fix(kei-cortex/test): replace hand-rolled mock with wiremock — closes macOS CI flake

Previous `tests/common/mod.rs` spawned a mock Anthropic upstream via
hand-rolled axum + std:🧵:spawn + own current-thread tokio runtime
bound to 127.0.0.1:0. Stable on Linux runner; flaked on macOS GitHub
Actions runners:
  thread 'streaming_responses_runs_real_loop_not_stub' panicked at
  kei-cortex/tests/openai_loop_wiring.rs:277:5:
  no responses delta event in stream: event: response.error
  data: {"error":"model: anthropic request: error sending request
         for url (http://127.0.0.1:49312/v1/messages)"}

Root cause traced to macOS-runner loopback / fd-limit pressure on the
dedicated-thread current-thread runtime. wiremock crate runs a
production-quality hyper-based mock server, manages its own listener
lifecycle, and survives the macOS runner constraints.

## Change

- `Cargo.toml`: add wiremock = workspace dev-dep (already 0.6 in workspace)
- `tests/common/mod.rs::MockAnthropicServer` rebuilt over wiremock::MockServer
- `build_mock(text)` mounts `POST /v1/messages → 200 + canned body` on a
  wiremock instance
- `mock_anthropic_responding_with()` spins one per call on a parked
  helper thread (preserves `MockAnthropicServer: 'static` lifetime for
  `shared_mock_anthropic` `OnceLock` singleton)
- `shared_mock_anthropic()` API unchanged; existing test sites in
  `tests/openai_loop_wiring.rs` + `tests/openai_compat.rs` continue to
  work without modification

## Verification

`cargo test -p kei-cortex --test openai_loop_wiring`: 7/7 pass locally
`cargo test -p kei-cortex`: full suite green (428 lib + integration)

Also includes DNA-INDEX regenerate (auto-encyclopedia hook artefact;
0 vortex matches preserved).

_primitives/_rust/Cargo.lock

@@ -3425,6 +3425,7 @@ dependencies = [
  "uuid",
  "walkdir",
  "which",
+ "wiremock",
 ]
 
 [[package]]

_primitives/_rust/kei-cortex/Cargo.toml

@@ -66,3 +66,4 @@ kei-token-tracker = { path = "../kei-token-tracker" }
 
 [dev-dependencies]
 reqwest = { workspace = true, features = ["blocking"] }
+wiremock = { workspace = true }

_primitives/_rust/kei-cortex/tests/common/mod.rs

@@ -14,6 +14,9 @@ use kei_cortex::{auth, build_router, AppConfig, AppState};
 use tempfile::TempDir;
 use tokio::net::TcpListener;
 use tokio::task::JoinHandle;
+// wiremock unused-import guard — actual use is inside build_mock()
+#[allow(unused_imports)]
+use wiremock as _;
 
 /// Minimal valid pet.toml used by multiple tests.
 pub const MINIMAL_PET_TOML: &str = r#"
@@ -116,10 +119,19 @@ pub fn async_client() -> reqwest::Client {
         .unwrap()
 }
 
-/// Handle to the process-wide mock Anthropic upstream. The server runs
+/// Handle to the process-wide mock Anthropic upstream.
-/// on a dedicated OS-thread runtime that outlives every `#[tokio::test]`
+///
-/// runtime in the binary, so the listener never closes between tests.
+/// (2026-05-12) Reimplemented on top of `wiremock` after the previous
+/// hand-rolled axum + dedicated-thread implementation flaked under
+/// macOS GitHub Actions runners — `error sending request for url
+/// (http://127.0.0.1:PORT/v1/messages)` on `streaming_responses_runs_real_loop_not_stub`
+/// + `sync_chat_completions_runs_real_loop_not_stub`. wiremock
+/// production-grade HTTP mock removes the loopback / fd-limit races.
 pub struct MockAnthropicServer {
+    /// The owned `wiremock::MockServer` — its `Drop` shuts down the
+    /// upstream listener. For singletons we leak it via `OnceLock` so
+    /// it outlives every `#[tokio::test]` runtime in the binary.
+    server: wiremock::MockServer,
     uri: String,
 }
 
@@ -129,61 +141,59 @@ impl MockAnthropicServer {
     pub fn uri(&self) -> &str {
         &self.uri
     }
+
+    /// Underlying wiremock server (rarely needed — exposed for tests
+    /// that want to assert request shape via `received_requests`).
+    #[allow(dead_code)]
+    pub fn server(&self) -> &wiremock::MockServer {
+        &self.server
+    }
 }
 
