@vierratale/ai 0.1.0-beta.7 → 0.1.0-beta.8

package/README.md

@@ -32,12 +32,12 @@ Local models run through the **Cortex** engine:
 
 | Model | Backend |
 |-------|---------|
-| vierratale-lite | qwen2.5-coder:0.5b |
-| vierratale-fast | qwen2.5-coder:1.5b |
-| vierratale-small | qwen2.5-coder:3b |
-| vierratale-balanced | qwen2.5-coder:7b |
-| vierratale-plus | qwen2.5-coder:14b |
-| vierratale-pro | qwen2.5-coder:32b |
+| vierratale-lite | qwen3:0.6b |
+| vierratale-fast | gemma3:1b |
+| vierratale-small | llama3.2:1b |
+| vierratale-balanced | qwen2.5:1.5b |
+| vierratale-plus | qwen3:1.7b |
+| vierratale-pro | qwen2.5:3b |
 
 Cloud models are available when the matching provider + API key is configured:
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vierratale/ai",
-  "version": "0.1.0-beta.7",
+  "version": "0.1.0-beta.8",
   "description": "VierrataleAI - Intelligent terminal assistant",
   "type": "module",
   "bin": {

package/src/catalog.js

@@ -1,10 +1,10 @@
 const MODELS = {
-  'qwen2.5-coder:0.5b': 'vierratale-lite',
-  'qwen2.5-coder:1.5b': 'vierratale-fast',
-  'qwen2.5-coder:3b': 'vierratale-small',
-  'qwen2.5-coder:7b': 'vierratale-balanced',
-  'qwen2.5-coder:14b': 'vierratale-plus',
-  'qwen2.5-coder:32b': 'vierratale-pro',
+  'qwen3:0.6b': 'vierratale-lite',
+  'gemma3:1b': 'vierratale-fast',
+  'llama3.2:1b': 'vierratale-small',
+  'qwen2.5:1.5b': 'vierratale-balanced',
+  'qwen3:1.7b': 'vierratale-plus',
+  'qwen2.5:3b': 'vierratale-pro',
   'gpt-4o-mini': 'vierratale-cloud-mini',
   'gpt-4o': 'vierratale-cloud',
   'claude-3-5-haiku-20241022': 'vierratale-cloud-fast',

package/src/cli.js

@@ -237,6 +237,7 @@ async function chat(provider, systemPrompt) {
           Config.save({ model: modelName });
           Config.setProviderModel(provider.name, modelName);
           Terminal.printSuccess(`Model: ${modelName}`);
+          await ensureLocalModel(modelName);
         } else {
           Terminal.printError(`Unknown model: ${modelName}`);
         }
@@ -308,6 +309,15 @@ async function chat(provider, systemPrompt) {
     }
 
     const intent = detectIntent(trimmed);
+    if (intent.type === 'self') {
+      const model = Config.getEffectiveModel(provider.name);
+      const info = Catalog.getModelInfo(model);
+      Terminal.printSuccess(
+        `I'm running on ${model} (${info ? info.realModel : model}) via the ${provider.displayName} engine.`
+      );
+      continue;
+    }
+
     if (intent.type === 'knowledge') {
       Terminal.printSuccess(`Auto-search: "${trimmed}"`);
       await answerWithSearch(messages, provider, trimmed, systemPrompt);
@@ -326,7 +336,10 @@ async function chat(provider, systemPrompt) {
 
     messages.push({ role: 'user', content: trimmed });
     Session.save(messages);
+    Terminal.printUser(trimmed);
     Terminal.printAIStart();
+    Terminal.beginThinking();
+    Terminal._thinking = true;
 
     let response = '';
     try {
@@ -349,6 +362,21 @@ async function chat(provider, systemPrompt) {
   }
 }
 
+async function ensureLocalModel(modelName) {
+  const info = Catalog.getModelInfo(modelName);
+  if (!info || !info.isLocal) return;
+  const host = Config.get('engineHost');
+  try {
+    const installed = await Installer.getInstalledModels(host);
+    if (installed.includes(info.realModel)) return;
+    Terminal.printInfo(`Downloading ${info.realModel} for "${modelName}"…`);
+    await Installer.pullModel(host, info.realModel);
+    Terminal.printSuccess(`Model ${info.realModel} ready.`);
+  } catch (err) {
+    Terminal.printError(`Could not download model: ${err.message}`);
+  }
+}
+
 export async function run() {
   const args = parseArgs(process.argv);
 
@@ -366,8 +394,12 @@ export async function run() {
   if (args.provider) Config.save({ provider: args.provider });
   if (args.model) Config.save({ model: args.model });
 
