diff --git a/app/app.py b/app/app.py
index c5fd748..b2d7633 100644
--- a/app/app.py
+++ b/app/app.py
@@ -400,6 +400,39 @@ def synthesize(cid):
     return send_file(path, mimetype="audio/mpeg", as_attachment=False)
 
 
+from PIL import Image
+
+@app.route("/api/vision/<cid>", methods=["POST"])
+@login_required
+@require_capability("can_use_vision")
+def vision(cid):
+    file = request.files.get("file")
+    if not file:
+        return jsonify({"error": "No image"}), 400
+
+    img_id = str(uuid.uuid4())
+    path = f"/data/img-{img_id}.png"
+    file.save(path)
+
+    # Here you would call a vision-capable model (pseudo-code).
+    # description = call_vision_model(path)
+
+    description = "Image description placeholder"
+
+    db = SessionLocal()
+    msg = Message(
+        id=str(uuid.uuid4()),
+        conversation_id=cid,
+        role="assistant",
+        content=description,
+    )
+    db.add(msg)
+    db.commit()
+    db.close()
+
+    return jsonify({"status": "ok", "description": description})
+
+
 # ***
 
 import json