app.py

import os
os.system('pip install curl_cffi tqdm bitsandbytes tiktoken g4f pinecone-client pandas datasets sentence-transformers')

# Setup and load your keys
import os
from g4f import ChatCompletion
#from google.colab import userdata
from pinecone import Pinecone
import pandas as pd
from datasets import Dataset
from sentence_transformers import SentenceTransformer
import gradio as gr

model_name = "BAAI/bge-m3"

# APIs personales
#PINECONE_ENVIRONMENT = us-east-1
#PINECONE_API_KEY = 3a3e9022-381d-436e-84cb-ba93464d283e

#os.environ["PINECONE_ENVIRONMENT"] = "us-east-1"
#os.environ["PINECONE_API_KEY"] = "3a3e9022-381d-436e-84cb-ba93464d283e"

# Retrieve the Pinecone API key from the user
PINECONE_API_KEY = "3a3e9022-381d-436e-84cb-ba93464d283e"  # Use the key you set in the secrets
PINECONE_ENVIRONMENT = "us-east-1"  # Use the environment you set in the secrets

# Initialize Pinecone with the API key
pc = Pinecone(api_key=PINECONE_API_KEY)

# Global variables to store the selected model and dimensions
EMBED_MODEL = 'BGE_M3-1024'
DIMENSIONS = 1024

# Confirm selection automatically
print(f"Model selected: {EMBED_MODEL}")
print(f"Dimensions set as: {DIMENSIONS}")

# Function to print current selection (can be used in other cells)
def print_current_selection():
    print(f"Currently selected model: {EMBED_MODEL}")
    print(f"Dimensions: {DIMENSIONS}")

# Establecer el nombre del índice automáticamente
INDEX_NAME = 'vestidos'

# Obtener la clave API de Pinecone
#PINECONE_API_KEY = userdata.get('PINECONE_API_KEY')

def connect_to_pinecone(index_name):
    global INDEX_NAME
    try:
        pc = Pinecone(api_key=PINECONE_API_KEY)
        index = pc.Index(index_name)

        # Asegurarse de que la conexión se establezca
        index_stats = index.describe_index_stats()
        print(f"Successfully connected to Pinecone index '{index_name}'!")
        print("Index Stats:", index_stats)

        # Actualizar la variable global INDEX_NAME
        INDEX_NAME = index_name
        print(f"Global INDEX_NAME updated to: {INDEX_NAME}")

    except Exception as e:
        print(f"Failed to connect to Pinecone index '{index_name}':", str(e))

# Conectar automáticamente al índice "vestidos"
connect_to_pinecone(INDEX_NAME)

# Función para imprimir el nombre del índice actual (puede ser usada en otras celdas)
def print_current_index():
    print(f"Current index name: {INDEX_NAME}")

# Verificar si las variables globales necesarias están configuradas
if 'INDEX_NAME' not in globals() or INDEX_NAME is None:
    raise ValueError("INDEX_NAME is not set. Please set the index name first.")

if 'EMBED_MODEL' not in globals() or EMBED_MODEL is None:
    raise ValueError("EMBED_MODEL is not set. Please select an embedding model first.")

# Inicializar cliente de Pinecone
#PINECONE_API_KEY = userdata.get('PINECONE_API_KEY')
pc = Pinecone(api_key=PINECONE_API_KEY)

# Inicializar el índice de Pinecone
index = pc.Index(INDEX_NAME)

# Obtener la dimensión del índice
index_stats = index.describe_index_stats()
vector_dim = index_stats['dimension']
print(f"Index dimension: {vector_dim}")

# Definir manualmente los campos de contexto y enlace
CONTEXT_FIELDS = ['Etiqueta', 'Pregunta 1', 'Pregunta 2', 'Pregunta 3', 'Respuesta Combinada']
LINK_FIELDS = ['Etiqueta', 'Respuesta Combinada']

# Imprimir confirmación de campos seleccionados
print(f"Context fields set to: {CONTEXT_FIELDS}")
print(f"Link fields set to: {LINK_FIELDS}")

# Función para obtener las selecciones actuales de campos (puede ser usada en otras celdas)
def get_field_selections():
    return {
        "CONTEXT_FIELDS": CONTEXT_FIELDS,
        "LINK_FIELDS": LINK_FIELDS
    }

#####################################

# Check if required global variables are set
if 'EMBED_MODEL' not in globals() or EMBED_MODEL is None:
    raise ValueError("EMBED_MODEL is not set. Please select an embedding model first.")
if 'INDEX_NAME' not in globals() or INDEX_NAME is None:
    raise ValueError("INDEX_NAME is not set. Please create or select an index first.")
if 'CONTEXT_FIELDS' not in globals() or 'LINK_FIELDS' not in globals():
    raise ValueError("CONTEXT_FIELDS and LINK_FIELDS are not set. Please run the field selection cell first.")