-  Terminal.printInfo('Initializing...');
-  await Installer.ensure();
+  await Terminal.showProgress(
+    ['Checking engine', 'Installing dependencies', 'Downloading model', 'Optimizing'],
+    () => Installer.ensure()
+  );
+
+  await ensureLocalModel(Config.get('model'));
 
   const provider = await ProviderFactory.autoDetect();
   if (!provider) {
@@ -377,6 +409,8 @@ export async function run() {
 
   Config.save({ provider: provider.name });
 
+  if (provider.warmup) provider.warmup();
+
   if (args.clear) Terminal.clear();
   showBanner(provider.name, Config.getEffectiveModel(provider.name), provider.displayName);
 

package/src/installer.js

@@ -55,6 +55,14 @@ async function getInstalledModels(host) {
 }
 
 export const Installer = {
+  async pullModel(host, model) {
+    return pullModel(host, model);
+  },
+
+  async getInstalledModels(host) {
+    return getInstalledModels(host);
+  },
+
   async ensure() {
     const host = Config.get('engineHost');
 
@@ -65,16 +73,23 @@ export const Installer = {
     const running = await isEngineRunning(host);
     if (!running) {
       try {
-        execSync('ollama serve &', { stdio: 'ignore' });
-        await new Promise((r) => setTimeout(r, 3000));
+        // No systemd available here, so launch the daemon directly, detached,
+        // and poll until it responds instead of relying on a service manager.
+        const child = exec('ollama serve', {
+          stdio: 'ignore',
+          detached: true,
+        });
+        if (child.unref) child.unref();
+        for (let i = 0; i < 30; i++) {
+          await new Promise((r) => setTimeout(r, 1000));
+          if (await isEngineRunning(host)) break;
+        }
       } catch {}
     }
 
     const realModel = Catalog.getRealModel(Config.get('model'));
     const installed = await getInstalledModels(host);
-    const needsPull = !installed.some(
-      (m) => m === realModel || m.startsWith(realModel.split(':')[0])
-    );
+    const needsPull = !installed.includes(realModel);
 
     if (needsPull) {
       await pullModel(host, realModel);

package/src/providers/cortex.js

@@ -95,4 +95,21 @@ export class CortexProvider extends BaseProvider {
       }
     }
   }
+
+  async warmup() {
+    // Fire-and-forget: send a tiny request so ollama loads the model into
+    // memory eagerly, so the first real query doesn't pay the cold-load cost.
+    const model = Catalog.getRealModel(Config.get('model'));
+    fetch(`${this.host}/api/chat`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({
+        model,
+        messages: [{ role: 'user', content: 'hi' }],
+        stream: false,
+        keep_alive: Config.get('keepAlive'),
+        options: { num_ctx: Config.get('numCtx') },
+      }),
+    }).catch(() => {});
+  }
 }

package/src/ui/branding.js

@@ -1,6 +1,6 @@
 export const Branding = {
   APP_NAME: 'VierrataleAI',
-  VERSION: '0.1.0-beta.6',
+  VERSION: '0.1.0-beta.8',
 
   colors: {
     primary: '\x1b[38;2;124;58;237m',

package/src/ui/terminal.js

@@ -9,25 +9,40 @@ export const Terminal = {
 
   printUser(text) {
     const time = new Date().toLocaleTimeString([], { hour: '2-digit', minute: '2-digit' });
+    process.stdout.write('\r\x1b[K');
+    console.log();
+    console.log(`  ${C.bold}${C.primary}You${C.reset}  ${C.dim}${time}${C.reset}`);
+    console.log(`  ${text}`);
     console.log();
-    console.log(`${this.divider('─', 4)} ${C.bold}${C.primary}YOU${C.reset} ${C.dim}${time}${C.reset} ${this.divider('─', 34)}`);
-    console.log(`${C.primary}┃${C.reset} ${text}`);
   },
 
   printAIStart() {
     const time = new Date().toLocaleTimeString([], { hour: '2-digit', minute: '2-digit' });
-    this.spinner = { frame: 0, timer: null };
     console.log();
-    console.log(`${this.divider('─', 4)} ${C.bold}${C.accent}${Branding.AI_PROMPT}${C.reset} ${C.dim}${time}${C.reset} ${this.divider('─', 34)}`);
+    console.log(`  ${C.bold}${C.accent}${Branding.AI_PROMPT}${C.reset}  ${C.dim}${time}${C.reset}`);
     this.curLine = '';
     this._printPrefix();
   },
 
