@hebo-ai/gateway 0.11.1 → 0.11.3

package/dist/endpoints/chat-completions/otel.js

@@ -160,6 +160,7 @@ export const getChatResponseAttributes = (completions, signalLevel) => {
             "gen_ai.usage.total_tokens": completions.usage?.total_tokens,
             "gen_ai.usage.input_tokens": completions.usage?.prompt_tokens,
             "gen_ai.usage.cache_read.input_tokens": completions.usage?.prompt_tokens_details?.cached_tokens,
+            "gen_ai.usage.cache_creation.input_tokens": completions.usage?.prompt_tokens_details?.cache_write_tokens,
             "gen_ai.usage.output_tokens": completions.usage?.completion_tokens,
             "gen_ai.usage.reasoning.output_tokens": completions.usage?.completion_tokens_details?.reasoning_tokens,
         });

package/dist/endpoints/conversations/storage/dialects/utils.d.ts

@@ -6,7 +6,7 @@ export declare function createParamsMapper(mappers: ((v: unknown) => unknown)[])
  * Normalizes an object (row) by applying a chain of atomic mappers.
  * Mappers are expected to mutate the object for performance and to avoid spreads.
  */
-export declare function createRowMapper<T>(mappers: ((row: Record<string, unknown>) => Record<string, unknown>)[]): (row: Record<string, unknown>) => T;
+export declare function createRowMapper<T>(mappers: ((row: T) => T)[]): (row: T) => T;
 /**
  * Atomic mappers for input parameters.
  */

package/dist/endpoints/messages/converters.js

@@ -541,17 +541,20 @@ export class MessagesTransformStream extends TransformStream {
                     }
                     case "tool-input-start": {
                         currentToolCallId = part.id;
+                        const contentBlock = {
+                            type: "tool_use",
+                            id: part.id,
+                            name: normalizeToolName(part.toolName),
+                            input: {},
+                        };
+                        if (part.providerMetadata)
+                            contentBlock.extra_content = part.providerMetadata;
                         controller.enqueue({
                             event: "content_block_start",
                             data: {
                                 type: "content_block_start",
                                 index: blockIndex,
-                                content_block: {
-                                    type: "tool_use",
-                                    id: part.id,
-                                    name: normalizeToolName(part.toolName),
-                                    input: {},
-                                },
+                                content_block: contentBlock,
                             },
                         });
                         break;
@@ -622,14 +625,12 @@ export class MessagesTransformStream extends TransformStream {
                     }
                     case "finish": {
                         const stopReason = mapStopReason(part.finishReason);
-                        const totalOutputTokens = part.totalUsage?.outputTokens ?? 0;
-                        const totalInputTokens = part.totalUsage?.inputTokens ?? 0;
                         controller.enqueue({
                             event: "message_delta",
                             data: {
                                 type: "message_delta",
                                 delta: { stop_reason: stopReason, stop_sequence: null },
-                                usage: { output_tokens: totalOutputTokens, input_tokens: totalInputTokens },
+                                usage: mapUsage(part.totalUsage),
                             },
                         });
                         controller.enqueue({

package/dist/endpoints/messages/schema.d.ts

@@ -647,10 +647,7 @@ export type MessageDeltaEvent = SseFrame<{
         stop_reason: MessagesStopReason;
         stop_sequence: string | null;
     };
-    usage: {
-        output_tokens: number;
-        input_tokens?: number;
-    };
+    usage: MessagesUsage;
 }, "message_delta">;
 export type MessageStopEvent = SseFrame<{
     type: "message_stop";

