Upload app.py

app.py

@@ -0,0 +1,678 @@
+import os
+import io
+import base64
+import json
+import requests
+import tempfile
+from PIL import Image
+import gradio as gr
+from huggingface_hub import InferenceClient
+from openai import OpenAI
+
+# ============================================================
+#   CONFIGURACIÓN
+# ============================================================
+
+# SambaNova
+SAMBA_API_KEY = os.getenv("REVE_API_KEY")
+SAMBA_BASE_URL = "https://api.sambanova.ai/v1"
+
+# OpenRouter
+OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")
+if not OPENROUTER_API_KEY:
+    raise ValueError("Falta la variable OPENROUTER_API_KEY.")
+
+openrouter_client = OpenAI(
+    base_url="https://openrouter.ai/api/v1",
+    api_key=OPENROUTER_API_KEY
+)
+
+# Hugging Face (para generación de imágenes)
+HF_TOKEN = os.getenv("HF_TOKEN")
+
+DEBUG = False  # Ponlo en True si quieres ver el payload enviado a SambaNova
+
+# ============================================================
+#   TODOS LOS MODELOS (SAMBA + OPENROUTER) CON ROLES
+# ============================================================
+
+MODELS = {
+    # ===================== SAMBANOVA =====================
+    "general_fast": {
+        "provider": "sambanova",
+        "name": "Meta-Llama-3.1-8B-Instruct",
+        "role": "🔄 Respuestas rápidas y generales. Conversación ligera y eficiente.",
+        "supports_images": False
+    },
+    "general_smart": {
+        "provider": "sambanova",
+        "name": "Meta-Llama-3.3-70B-Instruct",
+        "role": "🧠 Razonamiento profundo. Análisis detallado y avanzado.",
+        "supports_images": False
+    },
+    "coding_expert": {
+        "provider": "sambanova",
+        "name": "DeepSeek-V3.1",
+        "role": "💻 Programación y debugging. Ideal para desarrollo.",
+        "supports_images": False
+    },
+    "coding_alt": {
+        "provider": "sambanova",
+        "name": "DeepSeek-V3-0324",
+        "role": "⚡ Código rápido. Alternativa veloz.",
+        "supports_images": False
+    },
+    "massive_brain": {
+        "provider": "sambanova",
+        "name": "gpt-oss-120b",
+        "role": "🏛️ Sabiduría masiva. Problemas pesados y complejos.",
+        "supports_images": False
+    },
+    "specialized_1": {
+        "provider": "sambanova",
+        "name": "DeepSeek-V3.1-Terminus",
+        "role": "🎯 Especialista técnico. Tareas científicas y avanzadas.",
+        "supports_images": False
+    },
+    "specialized_2": {
+        "provider": "sambanova",
+        "name": "Llama-3.3-Swallow-70B-Instruct-v0.4",
+        "role": "🔥 Sin censura. Modelo sin restricciones.",
+        "supports_images": False
+    },
+    "multilingual": {
+        "provider": "sambanova",
+        "name": "Qwen3-32B",
+        "role": "🌍 Multilingüe. Manejo de múltiples idiomas.",
+        "supports_images": False
+    },
+    "vision_expert": {
+        "provider": "sambanova",
+        "name": "Llama-4-Maverick-17B-128E-Instruct",
+        "role": "👁️ Visión avanzada. Análisis de imágenes.",
+        "supports_images": True
+    },
+    "vision_light": {
+        "provider": "sambanova",
+        "name": "Llama-3.2-11B-Vision-Instruct",
+        "role": "👁️ Visión ligera. Modelo de visión eficiente y rápido.",
+        "supports_images": True
+    },
+    "boudoir_specialist": {
+        "provider": "sambanova",
+        "name": "ALLaM-7B-Instruct-preview",
+        "role": "🎭 Especialista en Fotografía Íntima Profesional. Experto en prompts para fotografía boudoir.",
+        "supports_images": False,
+        "specialties": [
+            "Fotografía Boudoir",
+            "Desnudo Artístico",
+            "Moda Sensual",
+            "Lencería y moda íntima"
+        ],
+        "technical_expertise": [
+            "Iluminación suave",
+            "Composición elegante",
+            "Dirección de poses",
+            "Edición fina",
+            "Escenografía íntima"
+        ],
+        "ethical_principles": [
+            "Consentimiento explícito",
