app.py

import os
import time
import shutil
from pydub import AudioSegment
from openai import OpenAI
import gradio as gr
from fastapi import HTTPException

# ======================================================
# 🔐 設定區
# ======================================================
PASSWORD = os.getenv("APP_PASSWORD", "defaultpass")
MAX_SIZE = 25 * 1024 * 1024
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

print("===== 🚀 啟動中 =====")
print(f"APP_PASSWORD: {'✅ 已載入' if PASSWORD else '❌ 未載入'}")
print(f"目前密碼內容：{PASSWORD}")

# ======================================================
# 🎧 音訊轉錄核心
# ======================================================
def split_audio_if_needed(path):
    size = os.path.getsize(path)
    if size <= MAX_SIZE:
        return [path]
    audio = AudioSegment.from_file(path)
    num = int(size / MAX_SIZE) + 1
    chunk_ms = len(audio) / num
    files = []
    for i in range(num):
        start, end = int(i * chunk_ms), int((i + 1) * chunk_ms)
        chunk = audio[start:end]
        fn = f"chunk_{i+1}.wav"
        chunk.export(fn, format="wav")
        files.append(fn)
    return files

def transcribe_core(path, model="whisper-1"):
    if path and path.lower().endswith(".mp4"):
        fixed_path = path[:-4] + ".m4a"
        try:
            shutil.copy(path, fixed_path)
            path = fixed_path
            print("🔧 已自動修正 mp4 → m4a")
        except Exception as e:
            print(f"⚠️ mp4→m4a 轉檔失敗：{e}")

    chunks = split_audio_if_needed(path)
    raw_parts = []
    for f in chunks:
        with open(f, "rb") as af:
            res = client.audio.transcriptions.create(
                model=model,
                file=af,
                response_format="text"
            )
            raw_parts.append(res)
    full_raw = "\n".join(raw_parts)

    # 簡轉繁
    conv_prompt = (
        "請將以下內容完整轉換為「繁體中文（台灣用語）」：\n"
        "規則：1) 僅做簡→繁字形轉換；2) 不要意譯或改寫；3) 不要添加任何前後綴。\n"
        "-----\n" + full_raw
    )
    trad_resp = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": "你是嚴格的繁體中文轉換器，只進行字形轉換。"},
            {"role": "user", "content": conv_prompt}
        ],
        temperature=0.0,
    )
    full_trad = trad_resp.choices[0].message.content.strip()

    # 摘要
    sum_prompt = (
        "請用台灣繁體中文撰寫摘要。若內容資訊多，可條列出重點；若內容簡短，請用一句話概述即可。\n\n"
        + full_trad
    )
    sum_resp = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {"role": "system", "content": "你是一位精準且嚴格使用台灣繁體中文的摘要助手。"},
            {"role": "user", "content": sum_prompt}
        ],
        temperature=0.2,
    )
    summ = sum_resp.choices[0].message.content.strip()
    return full_trad, summ

# ======================================================
# 💬 Gradio 介面
# ======================================================
def transcribe_with_password(password, file):
    if password.strip() != PASSWORD:
        raise HTTPException(status_code=403, detail="密碼錯誤 ❌")
    if not file:
        return "⚠️ 未選擇檔案", "", ""
    text, summary = transcribe_core(file.name)
    return "✅ 完成", text, summary

with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("## 🎧 LINE 語音轉錄與摘要（支援 .m4a / .mp4）")
    pw = gr.Textbox(label="輸入密碼", type="password")
    f = gr.File(label="上傳音訊檔")
    run = gr.Button("開始轉錄 🚀")
    s = gr.Textbox(label="狀態", interactive=False)
    t = gr.Textbox(label="轉錄結果", lines=10)
    su = gr.Textbox(label="AI 摘要", lines=8)
    run.click(transcribe_with_password, [pw, f], [s, t, su])

# ======================================================
# 🚀 啟動
# ======================================================
if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)
else:
    demo.launch()