Update app.py

app.py

@@ -14,6 +14,10 @@ import torch
 import cv2
 from gradio_client import Client, file
 
+# The image for the bot and human
+bot_image = "path_to_bot_image.png"
+human_image = "path_to_human_image.png"
+
 def image_gen(prompt):
     client = Client("KingNish/Image-Gen-Pro")
     return client.predict("Image Generation",None, prompt, api_name="/image_gen_pro")
@@ -25,7 +29,6 @@ processor = LlavaProcessor.from_pretrained(model_id)
 model = LlavaForConditionalGeneration.from_pretrained(model_id)
 model.to("cpu")
 
-
 def llava(message, history):
     if message["files"]:
         image = message["files"][0]    
@@ -38,7 +41,7 @@ def llava(message, history):
         
     gr.Info("Analyzing image")
     image = Image.open(image).convert("RGB")
-    prompt = f"<|im_start|>user <image>\n{txt}<|im_end|><|im_start|>assistant"
+    prompt = f"user <image>\n{txt}assistant"
     
     inputs = processor(prompt, image, return_tensors="pt")
     return inputs
@@ -138,15 +141,15 @@ 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|>"
+                messages = f"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."
                 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"\nuser\n{str(msg[0])}"
+                    messages += f"\nassistant\n{str(msg[1])}"
+                messages+=f"\nuser\n{message_text}\nweb_result\n{web2}\nassistant\n"
                 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 == "<|im_end|>":
+                    if not response.token.text == "":
                         output += response.token.text
                         yield output
             elif json_data["name"] == "image_generation":
@@ -168,44 +171,42 @@ def respond(message, history):
 
                 thread = Thread(target=model.generate, kwargs=generation_kwargs)
                 thread.start()
-    
+
                 buffer = ""
                 for new_text in streamer:
                     buffer += new_text
                     yield buffer
-            else:
-                messages = f"<|im_start|>system\nYou are OpenCHAT mini a helpful assistant made by Nithish. You answers users query like human friend. 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. 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|>assistant\n"
-                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
-                        yield output
-        except:
-            messages = f"<|start_header_id|>system\nYou are OpenCHAT mini a helpful assistant made by Nithish. You answers users query like human friend. 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. You also try to show emotions using Emojis and reply like human, use short forms, friendly tone and emotions.<|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"
-            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
-                    yield output
-
-# Create the Gradio interface
-demo = gr.ChatInterface(
-    fn=respond,
-    chatbot=gr.Chatbot(show_copy_button=True, likeable=True, layout="panel"),
-    description ="# OpenGPT 4o mini\n ### You can engage in chat, generate images, perform web searches, and Q&A with images.",
-    textbox=gr.MultimodalTextbox(),
-    multimodal=True,
-    concurrency_limit=200,
-    cache_examples=False,
-)
-demo.launch()
+        except Exception as ex:
+            gr.Error(str(ex))
+    
+# Gradio Blocks interface with a custom layout
+with gr.Blocks(css="#chatbot .chat-message.user { border-bottom: none; margin-bottom: 2px; }") as demo:
+    gr.Markdown("# OpenChat Mini 🚀")    
+    chatbot = gr.Chatbot(label="Nithish OpenChat").style(container=False).style(height=700)
+    with gr.Row():
+        with gr.Column(scale=0.95):
+            with gr.Row():
+                txt = gr.Textbox(show_label=False, placeholder="Type your message here...").style(container=False)
+            with gr.Row():
+                file_btn = gr.UploadButton("📁", file_types=["image", "video", "document"])
+            with gr.Row():
+                submit_btn = gr.Button("Send", size="small", variant="primary")
+        with gr.Column(scale=0.05, align="center"):
+            gr.Image.update(human_image, human_image)
+            
+    with gr.Row():
+        with gr.Column(scale=0.95):
+            with gr.Row():
+                file_btn2 = gr.UploadButton("📁", file_types=["image", "video", "document"]).style(container=False)
+            with gr.Row():
+                bot_img = gr.Image(bot_image, bot_image).style(container=False)
+            with gr.Row():
+                submit_btn2 = gr.Button("Send", size="small", variant="primary")
+
+    txt.submit(respond, [txt, chatbot], [txt, chatbot], scroll_to_output=True)
+    file_btn.upload(respond, [txt, chatbot], [txt, chatbot], scroll_to_output=True)
+    submit_btn.click(respond, [txt, chatbot], [txt, chatbot], scroll_to_output=True)
+    file_btn2.upload(respond, [txt, chatbot], [txt, chatbot], scroll_to_output=True)
+    submit_btn2.click(respond, [txt, chatbot], [txt, chatbot], scroll_to_output=True)
+
+demo.launch(debug=True)