turbollm 1.3.0 → 1.3.2

package/dist/cli.js

@@ -319,12 +319,19 @@ import { promisify as promisify2 } from "util";
 
 // src/engines/download.ts
 import { createWriteStream, existsSync as existsSync2, mkdirSync as mkdirSync2, readdirSync, rmSync } from "fs";
-import { execFile } from "child_process";
+import { execFile, execFileSync } from "child_process";
 import { join as join2 } from "path";
 import { Readable } from "stream";
 import { pipeline } from "stream/promises";
 import { promisify } from "util";
 var execFileP = promisify(execFile);
+function stripMacOsQuarantine(dir) {
+  if (process.platform !== "darwin") return;
+  try {
+    execFileSync("xattr", ["-r", "-d", "com.apple.quarantine", dir]);
+  } catch {
+  }
+}
 var LLAMA_BUILD = "b9608";
 var REPO = "ggml-org/llama.cpp";
 var CUDA_VER = "13.3";
@@ -335,7 +342,10 @@ function availableBackends(tag = LLAMA_BUILD) {
   const a = arch();
   const def = (id, label, ...assets) => ({ id, label, assets });
   if (plat() === "darwin") {
-    return [def("metal", "Metal (Apple GPU)", `llama-${tag}-bin-macos-${a}.tar.gz`)];
+    return [
+      def("metal", "Metal (Apple GPU)", `llama-${tag}-bin-macos-${a}.tar.gz`),
+      def("cpu", "CPU", `llama-${tag}-bin-macos-${a}.tar.gz`)
+    ];
   }
   if (plat() === "win32") {
     if (a === "arm64") return [def("cpu", "CPU", `llama-${tag}-bin-win-cpu-arm64.zip`)];
@@ -464,6 +474,7 @@ async function extractArchive(archive, destDir) {
   } else {
     await execFileP("tar", ["-xzf", archive, "-C", destDir]);
   }
+  stripMacOsQuarantine(destDir);
 }
 async function provisionBackend(enginesRoot, backend, tag, onProgress, signal) {
   const destDir = join2(enginesRoot, `llama.cpp-${tag}-${backend.id}`);
@@ -561,32 +572,6 @@ async function latestCommitSha(repo, branch = "", signal) {
   const data = await res.json();
   return data.sha ?? "";
 }
-async function provisionForkRelease(enginesRoot, repo, destName, onProgress, signal) {
-  const destDir = join2(enginesRoot, destName);
-  if (existsSync2(destDir)) {
-    const found = findServer(destDir);
-    if (found) return found;
-  }
-  const rel = await latestGithubRelease(repo, signal);
-  const asset = pickReleaseAsset(rel.assets ?? []);
-  if (!asset) throw new Error("no_release_asset");
-  mkdirSync2(destDir, { recursive: true });
-  const tmp = join2(enginesRoot, asset.name);
-  try {
-    onProgress?.({ phase: "downloading", pct: 0 });
-    await downloadFile(asset.browser_download_url, tmp, onProgress, signal);
-    onProgress?.({ phase: "extracting", pct: -1 });
-    await extractArchive(tmp, destDir);
-    rmSync(tmp, { force: true });
-  } catch (e) {
-    rmSync(tmp, { force: true });
-    rmSync(destDir, { recursive: true, force: true });
-    throw e;
-  }
-  const bin = findServer(destDir);
-  if (!bin) throw new Error("llama-server not found in extracted release archive");
-  return bin;
-}
 async function turboquantAssetUrl(repo, platform = process.platform, archStr = process.arch, signal) {
   if (platform === "win32") return null;
   const res = await fetch(`https://api.github.com/repos/${repo}/releases?per_page=100`, {
@@ -607,7 +592,10 @@ async function provisionTurboquant(enginesRoot, repo, onProgress, signal) {
   const destDir = join2(enginesRoot, "turboquant");
   if (existsSync2(destDir)) {
     const found = findServer(destDir);
-    if (found) return found;
+    if (found) {
+      stripMacOsQuarantine(destDir);
+      return found;
+    }
   }
   const url = await turboquantAssetUrl(repo, process.platform, process.arch, signal);
   if (!url) throw new Error("no_release_asset");
@@ -1783,7 +1771,8 @@ function isTimeout(err3) {
 function runCaptured(bin, arg) {
   return new Promise((resolve2) => {
     try {
-      execFile5(bin, [arg], { cwd: dirname3(bin), timeout: 1e4, windowsHide: true, maxBuffer: 4 * 1024 * 1024 }, (error, stdout, stderr) => {
+      const timeoutMs = process.platform === "darwin" ? 6e4 : 15e3;
+      execFile5(bin, [arg], { cwd: dirname3(bin), timeout: timeoutMs, windowsHide: true, maxBuffer: 4 * 1024 * 1024 }, (error, stdout, stderr) => {
         resolve2({ out: (stdout || "") + (stderr || ""), err: error });
       });
     } catch (e) {
@@ -1807,7 +1796,8 @@ async function probe(bin) {
   const h = await runCaptured(bin, "--help");
   if (v.err && h.err) {
     if (isTimeout(v.err) && isTimeout(h.err)) {
-      throw new ProbeError("probe_timeout", "The binary did not respond within 10 seconds.");
+      const timeoutSec = process.platform === "darwin" ? 60 : 15;
+      throw new ProbeError("probe_timeout", `The binary did not respond within ${timeoutSec} seconds.`);
     }
     let msg2 = "Could not run the binary (--version and --help both failed).";
     const tail = lastLine(v.out);
@@ -2504,7 +2494,7 @@ import { rmSync as rmSync5 } from "fs";
 import { join as join9 } from "path";
 
