Update app.py

app.py

@@ -14,29 +14,37 @@ import os
 import requests
 
 # upload image to google cloud storage
-def upload_image_to_gcs_blob(image):
+def upload_file_to_gcs_blob(file):
 
     google_creds = os.environ.get("GOOGLE_APPLICATION_CREDENTIALS_JSON")
 
     creds_json = json.loads(google_creds)
     credentials = service_account.Credentials.from_service_account_info(creds_json)
 
-    # 现在您可以使用这些凭证对Google Cloud服务进行认证
+    # Google Cloud credentials
     storage_client = storage.Client(credentials=credentials, project=creds_json['project_id'])
 
     bucket_name=os.environ.get('bucket_name')
     bucket = storage_client.bucket(bucket_name)
     
-    destination_blob_name = os.path.basename(image)
+    destination_blob_name = os.path.basename(file)
     blob = bucket.blob(destination_blob_name)
 
-    blob.upload_from_filename(image)
+    blob.upload_from_filename(file)
 
     public_url = blob.public_url
     
     return public_url
 
 
+from PIL import Image
+
+def is_image(file_path):
+    try:
+        Image.open(file_path)
+        return True
+    except IOError:
+        return False
 
 # def respond(
 #     message,
@@ -70,33 +78,33 @@ def upload_image_to_gcs_blob(image):
 #         response += token
 #         yield response
 
-def get_completion(message,history,system_message,max_tokens,temperature):
-    # base64_image = encode_image(image)
+def get_completion(message,history):
     if message["text"].strip() == "" and not message["files"]:
-        gr.Error("Please input a query and optionally image(s).")
+        raise gr.Error("Please input a query and optionally image(s).")
     
     if message["text"].strip() == "" and message["files"]:
-        gr.Error("Please input a text query along the image(s).")
+        raise gr.Error("Please input a text query along the image(s).")
     
     text = message['text']
     content = [
         {"type": "text", "text": text},
     ]
     if message['files']:
-        image = message['files'][0]
-        image_url = upload_image_to_gcs_blob(image)
-        content_image = {
-            "type": "image_url",
-            "image_url": {
-                "url": image_url,
-            },}
-        content.append(content_image)
+        file = message['files'][0]
+        public_url = upload_file_to_gcs_blob(file)
+        if is_image(file): # only support image file now
+            content_image = {
+                "type": "image_url",
+                "image_url": {
+                    "url": public_url,
+                },}
+            content.append(content_image)
+        else:
+            raise gr.Error("Only support image files now.")
 
-    init_message = [{"role": "system", "content": system_message}]
-    
     history_openai_format = []
     for human, assistant in history:
-        # 歷史消息中跳过包含图片的对话
+        # check if there is image info in the history message
         if isinstance(human, tuple):
             continue
         history_openai_format.append({"role": "user", "content": human })
@@ -104,47 +112,63 @@ def get_completion(message,history,system_message,max_tokens,temperature):
     history_openai_format.append({"role": "user", "content": content})
     print(history_openai_format)
     
+    system_message = '''You are GPT-4o("o" for omni), OpenAI's new flagship model that can reason across audio, vision, and text in real time. 
+    GPT-4o matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster. 
+    GPT-4o is especially better at vision and audio understanding compared to existing models.
+    GPT-4o's text and image capabilities are avaliable for users now. More capabilities like audio, and video will be rolled out iteratively in the future.
+    '''
+
     
-    # 请求头部信息
+    # headers
     openai_api_key = os.environ.get('openai_api_key')
     headers = {
       'Authorization': f'Bearer {openai_api_key}'
     }
+
+    temperature = 0.7
+    max_tokens = 2048
+
+    init_message = [{"role": "system", "content": system_message}]
+    messages = init_message + history_openai_format[-5:], #system message + latest 2 round dialogues + user input
     
-    # 请求体信息
+    # request body
     data = {
-        'model': 'gpt-4o',  # 可以根据需要更换其他模型
-        'messages': init_message + history_openai_format[-5:], #system message + 最近的2次對話 + 最新一條消息
-        'temperature': temperature,  # 可以根据需要调整
+        'model': 'gpt-4o',  # we use gpt-4o here
+        'messages': messages
+        'temperature': temperature, 
         'max_tokens':max_tokens,
         # 'stream':True，
     }
-    
+
+    # get response
     response = requests.post('https://burn.hair/v1/chat/completions', headers=headers, json=data)
-    
-    # 解析响应内容
     response_data = response.json()
-    response_content = response_data['choices'][0]['message']['content']
-    usage = response_data['usage']
+
+    if 'error' in response_data:
+        response_content = response_data['error']['message']
+    else:
+        response_content = response_data['choices'][0]['message']['content']
+        usage = response_data['usage']
     
     return response_content
 
 
 
-
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
 """
 demo = gr.ChatInterface(
     get_completion,
     multimodal=True,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-    ],
+    # additional_inputs=[
+    #     gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+    #     gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+    #     gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+    # ],
 )
 
+demo.queue(max_size = 10)
+
 
 if __name__ == "__main__":
     demo.launch()