package/dist/errors/ai-sdk.js

@@ -6,10 +6,19 @@ const normalizeApiCallError = (error) => {
     const statusText = `UPSTREAM_${STATUS_TEXT(status)}`;
     return new GatewayError(error, status, statusText, undefined, error.responseHeaders ?? undefined);
 };
+// `AbortError` / `TimeoutError` (raised by the AI SDK's internal `timeout` controller,
+// `AbortSignal.timeout`, or an aborted upstream `fetch`) reach us as plain DOMExceptions
+// that none of the AI SDK error classes match. Treat them as upstream gateway timeouts
+// so they surface as 504 with retry headers rather than defaulting to 500/502.
+// Inbound client disconnects are caught earlier in `lifecycle.ts` and overridden to 499.
+const isUpstreamAbortError = (error) => error instanceof Error && (error.name === "AbortError" || error.name === "TimeoutError");
 export const normalizeAiSdkError = (error) => {
     if (APICallError.isInstance(error)) {
         return normalizeApiCallError(error);
     }
+    if (isUpstreamAbortError(error)) {
+        return new GatewayError(error, 504, `UPSTREAM_${STATUS_TEXT(504)}`);
+    }
     if (RetryError.isInstance(error)) {
         if (APICallError.isInstance(error.lastError)) {
             return normalizeApiCallError(error.lastError);

package/dist/lifecycle.js

@@ -50,11 +50,12 @@ export const winterCgHandler = (run, config) => {
             else if (status === 200 && ctx.response?.status)
                 realStatus = ctx.response.status;
             if (realStatus !== 200) {
+                const err = reason ?? ctx.request.signal.reason;
                 logger[realStatus >= 500 ? "error" : "warn"]({
                     requestId: ctx.requestId,
-                    err: reason ?? ctx.request.signal.reason,
+                    err,
                 });
-                span.recordError(reason, true);
+                span.recordError(err, true);
             }
             span.setAttributes({ "http.response.status_code_effective": realStatus });
             if (ctx.operation === "chat" ||

package/dist/models/openai/presets.d.ts

@@ -249,6 +249,30 @@ export declare const gpt54Pro: import("../../utils").Preset<"openai/gpt-5.4-pro"
     capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
     providers: readonly ["openai", "azure"];
 }>;
+export declare const gpt55: import("../../utils").Preset<"openai/gpt-5.5", CatalogModel, {
+    name: string;
+    created: string;
+    knowledge: string;
+    context: number;
+    modalities: {
+        input: readonly ["text", "image"];
+        output: readonly ["text"];
+    };
+    capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+    providers: readonly ["openai", "azure"];
+}>;
+export declare const gpt55Pro: import("../../utils").Preset<"openai/gpt-5.5-pro", CatalogModel, {
+    name: string;
+    created: string;
+    knowledge: string;
+    context: number;
+    modalities: {
+        input: readonly ["text", "image"];
+        output: readonly ["text"];
+    };
+    capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+    providers: readonly ["openai", "azure"];
+}>;
 export declare const textEmbedding3Small: import("../../utils").Preset<"openai/text-embedding-3-small", CatalogModel, {
     name: string;
     created: string;
@@ -388,7 +412,18 @@ export declare const gptOss: {
     }>];
 };
 export declare const gpt: {
-    readonly latest: readonly [import("../../utils").Preset<"openai/gpt-5.4", CatalogModel, {
+    readonly latest: readonly [import("../../utils").Preset<"openai/gpt-5.5", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
+    }>, import("../../utils").Preset<"openai/gpt-5.5-pro", CatalogModel, {
         name: string;
         created: string;
         knowledge: string;
@@ -651,6 +686,28 @@ export declare const gpt: {
         };
         capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
         providers: readonly ["openai", "azure"];
+    }> | import("../../utils").Preset<"openai/gpt-5.5", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
+    }> | import("../../utils").Preset<"openai/gpt-5.5-pro", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
     }>)[];
     readonly "v5.x": readonly [import("../../utils").Preset<"openai/gpt-5", CatalogModel, {
         name: string;
@@ -870,6 +927,28 @@ export declare const gpt: {
         };
         capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
         providers: readonly ["openai", "azure"];
+    }>, import("../../utils").Preset<"openai/gpt-5.5", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
+    }>, import("../../utils").Preset<"openai/gpt-5.5-pro", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
     }>];
     readonly v5: readonly [import("../../utils").Preset<"openai/gpt-5", CatalogModel, {
         name: string;
@@ -1094,6 +1173,29 @@ export declare const gpt: {
         capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
         providers: readonly ["openai", "azure"];
     }>];
+    readonly "v5.5": readonly [import("../../utils").Preset<"openai/gpt-5.5", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
+    }>, import("../../utils").Preset<"openai/gpt-5.5-pro", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
+    }>];
     readonly codex: readonly [import("../../utils").Preset<"openai/gpt-5-codex", CatalogModel, {
         name: string;
         created: string;
@@ -1237,6 +1339,17 @@ export declare const gpt: {
         };
         capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
         providers: readonly ["openai", "azure"];
+    }>, import("../../utils").Preset<"openai/gpt-5.5-pro", CatalogModel, {
+        name: string;
+        created: string;
+        knowledge: string;
+        context: number;
+        modalities: {
+            input: readonly ["text", "image"];
+            output: readonly ["text"];
+        };
+        capabilities: readonly ["attachments", "reasoning", "tool_call", "structured_output", "temperature"];
+        providers: readonly ["openai", "azure"];
     }>];
 };
 export declare const textEmbeddings: {