+            "Positividad corporal"
+        ]
+    },
+
+    # ===================== OPENROUTER =====================
+    # META LLAMA
+    "llama_3.1_70b": {
+        "provider": "openrouter",
+        "name": "meta-llama/llama-3.1-70b-instruct",
+        "role": "Respondes con precisión técnica y claridad.",
+        "supports_images": False
+    },
+    "llama_3.1_405b": {
+        "provider": "openrouter",
+        "name": "meta-llama/llama-3.1-405b-instruct",
+        "role": "Eres experto en programación, ciencia y análisis avanzado.",
+        "supports_images": False
+    },
+
+    # LLAMA 3.2 VISION
+    "llama_3.2_11b_vision": {
+        "provider": "openrouter",
+        "name": "meta-llama/llama-3.2-11b-vision-instruct",
+        "role": "Eres un modelo experto en análisis visual detallado.",
+        "supports_images": True
+    },
+    "llama_3.2_90b_vision": {
+        "provider": "openrouter",
+        "name": "meta-llama/llama-3.2-90b-vision-instruct",
+        "role": "Eres un analista visual avanzado altamente preciso.",
+        "supports_images": True
+    },
+
+    # QWEN
+    "qwen_72b": {
+        "provider": "openrouter",
+        "name": "qwen/qwen2.5-72b-instruct",
+        "role": "Respondes de forma profesional, directa y clara.",
+        "supports_images": False
+    },
+    "qwen_110b": {
+        "provider": "openrouter",
+        "name": "qwen/qwen2.5-110b-instruct",
+        "role": "Asistente experto en razonamiento estructurado.",
+        "supports_images": False
+    },
+
+    # GPT / OPENAI
+    "gpt_4.1": {
+        "provider": "openrouter",
+        "name": "openai/gpt-4.1",
+        "role": "Asistente avanzado para cualquier tarea general.",
+        "supports_images": False
+    },
+    "gpt_4.1_mini": {
+        "provider": "openrouter",
+        "name": "openai/gpt-4.1-mini",
+        "role": "Modelo rápido y eficiente, ideal para respuestas concisas.",
+        "supports_images": False
+    },
+    "gpt_4o_mini": {
+        "provider": "openrouter",
+        "name": "openai/gpt-4o-mini",
+        "role": "Asistente veloz con buena comprensión general.",
+        "supports_images": False
+    },
+
+    # CLAUDE
+    "claude_3.5_sonnet": {
+        "provider": "openrouter",
+        "name": "anthropic/claude-3.5-sonnet",
+        "role": "Especialista en redacción, precisión y análisis profundo.",
+        "supports_images": False
+    },
+    "claude_3.5_haiku": {
+        "provider": "openrouter",
+        "name": "anthropic/claude-3.5-haiku",
+        "role": "Modelo rápido con buena comprensión general.",
+        "supports_images": False
+    },
+    "claude_3_opus": {
+        "provider": "openrouter",
+        "name": "anthropic/claude-3-opus",
+        "role": "Máxima capacidad de análisis y lenguaje.",
+        "supports_images": False
+    },
+
+    # GOOGLE GEMINI
+    "gemini_flash": {
+        "provider": "openrouter",
+        "name": "google/gemini-flash-1.5",
+        "role": "Especialista en escenarios visuales y respuestas rápidas.",
+        "supports_images": True
+    },
+    "gemini_pro": {
+        "provider": "openrouter",
+        "name": "google/gemini-pro-1.5",
+        "role": "Razonador general robusto y flexible.",
+        "supports_images": True
+    },
+    "gemini_thinking": {
+        "provider": "openrouter",
+        "name": "google/gemini-1.5-thinking",
+        "role": "Modelo de razonamiento profundo y detallado.",
+        "supports_images": True
+    },
+
+    # DEEPSEEK
+    "deepseek_r1": {
+        "provider": "openrouter",
+        "name": "deepseek/deepseek-r1",
+        "role": "Razonamiento profundo y cadena de pensamiento estructurada.",
+        "supports_images": False
+    },
+
+    # MISTRAL
+    "mistral_large": {
+        "provider": "openrouter",
+        "name": "mistral/mistral-large-latest",
+        "role": "Asistente técnico avanzado y preciso.",
+        "supports_images": False
+    },
+    "mixtral_8x7b": {
+        "provider": "openrouter",
+        "name": "mistral/mixtral-8x7b-instruct",
