import os import sys import argparse import time import re from pathlib import Path from types import SimpleNamespace from urllib.parse import urlparse from configparser import ConfigParser missing = [] for lib in [ "chardet", "bs4", "jinja2", "html2text", "pdf2docx", "docx", "pptx", "markitdown", ]: try: __import__(lib) except ImportError: missing.append(lib) if missing: print(f"Error: Missing libraries:\n{', '.join(missing)}") input("\nPress ENTER to terminate>>\n") sys.exit(1) import chardet from bs4 import BeautifulSoup from jinja2 import Environment, FileSystemLoader import html2text from pdf2docx import Converter from docx import Document from pptx import Presentation from pptx.enum.shapes import MSO_SHAPE_TYPE from markitdown import MarkItDown from tkai_lib import read_ai_config from tkai_lib import query_openai4, openai_response_to_json from tkai_lib import query_openai5, extract_openai5_text from tkai_lib import query_google2, google_response_to_json from tkai_lib import query_deepl, extract_deepl_text import openai #========================================================= # 初期化 #========================================================= parser = None def initialize(): cfg = SimpleNamespace() cfg.config_path = "translate5.env" cfg.common_config_path = "ai.env" # APIキーや共通設定は従来どおり .env / ai.env から read_ai_config(cfg.config_path, read_account_inf=False) read_ai_config(cfg.common_config_path) p = argparse.ArgumentParser(description="Translate/Revise .docx/.pptx/.pdf/.html/.md") p.add_argument("--mode", default="je", help="Translation mode (je, ej, jj, etc.)") p.add_argument( "--infile", "-i", default="translate_test.docx", help="Input file (.docx/.pptx/.pdf/.html/.md/.txt)", ) p.add_argument( "--output_html_path", default=None, help="Path for the output HTML file (comparison report, optional)", ) p.add_argument( "--html_template_path", default="template_translate.html", help="Path to the HTML template file for comparison report", ) p.add_argument( "--api", default="openai5", choices=["openai5", "openai", "google", "gemini", "deepl"], help="Translation API to use", ) p.add_argument("--model", default=None, help="Override API model") p.add_argument("--endpoint", default=os.getenv("endpoint"), help="API endpoint") p.add_argument( "--max_tokens", type=int, default=int(os.getenv("max_tokens", "2000")), help="Maximum number of tokens / output length", ) p.add_argument( "--openai_api_key", default=os.getenv("OPENAI_API_KEY"), help="OpenAI API key", ) p.add_argument( "--openai_model", default=os.getenv("openai_model", "gpt-4o"), help="OpenAI model to use", ) p.add_argument( "--temperature", type=float, default=float(os.getenv("temperature", "0.3")), help="Sampling temperature", ) p.add_argument( "--openai_model5", default=os.getenv("openai_model5", "gpt-5.2"), help="OpenAI GPT-5 model", ) p.add_argument( "--reasoning_effort", default=os.getenv("reasoning_effort", "low"), help="Reasoning effort level for GPT-5", ) p.add_argument( "--google_api_key", default=os.getenv("GOOGLE_API_KEY"), help="Google API key", ) p.add_argument( "--google_model", default=os.getenv("google_model", "gemini-2.5-flash"), help="Google/Gemini model to use", ) p.add_argument( "--deepl_api_key", default=os.getenv("DEEPL_API_KEY"), help="DeepL API key", ) p.add_argument( "--force_server_charcode", default=os.getenv("force_server_charcode", "utf-8"), help="force_server_charcode", ) p.add_argument( "--tsleep_rpm", type=float, default=0.5, help="Sleep time to avoid rpm (seconds)", ) p.add_argument("--use_md", action="store_true", help="Convert to markdown and process") p.add_argument( "--limit_to_multibyte_str", action="store_true", help="When set, only translate text containing multibyte chars (e.g. Japanese)", ) p.add_argument( "--process_unit", default="paragraph", choices=["paragraph", "run"], help="Unit of processing for docx/pptx/html", ) p.add_argument( "--min_translate_length", type=int, default=int(os.getenv("min_translate_length", "5")), help="Minimum length of text to translate", ) p.add_argument( "--allowed_translation_length_ratio", type=float, default=float(os.getenv("allowed_translation_length_ratio", "5.0")), help="Max allowed ratio (translated_len / original_len)", ) # ★ CLI からは prompt 本文のみ上書き可能（roleは統合済み） p.add_argument( "--translate_prompt", type=str, default="", help="Override translation prompt (full instructions incl. 