package/dist/models/openai/presets.js

@@ -194,6 +194,20 @@ export const gpt54Pro = presetFor()("openai/gpt-5.4-pro", {
     knowledge: "2025-08",
     context: 1050000,
 });
+export const gpt55 = presetFor()("openai/gpt-5.5", {
+    ...GPT_BASE,
+    name: "GPT-5.5",
+    created: "2026-04-22",
+    knowledge: "2025-08",
+    context: 1050000,
+});
+export const gpt55Pro = presetFor()("openai/gpt-5.5-pro", {
+    ...GPT_PRO_BASE,
+    name: "GPT-5.5 Pro",
+    created: "2026-04-24",
+    knowledge: "2025-12",
+    context: 1050000,
+});
 export const textEmbedding3Small = presetFor()("openai/text-embedding-3-small", {
     ...EMBEDDINGS_BASE,
     name: "Text Embedding 3 Small",
@@ -230,6 +244,7 @@ const gptAtomic = {
     "v5.2": [gpt52, gpt52Chat, gpt52Pro, gpt52Codex],
     "v5.3": [gpt53Codex, gpt53CodexSpark, gpt53Chat],
     "v5.4": [gpt54, gpt54Mini, gpt54Nano, gpt54Pro],
+    "v5.5": [gpt55, gpt55Pro],
     codex: [
         gpt5Codex,
         gpt51Codex,
@@ -240,7 +255,7 @@ const gptAtomic = {
         gpt53CodexSpark,
     ],
     chat: [gpt51Chat, gpt52Chat, gpt53Chat],
-    pro: [gpt5Pro, gpt52Pro, gpt54Pro],
+    pro: [gpt5Pro, gpt52Pro, gpt54Pro, gpt55Pro],
 };
 const gptGroups = {
     "v5.x": [
@@ -249,6 +264,7 @@ const gptGroups = {
         ...gptAtomic["v5.2"],
         ...gptAtomic["v5.3"],
         ...gptAtomic["v5.4"],
+        ...gptAtomic["v5.5"],
     ],
 };
 const textEmbeddingsAtomic = {
@@ -266,7 +282,8 @@ export const gptOss = {
 export const gpt = {
     ...gptAtomic,
     ...gptGroups,
-    latest: [gpt54, gpt54Mini, gpt54Nano],
+    // 5.5 Mini/Nano not released yet; keep 5.4 small variants in `latest` until they ship.
+    latest: [gpt55, gpt55Pro, gpt54Mini, gpt54Nano],
     all: Object.values(gptAtomic).flat(),
 };
 export const textEmbeddings = {