+        "role": "Modelo eficiente para tareas complejas sin alto costo.",
+        "supports_images": False
+    },
+
+    # REKA
+    "reka_core": {
+        "provider": "openrouter",
+        "name": "reka/core",
+        "role": "Asistente racional y estructurado.",
+        "supports_images": False
+    },
+
+    # SAMBANOVA EN OPENROUTER
+    "samba_allam_7b": {
+        "provider": "openrouter",
+        "name": "sambanova/ALLAM-1-7B",
+        "role": "Asistente optimizado para rendimiento y claridad.",
+        "supports_images": False
+    },
+
+    # FLUX
+    "flux_pro": {
+        "provider": "openrouter",
+        "name": "black-forest-labs/flux-1.1-pro",
+        "role": "Experto en generación y análisis de imágenes.",
+        "supports_images": True
+    }
+}
+
+# ============================================================
+#   HELPERS
+# ============================================================
+
+def encode_image_to_base64(image):
+    if image is None:
+        return None
+    buf = io.BytesIO()
+    # Usar JPEG para mejor compatibilidad
+    if image.mode in ('RGBA', 'LA', 'P'):
+        # Convertir imágenes con alpha channel a RGB
+        background = Image.new('RGB', image.size, (255, 255, 255))
+        if image.mode == 'P':
+            image = image.convert('RGBA')
+        background.paste(image, mask=image.split()[-1] if image.mode == 'RGBA' else None)
+        image = background
+    image.save(buf, format="JPEG", quality=95)
+    return base64.b64encode(buf.getvalue()).decode("utf-8")
+
+def build_messages(system_prompt, user_input, history, image_b64, supports_images):
+    messages = []
+
+    if system_prompt:
+        messages.append({"role": "system", "content": system_prompt})
+
+    # Procesar historial de Gradio
+    for entry in history:
+        if isinstance(entry, (list, tuple)) and len(entry) == 2:
+            user_msg, assistant_msg = entry
+            
+            # Solo agregar mensajes no vacíos
+            if user_msg and str(user_msg).strip():
+                messages.append({"role": "user", "content": str(user_msg).strip()})
+            
+            if assistant_msg and str(assistant_msg).strip():
+                messages.append({"role": "assistant", "content": str(assistant_msg).strip()})
+
+    # Manejar mensaje actual con imagen
+    current_content = []
+    
+    # Agregar texto si existe
+    if user_input and str(user_input).strip():
+        current_content.append({"type": "text", "text": str(user_input).strip()})
+    
+    # Agregar imagen si existe y es compatible
+    if image_b64 and supports_images:
+        current_content.append({
+            "type": "image_url", 
+            "image_url": {"url": f"data:image/jpeg;base64,{image_b64}"}
+        })
+    
+    # Solo agregar el mensaje si hay contenido
+    if current_content:
+        # Si solo hay texto, usar formato simple
+        if len(current_content) == 1 and current_content[0]["type"] == "text":
+            messages.append({"role": "user", "content": current_content[0]["text"]})
+        else:
+            messages.append({"role": "user", "content": current_content})
+    elif not user_input and not image_b64:
+        # Si no hay contenido, agregar mensaje vacío para mantener la conversación
+        messages.append({"role": "user", "content": ""})
+
+    return messages
+
+# ============================================================
+#   LLAMADAS A LOS MODELOS
+# ============================================================
+
+def call_sambanova(model_name, messages, temperature=0.7, top_p=1.0):
+    payload = {
+        "model": model_name,
+        "messages": messages,
+        "stream": False,
+        "temperature": temperature,
+        "top_p": top_p
+    }
+
+    if DEBUG:
+        print("=== DEBUG SAMBANOVA PAYLOAD ===")
+        print(json.dumps(payload, indent=2, ensure_ascii=False))
+        print("=== END DEBUG ===")
+
+    headers = {
+        "Authorization": f"Bearer {SAMBA_API_KEY}",
+        "Content-Type": "application/json"
