Spaces:
Running
Running
import gradio as gr | |
from transformers import pipeline | |
from langdetect import detect | |
from huggingface_hub import InferenceClient | |
import pandas as pd | |
import os | |
import asyncio | |
import nltk | |
from nltk.tokenize import sent_tokenize | |
# Téléchargement de punkt_tab avec gestion d'erreur | |
try: | |
nltk.download('punkt_tab', download_dir='/usr/local/share/nltk_data') | |
except Exception as e: | |
raise Exception(f"Erreur lors du téléchargement de punkt_tab : {str(e)}. Veuillez vérifier votre connexion réseau et les permissions du répertoire /usr/local/share/nltk_data.") | |
HF_TOKEN = os.getenv("HF_TOKEN") | |
# Fonction pour appeler l'API Zephyr avec des paramètres ajustés | |
async def call_zephyr_api(prompt, mode, hf_token=HF_TOKEN): | |
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=hf_token) | |
try: | |
if mode == "Rapide": | |
max_new_tokens = 50 | |
temperature = 0.3 | |
elif mode == "Équilibré": | |
max_new_tokens = 100 | |
temperature = 0.5 | |
else: # Précis | |
max_new_tokens = 150 | |
temperature = 0.7 | |
response = await asyncio.to_thread(client.text_generation, prompt, max_new_tokens=max_new_tokens, temperature=temperature) | |
return response | |
except Exception as e: | |
raise gr.Error(f"❌ Erreur d'appel API Hugging Face : {str(e)}") | |
# Chargement du modèle de sentiment pour analyser les réponses | |
classifier = pipeline("sentiment-analysis", model="mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis") | |
# Modèles de traduction | |
translator_to_en = pipeline("translation", model="Helsinki-NLP/opus-mt-mul-en") | |
translator_to_fr = pipeline("translation", model="Helsinki-NLP/opus-mt-en-fr") | |
# Traduction en français avec Helsinki-NLP | |
def safe_translate_to_fr(text, max_length=512): | |
try: | |
sentences = sent_tokenize(text) | |
translated_sentences = [] | |
for sentence in sentences: | |
translated = translator_to_fr(sentence, max_length=max_length)[0]['translation_text'] | |
translated_sentences.append(translated) | |
return " ".join(translated_sentences) | |
except Exception as e: | |
return f"Erreur de traduction : {str(e)}" | |
# Fonction pour suggérer le meilleur modèle | |
def suggest_model(text): | |
word_count = len(text.split()) | |
if word_count < 50: | |
return "Rapide" | |
elif word_count <= 200: | |
return "Équilibré" | |
else: | |
return "Précis" | |
# Fonction pour créer une jauge de sentiment | |
def create_sentiment_gauge(sentiment, score): | |
score_percentage = score * 100 | |
color = "#A9A9A9" | |
if sentiment.lower() == "positive": | |
color = "#2E8B57" | |
elif sentiment.lower() == "negative": | |
color = "#DC143C" | |
html = f""" | |
<div style='width: 100%; max-width: 300px; margin: 10px 0;'> | |
<div style='background-color: #D3D3D3; border-radius: 5px; height: 20px; position: relative;'> | |
<div style='background-color: {color}; width: {score_percentage}%; height: 100%; border-radius: 5px;'></div> | |
<span style='position: absolute; top: 0; left: 50%; transform: translateX(-50%); font-weight: bold;'>{score_percentage:.1f}%</span> | |
</div> | |
<div style='text-align: center; margin-top: 5px;'>Sentiment : {sentiment}</div> | |
</div> | |
""" | |
return html | |
# Fonction d'analyse | |
async def full_analysis(text, mode, detail_mode, count, history): | |
if not text: | |
yield "Entrez une phrase.", "", "", "", 0, history, "", "Aucune analyse effectuée." | |
return | |
yield "Analyse en cours... (Étape 1 : Détection de la langue)", "", "", "", count, history, "", "Détection de la langue" | |
try: | |
lang = detect(text) | |
except: | |
lang = "unknown" | |
if lang != "en": | |
text_en = translator_to_en(text, max_length=512)[0]['translation_text'] | |
else: | |
text_en = text | |
yield "Analyse en cours... (Étape 2 : Analyse du sentiment)", "", "", "", count, history, "", "Analyse du sentiment" | |
result = await asyncio.to_thread(classifier, text_en) | |
result = result[0] | |
sentiment_output = f"Sentiment prédictif : {result['label']} (Score: {result['score']:.2f})" | |