package/dist/models/types.d.ts

@@ -1,5 +1,5 @@
 import type { ProviderId } from "../providers/types";
-export declare const CANONICAL_MODEL_IDS: readonly ["anthropic/claude-opus-4.7", "anthropic/claude-opus-4.6", "anthropic/claude-sonnet-4.6", "anthropic/claude-haiku-4.5", "anthropic/claude-sonnet-4.5", "anthropic/claude-opus-4.5", "anthropic/claude-opus-4.1", "anthropic/claude-opus-4", "anthropic/claude-sonnet-4", "anthropic/claude-sonnet-3.7", "anthropic/claude-sonnet-3.5", "anthropic/claude-haiku-3.5", "anthropic/claude-haiku-3", "openai/gpt-oss-20b", "openai/gpt-oss-120b", "openai/gpt-5", "openai/gpt-5-pro", "openai/gpt-5.2", "openai/gpt-5.2-chat", "openai/gpt-5.2-pro", "openai/gpt-5.2-codex", "openai/gpt-5.3-codex", "openai/gpt-5.3-codex-spark", "openai/gpt-5.3-chat", "openai/gpt-5.4", "openai/gpt-5.4-mini", "openai/gpt-5.4-nano", "openai/gpt-5.4-pro", "openai/gpt-5-mini", "openai/gpt-5-nano", "openai/gpt-5-codex", "openai/gpt-5.1-codex", "openai/gpt-5.1-codex-max", "openai/gpt-5.1-codex-mini", "openai/gpt-5.1-chat", "openai/gpt-5.1", "openai/text-embedding-3-small", "openai/text-embedding-3-large", "amazon/nova-micro", "amazon/nova-lite", "amazon/nova-pro", "amazon/nova-premier", "amazon/nova-2-lite", "amazon/nova-2-multimodal-embeddings", "google/gemini-2.5-flash-lite", "google/gemini-2.5-flash", "google/gemini-2.5-pro", "google/gemini-3-flash-preview", "google/gemini-3.1-flash-lite-preview", "google/gemini-3.1-pro-preview", "google/gemini-embedding-2", "google/embedding-001", "google/gemma-3-1b", "google/gemma-3-4b", "google/gemma-3-12b", "google/gemma-3-27b", "google/gemma-4-e2b", "google/gemma-4-e4b", "google/gemma-4-26b-a4b", "google/gemma-4-31b", "meta/llama-3.1-8b", "meta/llama-3.1-70b", "meta/llama-3.1-405b", "meta/llama-3.2-1b", "meta/llama-3.2-3b", "meta/llama-3.2-11b", "meta/llama-3.2-90b", "meta/llama-3.3-70b", "meta/llama-4-scout", "meta/llama-4-maverick", "cohere/embed-v4.0", "cohere/embed-english-v3.0", "cohere/embed-english-light-v3.0", "cohere/embed-multilingual-v3.0", "cohere/embed-multilingual-light-v3.0", "cohere/command-a", "cohere/command-r7b", "cohere/command-a-translate", "cohere/command-a-reasoning", "cohere/command-a-vision", "cohere/command-r", "cohere/command-r-plus", "minimax/m2.5", "minimax/m2.7", "moonshot/kimi-k2.5", "moonshot/kimi-k2.6", "xai/grok-4.1-fast", "xai/grok-4.1-fast-reasoning", "xai/grok-4.2", "xai/grok-4.2-reasoning", "xai/grok-4.2-multi-agent", "deepseek/deepseek-v3.2", "voyage/voyage-2-code", "voyage/voyage-2-law", "voyage/voyage-2-finance", "voyage/voyage-3-code", "voyage/voyage-3-large", "voyage/voyage-3.5-lite", "voyage/voyage-3.5", "voyage/voyage-4-lite", "voyage/voyage-4", "voyage/voyage-4-large", "alibaba/qwen3-235b", "alibaba/qwen3-32b", "alibaba/qwen3.5-plus", "alibaba/qwen3.5-flash", "alibaba/qwen3.5-397b", "alibaba/qwen3.5-122b", "alibaba/qwen3.5-35b", "alibaba/qwen3.5-27b", "alibaba/qwen3.5-9b", "alibaba/qwen3.5-4b", "alibaba/qwen3.5-2b", "alibaba/qwen3.5-0.8b", "alibaba/qwen3.6-plus", "alibaba/qwen3.6-flash", "alibaba/qwen3.6-27b", "alibaba/qwen3.6-max-preview", "alibaba/qwen3-coder-next", "alibaba/qwen3-vl-235b", "alibaba/qwen3-embedding-0.6b", "alibaba/qwen3-embedding-4b", "alibaba/qwen3-embedding-8b", "zhipu/glm-5", "zhipu/glm-5-turbo", "zhipu/glm-5.1"];
