zevairouter 1.0.152 → 1.0.153

package/cli/app/CHANGELOG.md

@@ -1,3 +1,9 @@
+# v1.0.153 (2026-07-27)
+
+## Fixes
+- **Legacy Codex lock recovery**: accounts whose health checks already cleared `lastErrorAt` can now use a newer positive quota snapshot to bypass a stale model lock once, restoring accounts that still have usable quota.
+- **Durable lock evidence**: new model locks retain their own `lockedAt` timestamp, while successful authentication tests preserve active-lock error metadata instead of silently breaking future recovery decisions.
+
 # v1.0.152 (2026-07-27)
 
 ## Fixes

package/cli/app/package.json

@@ -1,6 +1,6 @@
 {
   "name": "zevairouter",
-  "version": "1.0.152",
+  "version": "1.0.153",
   "description": "Self-hosted AI router: many providers, one OpenAI-compatible /v1 endpoint and dashboard",
   "bin": {
     "zevai": "./cli/cli.js"

package/cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "zevairouter-cli",
-  "version": "1.0.152",
+  "version": "1.0.153",
   "description": "ZevaiRouter CLI - Start and manage ZevaiRouter server",
   "bin": {
     "zevai": "./cli.js"

package/open-sse/services/accountFallback.js

@@ -74,9 +74,10 @@ export function hasAvailableQuota(usage) {
 
 /**
  * A successful quota poll newer than the error which created a model lock is
- * authoritative evidence that a Codex account recovered. The caller may probe
- * the account once; a fresh upstream failure updates lastErrorAt and makes the
- * same snapshot too old to bypass the replacement lock.
+ * authoritative evidence that a Codex account recovered. New locks carry an
+ * explicit lockedAt timestamp. Legacy locks fall back to lastErrorAt, then the
+ * connection update timestamp when older health checks already cleared the
+ * error metadata.
  */
 export function isModelLockRecoveredByQuota(connection, model, snapshot) {
   if (!isModelLockActive(connection, model) || !hasAvailableQuota(snapshot?.usage)) {
@@ -84,7 +85,16 @@ export function isModelLockRecoveredByQuota(connection, model, snapshot) {
   }
 
   const fetchedAt = Date.parse(snapshot?.fetchedAt);
-  const errorAt = Date.parse(connection?.lastErrorAt);
+  const reason = connection?.modelLockReason;
+  const reasonAt = reason?.model === model ? Date.parse(reason?.lockedAt) : Number.NaN;
+  const lastErrorAt = Date.parse(connection?.lastErrorAt);
+  const legacyUpdatedAt = Date.parse(connection?.updatedAt);
+  const errorAt = Number.isFinite(reasonAt)
+    ? reasonAt
+    : Number.isFinite(lastErrorAt)
+      ? lastErrorAt
+      : legacyUpdatedAt;
+
   return Number.isFinite(fetchedAt)
     && Number.isFinite(errorAt)
     && fetchedAt > errorAt;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "zevairouter",
-  "version": "1.0.152",
+  "version": "1.0.153",
   "description": "Self-hosted AI router: many providers, one OpenAI-compatible /v1 endpoint and dashboard",
   "bin": {
     "zevai": "./cli/cli.js"

package/src/app/api/providers/[id]/test/testUtils.js

@@ -5,6 +5,7 @@ import { isOpenAICompatibleProvider, isAnthropicCompatibleProvider } from "@/sha
 import { PROVIDER_ENDPOINTS } from "@/shared/constants/config";
 import { getDefaultModel } from "open-sse/config/providerModels.js";
 import { resolveOllamaLocalHost } from "open-sse/config/providers.js";
+import { getEarliestModelLockUntil } from "open-sse/services/accountFallback.js";
 import {
   refreshProviderCredentials,
   shouldRefreshCredentials,
@@ -769,12 +770,20 @@ export async function testSingleConnection(id) {
 
   const latencyMs = Date.now() - start;
 
+  // A successful auth/health test does not prove that a model-specific quota
+  // lock recovered. Preserve its error timestamp so a newer quota snapshot can
+  // safely authorize one bounded recovery probe.
+  const preserveModelLockError = result.valid && !!getEarliestModelLockUntil(connection);
   const updateData = result.inconclusive
     ? {}
     : {
         testStatus: result.valid ? "active" : "error",
-        lastError: result.valid ? null : result.error,
-        lastErrorAt: result.valid ? null : new Date().toISOString(),
+        lastError: result.valid
+          ? (preserveModelLockError ? connection.lastError : null)
+          : result.error,
+        lastErrorAt: result.valid
+          ? (preserveModelLockError ? connection.lastErrorAt : null)
+          : new Date().toISOString(),
       };
 
   if (result.refreshed && result.newTokens) {

package/src/sse/services/auth.js

@@ -380,6 +380,7 @@ export async function markAccountUnavailable(connectionId, status, errorText, pr
 
   const reason = typeof errorText === "string" ? errorText.slice(0, 100) : "Provider error";
   const lockUpdate = buildModelLockUpdate(model, cooldownMs);
+  const lockedAt = new Date().toISOString();
 
   await updateProviderConnection(connectionId, {
     ...lockUpdate,
@@ -387,8 +388,8 @@ export async function markAccountUnavailable(connectionId, status, errorText, pr
     lastError: reason,
     errorCode: status,
     providerErrorCode,
-    modelLockReason: model ? { model, providerErrorCode, reason, resetsAt: new Date(Date.now() + cooldownMs).toISOString() } : null,
-    lastErrorAt: new Date().toISOString(),
+    modelLockReason: model ? { model, providerErrorCode, reason, lockedAt, resetsAt: new Date(Date.now() + cooldownMs).toISOString() } : null,
+    lastErrorAt: lockedAt,
     backoffLevel: newBackoffLevel ?? backoffLevel
   });