fix(lmstudio): cancel model discovery response body on non-ok (#109718)

* fix(lmstudio): cancel model discovery response body on non-ok

Wire the direct-fetch release hook to cancel unread bodies so failed
model discovery releases the TCP connection instead of leaving it open.

Co-authored-by: Cursor <cursoragent@cursor.com>

* refactor(lmstudio): unify response cleanup

Co-authored-by: Zeng Wen <27948732+ZengWen-DT@users.noreply.github.com>

---------

Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Co-authored-by: Zeng Wen <27948732+ZengWen-DT@users.noreply.github.com>

extensions/lmstudio/src/models.fetch.ts

@@ -48,6 +48,12 @@ type DiscoverLmstudioModelsParams = {
   fetchImpl?: typeof fetch;
 };
 
+async function cancelUnreadResponseBody(response: Response): Promise<void> {
+  if (!response.bodyUsed) {
+    await response.body?.cancel().catch(() => undefined);
+  }
+}
+
 async function fetchLmstudioEndpoint(params: {
   url: string;
   init?: RequestInit;
@@ -57,8 +63,10 @@ async function fetchLmstudioEndpoint(params: {
   auditContext: string;
 }): Promise<{ response: Response; release: () => Promise<void> }> {
   const timeoutMs = resolveTimerTimeoutMs(params.timeoutMs, 1);
+  let response: Response;
+  let release: () => Promise<void>;
   if (params.ssrfPolicy) {
-    return await fetchWithSsrFGuard({
+    const guarded = await fetchWithSsrFGuard({
       url: params.url,
       init: params.init,
       timeoutMs,
@@ -66,14 +74,22 @@ async function fetchLmstudioEndpoint(params: {
       policy: params.ssrfPolicy,
       auditContext: params.auditContext,
     });
-  }
-  const fetchFn = params.fetchImpl ?? fetch;
-  return {
-    response: await fetchFn(params.url, {
+    response = guarded.response;
+    release = guarded.release;
+  } else {
+    const fetchFn = params.fetchImpl ?? fetch;
+    response = await fetchFn(params.url, {
       ...params.init,
       signal: AbortSignal.timeout(timeoutMs),
-    }),
-    release: async () => {},
+    });
+    release = async () => undefined;
+  }
+  return {
+    response,
+    release: async () => {
+      await cancelUnreadResponseBody(response);
+      await release();
+    },
   };
 }
 

extensions/lmstudio/src/models.test.ts

@@ -392,6 +392,38 @@ describe("lmstudio-models", () => {
     });
   });
 
+  it("cancels the response body after a non-ok model discovery response", async () => {
+    const tracked = cancelTrackedResponse("unavailable", { status: 503 });
+    const fetchMock = vi.fn(async () => tracked.response);
+
+    const result = await fetchLmstudioModels({
+      baseUrl: "http://localhost:1234/v1",
+      fetchImpl: asFetch(fetchMock),
+    });
+
+    expect(result).toEqual({
+      reachable: true,
+      status: 503,
+      models: [],
+    });
+    expect(tracked.wasCanceled()).toBe(true);
+  });
+
+  it("cancels guarded non-ok discovery bodies before releasing the dispatcher", async () => {
+    const tracked = cancelTrackedResponse("unavailable", { status: 503 });
+    const release = vi.fn(async () => undefined);
+    fetchWithSsrFGuardMock.mockResolvedValue({ response: tracked.response, release });
+
+    const result = await fetchLmstudioModels({
+      baseUrl: "http://localhost:1234/v1",
+      ssrfPolicy: {},
+    });
+
+    expect(result).toMatchObject({ reachable: true, status: 503, models: [] });
+    expect(tracked.wasCanceled()).toBe(true);
+    expect(release).toHaveBeenCalledOnce();
+  });
+
   it("reports malformed model list JSON with an owned error", async () => {
     const fetchMock = vi.fn(async () => malformedJsonResponse());