+    }
+
+    try:
+        r = requests.post(
+            f"{SAMBA_BASE_URL}/chat/completions",
+            json=payload,
+            headers=headers,
+            timeout=60
+        )
+        r.raise_for_status()
+        
+        data = r.json()
+        return data["choices"][0]["message"]["content"]
+    except requests.exceptions.RequestException as e:
+        return f"Error en la conexión con SambaNova: {str(e)}"
+    except Exception as e:
+        return f"Error procesando respuesta de SambaNova: {str(e)}"
+
+def call_openrouter(model_name, messages, temperature=0.7, top_p=1.0):
+    try:
+        response = openrouter_client.chat.completions.create(
+            model=model_name,
+            messages=messages,
+            temperature=temperature,
+            top_p=top_p
+        )
+        return response.choices[0].message.content
+    except Exception as e:
+        return f"Error en OpenRouter: {str(e)}"
+
+# ============================================================
+#   LÓGICA DEL CHAT
+# ============================================================
+
+def chat_logic(user_text, user_image, model_key, history, temperature=0.7, top_p=1.0):
+    if history is None:
+        history = []
+
+    if model_key not in MODELS:
+        reply = "Error: modelo no encontrado."
+        history.append((user_text or "", reply))
+        return history, history
+
+    try:
+        model_cfg = MODELS[model_key]
+        image_b64 = encode_image_to_base64(user_image) if user_image else None
+
+        messages = build_messages(
+            system_prompt=model_cfg["role"],
+            user_input=user_text,
+            history=history,
+            image_b64=image_b64,
+            supports_images=model_cfg.get("supports_images", False)
+        )
+
+        if DEBUG:
+            print("=== FINAL MESSAGES ===")
+            for i, msg in enumerate(messages):
+                print(f"{i}: {msg['role']} - {type(msg['content'])}")
+                if isinstance(msg['content'], list):
+                    for item in msg['content']:
+                        print(f"   - {item['type']}")
+            print("=== END MESSAGES ===")
+
+        if model_cfg["provider"] == "sambanova":
+            reply = call_sambanova(model_cfg["name"], messages, temperature, top_p)
+        else:
+            reply = call_openrouter(model_cfg["name"], messages, temperature, top_p)
+
+        # Usar texto vacío si no hay entrada del usuario
+        display_text = user_text or ("[Imagen]" if user_image else "")
+        history.append((display_text, reply))
+        
+    except Exception as e:
+        error_msg = f"Error: {str(e)}"
+        history.append((user_text or "", error_msg))
+
+    return history, history
+
+# ============================================================
+#   GENERACIÓN DE IMÁGENES (HUGGING FACE, OPCIONAL)
+# ============================================================
+
+def generate_image_hf(prompt):
+    if not HF_TOKEN:
+        return None, "❌ Falta HF_TOKEN", gr.update(visible=False)
+
+    try:
+        client = InferenceClient(token=HF_TOKEN)
+        img = client.text_to_image(
+            prompt,
+            model="stabilityai/stable-diffusion-xl-base-1.0"
+        )
+
+        with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp:
+            img.save(tmp, format="PNG")
+            return img, "✅ Imagen generada", gr.update(value=tmp.name, visible=True)
+
+    except Exception as e:
+        return None, f"❌ Error: {e}", gr.update(visible=False)
+
+# ============================================================
+#   UI GRADIO UNIFICADA
+# ============================================================
+
+def create_ui():
+    with gr.Blocks(theme=gr.themes.Soft(), title="METASAMBA") as demo:
+        # Título principal
+        gr.Markdown("# 🚀 METASAMBA")
+        gr.Markdown("### Plataforma Multimodelo de Inteligencia Artificial")
+        
+        with gr.Row():
+            # Panel de configuración izquierdo