+export declare const CANONICAL_MODEL_IDS: readonly ["anthropic/claude-opus-4.7", "anthropic/claude-opus-4.6", "anthropic/claude-sonnet-4.6", "anthropic/claude-haiku-4.5", "anthropic/claude-sonnet-4.5", "anthropic/claude-opus-4.5", "anthropic/claude-opus-4.1", "anthropic/claude-opus-4", "anthropic/claude-sonnet-4", "anthropic/claude-sonnet-3.7", "anthropic/claude-sonnet-3.5", "anthropic/claude-haiku-3.5", "anthropic/claude-haiku-3", "openai/gpt-oss-20b", "openai/gpt-oss-120b", "openai/gpt-5", "openai/gpt-5-pro", "openai/gpt-5.2", "openai/gpt-5.2-chat", "openai/gpt-5.2-pro", "openai/gpt-5.2-codex", "openai/gpt-5.3-codex", "openai/gpt-5.3-codex-spark", "openai/gpt-5.3-chat", "openai/gpt-5.4", "openai/gpt-5.4-mini", "openai/gpt-5.4-nano", "openai/gpt-5.4-pro", "openai/gpt-5.5", "openai/gpt-5.5-pro", "openai/gpt-5-mini", "openai/gpt-5-nano", "openai/gpt-5-codex", "openai/gpt-5.1-codex", "openai/gpt-5.1-codex-max", "openai/gpt-5.1-codex-mini", "openai/gpt-5.1-chat", "openai/gpt-5.1", "openai/text-embedding-3-small", "openai/text-embedding-3-large", "amazon/nova-micro", "amazon/nova-lite", "amazon/nova-pro", "amazon/nova-premier", "amazon/nova-2-lite", "amazon/nova-2-multimodal-embeddings", "google/gemini-2.5-flash-lite", "google/gemini-2.5-flash", "google/gemini-2.5-pro", "google/gemini-3-flash-preview", "google/gemini-3.1-flash-lite-preview", "google/gemini-3.1-pro-preview", "google/gemini-embedding-2", "google/embedding-001", "google/gemma-3-1b", "google/gemma-3-4b", "google/gemma-3-12b", "google/gemma-3-27b", "google/gemma-4-e2b", "google/gemma-4-e4b", "google/gemma-4-26b-a4b", "google/gemma-4-31b", "meta/llama-3.1-8b", "meta/llama-3.1-70b", "meta/llama-3.1-405b", "meta/llama-3.2-1b", "meta/llama-3.2-3b", "meta/llama-3.2-11b", "meta/llama-3.2-90b", "meta/llama-3.3-70b", "meta/llama-4-scout", "meta/llama-4-maverick", "cohere/embed-v4.0", "cohere/embed-english-v3.0", "cohere/embed-english-light-v3.0", "cohere/embed-multilingual-v3.0", "cohere/embed-multilingual-light-v3.0", "cohere/command-a", "cohere/command-r7b", "cohere/command-a-translate", "cohere/command-a-reasoning", "cohere/command-a-vision", "cohere/command-r", "cohere/command-r-plus", "minimax/m2.5", "minimax/m2.7", "moonshot/kimi-k2.5", "moonshot/kimi-k2.6", "xai/grok-4.1-fast", "xai/grok-4.1-fast-reasoning", "xai/grok-4.2", "xai/grok-4.2-reasoning", "xai/grok-4.2-multi-agent", "deepseek/deepseek-v3.2", "voyage/voyage-2-code", "voyage/voyage-2-law", "voyage/voyage-2-finance", "voyage/voyage-3-code", "voyage/voyage-3-large", "voyage/voyage-3.5-lite", "voyage/voyage-3.5", "voyage/voyage-4-lite", "voyage/voyage-4", "voyage/voyage-4-large", "alibaba/qwen3-235b", "alibaba/qwen3-32b", "alibaba/qwen3.5-plus", "alibaba/qwen3.5-flash", "alibaba/qwen3.5-397b", "alibaba/qwen3.5-122b", "alibaba/qwen3.5-35b", "alibaba/qwen3.5-27b", "alibaba/qwen3.5-9b", "alibaba/qwen3.5-4b", "alibaba/qwen3.5-2b", "alibaba/qwen3.5-0.8b", "alibaba/qwen3.6-plus", "alibaba/qwen3.6-flash", "alibaba/qwen3.6-27b", "alibaba/qwen3.6-max-preview", "alibaba/qwen3-coder-next", "alibaba/qwen3-vl-235b", "alibaba/qwen3-embedding-0.6b", "alibaba/qwen3-embedding-4b", "alibaba/qwen3-embedding-8b", "zhipu/glm-5", "zhipu/glm-5-turbo", "zhipu/glm-5.1"];
 export type CanonicalModelId = (typeof CANONICAL_MODEL_IDS)[number];
 export type ModelId = CanonicalModelId | (string & {});
 export type CatalogModel = {

