test(ci): align release expectations with provider behavior

tests/unit/cc-compatible-provider.test.ts

@@ -782,6 +782,7 @@ test("handleChatCore preserves client cache markers for Claude Code requests to
   });
   assert.deepEqual(calls[0].body.messages[0].content[0].cache_control, {
     type: "ephemeral",
+    ttl: "1h",
   });
   assert.deepEqual(calls[0].body.messages[1].content[0].cache_control, {
     type: "ephemeral",

tests/unit/chatcore-translation-paths.test.ts

@@ -861,7 +861,10 @@ test("chatCore preserves Opus 5 mid-conversation system cache breakpoints", asyn
     call.body.messages.map((message: { role: string }) => message.role),
     ["user", "assistant", "system", "user"]
   );
-  assert.deepEqual(call.body.messages[2].content[0].cache_control, { type: "ephemeral" });
+  assert.deepEqual(call.body.messages[2].content[0].cache_control, {
+    type: "ephemeral",
+    ttl: "1h",
+  });
   assert.equal(
     call.body.system.some((block: { text?: string }) => block.text === "compact continuation"),
     false
@@ -1037,7 +1040,10 @@ test("chatCore preserves cache_control automatically for Claude Code single-mode
   assert.equal(hasCacheControl(call.body), true);
   // system[0] and system[1] are now the billing line and sentinel injected by base.ts for Claude Code
   assert.deepEqual(call.body.system[2].cache_control, { type: "ephemeral", ttl: "5m" });
-  assert.deepEqual(call.body.messages[0].content[0].cache_control, { type: "ephemeral" });
+  assert.deepEqual(call.body.messages[0].content[0].cache_control, {
+    type: "ephemeral",
+    ttl: "1h",
+  });
   // base.ts executor explicitly strips cache_control from tools for Claude Code clients
   assert.equal(call.body.tools[0].cache_control, undefined);
 });
@@ -1083,7 +1089,10 @@ test("chatCore supplements a missing message cache breakpoint for native Claude
     responseFormat: "claude",
   });
 
-  assert.deepEqual(call.body.messages[2].content[0].cache_control, { type: "ephemeral" });
+  assert.deepEqual(call.body.messages[2].content[0].cache_control, {
+    type: "ephemeral",
+    ttl: "1h",
+  });
   assert.equal(call.body.tools[0].cache_control, undefined);
 });
 test("chatCore auto cache policy becomes false for nondeterministic combos", async () => {
@@ -1171,7 +1180,10 @@ test("chatCore disables raw Claude passthrough when cache preservation is off an
     true
   );
   // Cache preservation is on for native Claude, so cache markers are intact
-  assert.deepEqual(call.body.messages[0].content[0].cache_control, { type: "ephemeral" });
+  assert.deepEqual(call.body.messages[0].content[0].cache_control, {
+    type: "ephemeral",
+    ttl: "1h",
+  });
   // Tools disable flag is applied
   assert.equal("_disableToolPrefix" in call.body, false);
 });
@@ -1314,7 +1326,7 @@ test("chatCore strips unsupported reasoning params and caps provider token field
   assert.equal(call.body.max_tokens, undefined);
   assert.equal(call.body.max_completion_tokens, 16384);
 });
-test("chatCore preserves reasoning_effort for assistant-prefill OpenAI-compatible requests", async () => {
+test("chatCore downgrades unsupported xhigh effort for assistant-prefill OpenAI-compatible requests", async () => {
   const { call, result } = await invokeChatCore({
     provider: "openai-compatible-aio",
     model: "glm-5.1",
@@ -1333,7 +1345,7 @@ test("chatCore preserves reasoning_effort for assistant-prefill OpenAI-compatibl
 
   assert.equal(result.success, true);
   assert.equal(call.body.model, "glm-5.1");
-  assert.equal(call.body.reasoning_effort, "xhigh");
+  assert.equal(call.body.reasoning_effort, "high");
 });
 test("chatCore logs chat completions endpoint as OpenAI protocol", async () => {
   const { call, result } = await invokeChatCore({

tests/unit/executor-xai.test.ts

@@ -23,11 +23,9 @@ test("XaiExecutor is registered under the 'xai' key and set as the registry exec
 test("XaiExecutor can target the separate xAI OAuth provider config", () => {
   const executor = new XaiExecutor("xai-oauth");
   assert.equal(executor.getProvider(), "xai-oauth");
-  // a367bf62f5 (#9994) scoped model-level targetFormat to the DECLARING provider's
-  // catalog: xai-oauth's own catalog lists grok-4.5 without a targetFormat tag, so
-  // it stays on /v1/chat/completions instead of importing a global entry's
-  // openai-responses tag (the transient #9612 global-fallback behavior).
-  assert.equal(executor.buildUrl("grok-4.5", false), "https://api.x.ai/v1/chat/completions");
+  // #10170 declares grok-4.5 as a native Responses model in xai-oauth's own
+  // catalog, so provider-scoped target-format resolution must select that URL.
+  assert.equal(executor.buildUrl("grok-4.5", false), "https://api.x.ai/v1/responses");
 });
 
 test("strips a -{level} suffix from an allow-listed model and sets reasoning_effort", () => {
@@ -88,7 +86,7 @@ test("strips reasoning_effort for the explicit -non-reasoning variant (already e
   assert.equal(out.reasoning_effort, undefined);
 });
 
-test("leaves a plain, unlisted model id and body unchanged (no suffix, not allow/deny listed)", () => {
+test("converts a Responses-tagged model while leaving its unclassified reasoning state alone", () => {
   const executor = new XaiExecutor();
   const body = { model: "grok-4.20-multi-agent-0309", messages: [{ role: "user", content: "hi" }] };
   const out = executor.transformRequest(
@@ -100,7 +98,13 @@ test("leaves a plain, unlisted model id and body unchanged (no suffix, not allow
 
   assert.equal(out.model, "grok-4.20-multi-agent-0309");
   assert.equal(out.reasoning_effort, undefined);
-  assert.deepEqual(out.messages, body.messages);
+  assert.equal(out.messages, undefined);
+  assert.deepEqual(out.input, [
+    {
+      role: "user",
+      content: [{ type: "input_text", text: "hi" }],
+    },
+  ]);
 });
 
 // Port of decolua/9router#2439 (author: @ryanngit): xAI ships a native