedCostUsd,
+                            ...(typeof routerMetaForCost.cost === 'number'
+                                ? { cost: routerMetaForCost.cost }
+                                : { cost: resolvedCostUsd })
+                        }
+                        : {}),
                     ...(traceEnabled
                         ? (() => {
                             const meta = routerResponse?.metadata || {};
@@ -531,18 +547,11 @@ export class AIGateway {
                                 : typeof mergedConfig?.maxTokens === 'number'
                                     ? mergedConfig.maxTokens
                                     : undefined;
-                            const costUsd = typeof meta.costUsd === 'number'
-                                ? meta.costUsd
-                                : typeof meta.cost === 'number'
-                                    ? meta.cost
-                                    : undefined;
                             return {
                                 provider,
                                 region,
                                 modelUsed,
                                 maxTokensRequested,
-                                cost: typeof meta.cost === 'number' ? meta.cost : undefined,
-                                costUsd,
                                 requestIds: traceRequestIds,
                                 retryCount: traceRetryCount,
                                 fallbackCount: traceFallbackCount,
@@ -568,6 +577,7 @@ export class AIGateway {
                         usage: tokens
                     };
                     await this.activityManager.logSuccess(activity, {
+                        ...(typeof resolvedCostUsd === 'number' ? { cost: resolvedCostUsd } : {}),
                         response: activityResponse,
                         endTime: Date.now(),
                         duration: Date.now() - startTime
@@ -581,6 +591,11 @@ export class AIGateway {
                     });
                 }
             }
+            warnIfSuccessfulInvokeReportsZeroUsageOrCost(this.logger, request.identity, {
+                tokens: enhancedResponse.metadata.tokens,
+                costUsd: enhancedResponse.metadata.costUsd,
+                cost: enhancedResponse.metadata.cost
+            }, 'invoke');
             this.logger.debug('gateway: enhancedResponse', withActivityIdentity(request.identity, {
                 latencyMs: enhancedResponse.metadata?.latencyMs,
                 contentType: enhancedResponse.metadata?.contentType,

package/dist/usage-tracker.js

@@ -8,6 +8,7 @@
  *
  * Note: x-models dependency removed - usage tracking functions are permanently disabled
  */
+import { extractTokenUsageFromRouterResponse } from './gateway-utils.js';
 /**
  * Manages usage tracking for LLM requests
  */
@@ -26,11 +27,7 @@ export class UsageTracker {
      * @returns Token usage breakdown
      */
     extractTokens(response) {
-        return {
-            prompt: response.usage?.promptTokens || 0,
-            completion: response.usage?.completionTokens || 0,
-            total: response.usage?.totalTokens || 0
-        };
+        return extractTokenUsageFromRouterResponse(response);
     }
     /**
      * Records usage for a request

