app.py

from sentence_transformers import SentenceTransformer
import chromadb
import pandas as pd
import gradio as gr
from datetime import datetime
import os 
from huggingface_hub import hf_hub_download, snapshot_download
import chromadb
import pandas as pd

#Código nuevo

# Download the entire dataset (gets all files/folders)
dataset_path = snapshot_download(
    repo_id="alexis07/oefa_21_26",
    repo_type="dataset"
)

# Load text data from pickle
textos = pd.read_pickle(f"{dataset_path}/oefa_21_26.pkl")
metadata = pd.read_csv(f"{dataset_path}/dfai_0421_0226.csv")
metadata = metadata.drop(columns=['Link', 'pdf_urls'])
metadata = pd.merge(textos, metadata, on="ID", how="inner")
metadata['ID'] = metadata['ID'].astype(str) + '_' + textos['Page'].astype(str)

# Set ID as index for faster lookup
metadata.set_index('ID', inplace=True)

# ===== 1. Conectar a la base de datos y cargar colecciones =====
print("Conectando a ChromaDB...")

# Load Chroma DB from the chroma_db subfolder
CHROMA_DIR = f"{dataset_path}/local_chroma_dir"
MODEL_NAME = "nomic-ai/nomic-embed-text-v2-moe"
COLLECTION_NAMES = ["oefa"]  # Example names

client = chromadb.PersistentClient(path=CHROMA_DIR)
collections = [client.get_collection(name) for name in COLLECTION_NAMES]

# ===== 2. Cargar el modelo =====
print("Cargando modelo...")
model = SentenceTransformer(MODEL_NAME, trust_remote_code=True)
print("✓ Sistema listo")

# ===== 3. Función de búsqueda y renderizado de Cards =====
def semantic_search(query: str, n_results: int = 15):
    print(f"Buscando: {query}")
    
    if not query.strip():
        return "### Por favor, introduce una consulta para comenzar.", ""
    
    query_embedding = model.encode(query, prompt_name="query").tolist()

    all_results = []
    for collection in collections:
        results = collection.query(
            query_embeddings=[query_embedding],
            n_results=n_results,
            include=["distances"]
        )
        cosine_similarities = [1 - dist for dist in results['distances'][0]]
        
        for i in range(len(results['ids'][0])):
            chroma_id = results['ids'][0][i]
            
            # Retrieve metadata from the external DataFrame using the ID
            try:
                if chroma_id in metadata.index:
                    row_metadata = metadata.loc[chroma_id]
                    
                    res = {
                        'id': chroma_id,
                        'similitud': cosine_similarities[i],
                        'enlace': row_metadata.get('Enlace', '#') if hasattr(row_metadata, 'get') else row_metadata['Enlace'],
                        'text': row_metadata.get('Text', 'N/A') if hasattr(row_metadata, 'get') else row_metadata['Text'],
                        'date': row_metadata.get('Date', 'N/A') if hasattr(row_metadata, 'get') else row_metadata['Date'],
                        'page': row_metadata.get('Page', 'N/A') if hasattr(row_metadata, 'get') else row_metadata['Page']
                    }
                    all_results.append(res)
                else:
                    print(f"Warning: ID {chroma_id} not found in metadata DataFrame")
            except Exception as e:
                print(f"Error retrieving metadata for ID {chroma_id}: {str(e)}")

    # Ordenar por similitud
    all_results = sorted(all_results, key=lambda x: x['similitud'], reverse=True)[:n_results]
    
    if not all_results:
        return "### No se encontraron resultados relevantes.", ""

    # Generar HTML de las Cards
    html_output = '<div class="results-container">'
    for item in all_results:
        similitud_pct = f"{item['similitud']*100:.1f}%"
        enlace = item.get('enlace', '#')
        fecha = item.get('date', 'N/A')
        texto = item.get('text', 'N/A').replace('\n', ' ')  # Remove newlines from text
        pagina = item.get('page', 'N/A')
        
        html_output += f"""
        <div class="legal-card">
            <div class="card-header">
                <span class="res-number">⚖️ Página: {pagina}</span>
                <span class="res-date">📅 {fecha}</span>
                <span class="res-score">🎯 Relevancia: {similitud_pct}</span>
            </div>
            <div class="card-body">
                <p class="res-summary">{texto}</p>
            </div>
            <div class="card-footer">
                <a href="{enlace}" target="_blank" class="view-link">🔗 Ver Documento Completo</a>
            </div>
        </div>
        """
    html_output += '</div>'
    
    return html_output, ""

