Update app.py

app.py

@@ -1,106 +1,51 @@
 from fastapi import FastAPI, UploadFile, File, Form
 from faster_whisper import WhisperModel
-import uvicorn
 import tempfile
 import shutil
-import torch
 import os
-import time
+import torch
 from fastapi.middleware.cors import CORSMiddleware
-from pydub import AudioSegment
-from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer
-from reportlab.lib.styles import getSampleStyleSheet
+import time
 
-# === Config GPU/CPU ===
 device = "cuda" if torch.cuda.is_available() else "cpu"
-compute_type = "float16" if device == "cuda" else "int8"
-
-origins = [
-    "https://cabane-data.fr" , "https://www.cabane-data.fr" # autoriser ton WordPress
-]
+compute_type = "float16" if device=="cuda" else "int8"
 
 app = FastAPI()
 
 app.add_middleware(
     CORSMiddleware,
-    allow_origins=origins,
+    allow_origins=["*"],  # à restreindre à ton domaine
     allow_credentials=True,
     allow_methods=["*"],
     allow_headers=["*"],
 )
 
-# === Dictionnaire des modèles dispo ===
 AVAILABLE_MODELS = ["tiny", "base", "small", "medium", "large-v2"]
 
 def load_model(model_name: str):
-    """Charger un modèle Whisper"""
     return WhisperModel(model_name, device=device, compute_type=compute_type)
 
-def split_audio(file_path, chunk_length_ms=300_000):
-    """Découpe l'audio en segments de 5 min max"""
-    audio = AudioSegment.from_file(file_path)
-    chunks = []
-    for i in range(0, len(audio), chunk_length_ms):
-        chunk = audio[i:i + chunk_length_ms]
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_chunk:
-            chunk.export(tmp_chunk.name, format="mp3")
-            chunks.append(tmp_chunk.name)
-    return chunks
-
-def generate_pdf(text: str, output_path: str):
-    """Créer un PDF avec la transcription"""
-    doc = SimpleDocTemplate(output_path)
-    styles = getSampleStyleSheet()
-    story = [Paragraph("Transcription Audio", styles["Title"]), Spacer(1, 12)]
-    story.append(Paragraph(text, styles["Normal"]))
-    doc.build(story)
-
-# === Endpoint API REST ===
 @app.post("/transcribe")
-async def transcribe(
-    file: UploadFile = File(...),
-    model_name: str = Form("base")  # par défaut "base"
-):
+async def transcribe(file: UploadFile = File(...), model_name: str = Form("base")):
     if model_name not in AVAILABLE_MODELS:
         return {"error": f"Modèle non reconnu. Choisissez parmi {AVAILABLE_MODELS}"}
 
-    # Chronomètre début
-    start_time = time.time()
-
-    # Charger modèle
+    start = time.time()
     model = load_model(model_name)
 
-    # Sauvegarder fichier temporaire
     with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp:
         shutil.copyfileobj(file.file, tmp)
         tmp_path = tmp.name
 
-    # Découpage en segments
-    chunks = split_audio(tmp_path)
-    full_text = ""
-
-    # Transcrire chaque segment
-    for chunk in chunks:
-        segments, info = model.transcribe(chunk, beam_size=5)
-        text_result = " ".join([segment.text for segment in segments])
-        full_text += text_result + "\n"
-        os.remove(chunk)
-
-    # Nettoyage
+    segments, info = model.transcribe(tmp_path, beam_size=5)
+    text_result = " ".join([s.text for s in segments])
     os.remove(tmp_path)
 
-    # Génération PDF
-    pdf_path = tempfile.NamedTemporaryFile(delete=False, suffix=".pdf").name
-    generate_pdf(full_text, pdf_path)
-
-    # Chronomètre fin
-    total_time = round(time.time() - start_time, 2)
+    end = time.time()
+    duration = round(end - start, 2)
 
     return {
-        "model_used": model_name,
-        "language": info.language,
-        "probability": info.language_probability,
-        "transcription": full_text,
-        "pdf_file": pdf_path,
-        "processing_time_sec": total_time
+        "text": text_result,
+        "duration": duration,
+        "model_used": model_name
     }