package/dist/models/types.js

@@ -29,6 +29,8 @@ export const CANONICAL_MODEL_IDS = [
     "openai/gpt-5.4-mini",
     "openai/gpt-5.4-nano",
     "openai/gpt-5.4-pro",
+    "openai/gpt-5.5",
+    "openai/gpt-5.5-pro",
     "openai/gpt-5-mini",
     "openai/gpt-5-nano",
     "openai/gpt-5-codex",

package/dist/telemetry/gen-ai.js

@@ -1,5 +1,6 @@
 import { metrics } from "@opentelemetry/api";
 import { STATUS_TEXT } from "../errors/utils";
+import { logger } from "../logger";
 const getMeter = () => metrics.getMeter("@hebo/gateway");
 let requestDurationHistogram;
 let timePerOutputTokenHistogram;
@@ -93,20 +94,59 @@ export const recordTimePerOutputToken = (start, ttft, tokenAttrs, metricAttrs, s
         return;
     getTimePerOutputTokenHistogram().record((performance.now() - start - ttft) / 1000 / (outputTokens - 1), metricAttrs);
 };
+// Partitioning follows OTel semconv PR #3624:
+// https://github.com/open-telemetry/semantic-conventions/pull/3624
+// When a cache or reasoning breakdown is reported, partitioned data points sum
+// to the total and a bare {type} point MUST NOT be emitted alongside them.
 // FUTURE: record unsuccessful calls
 export const recordTokenUsage = (tokenAttrs, metricAttrs, signalLevel) => {
-    if (!signalLevel || (signalLevel !== "recommended" && signalLevel !== "full"))
+    if (signalLevel !== "recommended" && signalLevel !== "full")
         return;
-    const record = (value, tokenType) => {
-        if (typeof value !== "number")
-            return;
-        getTokenUsageHistogram().record(value, Object.assign({}, metricAttrs, { "gen_ai.token.type": tokenType }));
+    const histogram = getTokenUsageHistogram();
+    const emit = (value, extra) => {
+        if (value > 0)
+            histogram.record(value, { ...metricAttrs, ...extra });
     };
-    record(tokenAttrs["gen_ai.usage.input_tokens"], "input");
-    record(tokenAttrs["gen_ai.usage.output_tokens"], "output");
-    // FUTURE: "cached" and "reasoning" token types are not yet in the OTel standard — monitor:
-    // https://github.com/open-telemetry/semantic-conventions/issues/1959
-    // https://github.com/open-telemetry/semantic-conventions/issues/3341
-    record(tokenAttrs["gen_ai.usage.cache_read.input_tokens"], "cached");
-    record(tokenAttrs["gen_ai.usage.reasoning.output_tokens"], "reasoning");
+    emitInputTokens(emit, tokenAttrs);
+    emitOutputTokens(emit, tokenAttrs);
+};
+const emitInputTokens = (emit, tokenAttrs) => {
+    const total = tokenAttrs["gen_ai.usage.input_tokens"];
+    if (total === undefined)
+        return;
+    const cacheRead = tokenAttrs["gen_ai.usage.cache_read.input_tokens"];
+    const cacheCreation = tokenAttrs["gen_ai.usage.cache_creation.input_tokens"];
+    if (cacheRead === undefined && cacheCreation === undefined) {
+        emit(total, { "gen_ai.token.type": "input" });
+        return;
+    }
+    const read = cacheRead ?? 0;
+    const creation = cacheCreation ?? 0;
+    let uncached = total - read - creation;
+    if (uncached < 0) {
+        logger.warn({ inputTokens: total, cacheRead: read, cacheCreation: creation }, "[telemetry] input token cache partitions exceed total; clamping uncached to 0");
+        uncached = 0;
+    }
+    emit(read, { "gen_ai.token.type": "input", "gen_ai.token.cache": "read" });
+    emit(creation, { "gen_ai.token.type": "input", "gen_ai.token.cache": "creation" });
+    emit(uncached, { "gen_ai.token.type": "input", "gen_ai.token.cache": "uncached" });
+};
+const emitOutputTokens = (emit, tokenAttrs) => {
+    const total = tokenAttrs["gen_ai.usage.output_tokens"];
+    if (total === undefined)
+        return;
+    const reasoning = tokenAttrs["gen_ai.usage.reasoning.output_tokens"];
+    if (reasoning === undefined) {
+        emit(total, { "gen_ai.token.type": "output" });
+        return;
+    }
+    let reasoned = reasoning;
+    let nonReasoning = total - reasoning;
+    if (nonReasoning < 0) {
+        logger.warn({ outputTokens: total, reasoningTokens: reasoning }, "[telemetry] reasoning tokens exceed output total; clamping non-reasoning to 0");
+        reasoned = total;
+        nonReasoning = 0;
+    }
+    emit(reasoned, { "gen_ai.token.type": "output", "gen_ai.token.reasoning": true });
+    emit(nonReasoning, { "gen_ai.token.type": "output", "gen_ai.token.reasoning": false });
 };

