Update app.py

app.py

@@ -142,7 +142,6 @@ def format_pretty_view(output):
 def process_single_image(model_name, image_file):
     runner = load_model_runner(model_name)
     image = Image.open(image_file.name).convert("RGB")
-
     result = runner(image)
 
     parsed_json = result.get("json")
@@ -151,11 +150,17 @@ def process_single_image(model_name, image_file):
     if parsed_json:
         json_output = format_result_json(parsed_json)
         pretty_output = format_pretty_view(parsed_json)
+
+        # Save JSON to a temporary file for download
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".json", mode="w", encoding="utf-8") as tmp:
+            json.dump(parsed_json, tmp, indent=2)
+            json_path = tmp.name
     else:
         json_output = "(No valid JSON extracted)"
         pretty_output = "(No structured content extracted)\n\n⚠️ Raw Model Output:\n" + raw_text
+        json_path = None
 
-    return image, json_output, pretty_output
+    return image, json_output, pretty_output, json_path
 
 # === Gradio UI
 iface = gr.Interface(
@@ -167,10 +172,11 @@ iface = gr.Interface(
     outputs=[
         gr.Image(label="Input Image"),
         gr.Textbox(label="Raw JSON Output (Technical)", lines=20),
-        gr.Textbox(label="Prettified View (User-Friendly)", lines=25)
+        gr.Textbox(label="Prettified View (User-Friendly)", lines=25),
+        gr.File(label="📥 Download JSON", visible=True)
     ],
     title="🖼️ Vision Model Extractor - JSON + Pretty View",
-    description="Upload a BPMN image and select a vision model to extract structured output. GPT-4o uses an API key from your Hugging Face Space Secret.",
+    description="Upload a BPMN image and select a vision model to extract structured output. Currenty supports only GPT-4o.",
     flagging_mode="never"
 )