Improve question detection, capture stealth, latency; stop tracking model binaries

- Detection: broaden request-starters, handle contractions, allow 1-word '?'
- Audio: PTT pre-roll (no clipped first word), drop Whisper silence hallucinations,
  greedy decoding + domain initial_prompt for faster/cleaner transcription
- AI: cap spoken-answer tokens so replies return at conversational speed
- Overlay: answers persist (no auto-hide); wire Ctrl+Shift+H show/hide toggle
- Screen capture: freeze-frame at hotkey press (immune to focus-blur lockouts),
  hide selector from screen-share (NSWindowSharingNone), higher capture resolution
- Stop tracking models/ and *.zip (large binaries; add to .gitignore)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

setup.py

@@ -17,32 +17,37 @@ def install_dependencies():
 
 
 def download_models():
-    """Download AI models"""
-    print("🤖 Downloading AI models...")
+    """Download the Qwen2.5-VL vision brain (model + mmproj) into models/.
 
-    # Create models directory
+    The Whisper STT model (distil-large-v3 by default) auto-downloads via
+    faster-whisper on first run, so it isn't fetched here.
+    """
+    print("🤖 Downloading AI models (Qwen2.5-VL-7B + vision projector)...")
     os.makedirs("models", exist_ok=True)
 
-    # Download TinyLlama (small, fast, works on CPU)
-    model_urls = {
-        "tinyllama-1.1b.Q4_K_M.gguf": "https://huggingface.co/TheBloke/TinyLlama-1.1B-GGUF/resolve/main/tinyllama-1.1b.Q4_K_M.gguf"
-    }
+    repo = "ggml-org/Qwen2.5-VL-7B-Instruct-GGUF"
+    files = [
+        "Qwen2.5-VL-7B-Instruct-Q4_K_M.gguf",        # ~4.7 GB
+        "mmproj-Qwen2.5-VL-7B-Instruct-f16.gguf",    # ~1.4 GB vision projector
+    ]
 
-    import urllib.request
+    try:
+        from huggingface_hub import hf_hub_download
+    except ImportError:
+        print("❌ huggingface_hub not installed. Run: pip install -r requirements.txt")
+        return
 
-    for model_name, url in model_urls.items():
-        model_path = os.path.join("models", model_name)
-
-        if not os.path.exists(model_path):
-            print(f"Downloading {model_name}...")
-            try:
-                urllib.request.urlretrieve(url, model_path)
-                print(f"✅ Downloaded {model_name}")
-            except Exception as e:
-                print(f"❌ Failed to download {model_name}: {e}")
-                print("⚠️ Will run in fallback mode without local LLM")
-        else:
-            print(f"✅ {model_name} already exists")
+    dest = os.path.abspath("models")
+    for fn in files:
+        if os.path.exists(os.path.join(dest, fn)):
+            print(f"✅ {fn} already exists")
+            continue
+        print(f"Downloading {fn} (this is large; behind a proxy it may be slow)...")
+        try:
+            hf_hub_download(repo_id=repo, filename=fn, local_dir=dest)
+            print(f"✅ Downloaded {fn}")
+        except Exception as e:
+            print(f"❌ Failed to download {fn}: {e}")
 
 
 def setup_audio():