fix: preserve generic DashScope streaming usage (#52395) (thanks @IVY-AI-gif)

src/agents/openai-transport-stream.test.ts

@@ -863,6 +863,33 @@ describe("openai transport stream", () => {
     expect(params.stream_options).toMatchObject({ include_usage: true });
   });
 
+  it("enables streaming usage compat for generic providers on native DashScope endpoints", () => {
+    const params = buildOpenAICompletionsParams(
+      {
+        id: "glm-5",
+        name: "GLM-5",
+        api: "openai-completions",
+        provider: "generic",
+        baseUrl: "https://coding.dashscope.aliyuncs.com/v1",
+        reasoning: true,
+        input: ["text"],
+        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+        contextWindow: 200000,
+        maxTokens: 8192,
+      } satisfies Model<"openai-completions">,
+      {
+        systemPrompt: "system",
+        messages: [],
+        tools: [],
+      } as never,
+      undefined,
+    ) as {
+      stream_options?: { include_usage?: boolean };
+    };
+
+    expect(params.stream_options).toMatchObject({ include_usage: true });
+  });
+
   it("disables developer-role-only compat defaults for configured custom proxy completions providers", () => {
     const params = buildOpenAICompletionsParams(
       {

src/agents/provider-attribution.test.ts

@@ -572,6 +572,19 @@ describe("provider attribution", () => {
       supportsNativeStreamingUsageCompat: true,
     });
 
+    expect(
+      resolveProviderRequestCapabilities({
+        provider: "generic",
+        api: "openai-completions",
+        baseUrl: "https://coding.dashscope.aliyuncs.com/v1",
+        capability: "llm",
+        transport: "stream",
+      }),
+    ).toMatchObject({
+      endpointClass: "modelstudio-native",
+      supportsNativeStreamingUsageCompat: true,
+    });
+
     expect(
       resolveProviderRequestCapabilities({
         provider: "ollama",
@@ -753,6 +766,28 @@ describe("provider attribution", () => {
           supportsNativeStreamingUsageCompat: true,
         },
       },
+      {
+        name: "generic provider on native DashScope completions",
+        input: {
+          provider: "generic",
+          api: "openai-completions",
+          baseUrl: "https://coding-intl.dashscope.aliyuncs.com/v1",
+          capability: "llm" as const,
+          transport: "stream" as const,
+        },
+        expected: {
+          knownProviderFamily: "generic",
+          endpointClass: "modelstudio-native",
+          isKnownNativeEndpoint: true,
+          allowsOpenAIServiceTier: false,
+          supportsOpenAIReasoningCompatPayload: false,
+          allowsResponsesStore: false,
+          supportsResponsesStoreField: false,
+          shouldStripResponsesPromptCache: false,
+          allowsAnthropicServiceTier: false,
+          supportsNativeStreamingUsageCompat: true,
+        },
+      },
       {
         name: "native Google Gemini api",
         input: {

src/agents/provider-attribution.ts

@@ -604,9 +604,10 @@ export function resolveProviderRequestCapabilities(
       policy.usesKnownNativeOpenAIEndpoint,
     shouldStripResponsesPromptCache:
       api !== undefined && OPENAI_RESPONSES_APIS.has(api) && policy.usesExplicitProxyLikeEndpoint,
+    // Native endpoint class is the real signal here. Users can point a generic
+    // provider key at Moonshot or DashScope and still need streaming usage.
     supportsNativeStreamingUsageCompat:
-      (provider === "moonshot" && endpointClass === "moonshot-native") ||
-      (provider === "modelstudio" && endpointClass === "modelstudio-native"),
+      endpointClass === "moonshot-native" || endpointClass === "modelstudio-native",
     compatibilityFamily,
   };
 }