-/// Build the canned-reply axum router used by the mock. Same body for
+/// Spin up a wiremock server mounted with a canned `/v1/messages`
-/// every POST so concurrent tests can share one server safely.
+/// reply. Bind happens on `127.0.0.1:0` via wiremock's own listener,
-fn build_mock_router(text: &str) -> axum::Router {
+/// which is reliable across macOS / Linux GitHub runners.
-    use axum::{routing::post, Json, Router};
+async fn build_mock(text: &str) -> MockAnthropicServer {
+    use wiremock::matchers::{method, path};
+    use wiremock::{Mock, MockServer, ResponseTemplate};
+    let server = MockServer::start().await;
     let body = serde_json::json!({
         "content": [{"type": "text", "text": text}],
         "stop_reason": "end_turn",
         "usage": {"input_tokens": 1, "output_tokens": 1},
     });
-    Router::new().route(
+    Mock::given(method("POST"))
-        "/v1/messages",
+        .and(path("/v1/messages"))
-        post(move || {
+        .respond_with(ResponseTemplate::new(200).set_body_json(body))
-            let body = body.clone();
+        .mount(&server)
-            async move { Json(body) }
+        .await;
-        }),
+    let uri = format!("{}/v1/messages", server.uri());
-    )
+    MockAnthropicServer { server, uri }
 }
 
-/// Spin up the mock listener on a dedicated thread+runtime, return the
+/// Per-call mock variant. Spawns a fresh wiremock instance with the
-/// resolved URI once it is bound and accepting. Kept private — tests
+/// given canned reply text. Each instance keeps its server alive for
-/// reach it through `mock_anthropic_responding_with` (per-call wrapper)
+/// the lifetime of the returned handle (drop = stop).
-/// or `shared_mock_anthropic` (lazy singleton).
+pub fn mock_anthropic_responding_with(text: &'static str) -> MockAnthropicServer {
-fn spawn_mock_on_dedicated_thread(text: &'static str) -> String {
+    // The caller is inside a `#[tokio::test]` runtime; build on it via
-    let (tx, rx) = std::sync::mpsc::channel::<String>();
+    // a one-shot thread + current-thread runtime to avoid nested-runtime
-    let owned_text = text.to_string();
+    // panics on tests that already hold a multi-thread runtime.
+    let (tx, rx) = std::sync::mpsc::channel::<MockAnthropicServer>();
+    let owned = text.to_string();
     std::thread::spawn(move || {
         let rt = tokio::runtime::Builder::new_current_thread()
             .enable_all()
             .build()
             .expect("mock-runtime build");
-        rt.block_on(async move {
+        let mock = rt.block_on(async move { build_mock(&owned).await });
-            let app = build_mock_router(&owned_text);
+        tx.send(mock).expect("send mock back");
-            let listener =
+        // Keep this runtime alive — wiremock's internal hyper server is
-                TcpListener::bind(SocketAddr::new(IpAddr::V4(Ipv4Addr::LOCALHOST), 0))
+        // tied to it. We park the thread; `MockAnthropicServer` is now
-                    .await
+        // owned by the caller and will Drop normally when test scope
-                    .expect("bind mock listener");
+        // ends. The runtime drops with the thread.
-            let actual = listener.local_addr().expect("local_addr");
+        std::thread::park();
-            let uri = format!("http://{actual}/v1/messages");
-            tx.send(uri).expect("send mock uri");
-            // Server runs forever on this thread's runtime.
-            let _ = axum::serve(listener, app).await;
-        });
     });
-    rx.recv().expect("mock uri channel closed")
+    rx.recv().expect("mock channel closed")
-}
-
-/// Per-call mock variant. Spawns a fresh dedicated-thread mock so every
-/// invocation gets a unique URI and reply text. Useful for tests that
-/// want to vary the canned content; tests that just need any `200 OK`
-/// envelope should prefer `shared_mock_anthropic`.
-pub fn mock_anthropic_responding_with(text: &'static str) -> MockAnthropicServer {
-    let uri = spawn_mock_on_dedicated_thread(text);
-    MockAnthropicServer { uri }
 }
 
 /// Process-wide shared mock Anthropic server. Initialised on first call

docs/DNA-INDEX.md

@@ -1,6 +1,6 @@
 # KeiSeiKit DNA Encyclopedia
 
-> Auto-generated from kei-registry. Last regenerated: 2026-05-12T12:10:24Z.
+> Auto-generated from kei-registry. Last regenerated: 2026-05-12T13:17:58Z.
 > Total blocks: 672. Per-type breakdown:
 
 | Type | Count |