Update app.py

app.py

@@ -649,17 +649,16 @@ api_interface = gr.Interface(
     api_name="/chat"
 )
 
-def http_api_infer(image, question):
-    """
-    Simple API endpoint that mimics InternVL logic with one image + text.
-    """
-    # Build a simplified version of `state` object here
+def http_api_infer(image, question, request: gr.Request):
+    token = request.headers.get("x-api-token", "")
+    if token != "123456789Rihem":
+        return "❌ Unauthorized. Missing or invalid API token."
+
     dummy_state = init_state()
     dummy_state.set_system_message("You are a vision-language assistant.")
     dummy_state.append_message(Conversation.USER, question, [image])
     dummy_state.skip_next = False
 
-    # Simulate inference (you can directly call your model function here instead)
     worker_addr = os.environ.get("WORKER_ADDR", "")
     api_token = os.environ.get("API_TOKEN", "")
     headers = {
@@ -670,8 +669,6 @@ def http_api_infer(image, question):
     if not worker_addr:
         return "⚠️ Model backend is not configured."
 
-    all_image_paths = [dummy_state.save_image(image)]
-
     pload = {
         "model": "InternVL2.5-78B",
         "messages": dummy_state.get_prompt_v2(inlude_image=True, max_dynamic_patch=12),
@@ -687,7 +684,7 @@ def http_api_infer(image, question):
         reply = response.json()["choices"][0]["message"]["content"]
         return reply
     except Exception as e:
-        return f"Error: {str(e)}"
+        return f"Error during model call: {str(e)}"
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
@@ -699,10 +696,28 @@ if __name__ == "__main__":
     args = parser.parse_args()
     logger.info(f"args: {args}")
 
-    logger.info(args)
+    # Build main UI and API endpoint interface
+    ui_demo = build_demo()
+
+    api_interface = gr.Interface(
+        fn=http_api_infer,
+        inputs=[gr.Image(type="pil"), gr.Textbox()],
+        outputs="text",
+        api_name="/chat"
+    )
+
+    # Serve both under tabs — API tab will be invisible in browser but callable
     demo = gr.TabbedInterface(
-    interface_list=[build_demo(), api_interface],
-    tab_names=["UI", "API"]
-)
-demo.queue(api_open=True).launch(...)
+        interface_list=[ui_demo, api_interface],
+        tab_names=["UI", "API"]
+    )
+
+    # ✅ Important: expose API via `api_open=True`
+    demo.queue(api_open=True).launch(
+        server_name=args.host,
+        server_port=args.port,
+        share=args.share,
+        max_threads=args.concurrency_count,
+    )
+