package/dist/utils/preset.js

@@ -31,6 +31,7 @@ export function deepMerge(base, override) {
     }
     return out;
 }
+// oxlint-disable-next-line no-unnecessary-type-parameters
 export function presetFor() {
     return function preset(id, base) {
         return (override) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hebo-ai/gateway",
-  "version": "0.11.1",
+  "version": "0.11.3",
   "description": "AI gateway as a framework. For full control over models, routing & lifecycle. OpenAI /chat/completions, OpenResponses /responses & Anthropic /messages.",
   "keywords": [
     "ai",
@@ -252,57 +252,57 @@
     "fix": "bun lint:staged && bun format:staged"
   },
   "dependencies": {
-    "@ai-sdk/provider": "^3.0.8",
-    "ai": "^6.0.168",
+    "@ai-sdk/provider": "^3.0.9",
+    "ai": "^6.0.169",
     "lru-cache": "^11.3.5",
-    "uuid": "^13.0.0",
+    "uuid": "^14.0.0",
     "zod": "^4.3.6"
   },
   "devDependencies": {
-    "@ai-sdk/alibaba": "^1.0.17",
-    "@ai-sdk/amazon-bedrock": "^4.0.96",
-    "@ai-sdk/anthropic": "^3.0.71",
-    "@ai-sdk/cohere": "^3.0.30",
-    "@ai-sdk/deepinfra": "^2.0.45",
-    "@ai-sdk/deepseek": "^2.0.29",
-    "@ai-sdk/fireworks": "^2.0.46",
-    "@ai-sdk/google-vertex": "^4.0.112",
-    "@ai-sdk/groq": "^3.0.35",
-    "@ai-sdk/moonshotai": "^2.0.16",
-    "@ai-sdk/openai": "^3.0.53",
-    "@ai-sdk/togetherai": "^2.0.45",
-    "@ai-sdk/xai": "^3.0.83",
-    "@anthropic-ai/sdk": "^0.88.0",
-    "@aws-sdk/credential-providers": "^3.1035.0",
+    "@ai-sdk/alibaba": "^1.0.18",
+    "@ai-sdk/amazon-bedrock": "^4.0.97",
+    "@ai-sdk/anthropic": "^3.0.72",
+    "@ai-sdk/cohere": "^3.0.31",
+    "@ai-sdk/deepinfra": "^2.0.46",
+    "@ai-sdk/deepseek": "^2.0.30",
+    "@ai-sdk/fireworks": "^2.0.47",
+    "@ai-sdk/google-vertex": "^4.0.113",
+    "@ai-sdk/groq": "^3.0.36",
+    "@ai-sdk/moonshotai": "^2.0.17",
+    "@ai-sdk/openai": "^3.0.54",
+    "@ai-sdk/togetherai": "^2.0.46",
+    "@ai-sdk/xai": "^3.0.84",