'あなたは〜').", ) p.add_argument( "--reformat_prompt", type=str, default="", help="Override reformat prompt (full instructions).", ) return cfg, p def load_prompt_config_from_ini(args, cfg): """ translate5.ini を以下の順で探す: 1. カレントディレクトリ 2. 入力ファイルディレクトリ 3. スクリプトディレクトリ [translate], [reformat] セクションから prompt を取得する。 - 「role と prompt をまとめる」方針: * 新しい書き方: prompt に全指示を書く（あなたは〜 + 条件 + {{text}} まで全部） * 互換性維持: role と prompt の両方があれば「role + 改行 + prompt」を結合して1つのpromptとして扱う * role だけあればそれを prompt として扱う 優先順位: - CLI (--translate_prompt / --reformat_prompt) があればそれが最優先 - なければ ini の結合済み prompt - それも無ければ組込デフォルト """ ini_name = "translate5.ini" candidate_paths = [] # 1. 作業ディレクトリ candidate_paths.append(os.path.join(os.getcwd(), ini_name)) # 2. 入力ファイルディレクトリ if getattr(args, "infile", None): infile_abs = os.path.abspath(args.infile) infile_dir = os.path.dirname(infile_abs) candidate_paths.append(os.path.join(infile_dir, ini_name)) # 3. スクリプトディレクトリ script_dir = os.path.dirname(os.path.abspath(__file__)) candidate_paths.append(os.path.join(script_dir, ini_name)) ini_path = None for p in candidate_paths: if os.path.isfile(p): ini_path = p break translate_prompt_ini = None reformat_prompt_ini = None if ini_path: print(f" Using prompt config file: {ini_path}") cp = ConfigParser() cp.read(ini_path, encoding="utf-8") # [translate] if cp.has_section("translate"): role_val = cp.get("translate", "role", fallback="").strip() prompt_val = cp.get("translate", "prompt", fallback="") prompt_val = prompt_val.lstrip("\n") if role_val and prompt_val: translate_prompt_ini = (role_val + "\n" + prompt_val).strip() elif prompt_val: translate_prompt_ini = prompt_val.strip() elif role_val: translate_prompt_ini = role_val # [reformat] if cp.has_section("reformat"): role_val = cp.get("reformat", "role", fallback="").strip() prompt_val = cp.get("reformat", "prompt", fallback="") prompt_val = prompt_val.lstrip("\n") if role_val and prompt_val: reformat_prompt_ini = (role_val + "\n" + prompt_val).strip() elif prompt_val: reformat_prompt_ini = prompt_val.strip() elif role_val: reformat_prompt_ini = role_val else: print(" translate5.ini not found. Using built-in default prompts.") # ---- 組込みデフォルト (role+prompt一体型) ---- if not translate_prompt_ini: translate_prompt_ini = ( "あなたは専門的な英語を正確かつプロフェッショナルに翻訳・校正するアシスタントです。\n" "*#翻訳してほしいテキスト*以降のテキストは学会プレゼンテーション等の一部です。\n" "以下の条件を守り、自然な米国英語に翻訳してください。\n" "条件1：元の意味を変えない。テキストに書かれていない解釈・内容は追加しない。\n" "条件2：文字数を大きく増やさない（目安 ±30%）。\n" "条件3：動詞を含まない短いフレーズはスライド見出しとして扱い、完全な文にしない。\n" "条件4：数式・記号・固有名詞は可能な限り保持する。\n" "{{additional_prompt}}\n\n" "#翻訳してほしいテキスト\n{{text}}" ) if not reformat_prompt_ini: reformat_prompt_ini = ( "あなたは専門的な英語文書を、内容を変えずに読みやすく整形するアシスタントです。\n" "以下の*＃テキスト*以降は PDF やスライドから抽出したテキストであり、" "改行位置や段落順が乱れたり、文が途中で分断されている可能性があります。\n" "元の意味・情報を一切削除・要約・追加せず、文や段落のつながりが自然になるように" "並び替えと改行のみを調整してください。\n" "箇条書きや見出しらしき部分は、その構造を保ってください。\n" "出力はプレーンテキストのみとし、余計な説明文は付けないでください。\n\n" "＃テキスト\n{{ text }}" ) # ---- 優先順位適用: CLI > ini/結合 > デフォルト ---- cfg.translate_prompt = args.translate_prompt if args.translate_prompt else translate_prompt_ini cfg.reformat_prompt = args.reformat_prompt if args.reformat_prompt else reformat_prompt_ini # role は内部的には空文字で扱う（全部入りプロンプトで運用） cfg.translate_role = "" cfg.reformat_role = "" return cfg def update_variables(cfg, p): args = p.parse_args() # モデル上書き if args.model is not None: if args.api == "openai5": args.openai_model5 = args.model elif args.api == "openai": args.openai_model = args.model elif args.api in ("google", "gemini"): args.google_model = args.model # mode から言語方向 if args.mode[0] == "j": args.source_lang = "JA" # GUI側で既に設定済みでも、CLI/デフォルトロジックに合わせて再度設定 if not args.limit_to_multibyte_str: args.limit_to_multibyte_str = True else: args.source_lang = "EN" if args.mode[0] != "j": # 英語ソースの場合は明示的にFalseにする args.limit_to_multibyte_str = False if args.mode[1] == "j": args.target_lang = "JA" else: args.target_lang = "EN" # cfg に基本情報コピー cfg.infile = args.infile cfg.api = args.api cfg.openai_model = args.openai_model cfg.openai_model5 = args.openai_model5 cfg.google_model = args.google_model cfg.deepl_api_key = args.deepl_api_key cfg.endpoint = args.endpoint cfg.max_tokens = args.max_tokens cfg.temperature = args.temperature cfg.reasoning_effort = args.reasoning_effort cfg.tsleep_rpm = args.tsleep_rpm cfg.use_md = args.use_md cfg.limit_to_multibyte_str = args.limit_to_multibyte_str cfg.process_unit = args.process_unit cfg.min_translate_length = args.min_translate_length cfg.allowed_translation_length_ratio = args.allowed_translation_length_ratio cfg.mode = args.mode cfg.source_lang = args.source_lang cfg.target_lang = args.target_lang cfg.html_template_path = args.html_template_path cfg.openai_api_key = args.openai_api_key cfg.google_api_key = args.google_api_key # ini + デフォルト + CLI 上書きで prompt を確定（roleは内部で空） cfg = load_prompt_config_from_ini(args, cfg) return cfg def usage(): global parser if parser: parser.print_usage() # ------------------------------------------------------ # ユーティリティ # ------------------------------------------------------ def process_template(template: str, context: dict) -> str: """ \t, \n, \r を展開し、{{ key }} を context で置換。 """ template = template.replace(r"\t", "\t") template = template.replace(r"\n", "\n") template = template.replace(r"\r", "\r") def replace_placeholder(match): key = match.group(1).strip() return str(context.get(key, f"{{{{ {key} }}}}")) return re.sub(r"\{\{\s*(.*?)\s*\}\}", replace_placeholder, template) def save(path, text): with open(path, "w", encoding="utf-8") as f: f.write(text) def read_file(path): if not os.path.exists(path): print(f"\nError in read_file(): File [{path}] does not exist\n") sys.exit(1) with open(path, "r", encoding="utf-8") as fp: return fp.read() def replace_path(path, ext): return os.path.splitext(path)[0] + ext def check_multibyte_str(text, limit_to_multibyte_str): if not limit_to_multibyte_str: return True pattern = re.compile(r"[\u0800-\uFFFF]") return bool(pattern.search(text)) def html_to_markdown(html_file_path): with open(html_file_path, "r", encoding="utf-8") as html_file: html_content = html_file.read() return html2text.html2text(html_content) def convert_to_md(infile): if ".html" in infile.lower(): return html_to_markdown(infile) else: print(f"Convert {infile} to markdown") md = MarkItDown() result = md.convert(infile) return result.text_content def n_leading_chars(s, c="#"): return len(s) - len(s.lstrip(c)) def pdf_to_docx(pdf_file, docx_file): cv = Converter(pdf_file) cv.convert(docx_file, start=0, end=None) cv.close() print(f"Converted '{pdf_file}' to '{docx_file}' successfully.") def get_filetype(path): ext = os.path.splitext(path)[1].lower() if ext == ".pdf": return "pdf" if ext == ".docx": return "docx" if ext == ".pptx": return "pptx" if ext in (".html", ".htm"): return "html" if ext in (".txt", ".text"): return "txt" if ext == ".md": return "md" return None def to_translate(text, min_translate_length, limit_to_multibyte_str): if text is None: return False text0 = text.strip() if text0 == "": return False if len(text0) < min_translate_length: return False is_mb = check_multibyte_str(text0, limit_to_multibyte_str) if not is_mb and not bool(re.search("[a-zA-Z]", text0)): return False return is_mb if limit_to_multibyte_str else True # ------------------------------------------------------ # 翻訳系ラッパ # ------------------------------------------------------ def revise_with_openai4(text, openai_model, prompt, temperature, max_tokens, cfg): prompt_text = process_template(prompt, {"text": text, "additional_prompt": ""}) response = query_openai4( prompt_text, openai_model, role="", # roleは統合済み。必要ならprompt側に書く。 