test(gui): cover the runtime-reported shadow-call model set

The GUI now renders whatever the runtime reports instead of a slug baked into
six locale files, but nothing tested that. A fallback or formatting regression
would have shipped green through the whole GUI suite.

Covers the three cases that actually differ: the runtime reported a list, an
operator configured an override, and a runtime too old to report the field at
all. The empty-array case earns its own test -- that is what a runtime sends
when every configured entry was rejected, and rendering nothing there would
read as 'nothing is intercepted', which is the opposite of the truth.

Author: lidge-jun

gui/tests/shadow-call-source.test.ts

@@ -0,0 +1,65 @@
+import { describe, expect, test } from "bun:test";
+import {
+  shadowSourceModelBadge,
+  shadowSourceModelLabel,
+  shadowSourceModelList,
+} from "../src/pages/shadow-call-source";
+
+/**
+ * The GUI used to spell the intercepted slug into six locale files, which went
+ * stale every time Codex changed its helper model. It now renders whatever the
+ * runtime reports, so the cases that matter are: the runtime reported a list,
+ * the runtime is too old to report one, and the operator configured an override.
+ * Without these, a fallback or formatting regression ships silently green.
+ */
+describe("shadowSourceModelList", () => {
+  test("renders the models the runtime reported", () => {
+    expect(shadowSourceModelList(["gpt-5.4-mini", "gpt-5.6-luna"]))
+      .toEqual(["gpt-5.4-mini", "gpt-5.6-luna"]);
+  });
+
+  test("renders a configured override rather than the defaults", () => {
+    expect(shadowSourceModelList(["gpt-6-helper"])).toEqual(["gpt-6-helper"]);
+  });
+
+  test("falls back when a runtime too old to report sourceModels omits it", () => {
+    expect(shadowSourceModelList(undefined)).toEqual(["gpt-5.4-mini", "gpt-5.6-luna"]);
+  });
+
+  // An empty array is what a runtime sends when every configured entry was
+  // rejected; showing nothing there would read as "nothing is intercepted",
+  // which is the opposite of the truth.
+  test("falls back on an empty list instead of rendering nothing", () => {
+    expect(shadowSourceModelList([])).toEqual(["gpt-5.4-mini", "gpt-5.6-luna"]);
+  });
+
+  test("drops blank entries and trims the rest", () => {
+    expect(shadowSourceModelList([" gpt-5.6-luna ", "", "   "])).toEqual(["gpt-5.6-luna"]);
+  });
+
+  test("falls back when the field is not an array at all", () => {
+    expect(shadowSourceModelList("gpt-5.6-luna" as unknown as string[]))
+      .toEqual(["gpt-5.4-mini", "gpt-5.6-luna"]);
+  });
+});
+
+describe("shadow source model rendering", () => {
+  test("label joins every reported model", () => {
+    expect(shadowSourceModelLabel(["gpt-5.4-mini", "gpt-5.6-luna"]))
+      .toBe("gpt-5.4-mini, gpt-5.6-luna");
+  });
+
+  test("badge drops the shared gpt- prefix to keep the row compact", () => {
+    expect(shadowSourceModelBadge(["gpt-5.4-mini", "gpt-5.6-luna"]))
+      .toBe("5.4-mini, 5.6-luna");
+  });
+
+  test("badge leaves a non-gpt id untouched", () => {
+    expect(shadowSourceModelBadge(["helper-x"])).toBe("helper-x");
+  });
+
+  test("both renderings use the fallback when the runtime reported nothing", () => {
+    expect(shadowSourceModelLabel(undefined)).toBe("gpt-5.4-mini, gpt-5.6-luna");
+    expect(shadowSourceModelBadge(undefined)).toBe("5.4-mini, 5.6-luna");
+  });
+});