Gateway: tolerate transient network failures(#4425)

src/infra/unhandled-rejections.test.ts

@@ -100,6 +100,24 @@ describe("isTransientNetworkError", () => {
     expect(isTransientNetworkError(error)).toBe(true);
   });
 
+  it("returns true for MediaFetchError with fetch_failed code", () => {
+    const error = Object.assign(new Error("media fetch failed"), {
+      name: "MediaFetchError",
+      code: "fetch_failed",
+    });
+    expect(isTransientNetworkError(error)).toBe(true);
+  });
+
+  it("returns true for MediaFetchError that wraps a network cause", () => {
+    const cause = Object.assign(new Error("timeout"), { code: "ETIMEDOUT" });
+    const error = Object.assign(new Error("media fetch failed"), {
+      name: "MediaFetchError",
+      code: "fetch_failed",
+      cause,
+    });
+    expect(isTransientNetworkError(error)).toBe(true);
+  });
+
   it("returns false for regular errors without network codes", () => {
     expect(isTransientNetworkError(new Error("Something went wrong"))).toBe(false);
     expect(isTransientNetworkError(new TypeError("Cannot read property"))).toBe(false);

src/infra/unhandled-rejections.ts

@@ -78,6 +78,16 @@ function isConfigError(err: unknown): boolean {
 export function isTransientNetworkError(err: unknown): boolean {
   if (!err) return false;
 
+  if (err && typeof err === "object") {
+    const name = "name" in err ? String(err.name) : "";
+    const code = extractErrorCodeWithCause(err);
+    if (name === "MediaFetchError" && code === "fetch_failed") {
+      const cause = getErrorCause(err);
+      if (cause && cause !== err) return isTransientNetworkError(cause);
+      return true;
+    }
+  }
+
   const code = extractErrorCodeWithCause(err);
   if (code && TRANSIENT_NETWORK_CODES.has(code)) return true;
 

src/media/fetch.ts

@@ -13,10 +13,17 @@ export type MediaFetchErrorCode = "max_bytes" | "http_error" | "fetch_failed";
 export class MediaFetchError extends Error {
   readonly code: MediaFetchErrorCode;
 
-  constructor(code: MediaFetchErrorCode, message: string) {
+  constructor(code: MediaFetchErrorCode, message: string, options?: { cause?: unknown }) {
     super(message);
     this.code = code;
     this.name = "MediaFetchError";
+    if (options?.cause !== undefined) {
+      try {
+        (this as { cause?: unknown }).cause = options.cause;
+      } catch {
+        // ignore if cause cannot be assigned
+      }
+    }
   }
 }
 
@@ -74,7 +81,9 @@ export async function fetchRemoteMedia(options: FetchMediaOptions): Promise<Fetc
   try {
     res = await fetcher(url);
   } catch (err) {
-    throw new MediaFetchError("fetch_failed", `Failed to fetch media from ${url}: ${String(err)}`);
+    throw new MediaFetchError("fetch_failed", `Failed to fetch media from ${url}: ${String(err)}`, {
+      cause: err,
+    });
   }
 
   if (!res.ok) {

src/telegram/monitor.ts

@@ -5,6 +5,7 @@ import { resolveAgentMaxConcurrent } from "../config/agent-limits.js";
 import { computeBackoff, sleepWithAbort } from "../infra/backoff.js";
 import { formatErrorMessage } from "../infra/errors.js";
 import { formatDurationMs } from "../infra/format-duration.js";
+import { registerUnhandledRejectionHandler } from "../infra/unhandled-rejections.js";
 import type { RuntimeEnv } from "../runtime.js";
 import { resolveTelegramAccount } from "./accounts.js";
 import { resolveTelegramAllowedUpdates } from "./allowed-updates.js";
@@ -108,6 +109,17 @@ export async function monitorTelegramProvider(opts: MonitorTelegramOpts = {}) {
     opts.proxyFetch ??
     (account.config.proxy ? makeProxyFetch(account.config.proxy as string) : undefined);
 
+  let unregisterUnhandled: (() => void) | null = null;
+  try {
+    unregisterUnhandled = registerUnhandledRejectionHandler((reason) => {
+      if (!isRecoverableTelegramNetworkError(reason)) return false;
+      const errMsg = formatErrorMessage(reason);
+      (opts.runtime?.error ?? console.warn)(
+        `telegram: suppressed unhandled network rejection: ${errMsg}`,
+      );
+      return true;
+    });
+
     let lastUpdateId = await readTelegramUpdateOffset({
       accountId: account.accountId,
     });
@@ -196,4 +208,7 @@ export async function monitorTelegramProvider(opts: MonitorTelegramOpts = {}) {
         opts.abortSignal?.removeEventListener("abort", stopOnAbort);
       }
     }
+  } finally {
+    unregisterUnhandled?.();
+  }
 }