# Initialize the Sentence-Transformer model
embedding_model = SentenceTransformer(model_name)

# Initialize Pinecone with the API key and connect to the index
pinecone_client = Pinecone(api_key=PINECONE_API_KEY)
index = pinecone_client.Index(INDEX_NAME)

# Constants
LIMIT = 3750

def vector_search(query):
    # Generate embedding using Sentence-Transformer model
    xq = embedding_model.encode(query)

    # Perform vector search on Pinecone index
    res = index.query(vector=xq.tolist(), top_k=3, include_metadata=True)
    if res['matches']:
        return [
            {
                'content': ' '.join(f"{k}: {v}" for k, v in match['metadata'].items() if k in CONTEXT_FIELDS and k != 'Etiqueta'),
                'metadata': match['metadata']
            }
            for match in res['matches']
            if 'metadata' in match
        ]
    return []

def create_prompt(query, contexts):
    prompt_start = "\n\nContexto:\n"
    prompt_end = f"\n\nPregunta: {query}\nRespuesta:"

    current_contexts = "\n\n---\n\n".join([context['content'] for context in contexts])
    if len(prompt_start + current_contexts + prompt_end) >= LIMIT:
        # Truncate contexts if they exceed the limit
        available_space = LIMIT - len(prompt_start) - len(prompt_end)
        truncated_contexts = current_contexts[:available_space]
        return prompt_start + truncated_contexts + prompt_end
    else:
        return prompt_start + current_contexts + prompt_end

def complete(prompt):
    return [f"Hola"]

def check_image_exists(filepath):
    return os.path.exists(filepath)

def chat_function(message, history):
    # Perform vector search
    search_results = vector_search(message)

    # Create prompt with relevant contexts
    query_with_contexts = create_prompt(message, search_results)

    # Generate response
    response = complete(query_with_contexts)

    partial_message = response[0].split("\n")[0]  # Solo tomar la primera línea de la respuesta

    # Handle the logic for processing tags and images internally
    relevant_links = [result['metadata'].get(field) for result in search_results for field in LINK_FIELDS if field in result['metadata']]
    full_response = partial_message
    image_url = None
    tags_detected = []

    filtered_links = []
    if relevant_links:
        for link in relevant_links:
            if any(tag in link for tag in ["lila_61", "lila_63", "lila_62", "lila_64", "fuxia_70", "fuxia_71", "fuxia_72", "fuxia_73", "fuxia_74", "melon_68", "melon_66", "melon_67", "melon_65", "vino_19", "vino_20", "barney_69", "loro_27", "lacre_02", "amarillo_03", "amarillo_04", "azulino_11", "azulino_14", "azulino_12", "azulino_13", "beigs_09", "beigs_10", "beigs_07", "beigs_06", "beigs_08", "beigs_05", "marina_32", "marina_29", "marina_28", "marina_31", "marina_30", "rojo_26", "rojo_23", "rojo_21", "rojo_22", "rojo_25", "rojo_24", "celeste_40", "celeste_38", "celeste_39", "celeste_33", "celeste_35", "celeste_37", "celeste_41", "celeste_42", "celeste_34", "celeste_36", "sirenita_01", "marino_18", "marino_17", "marino_16", "marino_15", "rosa_87", "rosa_86", "rosa_79", "rosa_82", "rosa_83", "rosa_78", "rosa_84", "rosa_85", "rosa_75", "rosa_80", "rosa_81", "rosa_77", "rosa_76", "blanco_55", "blanco_56", "blanco_53", "blanco_52", "blanco_57", "blanco_49", "blanco_51", "blanco_60", "blanco_47", "blanco_44", "blanco_50", "blanco_48", "blanco_59", "blanco_43", "blanco_58", "blanco_46", "blanco_45", "blanco_54"]):
                tags_detected.append(link)  # Save the tag but don't display it
            else:
                filtered_links.append(link)

        # Add the first relevant link under a single "Respuestas relevantes" section
        if filtered_links:
            full_response += f".\n\nTe detallamos nuestro contenido a continuación:\n" + filtered_links[0]