+            with gr.Column(scale=1):
+                with gr.Accordion("⚙️ CONFIGURACIÓN DEL MODELO", open=True):
+                    model_sel = gr.Dropdown(
+                        choices=list(MODELS.keys()),
+                        value="general_fast",
+                        label="Seleccionar Modelo",
+                        info="Elige el modelo que quieres usar"
+                    )
+                    
+                    # Mostrar detalles del modelo seleccionado
+                    model_info = gr.Markdown("")
+                    
+                    with gr.Row():
+                        temperature = gr.Slider(
+                            minimum=0.1,
+                            maximum=2.0,
+                            value=0.7,
+                            step=0.1,
+                            label="Temperatura",
+                            info="Controla la aleatoriedad (0.1=más determinista, 2.0=más creativo)"
+                        )
+                        
+                        top_p = gr.Slider(
+                            minimum=0.1,
+                            maximum=1.0,
+                            value=1.0,
+                            step=0.1,
+                            label="Top-p",
+                            info="Controla la diversidad del vocabulario"
+                        )
+                
+                with gr.Accordion("📊 INFORMACIÓN DEL MODELO", open=False):
+                    gr.Markdown("""
+                    ### Categorías de Modelos
+                    
+                    **SambaNova:**
+                    - 🚀 **Rápidos:** general_fast, coding_alt
+                    - 🧠 **Inteligentes:** general_smart, massive_brain
+                    - 💻 **Programación:** coding_expert, specialized_1
+                    - 👁️ **Visión:** vision_expert, vision_light
+                    - 🌍 **Multilingüe:** multilingual
+                    - 🎭 **Especializados:** boudoir_specialist
+                    
+                    **OpenRouter:**
+                    - 🦙 **Llama:** Variantes de 70B a 405B
+                    - 🤖 **GPT:** GPT-4.1 y variantes
+                    - 👻 **Claude:** Claude 3.5 Sonnet/Haiku/Opus
+                    - 🔷 **Gemini:** Flash, Pro, Thinking
+                    - 🏔️ **Otros:** DeepSeek, Mistral, Reka
+                    """)
+                    
+                with gr.Accordion("📎 ARCHIVOS ADJUNTOS", open=False):
+                    gr.Markdown("""
+                    ### Formatos soportados:
+                    - 📷 **Imágenes:** JPG, PNG, WebP
+                    - 📄 **Texto:** TXT, PDF, DOCX (próximamente)
+                    - 🎥 **Multimedia:** MP3, MP4 (próximamente)
+                    """)
+            
+            # Panel de chat principal
+            with gr.Column(scale=2):
+                chat = gr.Chatbot(
+                    height=500, 
+                    label="Conversación",
+                    show_copy_button=True,
+                    avatar_images=(None, "🤖")
+                )
+                
+                with gr.Row():
+                    img = gr.Image(
+                        type="pil", 
+                        label="📎 Adjuntar Imagen",
+                        height=150,
+                        show_label=True
+                    )
+                
+                with gr.Row():
+                    txt = gr.Textbox(
+                        label="✏️ Tu mensaje", 
+                        placeholder="Escribe tu mensaje aquí...", 
+                        lines=4,
+                        scale=5,
+                        show_label=True
+                    )
+                
+                with gr.Row():
+                    clear_btn = gr.Button("🧹 Limpiar Chat", variant="secondary", size="sm")
+                    attach_btn = gr.Button("📎 Adjuntar Archivo", variant="secondary", size="sm")
+                    send = gr.Button("🚀 Enviar", variant="primary", size="sm")
+
+        # Pestaña de generación de imágenes