package/dist-cjs/gateway-utils.cjs

@@ -40,6 +40,9 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.generateMD5Hash = generateMD5Hash;
 exports.ensureTaskTypeId = ensureTaskTypeId;
 exports.mergeConfig = mergeConfig;
+exports.normalizeRouterUsageTokens = normalizeRouterUsageTokens;
+exports.extractTokenUsageFromRouterResponse = extractTokenUsageFromRouterResponse;
+exports.extractCostUsdFromRouterResponse = extractCostUsdFromRouterResponse;
 const crypto = __importStar(require("crypto"));
 const gateway_instructions_js_1 = require("./gateway-instructions.cjs");
 const flex_md_loader_js_1 = require("./flex-md-loader.cjs");
@@ -217,3 +220,102 @@ async function mergeConfig(request, config, logger) {
     });
     return merged;
 }
+function firstFiniteNumber(...vals) {
+    for (const v of vals) {
+        if (typeof v === 'number' && Number.isFinite(v))
+            return v;
+    }
+    return undefined;
+}
+/**
+ * Maps provider/router usage objects to gateway token counts (`metadata.tokens`, Activix, trace attempts).
+ * Handles promptTokens/inputTokens, OpenAI-style snake_case, and missing total (sum prompt+completion).
+ */
+function normalizeRouterUsageTokens(usage) {
+    if (usage == null || typeof usage !== 'object')
+        return undefined;
+    const u = usage;
+    const prompt = firstFiniteNumber(u.promptTokens, u.inputTokens, u.prompt, u.prompt_tokens) ?? 0;
+    const completion = firstFiniteNumber(u.completionTokens, u.outputTokens, u.completion, u.completion_tokens) ?? 0;
+    let total = firstFiniteNumber(u.totalTokens, u.total_tokens) ?? 0;
+    if (!total && (prompt || completion))
+        total = prompt + completion;
+    return { prompt, completion, total };
+}
+/**
+ * Reads token usage from every stable location the router may populate (see docs/PROVIDERS_ROUTER_DIAGNOSTICS_TRACE_REQUIREMENTS.md).
+ */
+function extractTokenUsageFromRouterResponse(routerResponse) {
+    if (routerResponse == null || typeof routerResponse !== 'object') {
+        return { prompt: 0, completion: 0, total: 0 };
+    }
+    const r = routerResponse;
+    const meta = r.metadata != null && typeof r.metadata === 'object'
+        ? r.metadata
+        : undefined;
+    const buckets = [r.usage];
+    if (meta) {
+        buckets.push(meta.usage);
+        const nested = meta['ai-activities-response'];
+        if (nested != null && typeof nested === 'object') {
+            buckets.push(nested.usage);
+        }
+    }
+    const raw = r.rawResponse ?? r.raw;
+    if (raw != null && typeof raw === 'object') {
+        buckets.push(raw.usage);
+    }
+    for (const b of buckets) {
+        const n = normalizeRouterUsageTokens(b);
+        if (n && (n.prompt || n.completion || n.total))
+            return n;
+    }
+    return { prompt: 0, completion: 0, total: 0 };
+}
+/**
+ * Best-effort USD cost from router/sync AIResponse shape: metadata.costUsd (preferred),
+ * metadata.attempts[].costUsd, response root, then common raw payload locations.
+ * Does not compute cost from tokens — adapters must populate normalized fields or raw usage.cost-style keys.
+ */
+function extractCostUsdFromRouterResponse(routerResponse) {
+    if (routerResponse == null || typeof routerResponse !== 'object')
+        return undefined;
+    const r = routerResponse;
+    const meta = r.metadata != null && typeof r.metadata === 'object'
+        ? r.metadata
+        : undefined;
+    const pick = (...vals) => firstFiniteNumber(...vals);
+    const fromMeta = pick(meta?.costUsd, meta?.cost);
+    if (fromMeta !== undefined)
+        return fromMeta;
+    const fromRoot = pick(r.costUsd, r.cost);
+    if (fromRoot !== undefined)
+        return fromRoot;
+    const attempts = meta?.attempts;
+    if (Array.isArray(attempts)) {
+        for (let i = attempts.length - 1; i >= 0; i--) {
+            const a = attempts[i];
+            if (a != null && typeof a === 'object') {
+                const o = a;
+                const c = pick(o.costUsd, o.cost);
+                if (c !== undefined)
+                    return c;
+            }
+        }
+    }
+    const raw = r.rawResponse ?? r.raw;
+    if (raw != null && typeof raw === 'object') {
+        const rawObj = raw;
+        const usage = rawObj.usage;
+        if (usage != null && typeof usage === 'object') {
+            const u = usage;
+            const fromUsage = pick(u.cost, u.costUsd, u.total_cost, u.totalCost);
+            if (fromUsage !== undefined)
+                return fromUsage;
+        }
+        const fromRawTop = pick(rawObj.cost, rawObj.costUsd);
+        if (fromRawTop !== undefined)
+            return fromRawTop;
+    }
+    return undefined;
+}

