Update app.py

app.py

@@ -1,121 +1,90 @@
 import gradio as gr
 import os
 import sys
-import json 
-import requests
+import json
 import random
+from openai import OpenAI
 
-
-MODEL = "gpt-4o-2024-11-20"
-API_URL = os.getenv("API_URL")
-DISABLED = os.getenv("DISABLED") == 'True'
-OPENAI_API_KEYS = os.getenv("OPENAI_API_KEYS").split(',')
-print (API_URL)
-print (OPENAI_API_KEYS)
-NUM_THREADS = int(os.getenv("NUM_THREADS"))
-
-print (NUM_THREADS)
+MODEL = "gpt-4.1-mini"
+DISABLED = os.getenv("DISABLED", "False").lower() == 'true'
+OPENAI_API_KEYS_RAW = os.getenv("OPENAI_API_KEYS", "")
+OPENAI_API_KEYS = [key.strip() for key in OPENAI_API_KEYS_RAW.split(',') if key.strip()] if OPENAI_API_KEYS_RAW else []
+NUM_THREADS = int(os.getenv("NUM_THREADS", "4"))
 
 def exception_handler(exception_type, exception, traceback):
     print("%s: %s" % (exception_type.__name__, exception))
+
 sys.excepthook = exception_handler
 sys.tracebacklimit = 0
+
+def predict(inputs, top_p, temperature, chat_counter, chatbot, history, request: gr.Request):
+    if not OPENAI_API_KEYS or not OPENAI_API_KEYS[0]:
+        yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2)], history, chat_counter, "No API keys configured", gr.update(interactive=True), gr.update(interactive=True)
+        return
+        
+    api_key = random.choice(OPENAI_API_KEYS)
+    client = OpenAI(api_key=api_key)
     
-def predict(inputs, top_p, temperature, chat_counter, chatbot, history, request:gr.Request):
-    payload = {
-        "model": MODEL,
-        "messages": [{"role": "user", "content": f"{inputs}"}],
-        "temperature": temperature,
-        "top_p": top_p,
-        "n": 1,
-        "stream": True,
-        "presence_penalty":0,
-        "frequency_penalty":0,
-    }
-    OPENAI_API_KEY = random.choice(OPENAI_API_KEYS)
-    print (OPENAI_API_KEY)
     headers_dict = {key.decode('utf-8'): value.decode('utf-8') for key, value in request.headers.raw}
-    headers = {
-        "Content-Type": "application/json",
-        "Authorization": f"Bearer {OPENAI_API_KEY}",
-        "Headers": f"{headers_dict}"
-    }
-
-    # print(f"chat_counter - {chat_counter}")
-    if chat_counter != 0 :
-        messages = []
+    
+    messages = []
+    if chat_counter != 0:
         for i, data in enumerate(history):
-            if i % 2 == 0:
-                role = 'user'
-            else:
-                role = 'assistant'
-            message = {}
-            message["role"] = role
-            message["content"] = data
-            messages.append(message)
-        
-        message = {}
-        message["role"] = "user" 
-        message["content"] = inputs
-        messages.append(message)
-        payload = {
-            "model": MODEL,
-            "messages": messages,
-            "temperature": temperature,
-            "top_p": top_p,
-            "n": 1,
-            "stream": True,
-            "presence_penalty":0,
-            "frequency_penalty":0,
-        }
-
+            role = 'user' if i % 2 == 0 else 'assistant'
+            messages.append({"role": role, "content": data})
+    
+    messages.append({"role": "user", "content": inputs})
+    
     chat_counter += 1
-
     history.append(inputs)
-    token_counter = 0 
-    partial_words = "" 
-    counter = 0
-
+    token_counter = 0
+    partial_words = ""
+    
     try:
-        # make a POST request to the API endpoint using the requests.post method, passing in stream=True
-        response = requests.post(API_URL, headers=headers, json=payload, stream=True)
-        response_code = f"{response}"
-        #if response_code.strip() != "<Response [200]>":
-        #    #print(f"response code - {response}")
-        #    raise Exception(f"Sorry, hitting rate limit. Please try again later. {response}")
+        stream = client.chat.completions.create(
+            model=MODEL,
+            messages=messages,
+            temperature=temperature,
+            top_p=top_p,
+            stream=True,
+            presence_penalty=0,
+            frequency_penalty=0,
+            max_tokens=2048
+        )
         