temperature=temperature, max_tokens=max_tokens, openai_api_key=cfg.openai_api_key, ) if not response or response == {}: return "" usage = response.usage print(f" prompt_tokens : {usage.prompt_tokens}") print(f" completion_tokens: {usage.completion_tokens}") if cfg.tsleep_rpm > 0: time.sleep(cfg.tsleep_rpm) return response.choices[0].message.content.strip() def revise_with_openai5(text, openai_model5, prompt, effort, max_output_tokens, cfg): prompt_text = process_template(prompt, {"text": text, "additional_prompt": ""}) response = query_openai5( prompt_text, openai_model5, role="", # 統合済み effort=effort, max_output_tokens=max_output_tokens, openai_api_key=cfg.openai_api_key, ) if not response or response == {}: return "" text_out = extract_openai5_text(response) print(f" text: {text_out}") if cfg.tsleep_rpm > 0: time.sleep(cfg.tsleep_rpm) return text_out def revise_with_google(text, google_model, prompt, temperature, max_tokens, cfg): if cfg.google_api_key is None: cfg.google_api_key = os.getenv("GOOGLE_API_KEY") if not cfg.google_api_key: print( "\nError in tkai_lib.query_google(): Can not get GOOGLE_API_KEY" "\n define the environment variable GOOGLE_API_KEY\n" ) raise RuntimeError("GOOGLE_API_KEY is not set") prompt_text = process_template( prompt, { "text": text, "additional_prompt": "翻訳結果のみを出力してください。前置きや後付けのメッセージは不要です。", }, ) generation_config = { "temperature": temperature, "top_p": getattr(cfg, "top_p", 0.8), "top_k": getattr(cfg, "top_k", 40), "max_output_tokens": max_tokens, "candidate_count": 1, } messages = [ { "role": "user", "parts": [prompt_text], } ] safety_settings = [ {"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_MEDIUM_AND_ABOVE"}, {"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_MEDIUM_AND_ABOVE"}, { "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_MEDIUM_AND_ABOVE", }, { "category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_MEDIUM_AND_ABOVE", }, ] response = query_google2( messages, google_model=google_model, generation_config=generation_config, safety_settings=safety_settings, google_api_key=None, ) if cfg.tsleep_rpm > 0: time.sleep(cfg.tsleep_rpm) try: return response.text.strip() except ValueError as e: print("\n--- [WARNING] Google API returned invalid response (safety filter etc.).") print("--- Skipping this text. Error:", e) print(f"--- Original: '{text[:80]}...'") return "" def translate_with_deepl(text, deepl_api_key, endpoint, source_lang="JA", target_lang="EN", cfg = None): response = query_deepl( text, source_lang, target_lang, deepl_api_key=deepl_api_key, endpoint=endpoint, ) if cfg.tsleep_rpm > 0: time.sleep(cfg.tsleep_rpm) if response.status_code == 200: return extract_deepl_text(response) else: print(f"Error: {response.status_code}, {response.text}") return text def translate(text, api, prompt, cfg): if api == "openai5": return revise_with_openai5( text, cfg.openai_model5, prompt, cfg.reasoning_effort, cfg.max_tokens, cfg, ) elif api == "openai": return revise_with_openai4( text, cfg.openai_model, prompt, cfg.temperature, cfg.max_tokens, cfg, ) elif api in ("google", "gemini"): return revise_with_google( text, cfg.google_model, prompt, cfg.temperature, cfg.max_tokens, cfg, ) elif api == "deepl": return translate_with_deepl( text, cfg.deepl_api_key, cfg.endpoint, cfg.source_lang, cfg.target_lang, cfg, ) else: print(f"\nError in translate(): Invalid API [{api}]\n") sys.exit(1) # ------------------------------------------------------ # 翻訳結果チェック # ------------------------------------------------------ def check_translation(text, translated, allowed_translation_length_ratio): if translated is None: return False if translated.startswith("'''"): return False if translated.startswith("```"): return False ntranslated = n_leading_chars(translated, "#") ntext = n_leading_chars(text, "#") if ntranslated > ntext: return False ntranslated = n_leading_chars(translated, "*") ntext = n_leading_chars(text, "*") if ntranslated == ntext + 1: return False if len(text) > 0 and len(translated) > len(text) * allowed_translation_length_ratio: return False return True # ------------------------------------------------------ # 各フォーマット別処理 # ------------------------------------------------------ def translate_html(text, api, api_model, prompt, cfg): data = [] print() print(">>> Analyzing html...") soup = BeautifulSoup(text, "html.parser") print("Translating content...") for element in soup.find_all(string=True): if to_translate(element, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = element try: revised_text = translate(original_text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this element. Skipping. Error: {e}") revised_text = original_text # 失敗時は元のテキストを保持 continue