refactor(agents): keep overload matching provider-neutral

steipete · steipete · commit db79e94c89dc · 2026-06-20T23:41:43.000-04:00
diff --git a/src/agents/embedded-agent-helpers/failover-matches.test.ts b/src/agents/embedded-agent-helpers/failover-matches.test.ts
@@ -103,12 +103,10 @@ describe("Z.ai vendor error codes (#48988)", () => {
   });
 });
 
-describe("Zhipu (GLM) overload error (#93211)", () => {
-  // Zhipu returns HTTP 200 with the error in the body, so failover keys on the
-  // message phrase rather than a status code.
+describe("Chinese provider overload messages", () => {
   const ZHIPU_OVERLOAD = "[1305][该模型当前访问量过大，请您稍后再试]";
 
-  it("classifies the GLM overload body as overloaded so failover triggers", () => {
+  it("classifies the Zhipu GLM overload body as overloaded", () => {
     expect(isOverloadedErrorMessage(ZHIPU_OVERLOAD)).toBe(true);
   });
 
diff --git a/src/agents/embedded-agent-helpers/failover-matches.ts b/src/agents/embedded-agent-helpers/failover-matches.ts
@@ -107,9 +107,6 @@ const ERROR_PATTERNS = {
     // Chinese provider overloaded messages
     "服务过载",
     "当前负载过高",
-    // Zhipu (GLM) overload contract: code 1305 ships a 200 body
-    // "该模型当前访问量过大，请您稍后再试" (model traffic too high); the
-    // distinctive phrase is the overload signal failover keys on (#93211).
     "访问量过大",
   ],
   serverError: [
diff --git a/src/agents/model-fallback.test.ts b/src/agents/model-fallback.test.ts
@@ -746,11 +746,8 @@ describe("runWithModelFallback", () => {
     expect(result.attempts[0].reason).toBe("unknown");
   });
 
-  it("falls back on a Zhipu (GLM) 1305 overload body and classifies it as overloaded (#93211)", async () => {
+  it("falls back on a Zhipu GLM 1305 overload body and classifies it as overloaded", async () => {
     const cfg = makeCfg();
-    // Zhipu (GLM) signals overload as a 200 body "[1305][该模型当前访问量过大，请您稍后再试]";
-    // the runtime surfaces it as a thrown error. Without the overload pattern it
-    // classifies as "unknown"; with it, failover keys the recovery on "overloaded".
     const glmOverload = new Error("[1305][该模型当前访问量过大，请您稍后再试]");
     const run = vi.fn().mockRejectedValueOnce(glmOverload).mockResolvedValueOnce("ok");
 
