docs: update OCR description to reflect Ollama vision primary

docs/FEATURES.md

@@ -22,7 +22,7 @@ Three ways to add an asset:
 | Method | Description |
 |--------|-------------|
 | **Barcode** | Scan barcodes via device camera (ZXing library). Found assets redirect to detail view; unknown codes prompt to create new. |
-| **OCR** | Upload a photo of a machine ID sticker; text is extracted via Tesseract OCR. |
+|| **OCR** | Upload a photo of a machine ID sticker; text is extracted via Tesseract OCR or Ollama vision model (qwen2.5vl:3b on Windows PC). When Tesseract fails (dark/complex labels), falls back to Ollama for much more accurate text extraction. |
 | **Manual** | Full form with all fields. |
 
 ![Manual Entry Form](images/manual-entry.png)
@@ -199,6 +199,6 @@ Returns all geofence service areas assigned to a specific user — useful for fi
 | **Maps** | Leaflet 1.9.4 + OpenStreetMap |
 | **Geofences** | Leaflet Draw + point-in-polygon (Turf.js) |
 | **Scanner** | ZXing (CDN) |
-| **OCR** | Tesseract via pytesseract |
+| **OCR** | Ollama qwen2.5vl:3b (primary) + Tesseract (fallback) — via SSH tunnel to Windows PC |
 | **TLS** | Self-signed cert on port 8901 |
 | **Reverse Proxy** | Nginx Proxy Manager Plus |