package/dist-cjs/gateway-utils.d.ts

@@ -19,3 +19,26 @@ export declare function ensureTaskTypeId(request: ChatRequest, logger: Logxer):
 export declare function mergeConfig(request: ChatRequest & {
     useInternalDefaults?: 'skill' | 'audit';
 }, config: GatewayConfig, logger: Logxer): Promise<ChatRequest['config']>;
+/**
+ * Maps provider/router usage objects to gateway token counts (`metadata.tokens`, Activix, trace attempts).
+ * Handles promptTokens/inputTokens, OpenAI-style snake_case, and missing total (sum prompt+completion).
+ */
+export declare function normalizeRouterUsageTokens(usage: unknown): {
+    prompt: number;
+    completion: number;
+    total: number;
+} | undefined;
+/**
+ * Reads token usage from every stable location the router may populate (see docs/PROVIDERS_ROUTER_DIAGNOSTICS_TRACE_REQUIREMENTS.md).
+ */
+export declare function extractTokenUsageFromRouterResponse(routerResponse: unknown): {
+    prompt: number;
+    completion: number;
+    total: number;
+};
+/**
+ * Best-effort USD cost from router/sync AIResponse shape: metadata.costUsd (preferred),
+ * metadata.attempts[].costUsd, response root, then common raw payload locations.
+ * Does not compute cost from tokens — adapters must populate normalized fields or raw usage.cost-style keys.
+ */
+export declare function extractCostUsdFromRouterResponse(routerResponse: unknown): number | undefined;