 // src/sysinfo/sysinfo.ts
-import { execFileSync } from "child_process";
+import { execFileSync as execFileSync2 } from "child_process";
 import os from "os";
 import fs from "fs";
 var cached = null;
@@ -2535,7 +2525,7 @@ function classifyVendor(name) {
 }
 function detectGpus() {
   try {
-    const out = execFileSync("nvidia-smi", ["--query-gpu=name,memory.total", "--format=csv,noheader,nounits"], {
+    const out = execFileSync2("nvidia-smi", ["--query-gpu=name,memory.total", "--format=csv,noheader,nounits"], {
       timeout: 8e3,
       windowsHide: true
     }).toString();
@@ -2548,7 +2538,7 @@ function detectGpus() {
   }
   if (process.platform === "darwin") {
     try {
-      const out = execFileSync("system_profiler", ["SPDisplaysDataType"], { timeout: 8e3 }).toString();
+      const out = execFileSync2("system_profiler", ["SPDisplaysDataType"], { timeout: 8e3 }).toString();
       const m = out.match(/Chipset Model:\s*(.+)/);
       if (m && /Apple/.test(out)) {
         return [{ name: m[1].trim(), vramMb: Math.round(os.totalmem() / 1e6 * 0.65), vendor: "apple" }];
@@ -2565,7 +2555,7 @@ function detectGpus() {
 }
 function enumWindowsGpus() {
   const ps = 'Get-CimInstance Win32_VideoController | ForEach-Object { "$($_.Name)|$($_.AdapterRAM)" }';
-  const out = execFileSync("powershell", ["-NoProfile", "-NonInteractive", "-Command", ps], {
+  const out = execFileSync2("powershell", ["-NoProfile", "-NonInteractive", "-Command", ps], {
     timeout: 8e3,
     windowsHide: true
   }).toString();
@@ -2577,7 +2567,7 @@ function enumWindowsGpus() {
   }).filter((g) => g.name && g.vendor !== "unknown");
 }
 function enumLinuxGpus() {
-  const out = execFileSync("sh", ["-c", "lspci -mm 2>/dev/null | grep -iE 'VGA|3D|Display'"], {
+  const out = execFileSync2("sh", ["-c", "lspci -mm 2>/dev/null | grep -iE 'VGA|3D|Display'"], {
     timeout: 8e3
   }).toString();
   return out.trim().split("\n").filter(Boolean).map((line) => {
@@ -2660,10 +2650,9 @@ async function applyForkUpdate(d, engine, root, signal) {
     if (oldEng && d.registry.active()?.id === oldEng.id) await d.manager.stopAndWait();
     const tqDir = join9(root, "turboquant");
     rmSync5(tqDir, { recursive: true, force: true });
-    const bin = await provisionForkRelease(
+    const bin = await provisionTurboquant(
       root,
       "AtomicBot-ai/atomic-llama-cpp-turboquant",
-      "turboquant",
       (p) => d.provision.progress(p.phase, p.pct, p.part, p.parts),
       signal
     );

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "turbollm",
-  "version": "1.3.0",
+  "version": "1.3.2",
   "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",