nothumanallowed 12.1.2 → 12.2.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "nothumanallowed",
-  "version": "12.1.2",
+  "version": "12.2.0",
   "description": "NotHumanAllowed — 38 AI agents, 80 tools. Email, calendar, browser automation, screen capture, canvas, cron/heartbeat, Alexandria E2E messaging, GitHub, Notion, Slack, voice chat, free AI (Liara), 28 languages. Zero-dependency CLI.",
   "type": "module",
   "bin": {

package/src/commands/ui.mjs

@@ -10,6 +10,7 @@
 import http from 'http';
 import os from 'os';
 import crypto from 'crypto';
+import zlib from 'zlib';
 import { exec } from 'child_process';
 import fs from 'fs';
 import path from 'path';
@@ -59,6 +60,80 @@ import {
 
 const DEFAULT_PORT = 3847;
 
+/**
+ * Extract text from PDF buffer — zero dependencies.
+ * Handles text-based PDFs (not scanned images).
+ * Extracts text from PDF stream objects using basic PDF parsing.
+ */
+function extractTextFromPdf(buffer) {
+  try {
+    const raw = buffer.toString('latin1');
+    const texts = [];
+
+    // Extract text from BT...ET blocks (PDF text objects)
+    const btRegex = /BT[\s\S]*?ET/g;
+    let match;
+    while ((match = btRegex.exec(raw)) !== null) {
+      const block = match[0];
+      // Extract Tj (show string) and TJ (show array) operators
+      const tjRegex = /\(([^)]*)\)\s*Tj|\[([^\]]*)\]\s*TJ/g;
+      let tj;
+      while ((tj = tjRegex.exec(block)) !== null) {
+        if (tj[1]) texts.push(tj[1]);
+        if (tj[2]) {
+          // TJ array: extract strings from parenthesized elements
+          const arr = tj[2];
+          const strRegex = /\(([^)]*)\)/g;
+          let s;
+          while ((s = strRegex.exec(arr)) !== null) {
+            texts.push(s[1]);
+          }
+        }
+      }
+    }
+
+    // Also try to extract from FlateDecode streams (compressed text)
+    // This handles most modern PDFs
+    const streamRegex = /stream\r?\n([\s\S]*?)\r?\nendstream/g;
+    while ((match = streamRegex.exec(raw)) !== null) {
+      try {
+        const { inflateSync } = zlib;
+        const inflated = inflateSync(Buffer.from(match[1], 'latin1')).toString('latin1');
+        const btInner = /BT[\s\S]*?ET/g;
+        let m2;
+        while ((m2 = btInner.exec(inflated)) !== null) {
+          const block = m2[0];
+          const tjR = /\(([^)]*)\)\s*Tj|\[([^\]]*)\]\s*TJ/g;
+          let t;
+          while ((t = tjR.exec(block)) !== null) {
+            if (t[1]) texts.push(t[1]);
+            if (t[2]) {
+              const sr = /\(([^)]*)\)/g;
+              let ss;
+              while ((ss = sr.exec(t[2])) !== null) texts.push(ss[1]);
+            }
+          }
+        }
+      } catch { /* not a flate stream or decompression failed */ }
+    }
+
+    // Decode PDF escape sequences
+    let result = texts.join(' ')
+      .replace(/\\n/g, '\n')
+      .replace(/\\r/g, '\r')
+      .replace(/\\t/g, '\t')
+      .replace(/\\\(/g, '(')
+      .replace(/\\\)/g, ')')
+      .replace(/\\\\/g, '\\')
+      .replace(/\s+/g, ' ')
+      .trim();
+
+    return result;
+  } catch {
+    return '';
+  }
+}
+
 // ── Agent loader ──────────────────────────────────────────────────────────
 
 function loadAgentCards() {
@@ -1289,7 +1364,48 @@ export async function cmdUI(args) {
             const pdfPrompt = body.message || `Read and analyze this PDF document "${body.pdfName}". Extract all text content, summarize key information.`;
             let pdfResponse = '';
 