-        for chunk in response.iter_lines():
-            #print (chunk)
-            #sys.stdout.flush()
-            #Skipping first chunk
-            if counter == 0:
-                counter += 1
-                continue
-                #counter+=1
-            # check whether each line is non-empty
-            if chunk.decode() :
-                chunk = chunk.decode()
-                # decode each line as response data is in bytes
-                if len(chunk) > 12 and "content" in json.loads(chunk[6:])['choices'][0]['delta']:
-                    partial_words = partial_words + json.loads(chunk[6:])['choices'][0]["delta"]["content"]
-                    if token_counter == 0:
-                        history.append(" " + partial_words)
-                    else:
-                        history[-1] = partial_words
-                    token_counter += 1
-                    yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ], history, chat_counter, response, gr.update(interactive=False), gr.update(interactive=False)  # resembles {chatbot: chat, state: history}  
+        for chunk in stream:
+            if chunk.choices[0].delta.content is not None:
+                partial_words += chunk.choices[0].delta.content
+                if token_counter == 0:
+                    history.append(" " + partial_words)
+                else:
+                    history[-1] = partial_words
+                token_counter += 1
+                yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2)], history, chat_counter, "200", gr.update(interactive=False), gr.update(interactive=False)
+                
     except Exception as e:
-        print (f'error found: {e}')
-    yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ], history, chat_counter, response, gr.update(interactive=True), gr.update(interactive=True)
-    print(json.dumps({"chat_counter": chat_counter, "payload": payload, "partial_words": partial_words, "token_counter": token_counter, "counter": counter}))
-                   
+        print(f'error found: {e}')
+        yield [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2)], history, chat_counter, str(e), gr.update(interactive=True), gr.update(interactive=True)
+    
+    log_data = {
+        "chat_counter": chat_counter,
+        "model": MODEL,
+        "messages": messages,
+        "response": partial_words,
+        "headers": headers_dict,
+        "temperature": temperature,
+        "top_p": top_p,
+        "token_counter": token_counter
+    }
+    print(json.dumps(log_data))
 
 def reset_textbox():
     return gr.update(value='', interactive=False), gr.update(interactive=False)
 
-title = """<h1 align="center">GPT-4o: Research Preview (Short-Term Availability)</h1>"""
+title = """<h1 align="center">LatamGPT Data Collection: Research Preview</h1>"""
 if DISABLED:
-    title = """<h1 align="center" style="color:red">This app has reached OpenAI's usage limit. Please check back tomorrow.</h1>"""
+    title = """<h1 align="center" style="color:red">This app has reached usage limit. Please check back tomorrow.</h1>"""
+
 description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
 ```
 User: <utterance>
@@ -124,56 +93,51 @@ User: <utterance>
 Assistant: <utterance>
 ...
 ```
-In this app, you can explore the outputs of a gpt-4 turbo LLM.
+In this app, you can explore the outputs of GPT-4.1 mini while contributing to LatamGPT research.
 """
 
-theme = gr.themes.Default(primary_hue="green")                
+theme = gr.themes.Default(primary_hue="blue")
 
-with gr.Blocks(css = """#col_container { margin-left: auto; margin-right: auto;}
+with gr.Blocks(css="""#col_container { margin-left: auto; margin-right: auto;}
                 #chatbot {height: 520px; overflow: auto;}""",
               theme=theme) as demo:
     gr.HTML(title)
-    gr.HTML("""<h3 align="center" style="color: red;">If this app doesn't respond, consider trying our other GPT-4O app:<br/><a href="https://huggingface.co/spaces/yuntian-deng/ChatGPT4Turbo">https://huggingface.co/spaces/yuntian-deng/ChatGPT4Turbo</a></h3>""")
-
-    #gr.HTML('''<center><a href="https://huggingface.co/spaces/ysharma/ChatGPT4?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space and run securely with your OpenAI API Key</center>''')
-    with gr.Column(elem_id = "col_container", visible=False) as main_block:
-        #GPT4 API Key is provided by Huggingface 
-        #openai_api_key = gr.Textbox(type='password', label="Enter only your GPT4 OpenAI API key here")
-        chatbot = gr.Chatbot(elem_id='chatbot') #c
-        inputs = gr.Textbox(placeholder= "Hi there!", label= "Type an input and press Enter") #t
-        state = gr.State([]) #s
+    
+    with gr.Column(elem_id="col_container", visible=False) as main_block:
+        chatbot = gr.Chatbot(elem_id='chatbot')
+        inputs = gr.Textbox(placeholder="¡Hola! ¿En qué puedo ayudarte?", label="Escribe tu mensaje y presiona Enter")
+        state = gr.State([])
+        
         with gr.Row():
             with gr.Column(scale=7):
-                b1 = gr.Button(visible=not DISABLED) #.style(full_width=True)
+                b1 = gr.Button(visible=not DISABLED)
             with gr.Column(scale=3):
-                server_status_code = gr.Textbox(label="Status code from OpenAI server", )
-    
-        #inputs, top_p, temperature, top_k, repetition_penalty
+                server_status_code = gr.Textbox(label="Status code from server")
+        
         with gr.Accordion("Parameters", open=False):
-            top_p = gr.Slider( minimum=-0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p (nucleus sampling)",)
-            temperature = gr.Slider( minimum=-0, maximum=5.0, value=1.0, step=0.1, interactive=True, label="Temperature",)
-            #top_k = gr.Slider( minimum=1, maximum=50, value=4, step=1, interactive=True, label="Top-k",)
-            #repetition_penalty = gr.Slider( minimum=0.1, maximum=3.0, value=1.03, step=0.01, interactive=True, label="Repetition Penalty", )
+            top_p = gr.Slider(minimum=0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p (nucleus sampling)")
+            temperature = gr.Slider(minimum=0, maximum=2.0, value=0.7, step=0.1, interactive=True, label="Temperature")
             chat_counter = gr.Number(value=0, visible=False, precision=0)
     