+    "@anthropic-ai/sdk": "^0.91.1",
+    "@aws-sdk/credential-providers": "^3.1038.0",
     "@langfuse/otel": "^5.2.0",
-    "@libsql/client": "^0.17.2",
+    "@libsql/client": "^0.17.3",
     "@mjackson/node-fetch-server": "^0.7.0",
     "@opentelemetry/api": "^1.9.1",
     "@opentelemetry/context-async-hooks": "^2.7.0",
     "@opentelemetry/sdk-trace-base": "^2.7.0",
-    "@tanstack/react-router": "^1.168.23",
-    "@tanstack/react-start": "^1.167.42",
+    "@tanstack/react-router": "^1.168.25",
+    "@tanstack/react-start": "^1.167.50",
     "@types/better-sqlite3": "^7.6.13",
-    "@types/bun": "1.3.12",
+    "@types/bun": "1.3.13",
     "@types/pg": "^8.20.0",
     "@types/react": "^19.2.14",
     "@types/react-dom": "^19.2.3",
     "@types/uuid": "^11.0.0",
     "better-sqlite3": "^12.9.0",
     "elysia": "^1.4.28",
-    "hono": "^4.12.14",
+    "hono": "^4.12.15",
     "lefthook": "^2.1.6",
-    "mysql2": "^3.22.2",
+    "mysql2": "^3.22.3",
     "next": "^16.2.4",
-    "openai": "^6.34.0",
-    "oxfmt": "^0.44.0",
-    "oxlint": "^1.61.0",
-    "oxlint-tsgolint": "^0.20.0",
+    "openai": "^6.35.0",
+    "oxfmt": "^0.46.0",
+    "oxlint": "^1.62.0",
+    "oxlint-tsgolint": "^0.22.1",
     "pg": "^8.20.0",
     "pino": "^10.3.1",
     "postgres": "^3.4.9",
     "typescript": "^6.0.3",
-    "vite": "^7.3.2",
+    "vite": "^8.0.10",
     "vite-tsconfig-paths": "^6.1.1",
     "voyage-ai-provider": "^3.0.0",
     "zhipu-ai-provider": "^0.3.0"
@@ -315,12 +315,12 @@
     "@ai-sdk/deepinfra": "^2.0.45",
     "@ai-sdk/deepseek": "^2.0.29",
     "@ai-sdk/fireworks": "^2.0.46",
-    "@ai-sdk/google": "^3.0.64",
+    "@ai-sdk/google": "^3.0.65",
     "@ai-sdk/google-vertex": "^4.0.80",
     "@ai-sdk/groq": "^3.0.29",
     "@ai-sdk/moonshotai": "^2.0.16",
     "@ai-sdk/openai": "^3.0.41",
-    "@ai-sdk/openai-compatible": "^2.0.41",
+    "@ai-sdk/openai-compatible": "^2.0.42",
     "@ai-sdk/togetherai": "^2.0.45",
     "@ai-sdk/xai": "^3.0.83",
     "@libsql/client": "^0.14.0",