print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): element.replace_with(revised_text) data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") return str(soup), data def translate_pptx(ppt, api, api_model, prompt, cfg): cfg.translate_master = False data = [] print() print(">>> Translating PPTX (shapes, groups, tables, notes, charts)...") def translate_paragraph(paragraph): nonlocal data if cfg.process_unit == "paragraph": text = paragraph.text if to_translate(text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = text try: revised_text = translate(text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this paragraph. Skipping. Error: {e}") return # 失敗時は元のテキストを保持 print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): paragraph.text = revised_text data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") else: for run in paragraph.runs: if to_translate(run.text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = run.text try: revised_text = translate(run.text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this run. Skipping. Error: {e}") continue # 失敗時は元のテキストを保持 print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): run.text = revised_text data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") def translate_shape(shape): if shape.has_text_frame: for paragraph in shape.text_frame.paragraphs: translate_paragraph(paragraph) if shape.has_table: for row in shape.table.rows: for cell in row.cells: for paragraph in cell.text_frame.paragraphs: translate_paragraph(paragraph) if shape.shape_type == MSO_SHAPE_TYPE.GROUP: for shp in shape.shapes: translate_shape(shp) if hasattr(shape, "has_chart") and shape.has_chart: chart = shape.chart if chart.has_title and chart.chart_title.has_text_frame: for paragraph in chart.chart_title.text_frame.paragraphs: translate_paragraph(paragraph) if hasattr(chart, "category_axis"): ax = chart.category_axis if getattr(ax, "has_title", False) and ax.axis_title.has_text_frame: for paragraph in ax.axis_title.text_frame.paragraphs: translate_paragraph(paragraph) if hasattr(chart, "value_axis"): ax = chart.value_axis if getattr(ax, "has_title", False) and ax.axis_title.has_text_frame: for paragraph in ax.axis_title.text_frame.paragraphs: translate_paragraph(paragraph) for series in chart.series: for pt in getattr(series, "points", []): lbl = getattr(pt, "data_label", None) if lbl is not None and getattr(lbl, "has_text_frame", False): for paragraph in lbl.text_frame.paragraphs: translate_paragraph(paragraph) for slide_idx, slide in enumerate(ppt.slides, start=1): print(f"\n--- Slide {slide_idx} ---") for shape in slide.shapes: translate_shape(shape) if slide.has_notes_slide: notes_slide = slide.notes_slide ntf = getattr(notes_slide, "notes_text_frame", None) if ntf is not None: print(" Translating notes...") for paragraph in ntf.paragraphs: translate_paragraph(paragraph) # マスタースライド翻訳フラグ（無指定時は従来通り True 扱い） if getattr(cfg, "translate_master", True): print("\n>>> Translating slide masters and layouts (static text)...") for master in ppt.slide_masters: for shape in master.shapes: translate_shape(shape) for layout in master.slide_layouts: for shape in layout.shapes: translate_shape(shape) else: print("\n>>> Skipping translation of slide masters and layouts (static text).") return ppt, data def translate_docx(doc, api, api_model, prompt, cfg): data = [] print() print(">>> Processing paragraphs/runs...") if cfg.process_unit == "paragraph": for paragraph in doc.paragraphs: if to_translate(paragraph.text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = paragraph.text try: revised_text = translate(paragraph.text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this paragraph. Skipping. Error: {e}") continue print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): paragraph.text = revised_text data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") else: for paragraph in doc.paragraphs: for run in paragraph.runs: if to_translate(run.text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = run.text try: revised_text = translate(run.text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this run. Skipping. Error: {e}") continue print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): run.text = revised_text