@x12i/ai-gateway 9.0.7 → 9.0.8

package/dist/gateway-utils.d.ts

@@ -19,3 +19,20 @@ 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;
+};

package/dist/gateway-utils.js

@@ -179,3 +179,55 @@ 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 };
+}

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 { 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';
@@ -94,7 +94,7 @@ 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'
                 }
@@ -340,12 +340,9 @@ 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'
@@ -491,23 +488,7 @@ 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 enhancedResponse = {
                 content: content,
                 parsedContent: parsedContent,

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 +40,8 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.generateMD5Hash = generateMD5Hash;
 exports.ensureTaskTypeId = ensureTaskTypeId;
 exports.mergeConfig = mergeConfig;
+exports.normalizeRouterUsageTokens = normalizeRouterUsageTokens;
+exports.extractTokenUsageFromRouterResponse = extractTokenUsageFromRouterResponse;
 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 +219,55 @@ 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 };
+}

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

@@ -19,3 +19,20 @@ 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;
+};

package/dist-cjs/gateway.cjs

@@ -97,7 +97,7 @@ 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'
                 }
@@ -343,12 +343,9 @@ 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'
@@ -494,23 +491,7 @@ 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 enhancedResponse = {
                 content: content,
                 parsedContent: parsedContent,

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": "9.0.8",
   "description": "AI Gateway - Unified interface for LLM provider routing and management",
   "type": "module",
   "exports": {