sentiment_gauge = create_sentiment_gauge(result['label'], result['score']) | |
yield "Analyse en cours... (Étape 3 : Explication IA)", "", "", "", count, history, "", "Génération de l'explication" | |
explanation_prompt = f"""<|system|> | |
You are a professional financial analyst AI with expertise in economic forecasting. | |
</s> | |
<|user|> | |
Given the following question about a potential economic event: "{text}" | |
The predicted sentiment for this event is: {result['label'].lower()}. | |
Assume the event happens. Explain why this event would likely have a {result['label'].lower()} economic impact. | |
</s> | |
<|assistant|>""" | |
explanation_en = await call_zephyr_api(explanation_prompt, mode) | |
yield "Analyse en cours... (Étape 4 : Traduction en français)", "", "", "", count, history, "", "Traduction en français" | |
explanation_fr = safe_translate_to_fr(explanation_en) | |
count += 1 | |
history.append({ | |
"Texte": text, | |
"Sentiment": result['label'], | |
"Score": f"{result['score']:.2f}", | |
"Explication_EN": explanation_en, | |
"Explication_FR": explanation_fr | |
}) | |
yield sentiment_output, text, explanation_en, explanation_fr, count, history, sentiment_gauge, "✅ Analyse terminée." | |
# Historique CSV | |
def download_history(history): | |
if not history: | |
return None | |
df = pd.DataFrame(history) | |
file_path = "/tmp/analysis_history.csv" | |
df.to_csv(file_path, index=False) | |
return file_path | |
# Lancement Gradio avec l'interface restaurée | |
def launch_app(): | |
custom_css = """ | |
/* CSS restauré à la version précédente, avant les changements esthétiques non demandés */ | |
body { | |
background: linear-gradient(135deg, #0A1D37 0%, #1A3C34 100%); | |
font-family: 'Inter', sans-serif; | |
color: #E0E0E0; | |
padding: 20px; | |
} | |
.gr-box { | |
background: #2A4A43 !important; | |
border: 1px solid #FFD700 !important; | |
border-radius: 12px !important; | |
padding: 20px !important; | |
box-shadow: 0px 4px 12px rgba(255, 215, 0, 0.4); | |
} | |
.gr-button { | |
background: linear-gradient(90deg, #FFD700, #D4AF37); | |
color: #0A1D37; | |
font-weight: bold; | |
border: none; | |
border-radius: 8px; | |
padding: 12px 24px; | |
transition: transform 0.2s; | |
} | |
.gr-button:hover { | |
transform: translateY(-2px); | |
box-shadow: 0 6px 12px rgba(255, 215, 0, 0.5); | |
} | |
""" | |
with gr.Blocks(theme=gr.themes.Base(), css=custom_css) as iface: | |
gr.Markdown("# 📈 Analyse Financière Premium avec IA") | |
gr.Markdown("**Posez une question économique.** L'IA analyse et explique l'impact.") | |
count = gr.State(0) | |
history = gr.State([]) | |
with gr.Row(): | |
with gr.Column(scale=2): | |
input_text = gr.Textbox(lines=4, label="Votre question économique") | |
with gr.Column(scale=1): | |
mode_selector = gr.Dropdown(choices=["Rapide", "Équilibré", "Précis"], value="Équilibré", label="Mode de réponse") | |
detail_mode_selector = gr.Dropdown(choices=["Normal", "Expert"], value="Normal", label="Niveau de détail") | |
analyze_btn = gr.Button("Analyser") | |
download_btn = gr.Button("Télécharger l'historique") | |
with gr.Row(): | |
sentiment_output = gr.Textbox(label="Sentiment prédictif") | |
displayed_prompt = gr.Textbox(label="Votre question", interactive=False) | |
explanation_output_en = gr.Textbox(label="Explication en anglais") | |
explanation_output_fr = gr.Textbox(label="Explication en français") | |
sentiment_gauge = gr.HTML() | |
progress_message = gr.Textbox(label="Progression", interactive=False) | |
download_file = gr.File(label="Fichier CSV") | |
input_text.change(lambda t: gr.update(value=suggest_model(t)), inputs=[input_text], outputs=[mode_selector]) | |
analyze_btn.click( | |
full_analysis, | |
inputs=[input_text, mode_selector, detail_mode_selector, count, history], | |
outputs=[sentiment_output, displayed_prompt, explanation_output_en, explanation_output_fr, count, history, sentiment_gauge, progress_message] | |
) | |
download_btn.click( | |
download_history, | |
inputs=[history], | |
outputs=[download_file] | |
) | |
iface.launch(share=True) | |
if __name__ == "__main__": | |
launch_app() |