Update app.py

app.py

@@ -12,14 +12,13 @@ from urllib3.exceptions import HTTPError
 
 from pathlib import Path
 
-#os.environ['DASHSCOPE_HTTP_BASE_URL'] = 'https://dashscope.aliyuncs.com/api/v1'
+os.environ['DASHSCOPE_HTTP_BASE_URL'] = 'https://dashscope.aliyuncs.com/api/v1'
+# os.environ['DASHSCOPE_WEBSOCKET_BASE_URL'] = 'https://poc-dashscope.aliyuncs.com/api-ws/v1/inference'
 
 import dashscope
 from dashscope import MultiModalConversation
 API_KEY = os.environ['API_KEY']
-BASE_URL = os.environ['DASHSCOPE_HTTP_BASE_URL']
 dashscope.api_key = API_KEY
-dashscope.base_http_api_url = BASE_URL
 
 is_modelscope_studio = os.getenv('MODELSCOPE_ENVIRONMENT') == 'studio'
 
@@ -70,30 +69,6 @@ def process_image(image, shouldConvert=False):
 
     return filename
 
-
-
-def generate(image, query):
-    imageFile = process_image(image)
-    content = [
-        {'image': f'file://{imageFile}'},
-        {'text': query}
-    ]
-    messages = [
-        {'role': 'user', 'content': content},
-    ]
-    print('messages:', messages)
-    responses = MultiModalConversation.call(
-        model='qvq-72b-preview', messages=messages, stream=True,
-    )
-    for response in responses:
-        if not response.status_code == HTTPStatus.OK:
-            raise HTTPError(f'response.code: {response.code}\nresponse.message: {response.message}')
-        response = response.output.choices[0].message.content
-        if len(response) > 0 and response[0]['text']:
-            print(response[0]['text'])
-            yield response[0]['text']
-
-
 if __name__ == "__main__":
 
     def on_clear():
@@ -196,6 +171,9 @@ if __name__ == "__main__":
                             with antd.Card(title=get_text("Answer", "答案"),
                                            elem_style=dict(height="100%"),
                                            elem_id="output-container"):
+                                with ms.Slot("extra"):
+                                    cancel_btn = antd.Button(get_text("Stop", "停止"),
+                                                                block=True, disabled=True)
                                 output = gr.Markdown(
                                     show_copy_button=True,
                                     latex_delimiters=[{
@@ -239,12 +217,42 @@ if __name__ == "__main__":
                           fn=lambda: gr.update(props=dict(open=False)),
                           outputs=[tour])
 
-                    submit_btn.click(
+                    def generate(image, query):
+                        imageFile = process_image(image)
+                        content = [
+                            {'image': f'file://{imageFile}'},
+                            {'text': query}
+                        ]
+                        messages = [
+                            {'role': 'user', 'content': content},
+                        ]
+                        print('messages:', messages)
+                        responses = MultiModalConversation.call(
+                            model='qvq-72b-preview', messages=messages, stream=True,
+                        )
+                        yield {
+                            cancel_btn: gr.update(disabled=False)
+                        }
+                        for response in responses:
+                            if not response.status_code == HTTPStatus.OK:
+                                raise HTTPError(f'response.code: {response.code}\nresponse.message: {response.message}')
+                            response = response.output.choices[0].message.content
+                            if len(response) > 0 and response[0]['text']:
+                                print(response[0]['text'])
+                                yield {
+                                    output: response[0]['text']
+                                }
+                        yield {
+                            cancel_btn: gr.update(disabled=True)
+                        }
+
+                    output_process = submit_btn.click(
                         fn=generate,
                         inputs=[*input_image, input],
-                        outputs=[output])
+                        outputs=[output, cancel_btn])
                     clear_btn.click(
                         fn=on_clear,
                         outputs=[*input_image, input])
+                    cancel_btn.click(fn=lambda : gr.update(disabled=True), inputs=None, outputs=[cancel_btn], cancels=[output_process])
 
                 demo.queue(default_concurrency_limit=50).launch()