from fastapi import FastAPI, File, UploadFile, Form from fastapi.responses import JSONResponse from transformers import pipeline from typing import Optional import io from PIL import Image import tempfile import os import fitz # PyMuPDF import docx import pandas as pd import pptx from fastapi.middleware.cors import CORSMiddleware from langdetect import detect from fastapi.staticfiles import StaticFiles from fastapi.responses import HTMLResponse from fastapi import Request from fastapi.templating import Jinja2Templates app = FastAPI() app.add_middleware( CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) # Liste des langues supportées SUPPORTED_LANGUAGES = ["fr", "en", "de", "es", "it", "zh", "ar"] # Modèles de traduction valides (existants sur Hugging Face) translation_models = { "fr-en": "Helsinki-NLP/opus-mt-fr-en", "en-fr": "Helsinki-NLP/opus-mt-en-fr", "fr-de": "Helsinki-NLP/opus-mt-fr-de", "de-fr": "Helsinki-NLP/opus-mt-de-fr", "fr-es": "Helsinki-NLP/opus-mt-fr-es", "es-fr": "Helsinki-NLP/opus-mt-es-fr", "en-zh": "Helsinki-NLP/opus-mt-en-zh", "zh-en": "Helsinki-NLP/opus-mt-zh-en", "en-it": "Helsinki-NLP/opus-mt-en-it", "it-en": "Helsinki-NLP/opus-mt-it-en", "en-ar": "Helsinki-NLP/opus-mt-en-ar", "ar-en": "Helsinki-NLP/opus-mt-ar-en", "en-es": "Helsinki-NLP/opus-mt-en-es", "en-de": "Helsinki-NLP/opus-mt-en-de", "es-ar": "Helsinki-NLP/opus-mt-es-ar", "es-en": "Helsinki-NLP/opus-mt-es-en", "es-de": "Helsinki-NLP/opus-mt-es-de", "es-it": "Helsinki-NLP/opus-mt-es-it", "es-zh": "Helsinki-NLP/opus-mt-es-zh", "ar-fr": "Helsinki-NLP/opus-mt-ar-fr", "ar-de": "Helsinki-NLP/opus-mt-ar-de", "ar-es": "Helsinki-NLP/opus-mt-ar-es", "ar-it": "Helsinki-NLP/opus-mt-ar-it", "ar-zh": "Helsinki-NLP/opus-mt-ar-zh", "de-en": "Helsinki-NLP/opus-mt-de-en", "de-de": "Helsinki-NLP/opus-mt-de-de", "de-es": "Helsinki-NLP/opus-mt-de-es", "de-it": "Helsinki-NLP/opus-mt-de-it", "de-zh": "Helsinki-NLP/opus-mt-de-zh", "de-ar": "Helsinki-NLP/opus-mt-de-ar", "it-fr": "Helsinki-NLP/opus-mt-it-fr", "it-de": "Helsinki-NLP/opus-mt-it-de", "it-es": "Helsinki-NLP/opus-mt-it-es", "it-zh": "Helsinki-NLP/opus-mt-it-zh", "it-ar": "Helsinki-NLP/opus-mt-it-ar", "zh-fr": "Helsinki-NLP/opus-mt-zh-fr", "zh-de": "Helsinki-NLP/opus-mt-zh-en", "zh-it": "Helsinki-NLP/opus-mt-zh-it", "zh-es": "Helsinki-NLP/opus-mt-zh-es", "zh-ar": "Helsinki-NLP/opus-mt-zh-ar", } def extract_text_from_pdf(file_path): text = "" with fitz.open(file_path) as doc: for page in doc: text += page.get_text("text") + "\n" return text def extract_text_from_docx(file_path): doc = docx.Document(file_path) return "\n".join([p.text for p in doc.paragraphs]) def extract_text_from_pptx(file_path): presentation = pptx.Presentation(file_path) text = [] for slide in presentation.slides: for shape in slide.shapes: if hasattr(shape, "text"): text.append(shape.text) return "\n".join(text) def extract_text_from_excel(file_path): df = pd.read_excel(file_path, engine="openpyxl") return df.to_string(index=False) def chunk_text(text, max_length=512): words = text.split() chunks, current_chunk = [], [] for word in words: if len(" ".join(current_chunk) + " " + word) <= max_length: current_chunk.append(word) else: chunks.append(" ".join(current_chunk)) current_chunk = [word] if current_chunk: chunks.append(" ".join(current_chunk)) return chunks def translate_text(text, source_lang, target_lang): if source_lang not in SUPPORTED_LANGUAGES or target_lang not in SUPPORTED_LANGUAGES: return None # Langue non supportée model_key = f"{source_lang}-{target_lang}" if model_key in translation_models: model_name = translation_models[model_key] translator = pipeline("translation", model=model_name) translated_chunks = [translator(chunk)[0]["translation_text"] for chunk in chunk_text(text)] return " ".join(translated_chunks) # Si pas de traduction directe, utiliser l'anglais comme pivot model_to_en = f"{source_lang}-en" model_from_en = f"en-{target_lang}" if model_to_en in translation_models and model_from_en in translation_models: translator_to_en = pipeline("translation", model=translation_models[model_to_en]) translator_from_en = pipeline("translation", model=translation_models[model_from_en]) intermediate_texts = [translator_to_en(chunk)[0]["translation_text"] for chunk in chunk_text(text)] intermediate_text = " ".join(intermediate_texts) final_texts = [translator_from_en(chunk)[0]["translation_text"] for chunk in chunk_text(intermediate_text)] return " ".join(final_texts) return None # Pas de modèle disponible templates = Jinja2Templates(directory="templates") # Mount static files (CSS, JS, images) app.mount("/static", StaticFiles(directory="static"), name="static") @app.get("/") async def read_root(request: Request): return templates.TemplateResponse("prj.html", {"request": request}) @app.post("/translate") async def translate_document(file: UploadFile = File(...), language: str = Form(...)): try: suffix = file.filename.split(".")[-1].lower() temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=f".{suffix}") temp_file.write(await file.read()) temp_file.close() extractors = { "pdf": extract_text_from_pdf, "docx": extract_text_from_docx, "pptx": extract_text_from_pptx, "xls": extract_text_from_excel, "xlsx": extract_text_from_excel } if suffix not in extractors: return JSONResponse({"error": "Format non supporté"}, status_code=400) text = extractors[suffix](temp_file.name) os.remove(temp_file.name) if not text.strip(): return JSONResponse({"error": "Aucun texte détecté"}, status_code=400) detected_lang = detect(text) if detected_lang not in SUPPORTED_LANGUAGES: return JSONResponse({"error": f"Langue non supportée : {detected_lang}"}, status_code=400) if detected_lang == language: return JSONResponse({"translation": text, "note": "Déjà dans la langue choisie."}) translated_text = translate_text(text, detected_lang, language) if translated_text: return JSONResponse({"translation": translated_text}) else: return JSONResponse({"error": "Aucun modèle de traduction trouvé."}, status_code=400) except Exception as e: return JSONResponse({"error": str(e)}, status_code=500)