fix(agents): check billing errors before context overflow heuristics (#40409)

Merged via squash.

Prepared head SHA: c88f89c462
Co-authored-by: ademczuk <5212682+ademczuk@users.noreply.github.com>
Co-authored-by: altaywtf <9790196+altaywtf@users.noreply.github.com>
Reviewed-by: @altaywtf

src/agents/pi-embedded-helpers.isbillingerrormessage.test.ts

@@ -439,6 +439,18 @@ describe("isLikelyContextOverflowError", () => {
       expect(isLikelyContextOverflowError(sample)).toBe(false);
     }
   });
+
+  it("excludes billing errors even when text matches context overflow patterns", () => {
+    const samples = [
+      "402 Payment Required: request token limit exceeded for this billing plan",
+      "insufficient credits: request size exceeds your current plan limits",
+      "Your credit balance is too low. Maximum request token limit exceeded.",
+    ];
+    for (const sample of samples) {
+      expect(isBillingErrorMessage(sample)).toBe(true);
+      expect(isLikelyContextOverflowError(sample)).toBe(false);
+    }
+  });
 });
 
 describe("isTransientHttpError", () => {

src/agents/pi-embedded-helpers/errors.ts

@@ -138,6 +138,13 @@ export function isLikelyContextOverflowError(errorMessage?: string): boolean {
     return false;
   }
 
+  // Billing/quota errors can contain patterns like "request size exceeds" or
+  // "maximum token limit exceeded" that match the context overflow heuristic.
+  // Billing is a more specific error class — exclude it early.
+  if (isBillingErrorMessage(errorMessage)) {
+    return false;
+  }
+
   if (CONTEXT_WINDOW_TOO_SMALL_RE.test(errorMessage)) {
     return false;
   }