+        with gr.Tab("🎨 GENERADOR DE IMÁGENES"):
+            gr.Markdown("### Generación de imágenes con Stable Diffusion XL")
+            with gr.Row():
+                with gr.Column(scale=2):
+                    p = gr.Textbox(
+                        label="Prompt para la imagen",
+                        placeholder="Describe la imagen que quieres generar...",
+                        lines=3
+                    )
+                    generate_btn = gr.Button("🖼️ Generar Imagen", variant="primary")
+                
+                with gr.Column(scale=3):
+                    out = gr.Image(label="Imagen generada", height=400, show_label=True)
+                    status = gr.Textbox(label="Estado", interactive=False)
+                    d = gr.DownloadButton("📥 Descargar", visible=False)
+        
+        # Funciones para actualizar información del modelo
+        def update_model_info(model_key):
+            if model_key in MODELS:
+                model = MODELS[model_key]
+                info = f"""
+                ### **{model['name']}**
+                
+                **Proveedor:** {'SambaNova' if model['provider'] == 'sambanova' else 'OpenRouter'}
+                
+                **Rol:** {model['role']}
+                
+                **Soporte de imágenes:** {'✅ Sí' if model.get('supports_images', False) else '❌ No'}
+                """
+                
+                # Agregar información especializada para boudoir_specialist
+                if model_key == "boudoir_specialist":
+                    info += "\n**Especialidades:**\n"
+                    for specialty in model.get('specialties', []):
+                        info += f"- {specialty}\n"
+                    
+                    info += "\n**Expertise técnico:**\n"
+                    for expertise in model.get('technical_expertise', []):
+                        info += f"- {expertise}\n"
+                    
+                    info += "\n**Principios éticos:**\n"
+                    for principle in model.get('ethical_principles', []):
+                        info += f"- {principle}\n"
+                
+                return info
+            return "Selecciona un modelo para ver información detallada."
+        
+        # Conectar eventos
+        model_sel.change(
+            update_model_info,
+            inputs=[model_sel],
+            outputs=[model_info]
+        )
+        
+        send.click(
+            chat_logic,
+            inputs=[txt, img, model_sel, chat, temperature, top_p],
+            outputs=[chat, chat]
+        ).then(
+            lambda: ("", None),  # Limpiar inputs después de enviar
+            outputs=[txt, img]
+        )
+        
+        txt.submit(
+            chat_logic,
+            inputs=[txt, img, model_sel, chat, temperature, top_p],
+            outputs=[chat, chat]
+        ).then(
+            lambda: ("", None),
+            outputs=[txt, img]
+        )
+        
+        clear_btn.click(
+            lambda: ([], []),
+            outputs=[chat, chat]
+        )
+        
+        # Función para adjuntar archivo (placeholder)
+        def attach_file():
+            return "Funcionalidad de adjuntar archivo en desarrollo"
+        
+        attach_btn.click(
+            attach_file,
+            outputs=[txt]
+        )
+        
+        generate_btn.click(
+            generate_image_hf,
+            inputs=[p],
+            outputs=[out, status, d]
+        )
+        
+        # Inicializar información del modelo
+        demo.load(
+            update_model_info,
+            inputs=[model_sel],
+            outputs=[model_info]
+        )
+
+    return demo
+
+# ============================================================
+#   EJECUCIÓN
+# ============================================================
+
+demo = create_ui()
+
+if __name__ == "__main__":
+    demo.launch(
+        share=False,
+        show_error=True,
+        debug=False,
+        server_name="0.0.0.0",
+        server_port=7860
+    )