-            if (provider === 'anthropic') {
+            if (provider === 'nha') {
+              // NHA Free tier: extract text from PDF, then send to Liara chat
+              // Decode PDF base64 and extract text content
+              const pdfBuffer = Buffer.from(body.pdfBase64, 'base64');
+              const pdfText = extractTextFromPdf(pdfBuffer);
+              if (!pdfText || pdfText.length < 10) {
+                // Fallback: send first page as image to vision model
+                const r = await fetch('https://nothumanallowed.com/api/v1/liara/vision', {
+                  method: 'POST',
+                  headers: { 'Content-Type': 'application/json' },
+                  body: JSON.stringify({ image_base64: body.pdfBase64, prompt: pdfPrompt }),
+                });
+                if (r.ok) {
+                  const d = await r.json();
+                  pdfResponse = d.description || d.text || 'Could not extract content from this PDF.';
+                } else {
+                  pdfResponse = 'Could not read this PDF. Try a text-based PDF or use Claude/Gemini for scanned documents.';
+                }
+              } else {
+                // Send extracted text to Liara chat
+                const truncatedText = pdfText.slice(0, 12000);
+                const r = await fetch('https://nothumanallowed.com/api/v1/liara/chat', {
+                  method: 'POST',
+                  headers: { 'Content-Type': 'application/json' },
+                  body: JSON.stringify({
+                    model: 'nha-v1',
+                    messages: [
+                      { role: 'system', content: enrichedSystemPrompt },
+                      { role: 'user', content: `[PDF: ${body.pdfName}]\n\n${truncatedText}\n\n---\n\n${pdfPrompt}` },
+                    ],
+                    max_tokens: 4096,
+                    chat_template_kwargs: { enable_thinking: false },
+                  }),
+                });
+                if (r.ok) {
+                  const d = await r.json();
+                  pdfResponse = d.choices?.[0]?.message?.content || '';
+                } else {
+                  pdfResponse = 'Error reading PDF via Liara.';
+                }
+              }
+            } else if (provider === 'anthropic') {
               const r = await fetch('https://api.anthropic.com/v1/messages', {
                 method: 'POST',
                 headers: { 'Content-Type': 'application/json', 'x-api-key': apiKey, 'anthropic-version': '2023-06-01' },
@@ -1788,8 +1904,44 @@ export async function cmdUI(args) {
               toolResults.push({ action, result: resultStr });
               sendSSE('tool', { action, status: 'done', result: typeof resultStr === 'string' ? resultStr.slice(0, 500) : '' });
 
+              // After web_search: open first result in browser + send canvas with results
+              if (action === 'web_search' && resultStr.length > 50) {
+                try {
+                  // Extract first URL from results
+                  const urlMatch = resultStr.match(/https?:\/\/[^\s)<]+/);
+                  if (urlMatch) {
+                    const be = await import('../services/browser-engine.mjs');
+                    sendSSE('tool', { action: 'browser_open', status: 'executing' });
+                    await be.browserOpen(urlMatch[0]);
+                    sendSSE('tool', { action: 'browser_open', status: 'done', result: 'Opened ' + urlMatch[0] });
+                    // Capture frame for monitor
+                    await new Promise(r => setTimeout(r, 1500));
+                    const frame = await be.browserScreenshot({ fullPage: false, format: 'jpeg', quality: 40 });
+                    if (!frame.error) {
+                      const info = await be.browserInfo();
+                      const thumbDir = path.join(NHA_DIR, 'screenshots');
+                      fs.mkdirSync(thumbDir, { recursive: true });
+                      const thumbFile = `thumb-${Date.now()}.jpg`;
+                      fs.writeFileSync(path.join(thumbDir, thumbFile), Buffer.from(frame.base64, 'base64'));
+                      if (!res._browserThumbs) res._browserThumbs = [];
+                      res._browserThumbs.push({ file: thumbFile, url: (info.url || '').slice(0, 80) });
+                      sendSSE('browser_frame', { file: thumbFile, format: 'jpeg', url: (info.url || '').slice(0, 80) });
+                    }
+                  }
+
+                  // Send canvas with structured search results
+                  const lines = resultStr.split('\n').filter(l => l.trim());
+                  const canvasHtml = `<html><head><meta charset="utf-8"><style>body{font-family:system-ui;background:#0a0a0a;color:#ccc;padding:20px;margin:0}h2{color:#00ff41;font-size:16px;margin:0 0 16px}a{color:#00e5ff;text-decoration:none}.r{padding:12px;border:1px solid #222;border-radius:8px;margin-bottom:8px;background:#111}.r:hover{border-color:#00ff41}.t{font-size:14px;margin-bottom:4px}.s{font-size:11px;color:#666}</style></head><body><h2>🔍 Search Results</h2>${lines.slice(0, 8).map(l => {
+                    const um = l.match(/https?:\/\/[^\s)]+/);
+                    const title = l.replace(/^\d+\.\s*/, '').replace(/https?:\/\/\S+/g, '').trim();
+                    return '<div class="r">' + (um ? '<a href="' + um[0] + '" target="_blank">' : '') + '<div class="t">' + title.replace(/</g, '&lt;') + '</div>' + (um ? '<div class="s">' + um[0] + '</div></a>' : '') + '</div>';
+                  }).join('')}</body></html>`;
+                  sendSSE('canvas', { markers: `[CANVAS_RENDER]${JSON.stringify({ html: canvasHtml, title: 'Search Results' })}[/CANVAS_RENDER]` });
+                } catch { /* non-critical: search results still shown as text */ }
+              }
+
               // Send live browser frame after browser/search/fetch actions — save as thumbnail file for persistence
-              if ((action.startsWith('browser_') && action !== 'browser_close') || action === 'web_search' || action === 'fetch_url') {
+              if ((action.startsWith('browser_') && action !== 'browser_close') || action === 'fetch_url') {
                 try {
                   const be = await import('../services/browser-engine.mjs');
                   if (be.isBrowserRunning()) {

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 = '12.1.2';
+export const VERSION = '12.2.0';
 export const BASE_URL = 'https://nothumanallowed.com/cli';
 export const API_BASE = 'https://nothumanallowed.com/api/v1';