fix(ai): retry DNS transport failures (#6946)

Fixes #6904

packages/ai/src/utils/retry.ts

@@ -51,6 +51,9 @@ const RETRYABLE_PROVIDER_ERROR_PATTERN = buildProviderErrorPattern([
 	"connection.?lost",
 	"other side closed",
 	"fetch failed",
+	"getaddrinfo",
+	"ENOTFOUND",
+	"EAI_AGAIN",
 	"upstream.?connect",
 	"reset before headers",
 	"socket hang up",

packages/ai/test/retry.test.ts

@@ -10,6 +10,8 @@ const nvidiaNIMResourceExhaustedMessage = "ResourceExhausted: Worker local total
 const bunFetchSocketClosedMessage =
 	"The socket connection was closed unexpectedly. For more information, pass `verbose: true` in the second argument to fetch()";
 const openAIResponsesEarlyEofMessage = "OpenAI Responses stream ended before a terminal response event";
+const wrappedDnsLookupError =
+	"The pending stream has been canceled (caused by: getaddrinfo ENOTFOUND bedrock-runtime.us-east-1.amazonaws.com)";
 
 describe("provider retry classification", () => {
 	it("matches explicit provider retry guidance", () => {
@@ -38,6 +40,15 @@ describe("provider retry classification", () => {
 		).toBe(true);
 	});
 
+	it.each([
+		wrappedDnsLookupError,
+		"connect ENOTFOUND api.example.com",
+		"EAI_AGAIN api.example.com",
+		"getaddrinfo failed for api.example.com",
+	])("matches DNS transport failure wording: %s", (errorMessage) => {
+		expect(isRetryableAssistantError(fauxAssistantMessage("", { stopReason: "error", errorMessage }))).toBe(true);
+	});
+
 	it("matches OpenAI Responses streams that end before terminal events", () => {
 		expect(
 			isRetryableAssistantError(

packages/coding-agent/test/suite/regressions/6904-dns-transport-retry.test.ts

@@ -0,0 +1,28 @@
+import { fauxAssistantMessage } from "@earendil-works/pi-ai";
+import { describe, expect, it } from "vitest";
+import { createHarness } from "../harness.ts";
+
+const wrappedDnsLookupError =
+	"The pending stream has been canceled (caused by: getaddrinfo ENOTFOUND bedrock-runtime.us-east-1.amazonaws.com)";
+
+describe("issue #6904 DNS transport failure retry", () => {
+	it("retries a transient DNS lookup failure", async () => {
+		const harness = await createHarness({ settings: { retry: { enabled: true, maxRetries: 3, baseDelayMs: 1 } } });
+		try {
+			harness.setResponses([
+				fauxAssistantMessage("", { stopReason: "error", errorMessage: wrappedDnsLookupError }),
+				fauxAssistantMessage("recovered after DNS retry"),
+			]);
+
+			await harness.session.prompt("test");
+
+			expect(harness.faux.state.callCount).toBe(2);
+			expect(harness.eventsOfType("auto_retry_start").map((event) => event.errorMessage)).toEqual([
+				wrappedDnsLookupError,
+			]);
+			expect(harness.eventsOfType("auto_retry_end").map((event) => event.success)).toEqual([true]);
+		} finally {
+			harness.cleanup();
+		}
+	});
+});