package/dist-cjs/gateway.cjs

@@ -19,6 +19,25 @@ const gateway_retry_js_1 = require("./gateway-retry.cjs");
 /** Error message thrown by the router when no provider is registered or specified */
 const NO_PROVIDER_ERROR = 'No provider specified and no providers registered';
 const NO_PROVIDER_HINT = ' Set OPEN_ROUTER_KEY (or OPENROUTER_API_KEY) in the environment to use OpenRouter, or register a provider with the router (e.g. via autoRegisterProviders or gateway config).';
+/** Warn when a successful call reports no tokens and/or explicit zero cost (often missing adapter metadata). */
+function warnIfSuccessfulInvokeReportsZeroUsageOrCost(logger, identity, meta, invokeKind) {
+    const { tokens, costUsd, cost } = meta;
+    const zeroTokens = tokens.prompt === 0 && tokens.completion === 0 && tokens.total === 0;
+    const zeroCostUsd = typeof costUsd === 'number' && costUsd === 0;
+    const zeroCost = typeof cost === 'number' && cost === 0;
+    if (!zeroTokens && !zeroCostUsd && !zeroCost)
+        return;
+    logger.warn('Successful provider response reported zero token usage and/or zero cost; verify router adapter usage and billing metadata', (0, gateway_log_meta_js_1.withActivityIdentity)(identity, {
+        invokeKind,
+        zeroTokens,
+        zeroCostUsd,
+        zeroCostField: zeroCost,
+        tokens,
+        costUsd,
+        cost,
+        debugKind: gateway_log_meta_js_1.gatewayLogDebug.anomaly
+    }));
+}
 /**
  * Simplified AI Gateway - Clean proxy implementation
  */
@@ -90,6 +109,8 @@ class AIGateway {
                 },
                 mode: 'sync'
             });
+            const costUsdChat = (0, gateway_utils_js_1.extractCostUsdFromRouterResponse)(response);
+            const metaChat = response?.metadata || {};
             // Create enhanced response
             const enhancedResponse = {
                 content: response.content || '',
@@ -97,15 +118,22 @@ class AIGateway {
                     aiRequestId: request.aiRequestId,
                     identity: request.identity,
                     latencyMs: Date.now() - startTime,
-                    tokens: response.usage || { prompt: 0, completion: 0, total: 0 },
+                    tokens: (0, gateway_utils_js_1.extractTokenUsageFromRouterResponse)(response),
                     taskTypeId,
-                    agentType: 'chat'
+                    agentType: 'chat',
+                    ...(typeof costUsdChat === 'number'
+                        ? {
+                            costUsd: costUsdChat,
+                            ...(typeof metaChat.cost === 'number' ? { cost: metaChat.cost } : { cost: costUsdChat })
+                        }
+                        : {})
                 }
             };
             // Track activity success if activity was started
             if (activity) {
                 try {
                     await this.activityManager.logSuccess(activity, {
+                        ...(typeof costUsdChat === 'number' ? { cost: costUsdChat } : {}),
                         response: enhancedResponse,
                         endTime: Date.now(),
                         duration: Date.now() - startTime
@@ -119,6 +147,11 @@ class AIGateway {
                     });
                 }
             }
+            warnIfSuccessfulInvokeReportsZeroUsageOrCost(this.logger, request.identity, {
+                tokens: enhancedResponse.metadata.tokens,
+                costUsd: enhancedResponse.metadata.costUsd,
+                cost: enhancedResponse.metadata.cost
+            }, 'invokeChat');
             return enhancedResponse;
         }
         catch (error) {
@@ -343,12 +376,9 @@ class AIGateway {
                                 const respAny = tryResp;
                                 if (ok && respAny) {
                                     const meta = respAny.metadata || {};
-                                    const usage = respAny.usage || meta?.['ai-activities-response']?.usage;
-                                    const prompt = usage?.promptTokens ?? usage?.inputTokens ?? 0;
-                                    const completion = usage?.completionTokens ?? usage?.outputTokens ?? 0;
-                                    const total = usage?.totalTokens ?? 0;
+                                    const tokenCounts = (0, gateway_utils_js_1.extractTokenUsageFromRouterResponse)(respAny);
                                     a.usage = {
-                                        tokens: { prompt, completion, total },
+                                        tokens: tokenCounts,
                                         maxTokensRequested: typeof meta?.maxTokensRequested === 'number'
                                             ? meta.maxTokensRequested
                                             : typeof mergedConfig?.maxTokens === 'number'
@@ -375,17 +405,9 @@ class AIGateway {
                                     a.routing.requestIds = requestIds;
                                     a.modelUsed =
                                         meta?.modelUsed || meta?.model || respAny.model || candidate.model;
-                                    const costUsd = typeof meta?.costUsd === 'number'
-                                        ? meta.costUsd
-                                        : typeof meta?.cost === 'number'
-                                            ? meta.cost
-                                            : typeof respAny?.costUsd === 'number'
-                                                ? respAny.costUsd
-                                                : typeof respAny?.cost === 'number'
-                                                    ? respAny.cost
-                                                    : undefined;
-                                    if (typeof costUsd === 'number')
