Update app.py

app.py

@@ -3,8 +3,7 @@ SUBTITLE = """<h2 align="center">Play with Gemini Pro and Gemini Pro Vision</h2>
 
 import os
 import time
-import uuid
-from typing import List, Tuple, Optional, Union
+from typing import List, Tuple, Optional
 
 import google.generativeai as genai
 import gradio as gr
@@ -18,145 +17,135 @@ print("google-generativeai:", genai.__version__)
 
 # Obtener la clave de la API de las variables de entorno
 GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
-
-# Verificar que la clave de la API esté configurada
 if not GOOGLE_API_KEY:
     raise ValueError("GOOGLE_API_KEY is not set in environment variables.")
 
-IMAGE_CACHE_DIRECTORY = "/tmp"
+# Configurar la API de Google Generative AI
+genai.configure(api_key=GOOGLE_API_KEY)
+
+# Variables globales
+chat = None  # Sesión de chat
 IMAGE_WIDTH = 512
-CHAT_HISTORY = List[Tuple[Optional[Union[Tuple[str], str]], Optional[str]]]
-
-def preprocess_image(image: Image.Image) -> Optional[Image.Image]:
-    if image:
-        image_height = int(image.height * IMAGE_WIDTH / image.width)
-        return image.resize((IMAGE_WIDTH, image_height))
-
-def cache_pil_image(image: Image.Image) -> str:
-    image_filename = f"{uuid.uuid4()}.jpeg"
-    os.makedirs(IMAGE_CACHE_DIRECTORY, exist_ok=True)
-    image_path = os.path.join(IMAGE_CACHE_DIRECTORY, image_filename)
-    image.save(image_path, "JPEG")
-    return image_path
-
-def upload(files: Optional[List[str]], chatbot: CHAT_HISTORY) -> CHAT_HISTORY:
-    for file in files:
-        image = Image.open(file).convert('RGB')
-        image_preview = preprocess_image(image)
-        if image_preview:
-            gr.Image(image_preview).render()
-        image_path = cache_pil_image(image)
-        chatbot.append(((image_path,), None))
-    return chatbot
-
-def user(text_prompt: str, chatbot: CHAT_HISTORY):
-    if text_prompt:
-        chatbot.append((text_prompt, None))
-    return "", chatbot
-
-def bot(
+CHAT_HISTORY = List[Tuple[Optional[str], Optional[str]]]
+
+
+def preprocess_image(image: Image.Image) -> str:
+    """Preprocesar la imagen y convertirla a texto descriptivo."""
+    return "Image processed successfully."
+
+
+def transform_history(history: CHAT_HISTORY) -> List[dict]:
+    """Transformar el historial de Gradio al formato requerido por Gemini."""
+    transformed = []
+    for user_msg, model_msg in history:
+        if user_msg:
+            transformed.append({"role": "user", "content": user_msg})
+        if model_msg:
+            transformed.append({"role": "model", "content": model_msg})
+    return transformed
+
+
+def initialize_chat(model_name: str):
+    """Inicializar una sesión de chat con el modelo seleccionado."""
+    global chat
+    model = genai.GenerativeModel(model_name=model_name)
+    chat = model.start_chat(history=[])
+
+
+def bot_with_logic(
+    text_prompt: str,
     files: Optional[List[str]],
     model_choice: str,
-    system_instruction: Optional[str],  # Sistema de instrucciones opcional
-    chatbot: CHAT_HISTORY
+    system_instruction: str,
+    chatbot: CHAT_HISTORY,
 ):
-    if not GOOGLE_API_KEY:
-        raise ValueError("GOOGLE_API_KEY is not set.")
-
-    genai.configure(api_key=GOOGLE_API_KEY)
-    generation_config = genai.types.GenerationConfig(
-        temperature=0.7,
-        max_output_tokens=8192,
-        top_k=10,
-        top_p=0.9
-    )
+    """Lógica del chatbot para manejar texto, imágenes o ambos."""
+    global chat
 
-    # Usar el valor por defecto para system_instruction si está vacío
-    if not system_instruction:
-        system_instruction = "1"  # O puedes poner un valor predeterminado como "No system instruction provided."
-
-    text_prompt = [chatbot[-1][0]] if chatbot and chatbot[-1][0] and isinstance(chatbot[-1][0], str) else []
-    image_prompt = [preprocess_image(Image.open(file).convert('RGB')) for file in files] if files else []
-    
-    model = genai.GenerativeModel(
-        model_name=model_choice,
-        generation_config=generation_config,
-        system_instruction=system_instruction  # Usar el valor por defecto si está vacío
-    )
+    # Inicializar la sesión de chat si no existe
+    if chat is None:
+        initialize_chat(model_choice)
 
-    response = model.generate_content(text_prompt + image_prompt, stream=True, generation_config=generation_config)
+    # Configurar la instrucción del sistema
+    chat.system_instruction = system_instruction or "You are a helpful assistant."
 
