Update app.py

app.py

@@ -40,7 +40,7 @@ def llava(message, history):
 
     gr.Info("Analyzing image")
     image = Image.open(image).convert("RGB")
-    prompt = f"<|im_start|>user <image>\n{txt}<|im_start|>assistant"
+    prompt = f"<image>\n{txt}"
 
     inputs = processor(prompt, image, return_tensors="pt")
     return inputs
@@ -116,43 +116,47 @@ def respond(message, history):
             {"type": "function", "function": {"name": "web_search", "description": "Search query on google",
                                               "parameters": {"type": "object", "properties": {
                                                   "query": {"type": "string", "description": "web search query"}},
-                                                             "required": ["query"]}}},
+                                                             "required": ["query"]}},
             {"type": "function", "function": {"name": "general_query", "description": "Reply general query of USER",
                                               "parameters": {"type": "object", "properties": {
                                                   "prompt": {"type": "string", "description": "A detailed prompt"}},
-                                                             "required": ["prompt"]}}},
+                                                             "required": ["prompt"]}},
             {"type": "function", "function": {"name": "image_generation", "description": "Generate image for user",
                                               "parameters": {"type": "object", "properties": {
                                                   "query": {"type": "string",
                                                             "description": "image generation prompt"}},
-                                                             "required": ["query"]}}},
+                                                             "required": ["query"]}},
             {"type": "function",
              "function": {"name": "image_qna", "description": "Answer question asked by user related to image",
                           "parameters": {"type": "object",
                                          "properties": {"query": {"type": "string", "description": "Question by user"}},
-                                         "required": ["query"]}}},
+                                         "required": ["query"]}},
         ]
 
+
         for msg in history:
             func_caller.append({"role": "user", "content": f"{str(msg[0])}"})
             func_caller.append({"role": "assistant", "content": f"{str(msg[1])}"})
 
+
         message_text = message["text"]
         func_caller.append({"role": "user",
                             "content": f'[SYSTEM]You are a helpful assistant. You have access to the following functions: \n {str(functions_metadata)}\n\nTo use these functions respond with:\n<functioncall> {{ "name": "function_name", "arguments": {{ "arg_1": "value_1", "arg_1": "value_1", ... }} }}  </functioncall>  [USER] {message_text}'})
 
+
         response = client_gemma.chat_completion(func_caller, max_tokens=200)
         response = str(response)
         try:
             response = response[int(response.find("{")):int(response.rindex("</"))]
         except:
-            response = response[int(response.find("{")):(int(response.rfind("}")) + 1)]
+            response = response[int(response.find("{")):(int(response.find("}")) + 1)]
         response = response.replace("\\n", "")
         response = response.replace("\\'", "'")
         response = response.replace('\\"', '"')
         response = response.replace('\\', '')
         print(f"\n{response}")
 
