rag_hf.py

import streamlit as st
import datetime
import pickle
import numpy as np
import rdflib
import torch
import os
import requests
from rdflib import Graph as RDFGraph, Namespace
from sentence_transformers import SentenceTransformer

# === STREAMLIT UI CONFIG ===
st.set_page_config(
    page_title="Atlas de Lenguas: Lenguas Indígenas Sudamericanas",
    page_icon="🌍",
    layout="wide",
    initial_sidebar_state="expanded",
    menu_items={
        'About': "## Análisis con IA de lenguas indígenas en peligro\n"
                 "Esta aplicación integra grafos de conocimiento de Glottolog, Wikipedia y Wikidata."
    }
)

# === CONFIGURATION ===
ENDPOINT_URL = "https://api-inference.huggingface.co/models/HuggingFaceH4/zephyr-7b-beta"
HF_API_TOKEN = os.getenv("HF_API_TOKEN")
if not HF_API_TOKEN:
    st.error("⚠️ No se cargó el token HF_API_TOKEN desde los Secrets.")
else:
    st.success("✅ Token cargado correctamente.")
EMBEDDING_MODEL = "intfloat/multilingual-e5-base"
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
EX = Namespace("http://example.org/lang/")

# === CUSTOM CSS ===
st.markdown("""
<style>
    .tech-badge {
        background-color: #ecfdf5;
        color: #065f46;
        padding: 0.25rem 0.5rem;
        border-radius: 4px;
        font-size: 0.75rem;
        font-weight: 500;
    }
</style>
""", unsafe_allow_html=True)

# === CORE FUNCTIONS ===
@st.cache_resource(show_spinner="Cargando modelos de IA y grafos de conocimiento...")
def load_all_components():
    embedder = SentenceTransformer(EMBEDDING_MODEL, device=DEVICE)
    methods = {}
    label, suffix, ttl, matrix_path = ("LinkGraph", "_hybrid_graphsage", "grafo_ttl_hibrido_graphsage.ttl", "embed_matrix_hybrid_graphsage.npy")
    with open(f"id_map{suffix}.pkl", "rb") as f:
        id_map = pickle.load(f)
    with open(f"grafo_embed{suffix}.pickle", "rb") as f:
        G = pickle.load(f)
    matrix = np.load(matrix_path)
    rdf = RDFGraph()
    rdf.parse(ttl, format="ttl")
    methods[label] = (matrix, id_map, G, rdf)
    return methods, embedder

def get_top_k(matrix, id_map, query, k, embedder):
    vec = embedder.encode(f"query: {query}", convert_to_tensor=True, device=DEVICE)
    vec = vec.cpu().numpy().astype("float32")
    sims = np.dot(matrix, vec) / (np.linalg.norm(matrix, axis=1) * np.linalg.norm(vec) + 1e-10)
    top_k_idx = np.argsort(sims)[-k:][::-1]
    return [id_map[i] for i in top_k_idx]

def get_context(G, lang_id):
    node = G.nodes.get(lang_id, {})
    lines = [f"**Lengua:** {node.get('label', lang_id)}"]
    if node.get("wikipedia_summary"):
        lines.append(f"**Wikipedia:** {node['wikipedia_summary']}")
    if node.get("wikidata_description"):
        lines.append(f"**Wikidata:** {node['wikidata_description']}")
    if node.get("wikidata_countries"):
        lines.append(f"**Países:** {node['wikidata_countries']}")
    return "\n\n".join(lines)

def query_rdf(rdf, lang_id):
    q = f"""
    PREFIX ex: <http://example.org/lang/>
    SELECT ?property ?value WHERE {{ ex:{lang_id} ?property ?value }}
    """
    try:
        return [(str(row[0]).split("/")[-1], str(row[1])) for row in rdf.query(q)]
    except Exception as e:
        return [("error", str(e))]

def query_llm(prompt):
    try:
        res = requests.post(
            ENDPOINT_URL,
            headers={"Authorization": f"Bearer {HF_API_TOKEN}", "Content-Type": "application/json"},
            json={"inputs": prompt}, timeout=60
        )
        res.raise_for_status()
        out = res.json()
        if isinstance(out, list):
            if len(out) > 0 and isinstance(out[0], dict) and "generated_text" in out[0]:
                return out[0]["generated_text"].strip()
        elif isinstance(out, dict) and "generated_text" in out:
            return out["generated_text"].strip()
        return "Sin respuesta del modelo."
    except Exception as e:
        return f"Error al consultar el modelo: {str(e)}"

def generate_response(matrix, id_map, G, rdf, user_question, k, embedder):
    ids = get_top_k(matrix, id_map, user_question, k, embedder)
    context = [get_context(G, i) for i in ids]
    rdf_facts = []
    for i in ids:
        rdf_facts.extend([f"{p}: {v}" for p, v in query_rdf(rdf, i)])

    prompt_es = (
        "Eres un experto en lenguas indígenas sudamericanas.\n"
        "Usa solo la información del contexto y hechos RDF siguientes.\n\n"
        + "### CONTEXTO:\n" + "\n".join(context) + "\n\n"
        + "### RELACIONES RDF:\n" + "\n".join(rdf_facts) + "\n\n"
        + f"### PREGUNTA:\n{user_question}\n\nRespuesta breve en español:"
    )

    prompt_en = (
        "You are an expert in South American indigenous languages.\n"
        "Use only the following context and RDF facts to answer.\n\n"
        + "### CONTEXT:\n" + "\n".join(context) + "\n\n"
        + "### RDF RELATIONS:\n" + "\n".join(rdf_facts) + "\n\n"
        + f"### QUESTION:\n{user_question}\n\nShort answer in English:"
    )

    response_es = query_llm(prompt_es)
    response_en = query_llm(prompt_en)

    full_response = (
        f"<b>Respuesta en español:</b><br>{response_es}<br><br>"
        f"<b>Answer in English:</b><br>{response_en}"
    )
    return full_response, ids, context, rdf_facts

def main():
    methods, embedder = load_all_components()
    st.title("Atlas de Lenguas: Lenguas Indígenas Sudamericanas")
    st.markdown("<span class='tech-badge'>Correo: jxvera@gmail.com</span>", unsafe_allow_html=True)
    query = st.text_input("Escribe tu pregunta sobre lenguas indígenas:")
    k = st.slider("Número de lenguas similares a recuperar", min_value=1, max_value=10, value=3)
    if st.button("Analizar"):
        method = methods["LinkGraph"]
        start = datetime.datetime.now()
        response, lang_ids, context, rdf_data = generate_response(*method, query, k, embedder)
        duration = (datetime.datetime.now() - start).total_seconds()
        st.markdown(response, unsafe_allow_html=True)
        st.caption(f"⏱️ {duration:.2f} segundos | 🌐 {len(lang_ids)} idiomas analizados")
        with st.expander("📖 Contexto"):
            for ctx in context:
                st.markdown(ctx)
        with st.expander("🔗 Hechos RDF"):
            st.code("\n".join(rdf_data))

if __name__ == "__main__":
    main()