Confiance : {conf*100:.1f}%
" ) return transcription, summary_html, seg_df, history_entry except Exception as e: error_msg = f"Erreur lors de l'analyse: {str(e)}" return error_msg, "", pd.DataFrame(), {} # ===== API FASTAPI ===== app = FastAPI( title="API Multimodale de Transcription & Sentiment", description="API pour l'analyse de sentiment audio en français", version="1.0", docs_url="/api/docs", redoc_url="/api/redoc" ) # Configuration CORS app.add_middleware( CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) @app.get("/api/") async def root(): """Endpoint racine avec informations sur l'API""" return { "message": "API Multimodale de Transcription & Sentiment", "version": "1.0", "endpoints": { "docs": "/api/docs", "predict": "/api/predict", "health": "/api/health" }, "supported_formats": ["wav", "flac", "mp3"] } @app.get("/api/health") async def health_check(): """Vérification de l'état de l'API""" return { "status": "healthy", "models_loaded": True, "timestamp": "2024-01-01T00:00:00Z" } @app.post("/api/predict") async def predict(file: UploadFile = File(...)): """Analyse de sentiment audio""" # 1. Vérifier le type de fichier if not file.filename or not file.filename.lower().endswith((".wav", ".flac", ".mp3")): raise HTTPException( status_code=400, detail="Seuls les fichiers audio WAV/FLAC/MP3 sont acceptés." ) # 2. Vérifier la taille du fichier (max 50MB) content = await file.read() if len(content) > 50 * 1024 * 1024: # 50MB raise HTTPException( status_code=400, detail="Fichier trop volumineux. Taille maximale: 50MB" ) # 3. Sauvegarder temporairement import tempfile suffix = os.path.splitext(file.filename)[1] with tempfile.NamedTemporaryFile(suffix=suffix, delete=False) as tmp: tmp.write(content) tmp_path = tmp.name try: # 4. Transcription transcription = transcribe_ctc(tmp_path) if not transcription.strip(): return JSONResponse({ "transcription": "", "sentiment": {"négatif": 0.33, "neutre": 0.34, "positif": 0.33}, "warning": "Aucune transcription détectée" }) # 5. Features multimodales try: audio_feat = speech_enc.extract_features(tmp_path) text_feat = text_enc.extract_features([transcription]) # 6. Classification logits = model_mm.classifier(torch.cat([audio_feat, text_feat], dim=1)) probs = F.softmax(logits, dim=1).squeeze().tolist() labels = ["négatif", "neutre", "positif"] sentiment = {labels[i]: round(probs[i], 3) for i in range(len(labels))} except Exception as e: # Fallback vers analyse textuelle uniquement print(f"Erreur multimodal, fallback textuel: {e}") sent_dict = TextEncoder.analyze_sentiment(transcription) sentiment = {k: round(v, 3) for k, v in sent_dict.items()} return JSONResponse({ "transcription": transcription, "sentiment": sentiment, "filename": file.filename, "file_size": len(content) }) except Exception as e: raise HTTPException(status_code=500, detail=f"Erreur lors de l'analyse: {str(e)}") finally: # Nettoyage fichier temporaire try: os.remove(tmp_path) except: pass @app.post("/api/predict_text") async def predict_text(text: str): """Analyse de sentiment textuel uniquement""" try: sent_dict = TextEncoder.analyze_sentiment(text) sentiment = {k: round(v, 3) for k, v in sent_dict.items()} return JSONResponse({ "text": text, "sentiment": sentiment }) except Exception as e: raise HTTPException(status_code=500, detail=f"Erreur analyse textuelle: {str(e)}") # ===== INTERFACE GRADIO ===== def export_history_csv(history): if not history: return None df = pd.DataFrame(history) path = "history.csv" df.to_csv(path, index=False) return path # Interface Gradio demo = gr.Blocks( theme=gr.themes.Monochrome(primary_hue="purple"), title="Analyse de Sentiment Audio - Hugging Face Space" ) with demo: gr.Markdown(""" # 🎤 Analyse de Sentiment Audio Ce Space permet d'analyser le sentiment d'extraits audio en français en combinant : - **Transcription audio** avec Wav2Vec2 - **Analyse de sentiment** avec BERT multilingue - **API REST** pour intégration """) gr.HTML("""