@hebo-ai/gateway 0.4.0-beta.0 → 0.4.0-beta.1

package/README.md

@@ -620,7 +620,7 @@ const gw = gateway({
     // "required" = minimal baseline attributes
     // "recommended" = practical operational attributes (request/response metadata, genai model/usage fields)
     // "full" = also include body fields (e.g. genai input/output messages)
-    attributes: "recommended",
+    attributes: "full",
   },
 });
 ```
@@ -628,10 +628,10 @@ const gw = gateway({
 Attribute names and span semantics follow OpenTelemetry GenAI semantic conventions:
 https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-spans/
 
-To populate custom span attributes, the inbound W3C `baggage` header is supported. Keys in the `hebo.` namespace are mapped to span attributes, with the namespace stripped. For example: `baggage: hebo.user_id=u-123` becomes span attribute `user_id=u-123`.
-
 > [!TIP]
-> For observability integration that is not otel compliant (for example, Langfuse), you can disable built-in telemetry and manually instrument requests during `before` / `after` hooks.
+> To populate custom span attributes, the inbound W3C `baggage` header is supported. Keys in the `hebo.` namespace are mapped to span attributes, with the namespace stripped. For example: `baggage: hebo.user_id=u-123` becomes span attribute `user_id=u-123`.
+
+For observability integration that is not otel compliant, you can disable built-in telemetry and manually instrument requests during `before` / `after` hooks.
 
 ### Passing Framework State to Hooks
 

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

@@ -93,8 +93,8 @@ export const chatCompletions = (config) => {
                 onAbort: () => {
                     throw new DOMException("Upstream failed", "AbortError");
                 },
-                onFinish: (event) => {
-                    ctx.streamResult = toChatCompletions(event, ctx.resolvedModelId);
+                onFinish: (result) => {
+                    ctx.streamResult = toChatCompletions(result, ctx.resolvedModelId);
                 },
                 timeout: {
                     totalMs: 5 * 60 * 1000,

package/dist/middleware/matcher.js

@@ -1,4 +1,5 @@
 import { logger } from "../logger";
+import { addSpanEvent } from "../telemetry/span";
 import { forwardParamsEmbeddingMiddleware, forwardParamsMiddleware } from "./common";
 class SimpleMatcher {
     rules = [];
@@ -82,6 +83,7 @@ class ModelMiddlewareMatcher {
                     break;
             }
             logger.warn(`[middleware] cache eviction`);
+            addSpanEvent("hebo.middelware.cache.evicted");
         }
         this.cache.set(key, out);
         return out;

package/dist/telemetry/span.js

@@ -59,5 +59,6 @@ export const withSpan = async (name, run, options) => {
     }
 };
 export const addSpanEvent = (name, attributes) => {
-    trace.getActiveSpan()?.addEvent(name, attributes);
+    const allAttributes = Object.assign(attributes ?? {}, getMemoryAttributes());
+    trace.getActiveSpan()?.addEvent(name, allAttributes);
 };

package/dist/telemetry/utils.js

@@ -1,4 +1,4 @@
-import { resolveRequestId } from "#/utils/headers";
+import { resolveRequestId } from "../utils/headers";
 const DEFAULT_ATTRIBUTES_LEVEL = "recommended";
 const HEBO_BAGGAGE_PREFIX = "hebo.";
 const toTextPart = (content) => ({ type: "text", content });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hebo-ai/gateway",
-  "version": "0.4.0-beta.0",
+  "version": "0.4.0-beta.1",
   "description": "AI gateway as a framework. For full control over models, routing & lifecycle. OpenAI-compatible /chat/completions, /embeddings & /models.",
   "keywords": [
     "ai",

package/src/endpoints/chat-completions/handler.ts

@@ -127,9 +127,9 @@ export const chatCompletions = (config: GatewayConfig): Endpoint => {
         onAbort: () => {
           throw new DOMException("Upstream failed", "AbortError");
         },
-        onFinish: (event) => {
+        onFinish: (result) => {
           ctx.streamResult = toChatCompletions(
-            event as unknown as GenerateTextResult<ToolSet, Output.Output>,
+            result as unknown as GenerateTextResult<ToolSet, Output.Output>,
             ctx.resolvedModelId!,
           );
         },

package/src/middleware/matcher.ts

@@ -4,6 +4,7 @@ import type { ModelId } from "../models/types";
 import type { ProviderId } from "../providers/types";
 
 import { logger } from "../logger";
+import { addSpanEvent } from "../telemetry/span";
 import { forwardParamsEmbeddingMiddleware, forwardParamsMiddleware } from "./common";
 
 type MiddlewareEntries = {
@@ -117,6 +118,7 @@ class ModelMiddlewareMatcher {
         if (--n === 0) break;
       }
       logger.warn(`[middleware] cache eviction`);
+      addSpanEvent("hebo.middelware.cache.evicted");
     }
 
     this.cache.set(key, out);

package/src/telemetry/span.ts

@@ -83,5 +83,6 @@ export const withSpan = async <T>(
 };
 
 export const addSpanEvent = (name: string, attributes?: Attributes) => {
-  trace.getActiveSpan()?.addEvent(name, attributes);
+  const allAttributes = Object.assign(attributes ?? {}, getMemoryAttributes());
+  trace.getActiveSpan()?.addEvent(name, allAttributes);
 };

package/src/telemetry/utils.ts

@@ -1,13 +1,12 @@
-import type { Embeddings, EmbeddingsBody } from "#/endpoints/embeddings";
-
-import { resolveRequestId } from "#/utils/headers";
-
 import type {
   ChatCompletions,
   ChatCompletionsBody,
   ChatCompletionsContentPart,
   ChatCompletionsMessage,
 } from "../endpoints/chat-completions/schema";
+import type { Embeddings, EmbeddingsBody } from "../endpoints/embeddings";
+
+import { resolveRequestId } from "../utils/headers";
 
 type GenAIPart = Record<string, unknown>;
 const DEFAULT_ATTRIBUTES_LEVEL = "recommended";