   _printPrefix() {
-    process.stdout.write(`${C.accent}┃ ${C.reset}`);
+    process.stdout.write(`  ${C.accent}│${C.reset} `);
+  },
+
+  beginThinking() {
+    process.stdout.write(`${C.dim}· · ·${C.reset}`);
+  },
+
+  endThinking() {
+    process.stdout.write('\r\x1b[2K');
+    this._printPrefix();
   },
 
   printAIChunk(text) {
+    this._thinking = this._thinking || false;
+    if (this._thinking) {
+      this.endThinking();
+      this._thinking = false;
+    }
     this.curLine = (this.curLine || '') + text;
     process.stdout.write(text);
     this._maybeWrap();
@@ -39,7 +54,7 @@ export const Terminal = {
       if (words.length > 1) {
         const lastWord = words[words.length - 1];
         const rest = this.curLine.slice(0, this.curLine.length - lastWord.length).trimEnd();
-        process.stdout.write(`\n${C.accent}┃ ${C.reset}`);
+        process.stdout.write(`\n${C.accent}│${C.reset} `);
         this.curLine = lastWord;
       }
     }
@@ -47,36 +62,54 @@ export const Terminal = {
 
   printAIEnd() {
     process.stdout.write('\n');
-    console.log(`${C.accent}┗${C.reset}${this.divider('─', 44)}`);
     console.log();
   },
 
   printError(msg) {
-    console.log(`${C.error}✖ ${msg}${C.reset}`);
+    console.log(`  ${C.error}✖${C.reset} ${msg}`);
   },
 
   printInfo(msg) {
-    console.log(`${C.dim}${msg}${C.reset}`);
+    console.log(`  ${C.dim}${msg}${C.reset}`);
   },
 
   printSuccess(msg) {
-    console.log(`${C.success}✔ ${msg}${C.reset}`);
+    console.log(`  ${C.success}✔${C.reset} ${msg}`);
   },
 
   printWarning(msg) {
-    console.log(`${C.warning}⚠ ${msg}${C.reset}`);
+    console.log(`  ${C.warning}⚠${C.reset} ${msg}`);
   },
 
   printSearchResults(results) {
-    console.log(`${C.warning}┌── Web Search Results ──${C.reset}`);
+    console.log(`  ${C.warning}Web Search Results${C.reset}`);
     results.forEach((r, i) => {
-      console.log(
-        `  ${C.warning}${i + 1}.${C.reset} ${C.bold}${r.title}${C.reset}`
-      );
+      console.log(`  ${C.warning}${i + 1}.${C.reset} ${C.bold}${r.title}${C.reset}`);
       console.log(`     ${C.dim}${r.url}${C.reset}`);
       if (r.snippet) console.log(`     ${r.snippet.slice(0, 120)}`);
     });
-    console.log(`${C.warning}└──${C.reset}`);
+  },
+
+  async showProgress(steps, work) {
+    // Hidden real progress: the work() task runs silently while we paint a
+    // fake animated progress line so the install feels alive.
+    const frames = ['▁', '▂', '▃', '▄', '▅', '▆', '▇', '█'];
+    let i = 0;
+    const timer = setInterval(() => {
+      i++;
+      const stage = steps[Math.floor((i / 8) % steps.length)];
+      const bar = frames[i % frames.length];
+      const fill = '='.repeat(1 + (i % 18)) + '>' + ' '.repeat(18 - (i % 18));
+      process.stdout.write(`\r  ${C.accent}${bar}${C.reset} ${C.dim}${stage}${C.reset} [${C.accent}${fill}${C.reset}]`);
+    }, 100);
+    try {
+      await work();
+    } finally {
+      clearInterval(timer);
+      process.stdout.write('\r\x1b[2K');
+      console.log(`  ${C.success}✔${C.reset} ${C.dim}Ready${C.reset}`);
+      console.log();
+    }
   },
 
   clear() {

package/src/utils/intents.js

@@ -11,10 +11,22 @@ const KNOWLEDGE_PREFIXES = [
   'latest', 'news', 'current',
 ];
 
+const SELF_PHRASES = [
+  'what model', 'which model', 'what model are you', 'which model are you',
+  'what are you', 'who are you', "what's your name", 'what is your name',
+  'what language model', 'what llm', 'what ai model', 'what ai are you',
+  'your model', 'what model do you use', 'what model you use',
+  'what engine', 'which engine', 'are you gpt', 'are you chatgpt',
+];
+
 export function detectIntent(text) {
   const lower = text.toLowerCase().trim();
   if (lower.startsWith('/')) return { type: 'command' };
 
+  if (SELF_PHRASES.some((p) => lower.includes(p))) {
+    return { type: 'self' };
+  }
+
   const isQuestion = /\?$/.test(lower);
   const firstWord = lower.split(/\s+/)[0] || '';