feat(api): wire shared admission across LLM routes

Acquire admission once after API-key policy, preserve lazy raw-request snapshots, and bind lease settlement to JSON, SSE, abort, deadline, and failure lifecycles. Expose a low-cardinality health summary and preserve non-SSE Ollama errors unchanged.

src/app/api/monitoring/health/route.ts

@@ -56,6 +56,7 @@ export async function GET() {
       sessionManagerModule,
       credentialHealthModule,
       localHealthModule,
+      adaptiveAdmissionModule,
       settingsResult,
       connectionsResult,
     ] = await Promise.allSettled([
@@ -67,6 +68,7 @@ export async function GET() {
       import("@omniroute/open-sse/services/sessionManager.ts"),
       import("@/lib/credentialHealth/cache"),
       import("@/lib/localHealthCheck"),
+      import("@omniroute/open-sse/services/admission/runtime.ts"),
       getCachedSettings(),
       getProviderConnections(),
     ]);
@@ -145,6 +147,14 @@ export async function GET() {
         : {};
     const settings = settingsResult.status === "fulfilled" ? settingsResult.value : {};
     const connections = connectionsResult.status === "fulfilled" ? connectionsResult.value : [];
+    const adaptiveAdmission =
+      adaptiveAdmissionModule.status === "fulfilled"
+        ? readHealthValue(
+            "adaptive admission",
+            () => adaptiveAdmissionModule.value.getAdaptiveAdmissionRuntime().snapshot(),
+            null
+          )
+        : null;
 
     const payload = buildHealthPayload({
       appVersion: APP_CONFIG.version,
@@ -169,6 +179,7 @@ export async function GET() {
       activeSessions,
       activeSessionsByKey,
       credentialHealth,
+      adaptiveAdmission,
     });
 
     healthPayloadCache = { payload, expiresAt: Date.now() + HEALTH_PAYLOAD_TTL_MS };
@@ -186,6 +197,7 @@ export async function GET() {
       lockouts: [],
       quotaMonitor: { ...fallbackQuotaMonitorSummary, monitors: [] },
       sessions: { activeCount: 0, stickyBoundCount: 0, byApiKey: {}, top: [] },
+      adaptiveAdmission: null,
       dedup: { inflightRequests: 0 },
     });
   }

src/app/api/v1/completions/route.ts

@@ -82,6 +82,7 @@ export async function POST(request: Request) {
           method: request.method,
           headers: request.headers,
           body: JSON.stringify(normalized),
+          signal: request.signal,
         });
         // #3571 — translate the chat-pipeline response back to the legacy
         // text-completion shape so OpenAI Completion clients (e.g. TabbyML) work.
@@ -90,7 +91,7 @@ export async function POST(request: Request) {
         // echo the compression header on the way out.
         return withCompressionHeaderEcho(
           await asTextCompletionResponse(
-            await handleChat(newRequest, buildClientRawRequest(request, body)),
+            await handleChat(newRequest, () => buildClientRawRequest(request, body)),
             typeof body.model === "string" ? body.model : undefined
           ),
           compressionRequestHeader
@@ -106,7 +107,10 @@ export async function POST(request: Request) {
   // Re-read body.model so the response echoes the caller's requested identifier.
   let requestedModel: string | undefined;
   try {
-    const bodyForModel = await request.clone().json().catch(() => null);
+    const bodyForModel = await request
+      .clone()
+      .json()
+      .catch(() => null);
     if (bodyForModel && typeof bodyForModel.model === "string") {
       requestedModel = bodyForModel.model;
     }

src/app/api/v1/providers/[provider]/chat/completions/route.ts

@@ -98,7 +98,8 @@ export async function POST(request, { params }) {
     method: request.method,
     headers: request.headers,
     body: JSON.stringify(body),
+    signal: request.signal,
   });
 
-  return await handleChat(newRequest, buildClientRawRequest(request, rawBody));
+  return await handleChat(newRequest, () => buildClientRawRequest(request, rawBody));
 }

src/app/api/v1beta/models/[...path]/route.ts

@@ -89,9 +89,7 @@ export async function POST(request, { params }) {
       action = modelAction.includes(":streamGenerateContent")
         ? ":streamGenerateContent"
         : ":generateContent";
-      model = modelAction
-        .replace(":streamGenerateContent", "")
-        .replace(":generateContent", "");
+      model = modelAction.replace(":streamGenerateContent", "").replace(":generateContent", "");
     }
 
     const validation = validateBody(v1betaGeminiGenerateSchema, rawBody);
@@ -113,9 +111,10 @@ export async function POST(request, { params }) {
       method: "POST",
       headers: request.headers,
       body: JSON.stringify(convertedBody),
+      signal: request.signal,
     });
 
-    const response = await handleChat(newRequest, buildClientRawRequest(request, rawBody));
+    const response = await handleChat(newRequest, () => buildClientRawRequest(request, rawBody));
 
     if (stream) {
       // Transform OpenAI SSE => Gemini SSE on the fly. The @google/genai SDK