        # Now handle the images based on the detected tags
        tags_to_images = {
            "lila_61": "images/lila_61.jpeg",
            "lila_63": "images/lila_63.jpeg",
            "lila_62": "images/lila_62.jpeg",
            "lila_64": "images/lila_64.jpeg",
            "fuxia_70": "images/fuxia_70.jpeg",
            "fuxia_71": "images/fuxia_71.jpeg",
            "fuxia_72": "images/fuxia_72.jpeg",
            "fuxia_73": "images/fuxia_73.jpeg",
            "fuxia_74": "images/fuxia_74.jpeg",
            "melon_68": "images/melon_68.jpeg",
            "melon_66": "images/melon_66.jpeg",
            "melon_67": "images/melon_67.jpeg",
            "melon_65": "images/melon_65.jpeg",
            "vino_19": "images/vino_19.jpeg",
            "vino_20": "images/vino_20.jpeg",
            "barney_69": "images/barney_69.jpeg",
            "loro_27": "images/loro_27.png",
            "lacre_02": "images/lacre_02.jpeg",
            "amarillo_03": "images/amarillo_03.jpeg",
            "amarillo_04": "images/amarillo_04.jpeg",
            "azulino_11": "images/azulino_11.jpeg",
            "azulino_14": "images/azulino_14.jpeg",
            "azulino_12": "images/azulino_12.jpeg",
            "azulino_13": "images/azulino_13.jpeg",
            "beigs_09": "images/beigs_09.jpeg",
            "beigs_10": "images/beigs_10.jpeg",
            "beigs_07": "images/beigs_07.jpeg",
            "beigs_06": "images/beigs_06.jpeg",
            "beigs_08": "images/beigs_08.jpeg",
            "beigs_05": "images/beigs_05.jpeg",
            "marina_32": "images/marina_32.jpeg",
            "marina_29": "images/marina_29.jpeg",
            "marina_28": "images/marina_28.jpeg",
            "marina_31": "images/marina_31.jpeg",
            "marina_30": "images/marina_30.jpeg",
            "rojo_26": "images/rojo_26.jpeg",
            "rojo_23": "images/rojo_23.jpeg",
            "rojo_21": "images/rojo_21.jpeg",
            "rojo_22": "images/rojo_22.jpeg",
            "rojo_25": "images/rojo_25.jpeg",
            "rojo_24": "images/rojo_24.jpeg",
            "celeste_40": "images/celeste_40.jpeg",
            "celeste_38": "images/celeste_38.jpeg",
            "celeste_39": "images/celeste_39.jpeg",
            "celeste_33": "images/celeste_33.jpeg",
            "celeste_35": "images/celeste_35.jpeg",
            "celeste_37": "images/celeste_37.jpeg",
            "celeste_41": "images/celeste_41.jpeg",
            "celeste_42": "images/celeste_42.jpeg",
            "celeste_34": "images/celeste_34.jpeg",
            "celeste_36": "images/celeste_36.jpeg",
            "sirenita_01": "images/sirenita_01.png",
            "marino_18": "images/marino_18.jpeg",
            "marino_17": "images/marino_17.jpeg",
            "marino_16": "images/marino_16.jpeg",
            "marino_15": "images/marino_15.jpeg",
            "rosa_87": "images/rosa_87.jpeg",
            "rosa_86": "images/rosa_86.png",
            "rosa_79": "images/rosa_79.jpeg",
            "rosa_82": "images/rosa_82.png",
            "rosa_83": "images/rosa_83.jpeg",
            "rosa_78": "images/rosa_78.jpeg",
            "rosa_84": "images/rosa_84.jpeg",
            "rosa_85": "images/rosa_85.jpeg",
            "rosa_75": "images/rosa_75.jpeg",
            "rosa_80": "images/rosa_80.png",
            "rosa_81": "images/rosa_81.png",
            "rosa_77": "images/rosa_77.jpeg",
            "rosa_76": "images/rosa_76.png",
            "blanco_55": "images/blanco_55.jpeg",
            "blanco_56": "images/blanco_56.jpeg",
            "blanco_53": "images/blanco_53.jpeg",
            "blanco_52": "images/blanco_52.jpeg",
            "blanco_57": "images/blanco_57.jpeg",
            "blanco_49": "images/blanco_49.jpeg",
            "blanco_51": "images/blanco_51.jpeg",
            "blanco_60": "images/blanco_60.jpeg",
            "blanco_47": "images/blanco_47.jpeg",
            "blanco_44": "images/blanco_44.jpeg",
            "blanco_50": "images/blanco_50.jpeg",
            "blanco_48": "images/blanco_48.jpeg",
            "blanco_59": "images/blanco_59.jpeg",
            "blanco_43": "images/blanco_43.jpeg",
            "blanco_58": "images/blanco_58.png",
            "blanco_46": "images/blanco_46.jpeg",
            "blanco_45": "images/blanco_45.jpeg",
            "blanco_54": "images/blanco_54.jpeg",
        }



        for tag in tags_detected:
            for key, path in tags_to_images.items():
                if key in tag and check_image_exists(path):
                    image_url = path
                    break

    return full_response, image_url


def update_image(image_url):
    if image_url:
        return image_url
    else:
        return None

# Gradio layout setup
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column(scale=1):
            chatbot_input = gr.Textbox(label="Tu mensaje")
            chatbot_output = gr.Chatbot(label="ChatBot")
            chatbot_history = gr.State(value=[])
            image_url = gr.State(value=None)
            submit_button = gr.Button("Enviar")
        with gr.Column(scale=1):
            image_output = gr.Image(label="Imagen asociada")

    def process_input(message, history):
        full_response, image = chat_function(message, history)
        history.append((message, full_response))
        return history, history, image

    submit_button.click(process_input, inputs=[chatbot_input, chatbot_history], outputs=[chatbot_output, chatbot_history, image_url])
    image_url.change(fn=update_image, inputs=image_url, outputs=image_output)

# Launch the interface
demo.launch(debug=True)