-    with gr.Column(elem_id = "user_consent_container") as user_consent_block:
-        # Get user consent
+    with gr.Column(elem_id="user_consent_container") as user_consent_block:
         accept_checkbox = gr.Checkbox(visible=False)
-        js = "(x) => confirm('By clicking \"OK\", I agree that my data may be published or shared.')"
-        with gr.Accordion("User Consent for Data Collection, Use, and Sharing", open=True):
+        js = "(x) => confirm('Al hacer clic en \"Acepto\", acepto que mis datos pueden ser publicados o compartidos para investigación.')"
+        
+        with gr.Accordion("Consentimiento de Usuario para Recolección, Uso y Compartición de Datos", open=True):
             gr.HTML("""
             <div>
-                <p>By using our app, which is powered by OpenAI's API, you acknowledge and agree to the following terms regarding the data you provide:</p>
+                <p>Al usar nuestra aplicación, que funciona con la API de OpenAI, reconoces y aceptas los siguientes términos sobre los datos que proporcionas:</p>
                 <ol>
-                    <li><strong>Collection:</strong> We may collect information, including the inputs you type into our app, the outputs generated by OpenAI's API, and certain technical details about your device and connection (such as browser type, operating system, and IP address) provided by your device's request headers.</li>
-                    <li><strong>Use:</strong> We may use the collected data for research purposes, to improve our services, and to develop new products or services, including commercial applications, and for security purposes, such as protecting against unauthorized access and attacks.</li>
-                    <li><strong>Sharing and Publication:</strong> Your data, including the technical details collected from your device's request headers, may be published, shared with third parties, or used for analysis and reporting purposes.</li>
-                    <li><strong>Data Retention:</strong> We may retain your data, including the technical details collected from your device's request headers, for as long as necessary.</li>
+                    <li><strong>Recolección:</strong> Podemos recopilar información, incluyendo las entradas que escribes en nuestra aplicación, las salidas generadas por la API de OpenAI, y ciertos detalles técnicos sobre tu dispositivo y conexión (como tipo de navegador, sistema operativo e dirección IP) proporcionados por los headers de solicitud de tu dispositivo.</li>
+                    <li><strong>Uso:</strong> Podemos usar los datos recopilados para propósitos de investigación, para mejorar nuestros servicios, y para desarrollar nuevos productos o servicios, incluyendo aplicaciones comerciales, y para propósitos de seguridad, como proteger contra acceso no autorizado y ataques.</li>
+                    <li><strong>Compartición y Publicación:</strong> Tus datos, incluyendo los detalles técnicos recopilados de los headers de solicitud de tu dispositivo, pueden ser publicados, compartidos con terceros, o usados para análisis y propósitos de reportes.</li>
+                    <li><strong>Retención de Datos:</strong> Podemos retener tus datos, incluyendo los detalles técnicos recopilados de los headers de solicitud de tu dispositivo, por el tiempo que sea necesario.</li>
                 </ol>
-                <p>By continuing to use our app, you provide your explicit consent to the collection, use, and potential sharing of your data as described above. If you do not agree with our data collection, use, and sharing practices, please do not use our app.</p>
+                <p>Al continuar usando nuestra aplicación, proporcionas tu consentimiento explícito para la recolección, uso y potencial compartición de tus datos como se describe arriba. Si no estás de acuerdo con nuestras prácticas de recolección, uso y compartición de datos, por favor no uses nuestra aplicación.</p>
+                <p><strong>Este proyecto contribuye al desarrollo de LatamGPT, un modelo de lenguaje para América Latina.</strong></p>
             </div>
             """)
-            accept_button = gr.Button("I Agree")
+            accept_button = gr.Button("Acepto / I Agree")
 
         def enable_inputs():
             return gr.update(visible=False), gr.update(visible=True)
@@ -182,8 +146,8 @@ with gr.Blocks(css = """#col_container { margin-left: auto; margin-right: auto;}
     accept_checkbox.change(fn=enable_inputs, inputs=[], outputs=[user_consent_block, main_block], queue=False)
 
     inputs.submit(reset_textbox, [], [inputs, b1], queue=False)
-    inputs.submit(predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code, inputs, b1],)  #openai_api_key
+    inputs.submit(predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code, inputs, b1])
     b1.click(reset_textbox, [], [inputs, b1], queue=False)
-    b1.click(predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code, inputs, b1],)  #openai_api_key
-             
-    demo.queue(max_size=10, default_concurrency_limit=NUM_THREADS, api_open=False).launch(share=False)
+    b1.click(predict, [inputs, top_p, temperature, chat_counter, chatbot, state], [chatbot, state, chat_counter, server_status_code, inputs, b1])
+
+demo.queue(max_size=20, default_concurrency_limit=NUM_THREADS, api_open=False).launch()