+
         try:
             json_data = json.loads(str(response))
             if json_data["name"] == "web_search":
@@ -161,17 +165,13 @@ def respond(message, history):
                 web_results = search(query)
                 gr.Info("Extracting relevant Info")
                 web2 = ' '.join([f"Link: {res['link']}\nText: {res['text']}\n\n" for res in web_results])
-                messages = f"<|im_start|>system\nYou are OpenCHAT mini a helpful assistant made by Nithish. You are provided with WEB results from which you can find informations to answer users query in Structured and More better way. You do not say Unnecesarry things Only say thing which is important and relevant. You also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user. You also try to show emotions using Emojis and reply like human, use short forms, friendly tone and emotions.<|im_end|>"
-                for msg in history:
-                    messages += f"\n<|im_start|>user\n{str(msg[0])}<|im_end|>"
-                    messages += f"\n<|im_start|>assistant\n{str(msg[1])}<|im_end|>"
-                messages += f"\n<|im_start|>user\n{message_text}<|im_end|>\n<|im_start|>web_result\n{web2}<|im_end|>\n<|im_start|>assistant\n"
+                messages = f"[SYSTEM]You are OpenCHAT mini a helpful assistant made by Nithish. You are provided with WEB results from which you can find informations to answer users query in Structured and More better way. You do not say Unnecesary things Only say thing which is important and relevant. You also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user. You also try to show emotions using Emojis and reply like human, use short forms, friendly tone and emotions. [USER]\n{message_text}[WEB_RESULT]\n{web2}[ASSISTANT]"
                 stream = client_mixtral.text_generation(messages, max_new_tokens=2000, do_sample=True, stream=True,
                                                         details=True, return_full_text=False)
                 output = ""
                 for response in stream:
                     if not response.token.text == "hello":
-                        output += response.token.text.replace("<|im_end|>", "")
+                        output += response.token.text.replace("]", "")
                         yield output
             elif json_data["name"] == "image_generation":
                 query = json_data["arguments"]["query"]
@@ -179,7 +179,7 @@ def respond(message, history):
                 yield "Generating Image, Please wait 10 sec..."
                 try:
                     client_sd3 = InferenceClient("stabilityai/stable-diffusion-3-medium-diffusers")
-                    seed = random.randint(0, 999999)
+                    seed = random.randint(0, 99999)
                     negativeprompt = ""
                     image = client_sd3.text_to_image(query, negative_prompt=f"{seed},{negativeprompt}")
                     yield gr.Image(image)
@@ -199,33 +199,24 @@ def respond(message, history):
                     buffer += new_text
                     yield buffer
             else:
-                messages = f"<|im_start|>system\nYou are OpenGPT a Expert AI Chat bot made by Nithish. You answers users query like professional AI. You are also Mastered in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user.<|im_end|>"
-                for msg in history:
-                    messages += f"\n<|im_start|>user\n{str(msg[0])}<|im_end|>"
-                    messages += f"\n<|im_start|>assistant\n{str(msg[1])}<|im_end|>"
-                messages += f"\n<|im_start|>user\n{message_text}<|im_end|>\n<|im_start|>assistant\n"
+                messages = f"[SYSTEM]You are OpenGPT a Expert AI Chat bot made by Nithish. You answers users query like professional . You are also Mastered in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user. [USER]\n{message_text}[ASSISTANT]"
                 stream = client_yi.text_generation(messages, max_new_tokens=2000, do_sample=True, stream=True,
                                                    details=True, return_full_text=False)
                 output = ""
                 for response in stream:
-                    if not response.token.text == "<|endoftext|>":
-                        output += response.token.text.replace("<|im_end|>", "")
+                    if not response.token.text == " ":
+                        output += response.token.text
                         yield output
         except:
-            messages = f"<|start_header_id|>system\nYou are OpenGPT a helpful AI CHAT BOT made by Nithish. You answers users query like professional . You are also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user.<|end_header_id|>"
-            for msg in history:
-                messages += f"\n<|start_header_id|>user\n{str(msg[0])}<|end_header_id|>"
-                messages += f"\n<|start_header_id|>assistant\n{str(msg[1])}<|end_header_id|>"
-            messages += f"\n<|start_header_id|>user\n{message_text}<|end_header_id|>\n<|start_header_id|>assistant\n"
+            messages = f"[SYSTEM]You are OpenGPT a helpful AI CHAT BOT made by Nithish. You answers users query like professional . You are also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user.[USER]\n{message_text}[ASSISTANT]"
             stream = client_llama.text_generation(messages, max_new_tokens=2000, do_sample=True, stream=True,
                                                   details=True, return_full_text=False)
             output = ""
             for response in stream:
                 if not response.token.text == "<|eot_id|>":
-                    output += response.token.text
+                    output += response.token
                     yield output
 
-
 demo = gr.ChatInterface(
     fn=respond,
     chatbot=gr.Chatbot(
@@ -236,11 +227,11 @@ demo = gr.ChatInterface(
     description="# OpenGPT 4o \n ### chat, generate images, perform web searches, and Q&A with images.",
     textbox=gr.MultimodalTextbox(),
     multimodal=True,
-    concurrency_limit=200,
+    concurrency_limit=20,
     cache_examples=False,
     theme="default",
     css=
-       """.chat-container {
+       .chat-container {
             border: 1px solid #ccc;
             border-radius: 5px;
             padding: 10px;
@@ -284,6 +275,6 @@ demo = gr.ChatInterface(
             content: 'Bot';
             font-weight: bold;
         }
-    """,
+    ,
 )
 demo.launch()