app.py

# app.py

import streamlit as st
from transformers import pipeline, AutoTokenizer
import pandas as pd
import sqlite3
import os

# --- CONFIGURATION
st.set_page_config(page_title="Détection de Fake News", layout="wide")
MODEL = "jy46604790/Fake-News-Bert-Detect"
DB_FILE = "historique.db"

# --- Chargement modèle et tokenizer
@st.cache_resource
def load_pipeline():
    return pipeline("text-classification", model=MODEL, tokenizer=MODEL), AutoTokenizer.from_pretrained(MODEL)

clf, tokenizer = load_pipeline()

# --- Initialisation base de données
def init_db():
    if not os.path.exists(DB_FILE):
        conn = sqlite3.connect(DB_FILE)
        c = conn.cursor()
        c.execute('''
            CREATE TABLE IF NOT EXISTS historique (
                id INTEGER PRIMARY KEY AUTOINCREMENT,
                texte TEXT,
                prediction TEXT,
                score REAL
            )
        ''')
        conn.commit()
        conn.close()

init_db()

# --- Fonctions principales
def detect(text):
    tokens = tokenizer.encode(text, truncation=True, max_length=512, add_special_tokens=True)
    text = tokenizer.decode(tokens, skip_special_tokens=True)
    result = clf(text)[0]
    label = "Fake news" if result["label"] == "LABEL_0" else "Real news"
    return label, result["score"]

def save_to_db(texte, prediction, score):
    conn = sqlite3.connect(DB_FILE)
    c = conn.cursor()
    c.execute('INSERT INTO historique (texte, prediction, score) VALUES (?, ?, ?)', (texte, prediction, score))
    conn.commit()
    conn.close()

def load_history():
    conn = sqlite3.connect(DB_FILE)
    df = pd.read_sql_query('SELECT * FROM historique ORDER BY id DESC', conn)
    conn.close()
    return df

# --- Interface principale
st.title("Détection de Fake News")

# Barre latérale : historique
with st.sidebar:
    st.header("Historique des analyses")
    history = load_history()
    if not history.empty:
        options = [f"{row['id']}: {row['prediction']} (Score: {row['score']:.2f})" for idx, row in history.iterrows()]
        selection = st.selectbox("Sélectionner une analyse :", options)
        selected_id = int(selection.split(":")[0])
        selected_row = history[history['id'] == selected_id].iloc[0]
        
        st.subheader("Détail de l'analyse")
        color = "red" if selected_row["prediction"] == "Fake news" else "green"
        st.markdown(f"<h4 style='color:{color};'>{selected_row['prediction']} (Confiance: {selected_row['score']:.2f})</h4>", unsafe_allow_html=True)
        st.markdown(f"<div style='border:1px solid #ccc; padding:10px; border-radius:5px;'>{selected_row['texte']}</div>", unsafe_allow_html=True)
    else:
        st.info("Aucune analyse enregistrée.")

# Champ d'analyse
text = st.text_area("Entrez un nouveau texte à analyser :", height=150)

if st.button("Analyser") and text.strip():
    label, score = detect(text)
    color = "red" if label == "Fake news" else "green"
    st.markdown(f"<h3 style='color:{color};'>{label} (Confiance: {score:.2f})</h3>", unsafe_allow_html=True)
    save_to_db(text, label, score)