data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") print(">>> Processing tables...") for table in doc.tables: for row in table.rows: for cell in row.cells: for paragraph in cell.paragraphs: if cfg.process_unit == "paragraph": if to_translate(paragraph.text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = paragraph.text try: revised_text = translate(paragraph.text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this table cell (paragraph). Skipping. Error: {e}") continue print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): paragraph.text = revised_text data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") else: for run in paragraph.runs: if to_translate(run.text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = run.text try: revised_text = translate(run.text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for this table cell (run). Skipping. Error: {e}") continue print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") if check_translation(original_text, revised_text, cfg.allowed_translation_length_ratio): run.text = revised_text data.append({"original": original_text, "translated": revised_text}) else: print(" *** This translation is rejected") return doc, data def translate_text(text, api, api_model, prompt, cfg): data = [] print(">>> Translating plain text...") revised_text = text if to_translate(text, cfg.min_translate_length, cfg.limit_to_multibyte_str): original_text = text try: revised_text = translate(text, api, prompt, cfg) except Exception as e: print(f" *** Translation failed for the entire text block. Skipping. Error: {e}") revised_text = original_text # 失敗時は元のテキストを保持 print(f"[Original] {original_text}") print(f" -> [Revised] {revised_text}") data.append({"original": original_text, "translated": revised_text}) return revised_text, data # ------------------------------------------------------ # メイン実行 # ------------------------------------------------------ def execute(cfg): filetype = get_filetype(cfg.infile) if filetype is None: print(f"\nError in execute(): Invalid extension in [{cfg.infile}]\n") sys.exit(1) # ------------------ ファイルパス設定 ------------------ # NOTE: GUI側のロジックと重複するが、ここではCUI実行時のパスを決定する if filetype == "md": outfile = replace_path(cfg.infile, "_revised.md") elif cfg.use_md: mdfile = replace_path(cfg.infile, ".md") outfile = replace_path(cfg.infile, "_revised.md") elif filetype == "pdf": outfile = replace_path(cfg.infile, "_revised.md" if cfg.use_md else "_revised.docx") elif filetype == "docx": outfile = replace_path(cfg.infile, "_revised.docx") elif filetype == "html": outfile = replace_path(cfg.infile, "_revised.html") elif filetype == "pptx": outfile = replace_path(cfg.infile, "_revised.pptx") elif filetype == "txt": outfile = replace_path(cfg.infile, "_revised.txt") else: print(f"\nError in execute(): Unsupported extension in [{cfg.infile}]\n") sys.exit(1) output_comparison_html_path = replace_path(cfg.infile, "_compare.html") output_reformat_md = replace_path(cfg.infile, "_reformat.md") output_docx = replace_path(cfg.infile, ".docx") print("=== Translate and revise file ===") print(f" Input file : {cfg.infile}") # ... (設定情報の出力は省略) ... print(f" Output file : {outfile}") if not os.path.isfile(cfg.infile): print(f"Error: File [{cfg.infile}] does not exist") usage() sys.exit(1) data = [] translated_document = None # 翻訳済みドキュメントオブジェクト/テキストを保持する変数 # ------------------ 翻訳処理本体 (Tryブロックで囲む) ------------------ try: # ---- md/txt ---- if filetype in ("md", "txt"): print(f"Read [{cfg.infile}]") text = read_file(cfg.infile) text, data = translate_text(text, cfg.api, cfg.openai_model, cfg.translate_prompt, cfg) translated_document = text if text is None: print(f"Error for filetype = {filetype}: Could not get text") # ---- use_md モード ---- elif cfg.use_md: print(f"Read [{cfg.infile}] and convert to markdown") text = convert_to_md(cfg.infile) if text is None: print("Error for md: Could not get text") print(f"=== Saving markdown to {mdfile} ===") save(mdfile, text) if filetype == "pdf": print(f"Reformatting [{mdfile}] by API...") # reformat → translate text2 = translate(text, cfg.api, cfg.reformat_prompt, cfg) if not text2: print("Error: Could not get reformatted