feat: separate Ollama text/image/think models + add YouTube downloader

Support OLLAMA_IMAGE_MODEL and new OLLAMA_THINK_MODEL (both default to OLLAMA_MODEL)

Add /ollamathink command; validate thinking capability; disable image-analysis flow in think mode

Make /ollama-get-model show Text/Image/Think blocks only when models differ

Add /ytdl (/youtube) command + auto-detect YouTube URLs in messages

Cache downloaded videos to data/video and schedule daily cleanup

Move photo storage from data/temp to data/photo; improve env boolean parsing via ifTrue

Update deps (youtubei.js, puppeteer*) and TS config (allowJs, allowSyntheticDefaultImports)

package.json

@@ -17,14 +17,18 @@
     "drizzle-orm": "^1.0.0-beta.12-5845444",
     "emoji-regex": "^10.6.0",
     "ollama": "^0.6.3",
+    "puppeteer": "^24.36.1",
+    "puppeteer-extra": "^3.3.6",
+    "puppeteer-extra-plugin-stealth": "^2.11.2",
     "qrcode": "^1.5.4",
     "sharp": "^0.34.5",
     "systeminformation": "^5.30.6",
     "twemoji": "^14.0.2",
-    "typescript-telegram-bot-api": "^0.11.0"
+    "typescript-telegram-bot-api": "^0.11.0",
+    "youtubei.js": "^16.0.1"
   },
   "devDependencies": {
-    "@types/bun": "^1.3.7",
+    "@types/bun": "^1.3.8",
     "@types/node": "^25.1.0",
     "@types/qrcode": "^1.5.6",
     "@typescript-eslint/eslint-plugin": "^8.54.0",