-                                        a.costUsd = costUsd;
+                                    const attemptCostUsd = (0, gateway_utils_js_1.extractCostUsdFromRouterResponse)(respAny);
+                                    if (typeof attemptCostUsd === 'number')
+                                        a.costUsd = attemptCostUsd;
                                     if (includeRawProviderPayload) {
                                         // Size-capped preview only.
                                         const raw = respAny.rawResponse ?? respAny.raw ?? respAny;
@@ -494,23 +516,9 @@ class AIGateway {
             }
             contentType = 'structured';
             parsingMethod = 'flex-md';
-            // Extract token usage properly
-            let tokens = { prompt: 0, completion: 0, total: 0 };
-            if (routerResponse.usage) {
-                tokens = {
-                    prompt: routerResponse.usage.promptTokens || routerResponse.usage.inputTokens || 0,
-                    completion: routerResponse.usage.completionTokens || routerResponse.usage.outputTokens || 0,
-                    total: routerResponse.usage.totalTokens || 0
-                };
-            }
-            else if (routerResponse.metadata?.['ai-activities-response']?.usage) {
-                const usage = routerResponse.metadata['ai-activities-response'].usage;
-                tokens = {
-                    prompt: usage.promptTokens || usage.inputTokens || 0,
-                    completion: usage.completionTokens || usage.outputTokens || 0,
-                    total: usage.totalTokens || 0
-                };
-            }
+            const tokens = (0, gateway_utils_js_1.extractTokenUsageFromRouterResponse)(routerResponse);
+            const resolvedCostUsd = (0, gateway_utils_js_1.extractCostUsdFromRouterResponse)(routerResponse);
+            const routerMetaForCost = routerResponse?.metadata || {};
             const enhancedResponse = {
                 content: content,
                 parsedContent: parsedContent,
@@ -523,6 +531,14 @@ class AIGateway {
                     agentType: 'ai',
                     contentType,
                     parsingMethod,
+                    ...(typeof resolvedCostUsd === 'number'
+                        ? {
+                            costUsd: resolvedCostUsd,
+                            ...(typeof routerMetaForCost.cost === 'number'
+                                ? { cost: routerMetaForCost.cost }
+                                : { cost: resolvedCostUsd })
+                        }
+                        : {}),
                     ...(traceEnabled
                         ? (() => {
                             const meta = routerResponse?.metadata || {};
@@ -534,18 +550,11 @@ class AIGateway {
                                 : typeof mergedConfig?.maxTokens === 'number'
                                     ? mergedConfig.maxTokens
                                     : undefined;
-                            const costUsd = typeof meta.costUsd === 'number'
-                                ? meta.costUsd
-                                : typeof meta.cost === 'number'
-                                    ? meta.cost
-                                    : undefined;
                             return {
                                 provider,
                                 region,
                                 modelUsed,
                                 maxTokensRequested,
-                                cost: typeof meta.cost === 'number' ? meta.cost : undefined,
-                                costUsd,
                                 requestIds: traceRequestIds,
                                 retryCount: traceRetryCount,
                                 fallbackCount: traceFallbackCount,
@@ -571,6 +580,7 @@ class AIGateway {
                         usage: tokens
                     };
                     await this.activityManager.logSuccess(activity, {
+                        ...(typeof resolvedCostUsd === 'number' ? { cost: resolvedCostUsd } : {}),
                         response: activityResponse,
                         endTime: Date.now(),
                         duration: Date.now() - startTime
@@ -584,6 +594,11 @@ class AIGateway {
                     });
                 }
             }
+            warnIfSuccessfulInvokeReportsZeroUsageOrCost(this.logger, request.identity, {
+                tokens: enhancedResponse.metadata.tokens,
+                costUsd: enhancedResponse.metadata.costUsd,
+                cost: enhancedResponse.metadata.cost
+            }, 'invoke');
             this.logger.debug('gateway: enhancedResponse', (0, gateway_log_meta_js_1.withActivityIdentity)(request.identity, {
                 latencyMs: enhancedResponse.metadata?.latencyMs,
                 contentType: enhancedResponse.metadata?.contentType,

package/dist-cjs/usage-tracker.cjs

@@ -11,6 +11,7 @@
  */
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.UsageTracker = void 0;
+const gateway_utils_js_1 = require("./gateway-utils.cjs");
 /**
  * Manages usage tracking for LLM requests
  */
@@ -29,11 +30,7 @@ class UsageTracker {
      * @returns Token usage breakdown
      */
     extractTokens(response) {
-        return {
-            prompt: response.usage?.promptTokens || 0,
-            completion: response.usage?.completionTokens || 0,
-            total: response.usage?.totalTokens || 0
-        };
+        return (0, gateway_utils_js_1.extractTokenUsageFromRouterResponse)(response);
     }
     /**
      * Records usage for a request

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@x12i/ai-gateway",
-  "version": "9.0.7",
+  "version": "9.0.9",
   "description": "AI Gateway - Unified interface for LLM provider routing and management",
   "type": "module",
   "exports": {