# ===== 4. Interfaz Gradio =====

# Definición del CSS con soporte para Dark Mode
custom_css = """
    /* Light Mode (Default) */
    .results-container {
        display: flex;
        flex-direction: column;
        gap: 20px;
        padding: 10px 0;
    }
    .legal-card {
        background: white;
        border-radius: 12px;
        border: 1px solid #e5e7eb;
        box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);
        transition: transform 0.2s ease, box-shadow 0.2s ease;
        overflow: hidden;
    }
    .legal-card:hover {
        transform: translateY(-3px);
        box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.1);
        border-color: #3b82f6;
    }
    .card-header {
        background: #f8fafc;
        padding: 12px 20px;
        border-bottom: 1px solid #e5e7eb;
        display: flex;
        justify-content: space-between;
        align-items: center;
        flex-wrap: wrap;
        gap: 15px;
    }
    .res-number, .res-date {
        font-size: 14px;
        color: #334155;
    }
    .res-score {
        background: #3b82f6;
        color: #ffffff;
        padding: 4px 12px;
        border-radius: 20px;
        font-weight: 600;
        font-size: 13px;
    }
    .card-body {
        padding: 20px;
    }
    .res-summary {
        color: #334155;
        line-height: 1.7;
        font-size: 15px;
        margin: 0;
        text-align: justify;
        white-space: pre-wrap;
        background-color: #f9fafb;
        padding: 12px;
        border-left: 3px solid #3b82f6;
        border-radius: 4px;
    }
    .card-footer {
        padding: 12px 20px;
        background: #ffffff;
        border-top: 1px dashed #e5e7eb;
        text-align: right;
    }
    .view-link {
        color: #2563eb;
        text-decoration: none;
        font-weight: 600;
        font-size: 14px;
    }
    .view-link:hover {
        text-decoration: underline;
        color: #1d4ed8;
    }

    /* Dark Mode */
    @media (prefers-color-scheme: dark) {
        .legal-card {
            background: #1f2937;
            border: 1px solid #374151;
        }
        .card-header {
            background: #111827;
            border-bottom: 1px solid #374151;
        }
        .res-number, .res-date {
            color: #e5e7eb;
        }
        .res-score {
            background: #3b82f6;
            color: #ffffff;
        }
        .res-summary {
            color: #e5e7eb;
            background-color: #111827;
            border-left: 3px solid #60a5fa;
        }
        .card-footer {
            background: #1f2937;
            border-top: 1px dashed #374151;
        }
        .view-link {
            color: #60a5fa;
        }
        .view-link:hover {
            color: #93c5fd;
        }
    }
"""
               
with gr.Blocks(title="Buscador Jurisprudencia Pro", theme=gr.themes.Soft(primary_hue="blue"), css=custom_css) as demo:
    
    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("""
            # ⚖️ Buscador de Jurisprudencia Profesional
            ### Inteligencia Artificial aplicada al Derecho Ambiental
            """)
    
    with gr.Row():
        query_box = gr.Textbox(
            show_label=False, 
            placeholder="Ej: ¿Cuáles son los requisitos para la subsanación voluntaria?",
            lines=1,
            container=False
        )
    
    with gr.Row():
        # Usamos un componente HTML para mostrar las Cards
        results_display = gr.HTML(
            label="Resultados",
            value='<div style="text-align: center; color: #64748b; padding: 40px;">Los resultados aparecerán aquí...</div>'
        )
    
    # Eventos
    query_box.submit(
        semantic_search, 
        [query_box], 
        [results_display, query_box]
    ).then(lambda: "", None, query_box)

if __name__ == "__main__":
    demo.launch()