groove-dev 0.27.152 → 0.27.153

package/node_modules/@groove-dev/cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@groove-dev/cli",
-  "version": "0.27.152",
+  "version": "0.27.153",
   "description": "GROOVE CLI — manage AI coding agents from your terminal",
   "license": "FSL-1.1-Apache-2.0",
   "type": "module",

package/node_modules/@groove-dev/daemon/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@groove-dev/daemon",
-  "version": "0.27.152",
+  "version": "0.27.153",
   "description": "GROOVE daemon — agent orchestration engine",
   "license": "FSL-1.1-Apache-2.0",
   "type": "module",

package/node_modules/@groove-dev/daemon/src/llama-server.js

@@ -5,7 +5,10 @@
 // Each model gets its own server on a unique port.
 // Auto-starts when an agent needs a GGUF model, auto-stops when idle.
 
-import { spawn, execSync } from 'child_process';
+import { spawn, execSync, execFileSync } from 'child_process';
+import { existsSync, mkdirSync, chmodSync } from 'fs';
+import { resolve } from 'path';
+import { homedir } from 'os';
 
 const BASE_PORT = 8081;
 const MAX_SERVERS = 5;
@@ -25,10 +28,98 @@ export class LlamaServerManager {
       execSync('which llama-server', { stdio: 'ignore' });
       return true;
     } catch {
-      return false;
+      // Check common manual install locations
+      const paths = [
+        resolve(homedir(), '.local', 'bin', 'llama-server'),
+        resolve(homedir(), '.groove', 'bin', 'llama-server'),
+        '/usr/local/bin/llama-server',
+      ];
+      return paths.some(p => existsSync(p));
     }
   }
 
+  static getLlamaServerPath() {
+    try {
+      return execSync('which llama-server', { stdio: 'pipe', encoding: 'utf8' }).trim();
+    } catch {
+      const paths = [
+        resolve(homedir(), '.local', 'bin', 'llama-server'),
+        resolve(homedir(), '.groove', 'bin', 'llama-server'),
+        '/usr/local/bin/llama-server',
+      ];
+      return paths.find(p => existsSync(p)) || 'llama-server';
+    }
+  }
+
+  static async install() {
+    const platform = process.platform;
+
+    if (platform === 'darwin') {
+      try {
+        execSync('which brew', { stdio: 'ignore' });
+      } catch {
+        throw new Error('Homebrew not found. Install it from https://brew.sh then retry.');
+      }
+      execSync('brew install llama.cpp', { stdio: 'pipe', timeout: 600000 });
+      return { method: 'brew', path: execSync('which llama-server', { encoding: 'utf8', stdio: 'pipe' }).trim() };
+    }
+
+    if (platform === 'linux') {
+      const installDir = resolve(homedir(), '.local', 'bin');
+      mkdirSync(installDir, { recursive: true });
+
+      const arch = process.arch === 'arm64' ? 'arm64' : 'x64';
+      const hasCuda = (() => { try { execSync('which nvidia-smi', { stdio: 'ignore' }); return true; } catch { return false; } })();
+
+      const resp = await fetch('https://api.github.com/repos/ggml-org/llama.cpp/releases/latest', {
+        headers: { 'User-Agent': 'groove-dev' },
+      });
+      if (!resp.ok) throw new Error(`GitHub API error: ${resp.status}`);
+      const release = await resp.json();
+
+      const suffix = hasCuda ? `ubuntu-${arch}-cuda` : `ubuntu-${arch}`;
+      let asset = release.assets.find(a => a.name.includes(suffix) && a.name.endsWith('.zip'));
+      if (!asset && hasCuda) {
+        asset = release.assets.find(a => a.name.includes(`ubuntu-${arch}`) && a.name.endsWith('.zip'));
+      }
+      if (!asset) {
+        asset = release.assets.find(a => a.name.includes('ubuntu') && a.name.includes(arch) && a.name.endsWith('.zip'));
+      }
+      if (!asset) throw new Error(`No pre-built binary found for linux-${arch}. Build from source: https://github.com/ggml-org/llama.cpp#build`);
+
+      const tmpZip = `/tmp/groove-llama-${Date.now()}.zip`;
+      const tmpDir = `/tmp/groove-llama-extract-${Date.now()}`;
+
+      execSync(`curl -fSL "${asset.browser_download_url}" -o "${tmpZip}"`, { stdio: 'pipe', timeout: 600000 });
+      execSync(`unzip -o "${tmpZip}" -d "${tmpDir}"`, { stdio: 'pipe', timeout: 60000 });
+
+      const findResult = execSync(`find "${tmpDir}" -name llama-server -type f`, { encoding: 'utf8', stdio: 'pipe' }).trim();
+      const binPath = findResult.split('\n')[0];
+      if (!binPath) throw new Error('llama-server binary not found in release archive');
+
+      const destPath = resolve(installDir, 'llama-server');
+      execSync(`cp "${binPath}" "${destPath}"`, { stdio: 'pipe' });
+      chmodSync(destPath, 0o755);
+
+      // Copy shared libraries if present
+      try {
+        const libDir = resolve(binPath, '..', '..', 'lib');
+        if (existsSync(libDir)) {
+          const userLibDir = resolve(homedir(), '.local', 'lib');
+          mkdirSync(userLibDir, { recursive: true });
+          execSync(`cp -r "${libDir}/"* "${userLibDir}/"`, { stdio: 'pipe' });
+        }
+      } catch { /* libs are optional */ }
+
+      // Cleanup
+      try { execSync(`rm -rf "${tmpZip}" "${tmpDir}"`, { stdio: 'ignore' }); } catch { /* best-effort */ }
+
+      return { method: 'github-release', path: destPath, cuda: hasCuda, release: release.tag_name };
+    }
+
+    throw new Error(`Automatic install not supported on ${platform}. Install llama-server manually: https://github.com/ggml-org/llama.cpp#build`);
+  }
+
   // --- Server Lifecycle ---
 
