@khanglvm/llm-router 2.0.0 → 2.0.1

package/CHANGELOG.md

@@ -10,6 +10,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Fixed
 - Raised the default inbound JSON body limit for OpenAI `/responses` requests from `1 MiB` to `8 MiB` while keeping other JSON routes at `1 MiB`. This prevents local `413 Request body too large` failures for Codex CLI and other Responses API clients carrying larger conversation state.
 
+## [2.0.1] - 2026-03-15
+
+### Fixed
+- Fixed alias-route failover after transient upstream failures. When every candidate on a route was only in cooldown, the balancer now retries the earliest-recovering candidate instead of returning `No eligible providers remain for route ...`.
+
 ## [2.0.0] - 2026-03-15
 
 ### Changed

package/README.md

@@ -14,7 +14,7 @@ The primary CLI command is now:
 llr
 ```
 
-`2.0.0` is the current public release. It includes the Web UI, AMP routing, and coding-tool integrations introduced in the 2.x line.
+`2.0.1` is the current public release. It includes the Web UI, AMP routing, and coding-tool integrations introduced in the 2.x line.
 
 ## Install
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@khanglvm/llm-router",
-  "version": "2.0.0",
+  "version": "2.0.1",
   "description": "LLM Router: single gateway endpoint for multi-provider LLMs with unified OpenAI+Anthropic format and seamless fallback",
   "keywords": [
     "llm-router",

package/src/runtime/balancer.js

@@ -186,6 +186,19 @@ function sortEntriesByOriginalOrder(left, right) {
   return left.originalIndex - right.originalIndex;
 }
 
+function sortCooldownEntries(left, right) {
+  if (left.openUntil !== right.openUntil) {
+    return left.openUntil - right.openUntil;
+  }
+  return sortEntriesByOriginalOrder(left, right);
+}
+
+function isCooldownOnlyEntry(entry) {
+  if (!entry || entry.eligible) return false;
+  const reasons = Array.isArray(entry.skipReasons) ? entry.skipReasons : [];
+  return reasons.length > 0 && reasons.every((reason) => reason === "cooldown");
+}
+
 async function buildCandidateEntries({
   candidates,
   stateStore,
@@ -315,6 +328,19 @@ export async function rankRouteCandidates({
   const ineligibleEntries = entries
     .filter((entry) => !entry.eligible)
     .sort(sortEntriesByOriginalOrder);
+  const fallbackCooldownEntries = eligibleEntries.length === 0
+    ? ineligibleEntries
+      .filter((entry) => isCooldownOnlyEntry(entry))
+      .sort(sortCooldownEntries)
+      .map((entry) => ({
+        ...entry,
+        eligible: true,
+        skipReasons: [...entry.skipReasons, "cooldown-overridden"]
+      }))
+    : [];
+  const skippedIneligibleEntries = fallbackCooldownEntries.length > 0
+    ? ineligibleEntries.filter((entry) => !isCooldownOnlyEntry(entry))
+    : ineligibleEntries;
   const estimatedRequiredTokens = normalizeNonNegativeInteger(
     requestContext?.estimatedRequiredTokens ??
     requestContext?.requiredTokens ??
@@ -324,10 +350,13 @@ export async function rankRouteCandidates({
   const routeCursor = stateStore
     ? await stateStore.getRouteCursor(resolvedRouteKey)
     : 0;
+  const rankableEntries = fallbackCooldownEntries.length > 0
+    ? fallbackCooldownEntries
+    : eligibleEntries;
   const contextAwareGroups = shouldApplyContextAwareOrdering(route, estimatedRequiredTokens)
-    ? partitionEligibleEntriesByContextWindow(eligibleEntries, estimatedRequiredTokens)
+    ? partitionEligibleEntriesByContextWindow(rankableEntries, estimatedRequiredTokens)
     : {
-        prioritizedEntries: eligibleEntries,
+        prioritizedEntries: rankableEntries,
         deferredEntries: []
       };
   const ranking = rankEligibleEntries(
@@ -339,7 +368,7 @@ export async function rankRouteCandidates({
   const rankedEntries = [
     ...ranking.orderedEligible,
     ...contextAwareGroups.deferredEntries,
-    ...ineligibleEntries
+    ...skippedIneligibleEntries
   ];
 
   return {
@@ -351,7 +380,7 @@ export async function rankRouteCandidates({
     shouldAdvanceCursor: ranking.shouldAdvanceCursor,
     entries: rankedEntries,
     selectedEntry: ranking.orderedEligible[0] || null,
-    skippedEntries: ineligibleEntries,
+    skippedEntries: skippedIneligibleEntries,
     rankedCandidates: rankedEntries.map((entry) => entry.candidate)
   };
 }