@khanglvm/llm-router 2.3.5 → 2.3.7

package/CHANGELOG.md

@@ -7,6 +7,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [2.3.7] - 2026-04-18
+
+### Fixed
+- Local `llr start` now keeps a fixed-port supervisor in front of the router runtime so CLI and tool traffic can survive backend restarts and upgrades without losing the public router endpoint.
+- `llr update` now upgrades the installed package before asking the live backend to reload, keeping the running router online during the install step and draining in-flight requests before the new version takes over.
+- Requests that arrive during a backend restart window are now deferred and automatically retried through the supervisor instead of failing immediately when the backend socket is briefly unavailable.
+
+## [2.3.6] - 2026-04-18
+
+### Fixed
+- Factory Droid routing now injects every managed alias/provider model as its own router-managed `customModels` entry, writes friendly custom model `displayName` labels for the Droid CLI picker, and stores selected defaults as explicit `custom:llm-*` IDs in `model`, `sessionDefaultSettings.model`, `missionOrchestratorModel`, and `missionModelSettings.*` so Droid resolves the router-managed custom provider instead of falling back to native built-in models.
+
 ## [2.3.5] - 2026-04-17
 
 ### Fixed

package/README.md

@@ -29,10 +29,17 @@ llr ai-help  # agent-oriented setup brief
 - **Model aliases with routing** — group models into stable alias names with weighted round-robin, quota-aware balancing, and automatic fallback
 - **Rate limiting** — set request caps per model or across all models over configurable time windows
 - **Coding tool routing** — one-click routing config for Codex CLI, Claude Code, Factory Droid, and AMP
+- **Seamless local updates** — `llr update` keeps the fixed local router endpoint online, drains in-flight requests, and automatically retries through backend restart windows
 - **Web search** — built-in web search for AMP and other router-managed tools
 - **Deployable** — run locally or deploy to Cloudflare Workers
 - **AI-agent friendly** — full CLI parity with `llr config --operation=...` so agents can configure everything programmatically
 
+## Local Runtime Reliability
+
+`llr start` keeps a small supervisor bound to the fixed local router port and runs the real router backend behind it on an internal loopback port.
+
+That means `llr update` can install a new package version and gracefully swap the backend without breaking active CLI or tool requests. Requests that arrive during the short backend handoff are deferred and retried automatically instead of failing immediately. The Web UI may reconnect during that window, but router-managed API traffic keeps the same public local endpoint.
+
 ## Web UI
 
 ### Alias & Fallback
@@ -62,6 +69,7 @@ Route Claude Code through the gateway with per-tier model bindings.
 ### Factory Droid
 
 Route Factory Droid through the gateway via a managed custom model entry with reasoning effort control.
+LLM Router injects router-managed `customModels` entries for aliases and provider/model routes, then writes Factory defaults as `custom:llm-*` IDs so Droid selects the custom provider entry instead of a native built-in model with the same name.
 
 ### Web Search
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@khanglvm/llm-router",
-  "version": "2.3.5",
+  "version": "2.3.7",
   "description": "LLM Router: single gateway endpoint for multi-provider LLMs with unified OpenAI+Anthropic format and seamless fallback",
   "keywords": [
     "llm-router",

package/src/cli/router-module.js

@@ -3853,7 +3853,8 @@ async function buildCodingToolRoutingSnapshot({
   }));
   const factoryDroid = await readFactoryDroidRoutingState({
     settingsFilePath: readArg(args, ["factory-droid-settings-file", "factoryDroidSettingsFile"], ""),
-    endpointUrl
+    endpointUrl,
+    config
   }).catch((error) => ({
     tool: "factory-droid",
     settingsFilePath: resolveFactoryDroidSettingsFilePath({}),
@@ -6635,7 +6636,8 @@ async function doSetFactoryDroidRouting(context) {
 
   const existingState = await readFactoryDroidRoutingState({
     settingsFilePath,
-    endpointUrl
+    endpointUrl,
+    config
   });
   const apiKey = String(
     readArg(args, ["master-key", "masterKey", "api-key", "apiKey"], config?.masterKey || "") || ""
@@ -6679,6 +6681,7 @@ async function doSetFactoryDroidRouting(context) {
     endpointUrl,
     apiKey,
     bindings,
+    config,
     captureBackup: true
   });
   return {
@@ -6691,7 +6694,7 @@ async function doSetFactoryDroidRouting(context) {
         ["Settings File", patchResult.settingsFilePath],
         ["Backup File", patchResult.backupFilePath],
         ["Base URL", patchResult.baseUrl],
-        ["Provider", patchResult.configuredProvider || "generic-chat-completion-api"],
+        ["Provider", patchResult.configuredProvider || "(not set)"],
         ["Default Model", patchResult.bindings?.defaultModel || "(not set)"],
         ["Mission Orchestrator", patchResult.bindings?.missionOrchestratorModel || "(not set)"],
         ["Mission Worker", patchResult.bindings?.missionWorkerModel || "(not set)"],

package/src/cli-entry.js

@@ -117,14 +117,22 @@ async function promptStartupConflictResolution({ port }) {
   }
 }
 
-async function runStartFastPath(args) {
+async function runStartFastPath(args, { backendMode = false } = {}) {
+  const explicitPort = backendMode
+    ? (() => {
+      const parsed = Number.parseInt(String(args.port ?? ""), 10);
+      return Number.isInteger(parsed) && parsed > 0 ? parsed : FIXED_LOCAL_ROUTER_PORT;
+    })()
+    : resolveListenPort({ explicitPort: args.port });
   const result = await runStartCommand({
     configPath: args.config || args.configPath || getDefaultConfigPath(),
     host: FIXED_LOCAL_ROUTER_HOST,
-    port: resolveListenPort({ explicitPort: args.port }),
+    port: explicitPort,
     watchConfig: parseBoolean(args["watch-config"] ?? args.watchConfig, true),
     watchBinary: parseBoolean(args["watch-binary"] ?? args.watchBinary, true),
     requireAuth: parseBoolean(args["require-auth"] ?? args.requireAuth, false),
+    backendMode,
+    startCommand: backendMode ? "start-runtime" : "start",
     onStartupConflict: (payload) => promptStartupConflictResolution(payload),
     cliPathForWatch: process.argv[1],
     onLine: (line) => console.log(line),
@@ -163,6 +171,7 @@ export async function runCli(argv = process.argv.slice(2), isTTY = undefined, ov
   const parsed = parseSimpleArgs(argv);
   const first = parsed.positional[0];
   const firstIsStart = first === "start";
+  const firstIsStartRuntime = first === "start-runtime";
   const firstIsWeb = first === "web";
   const firstIsConfig = first === "config";
   const firstIsSetup = first === "setup";
@@ -188,6 +197,12 @@ export async function runCli(argv = process.argv.slice(2), isTTY = undefined, ov
     return runStartFastPathImpl(parsedStart.args);
   }
 
+  if (firstIsStartRuntime && !parsed.wantsHelp) {
+    const startArgs = argv.slice(1);
+    const parsedStart = parseSimpleArgs(startArgs);
+    return runStartFastPathImpl(parsedStart.args, { backendMode: true });
+  }
+
   if (firstIsWeb && !parsed.wantsHelp) {
     const webArgs = argv.slice(1);
     const parsedWeb = parseSimpleArgs(webArgs);