   /**
@@ -74,9 +165,11 @@ export class LlamaServerManager {
       args.push('--flash-attn', 'auto');
     }
 
-    const proc = spawn('llama-server', args, {
+    const serverBin = LlamaServerManager.getLlamaServerPath();
+    const proc = spawn(serverBin, args, {
       stdio: ['ignore', 'pipe', 'pipe'],
       detached: false,
+      env: { ...process.env, LD_LIBRARY_PATH: [resolve(homedir(), '.local', 'lib'), process.env.LD_LIBRARY_PATH].filter(Boolean).join(':') },
     });
 
     if (!proc.pid) {

package/node_modules/@groove-dev/daemon/src/model-manager.js

@@ -69,7 +69,6 @@ export class ModelManager {
   async search(query, { limit = 20, sort = 'downloads' } = {}) {
     const params = new URLSearchParams({
       search: query,
-      filter: 'gguf',
       sort,
       direction: '-1',
       limit: String(limit),
@@ -83,15 +82,20 @@ export class ModelManager {
     if (!res.ok) throw new Error(`HuggingFace API error: ${res.status}`);
     const models = await res.json();
 
-    return models.map((m) => ({
-      id: m.modelId || m.id,
-      name: m.modelId?.split('/').pop() || m.id,
-      author: m.modelId?.split('/')[0] || '',
-      downloads: m.downloads || 0,
-      likes: m.likes || 0,
-      tags: m.tags || [],
-      lastModified: m.lastModified,
-    }));
+    return models.map((m) => {
+      const id = m.modelId || m.id;
+      const tags = m.tags || [];
+      return {
+        id,
+        name: id.split('/').pop() || id,
+        author: id.split('/')[0] || '',
+        downloads: m.downloads || 0,
+        likes: m.likes || 0,
+        tags,
+        lastModified: m.lastModified,
+        recommendedRuntimes: inferRuntimes(id, tags),
+      };
+    });
   }
 
   async getModelFiles(repoId) {
@@ -409,3 +413,41 @@ function classifyTier(params, quant) {
   if (billions >= 10) return 'medium';
   return 'light';
 }
+
+function inferRuntimes(repoId, tags) {
+  const lower = repoId.toLowerCase();
+  const tagSet = new Set(tags.map((t) => t.toLowerCase()));
+  const runtimes = new Set();
+
+  // GGUF → llama.cpp and (implicitly) Ollama
+  if (tagSet.has('gguf') || lower.includes('-gguf') || lower.includes('_gguf')) {
+    runtimes.add('llama.cpp');
+  }
+
+  // MLX-optimized models
+  if (tagSet.has('mlx') || lower.includes('-mlx') || lower.includes('_mlx')) {
+    runtimes.add('MLX');
+  }
+
+  // GPTQ / AWQ quantized → vLLM handles these well
+  if (tagSet.has('gptq') || tagSet.has('awq') || lower.includes('-gptq') || lower.includes('-awq')) {
+    runtimes.add('vLLM');
+  }
+
+  // SafeTensors / standard transformer weights → vLLM, TGI, MLX
+  if (tagSet.has('safetensors') || tagSet.has('transformers')) {
+    runtimes.add('vLLM');
+    runtimes.add('TGI');
+    if (!runtimes.has('MLX')) runtimes.add('MLX');
+  }
+
+  // If nothing matched, infer from general model traits
+  if (runtimes.size === 0) {
+    if (tagSet.has('pytorch') || tagSet.has('tf') || tagSet.has('jax')) {
+      runtimes.add('vLLM');
+      runtimes.add('TGI');
+    }
+  }
+
+  return [...runtimes];
+}

package/node_modules/@groove-dev/daemon/src/routes/providers.js

@@ -713,6 +713,17 @@ export function registerProviderRoutes(app, daemon) {
     res.json(daemon.llamaServer.getStatus());
   });
 
+  app.post('/api/llama/install', async (req, res) => {
+    try {
+      const { LlamaServerManager } = await import('../llama-server.js');
+      const result = await LlamaServerManager.install();
+      daemon.modelLab.refreshInstalledTools();
+      res.json({ success: true, ...result });
+    } catch (err) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
   app.get('/api/mlx/status', (req, res) => {
     res.json(daemon.mlxServer.getStatus());
   });