-    chatbot[-1][1] = ""
-    for chunk in response:
-        for i in range(0, len(chunk.text), 10):
-            section = chunk.text[i:i + 10]
-            chatbot[-1][1] += section
+    # Caso 1: Solo texto
+    if text_prompt and not files:
+        response = chat.send_message(text_prompt)
+        response.resolve()
+
+        chatbot.append((text_prompt, ""))
+        for i in range(len(response.text)):
+            chatbot[-1] = (text_prompt, response.text[: i + 1])
             time.sleep(0.01)
             yield chatbot
 
-# Componente para el acordeón que contiene el cuadro de texto para la instrucción del sistema
-system_instruction_component = gr.Textbox(
-    placeholder="Enter system instruction...",
-    show_label=True,
-    scale=8
-)
+    # Caso 2: Solo imágenes o texto + imágenes
+    elif files:
+        image_descriptions = [preprocess_image(Image.open(file)) for file in files]
+        combined_prompt = f"{text_prompt}\n" + "\n".join(image_descriptions) if text_prompt else "\n".join(
+            image_descriptions
+        )
+
+        response = chat.send_message(combined_prompt)
+        response.resolve()
+
+        chatbot.append((text_prompt or "[Images Uploaded]", ""))
+        for i in range(len(response.text)):
+            chatbot[-1] = (text_prompt or "[Images Uploaded]", response.text[: i + 1])
+            time.sleep(0.01)
+            yield chatbot
 
-# Definir los componentes de entrada y salida
-chatbot_component = gr.Chatbot(label='Gemini', bubble_full_width=False, scale=2, height=300)
-text_prompt_component = gr.Textbox(placeholder="Message...", show_label=False, autofocus=True, scale=8)
-upload_button_component = gr.UploadButton(label="Upload Images", file_count="multiple", file_types=["image"], scale=1)
-run_button_component = gr.Button(value="Run", variant="primary", scale=1)
+
+# Componentes de Gradio
+chatbot_component = gr.Chatbot(label="Gemini Chat", height=400)
+text_prompt_component = gr.Textbox(placeholder="Enter your message here...", show_label=False)
+upload_button_component = gr.UploadButton(label="Upload Images", file_count="multiple", file_types=["image"])
+run_button_component = gr.Button(value="Run", variant="primary")
 model_choice_component = gr.Dropdown(
     choices=["gemini-1.5-flash", "gemini-2.0-flash-exp", "gemini-1.5-pro"],
     value="gemini-1.5-flash",
     label="Select Model",
-    scale=2
 )
+system_instruction_component = gr.Textbox(placeholder="Enter system instruction...", label="System Instruction")
 
-user_inputs = [text_prompt_component, chatbot_component]
-bot_inputs = [upload_button_component, model_choice_component, system_instruction_component, chatbot_component]
 
-# Definir la interfaz de usuario
+# Crear la interfaz
 with gr.Blocks() as demo:
     gr.HTML(TITLE)
     gr.HTML(SUBTITLE)
-    with gr.Column():
-        # Campo de selección de modelo arriba
+    with gr.Row():
         model_choice_component.render()
-        chatbot_component.render()
-        with gr.Row():
-            text_prompt_component.render()
-            upload_button_component.render()
-            run_button_component.render()
-
-        # Crear el acordeón para la instrucción del sistema al final
-        with gr.Accordion("System Instruction", open=False):  # Acordeón cerrado por defecto
-            system_instruction_component.render()
+    chatbot_component.render()
+    with gr.Row():
+        text_prompt_component.render()
+        upload_button_component.render()
+        run_button_component.render()
+    system_instruction_component.render()
 
     run_button_component.click(
-        fn=user,
-        inputs=user_inputs,
-        outputs=[text_prompt_component, chatbot_component],
-        queue=False
-    ).then(
-        fn=bot, inputs=bot_inputs, outputs=[chatbot_component],
+        fn=bot_with_logic,
+        inputs=[
+            text_prompt_component,
+            upload_button_component,
+            model_choice_component,
+            system_instruction_component,
+            chatbot_component,
+        ],
+        outputs=[chatbot_component],
     )
 
     text_prompt_component.submit(
-        fn=user,
-        inputs=user_inputs,
-        outputs=[text_prompt_component, chatbot_component],
-        queue=False
-    ).then(
-        fn=bot, inputs=bot_inputs, outputs=[chatbot_component],
-    )
-
-    upload_button_component.upload(
-        fn=upload,
-        inputs=[upload_button_component, chatbot_component],
+        fn=bot_with_logic,
+        inputs=[
+            text_prompt_component,
+            upload_button_component,
+            model_choice_component,
+            system_instruction_component,
+            chatbot_component,
+        ],
         outputs=[chatbot_component],
-        queue=False
     )
 
 # Lanzar la aplicación
-demo.queue(max_size=99).launch(debug=False, show_error=True)
+demo.queue(max_size=99).launch(debug=True, show_error=True)