turbollm 1.9.4 → 1.9.5

package/dist/cli.js

@@ -1460,11 +1460,11 @@ var Manager = class _Manager {
     }
   }
   /** Wait until the engine leaves the 'starting' state (→ running or error/stopped),
-   *  bounded by the engine kind's readiness window plus a small grace. The internal
-   *  readiness loop flips the state and surfaces errors; this just keeps the load
-   *  gate held until that resolves. */
+   *  bounded by the readiness window plus a small grace. The internal readiness loop
+   *  flips the state and surfaces errors; this just keeps the load gate held until
+   *  that resolves. */
   async awaitNotStarting() {
-    const deadline = Date.now() + readinessTimeoutMs(this.opts?.engine.kind ?? "llama-server") + 5e3;
+    const deadline = Date.now() + READINESS_TIMEOUT_MS + 5e3;
     while (this.state === "starting" && Date.now() < deadline) await sleep(200);
   }
   async startInternal(opts) {
@@ -1694,10 +1694,13 @@ var Manager = class _Manager {
     if (this.child !== child) return;
     if (child.pid) clearEnginePid(this.store.dir(), child.pid);
     const cleanStop = this.state === "stopping" || this.state === "stopped";
+    const timedOut = this.state === "error" && this.errInfo?.code === "readiness_timeout";
     try {
       logStream.write(
         cleanStop ? `
 [turbollm] engine stopped \u2014 the model is no longer loaded.
+` : timedOut ? `
+[turbollm] engine process killed after the readiness timeout above. The model did NOT load.
 ` : `
 [turbollm] engine process exited unexpectedly (exit ${code})${errMsg ? ` \u2014 ${errMsg}` : ""}. The model did NOT load / is no longer loaded.
 `
@@ -1705,9 +1708,9 @@ var Manager = class _Manager {
     } catch {
     }
     logStream.end();
-    if (this.state === "stopping" || this.state === "stopped") {
+    if (cleanStop) {
       this.state = "stopped";
-    } else {
+    } else if (!timedOut) {
       this.state = "error";
       this.errInfo = {
         code: errMsg ? "engine_spawn_failed" : "engine_exited",
@@ -1723,7 +1726,7 @@ var Manager = class _Manager {
   }
   async readiness(child, port2) {
     const kind = this.opts?.engine.kind ?? "llama-server";
-    const deadline = Date.now() + readinessTimeoutMs(kind);
+    const deadline = Date.now() + READINESS_TIMEOUT_MS;
     for (; ; ) {
       await sleep(500);
       if (this.child !== child || this.state !== "starting") return;
@@ -1750,7 +1753,7 @@ var Manager = class _Manager {
           this.state = "error";
           this.errInfo = {
             code: "readiness_timeout",
-            message: `The model did not become ready within ${Math.round(readinessTimeoutMs(this.opts?.engine.kind ?? "llama-server") / 1e3)} seconds.`,
+            message: `The model did not become ready within ${Math.round(READINESS_TIMEOUT_MS / 1e3)} seconds.`,
             exitCode: -1,
             logTail: readTail(this.logPathStr, 20)
           };
@@ -1794,9 +1797,7 @@ function engineCommand(opts, port2, slotSavePath) {
   }
   return { cmd: opts.engine.binPath, args: buildArgs(opts, port2, slotSavePath) };
 }
-function readinessTimeoutMs(kind) {
-  return kind === "vllm" || kind === "sglang" ? 6e5 : 12e4;
-}
+var READINESS_TIMEOUT_MS = 6e5;
 function pyEngineEnv(kind, dataDir, binPath) {
   if (kind !== "mlx" && kind !== "rapid-mlx" && kind !== "vllm" && kind !== "sglang") {
     if (process.platform === "win32") return void 0;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "turbollm",
-  "version": "1.9.4",
+  "version": "1.9.5",
   "description": "TurboLLM — local LLM platform: run any inference engine auto-tuned to your GPU, with a web UI and OpenAI/Anthropic-compatible API. Point Claude Code at your own machine in one command.",
   "license": "FSL-1.1-ALv2",
   "author": "Mohit Soni",