text") else: print(f"Reformatted MD file is saved to [{output_reformat_md}]") save(output_reformat_md, text2) text = text2 text, data = translate_text(text, cfg.api, cfg.openai_model, cfg.translate_prompt, cfg) translated_document = text if text is None: print("Error in getting translated text: Could not get text") # ---- 通常モード (docx, pptx, html, pdf -> docx) ---- else: if filetype == "pdf": print(f"Converting [{cfg.infile}] to [{output_docx}]") pdf_to_docx(cfg.infile, output_docx) print(f"Read [{output_docx}]") doc = Document(output_docx) translated_document, data = translate_docx(doc, cfg.api, cfg.openai_model, cfg.translate_prompt, cfg) elif filetype == "pptx": ppt = Presentation(cfg.infile) translated_document, data = translate_pptx(ppt, cfg.api, cfg.openai_model, cfg.translate_prompt, cfg) elif filetype == "docx": print(f"Read [{cfg.infile}]") doc = Document(cfg.infile) translated_document, data = translate_docx(doc, cfg.api, cfg.openai_model, cfg.translate_prompt, cfg) elif filetype == "html": print(f"Read [{cfg.infile}]") html = read_file(cfg.infile) translated_document, data = translate_html(html, cfg.api, cfg.openai_model, cfg.translate_prompt, cfg) except Exception as e: # 翻訳中にエラーが発生した場合、メッセージを出力して保存処理へ進む print(f"\n[ERROR] Translation process interrupted by an exception: {e}") # スタックトレースはGUI側で表示されるため、ここでは簡易的なメッセージに留める # ここで処理を終了せずに、以下の保存処理に進むのが重要 # ------------------ 翻訳結果の保存 (エラーに関わらず実行) ------------------ print() if translated_document is not None: try: if filetype in ("md", "txt") or cfg.use_md: # md, txt, または use_mdモードの場合 print(f"=== Saving revised text to {outfile} ===") save(outfile, translated_document) elif filetype == "html" and not cfg.use_md: # html の場合 print(f"=== Saving revised html to {outfile} ===") save(outfile, translated_document) elif filetype in ("docx", "pdf", "pptx") and not cfg.use_md: # docx, pdf (docx), pptx のドキュメントオブジェクトの場合 save_doc_type = "docx" if filetype in ("docx", "pdf") else "pptx" print(f"=== Saving revised {save_doc_type} to {outfile} ===") translated_document.save(outfile) except Exception as save_e: print(f"[ERROR] Failed to save the partially translated file to {outfile}: {save_e}") # ------------------------------------------------------ # 比較HTML生成 # ------------------------------------------------------ print() print(f"Creating comparison HTML file from template [{cfg.html_template_path}]") if not data and translated_document is None: print("--- [WARNING] No translation data or document object available. Skipping HTML report.") return # 保存対象がないためここで終了 context = cfg.__dict__.copy() context["data"] = data current_dir = os.getcwd() script_dir = os.path.dirname(os.path.abspath(__file__)) # テンプレートファイルの探索ロジック（そのまま維持） if not os.path.isfile(cfg.html_template_path): template_filename = os.path.basename(cfg.html_template_path) cand = os.path.join(current_dir, template_filename) if os.path.isfile(cand): cfg.html_template_path = cand if not os.path.isfile(cfg.html_template_path): template_filename = os.path.basename(cfg.html_template_path) cand = os.path.join(script_dir, template_filename) if os.path.isfile(cand): cfg.html_template_path = cand print(f" Template path: [{cfg.html_template_path}]") if os.path.isfile(cfg.html_template_path): try: template_dir = os.path.dirname(cfg.html_template_path) template_file = os.path.basename(cfg.html_template_path) env = Environment(loader=FileSystemLoader(template_dir)) template = env.get_template(template_file) rendered_html = template.render(context) print() print(f"Save translation data to [{output_comparison_html_path}]") with open(output_comparison_html_path, "w", encoding="utf-8") as f: f.write(rendered_html) print(f" Finished.\n") except Exception as html_e: print(f"[ERROR] Failed to generate comparison HTML report: {html_e}") else: print() print(f"--- [WARNING] Template file '{cfg.html_template_path}' not found.") print(f"--- Looked in current and script directories.") print("--- Skipping generation of comparison HTML report.") print() def main(): global parser cfg, parser = initialize() cfg = update_variables(cfg, parser) execute(cfg) if __name__ == "__main__": main() print() usage() input("\nPress ENTER to terminate>>\n")