fix(embeddings): cool down account on hard errors (402/401/5xx) (#10529)

Merged — locally validated (focused embedding-cooldown tests green, gates green). Good parity with the chat path's existing pattern. Thanks!

src/lib/embeddings/service.ts

@@ -10,13 +10,14 @@ import { errorResponse, unavailableResponse } from "@omniroute/open-sse/utils/er
 import { HTTP_STATUS } from "@omniroute/open-sse/config/constants.ts";
 import * as log from "@/sse/utils/logger";
 import { toJsonErrorPayload } from "@/shared/utils/upstreamError";
-import { getProviderCredentials, clearRecoveredProviderState } from "@/sse/services/auth";
 import {
-  getCachedProviderNodes,
-  getComboByName,
-  getCombos,
-  getDatabaseSettings,
-} from "@/lib/localDb";
+  getProviderCredentials,
+  clearRecoveredProviderState,
+  markAccountUnavailable,
+} from "@/sse/services/auth";
+import { getCachedProviderNodes } from "@/lib/db/readCache";
+import { getComboByName, getCombos } from "@/lib/db/combos";
+import { getDatabaseSettings } from "@/lib/db/databaseSettings";
 import { resolveProxyForConnection } from "@/lib/db/settings";
 import { runWithProxyContext } from "@omniroute/open-sse/utils/proxyFetch.ts";
 import { handleComboChat } from "@omniroute/open-sse/services/combo.ts";
@@ -309,7 +310,7 @@ export async function createEmbeddingResponse(
       // #10347 — thread the selected connection id so handleEmbedding can cool the
       // account on a hard upstream failure (previously always null on /v1/embeddings).
       connectionId:
-        ((credentials as { connectionId?: string } | null)?.connectionId) ||
+        (credentials as { connectionId?: string } | null)?.connectionId ||
         options.connectionId ||
         connectionIdForProxy ||
         null,
@@ -340,6 +341,33 @@ export async function createEmbeddingResponse(
     });
   }
 
+  // #10347: cool down the account on hard errors (402 subscription expired,
+  // 401 revoked, 403 forbidden, 404 model gone, 429 rate limit, 5xx server
+  // errors) so the next embedding request skips this account. Mirrors chat.ts
+  // behavior.
+  // Skip for 400 (bad request) — the account is fine, the request was wrong.
+  // Best-effort: don't block the error response on the DB write.
+  const HARD_ERROR_STATUSES = new Set([401, 402, 403, 404, 429, 500, 502, 503, 504]);
+  if (
+    credentials &&
+    "connectionId" in credentials &&
+    typeof credentials.connectionId === "string" &&
+    HARD_ERROR_STATUSES.has(result.status)
+  ) {
+    markAccountUnavailable(
+      credentials.connectionId,
+      result.status,
+      result.error || "Embedding provider error",
+      provider,
+      resolvedModel || null
+    ).catch((err) => {
+      log.debug(
+        "EMBED",
+        `Cooldown write failed for ${provider}/${credentials.connectionId?.slice(0, 8)}: ${err}`
+      );
+    });
+  }
+
   responseHeaders.set("Content-Type", "application/json");
   const errorPayload = toJsonErrorPayload(result.error, "Embedding provider error");
   return new Response(JSON.stringify(errorPayload), {