nothumanallowed 11.3.0 → 11.4.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "nothumanallowed",
-  "version": "11.3.0",
+  "version": "11.4.0",
   "description": "NotHumanAllowed — 38 AI agents, 53 tools. Email, calendar, browser automation, screen capture, canvas, cron/heartbeat, GitHub, Notion, Slack, voice chat, 28 languages. Zero-dependency CLI.",
   "type": "module",
   "bin": {

package/src/constants.mjs

@@ -5,7 +5,7 @@ import { fileURLToPath } from 'url';
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = path.dirname(__filename);
 
-export const VERSION = '11.3.0';
+export const VERSION = '11.4.0';
 export const BASE_URL = 'https://nothumanallowed.com/cli';
 export const API_BASE = 'https://nothumanallowed.com/api/v1';
 

package/src/services/llm.mjs

@@ -237,15 +237,22 @@ export async function streamSSE(res, format) {
  * Connects to the NHA-hosted Qwen3 + LoRA on Hetzner GPU.
  * OpenAI-compatible format. Slower than paid providers (5-15 seconds).
  */
+/**
+ * NHA Free (Liara) — free LLM tier, no API key required.
+ * Connects to the NHA-hosted Qwen3 32B on Hetzner GPU.
+ * OpenAI-compatible format. Slower than paid providers (5-15 seconds).
+ * Supports thinking mode (Qwen3 native reasoning).
+ */
 export async function callNHA(apiKey, model, systemPrompt, userMessage, stream = false) {
   const body = {
-    model: model || 'nha-v1',
+    model: model || '/opt/models/qwen3-32b',
     max_tokens: 4096,
     messages: [
       { role: 'system', content: systemPrompt },
       { role: 'user', content: userMessage },
     ],
     stream,
+    chat_template_kwargs: { enable_thinking: false }, // OFF by default for speed
   };
   const res = await fetch('https://liara.nothumanallowed.com/v1/chat/completions', {
     method: 'POST',
@@ -258,9 +265,12 @@ export async function callNHA(apiKey, model, systemPrompt, userMessage, stream =
     const err = await res.text();
     throw new Error(`NHA Free ${res.status}: ${err}`);
   }
-  if (stream) return streamSSE(res, 'openai'); // OpenAI-compatible SSE format
+  if (stream) return streamSSE(res, 'openai');
   const data = await res.json();
-  return data.choices?.[0]?.message?.content || '';
+  let content = data.choices?.[0]?.message?.content || '';
+  // Strip thinking tags if present
+  content = content.replace(/<think>[\s\S]*?<\/think>/g, '').trim();
+  return content;
 }
 
 const PROVIDERS = {
@@ -452,7 +462,7 @@ function buildRequestBody(provider, model, systemPrompt, userMessage, stream) {
   }
   // OpenAI-compatible format (OpenAI, DeepSeek, Grok, Mistral)
   const modelDefaults = {
-    nha: 'nha-v1',
+    nha: '/opt/models/qwen3-32b',
     openai: 'gpt-4o',
     deepseek: 'deepseek-chat',
     grok: 'grok-3-latest',