Update app.py

app.py

@@ -1,13 +1,12 @@
 import os
 import shutil
 import hashlib
-from fastapi import FastAPI, UploadFile, File, Form
+from fastapi import FastAPI, Form
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.staticfiles import StaticFiles
 from fastapi.responses import JSONResponse
-from pydub import AudioSegment
 import google.generativeai as genai
-from gradio_client import Client, handle_file  # 👈 迎回完全正常的 handle_file
+from gradio_client import Client
 import uvicorn
 
 app = FastAPI()
@@ -26,27 +25,21 @@ if not os.path.exists("static"):
     os.makedirs("static")
 app.mount("/static", StaticFiles(directory="static"), name="static")
 
-# 💡 3. API Keys & Models 初始化 (對對接 2026 最新 Gemini 3.5 Flash)
+# 💡 3. API Keys & Models 初始化 (對接 2026 最新 Gemini 3.5 Flash)
 GOOGLE_API_KEY = os.getenv('GEMINI_KEY', '請填入您的GEMINI_KEY')
 genai.configure(api_key=GOOGLE_API_KEY)
 MODEL_NAME = 'gemini-3.5-flash' 
 model = genai.GenerativeModel(MODEL_NAME) 
 
-# 💡 4. 【完全複製你最原始、完全不加額外參數的初始化寫法】
-asr_client = Client("https://ai-labs.ilrdf.org.tw/sapolita-kaldi/")
+# 💡 4. 初始化原語會的翻譯與 TTS 伺服器
 tts_client = Client("https://ai-labs.ilrdf.org.tw/hnang-kari-ai-asi-sluhay/")
 mt_client  = Client("https://ai-labs.ilrdf.org.tw/kari-seejiq-tnpusu-ai-hmjil/")
 
-# 【完全複製你原本的族語配置】
 TRIBE_CONFIG = {
-    "阿美": {"asr": "formosan_ami", "mt": "阿美"}, "泰雅": {"asr": "formosan_tay", "mt": "泰雅"},
-    "排灣": {"asr": "formosan_pwn", "mt": "排灣"}, "布農": {"asr": "formosan_bnn", "mt": "布農"},
-    "卑南": {"asr": "formosan_pyu", "mt": "卑南"}, "魯凱": {"asr": "formosan_dru", "mt": "魯凱"},
-    "鄒": {"asr": "formosan_tsu", "mt": "鄒"}, "賽夏": {"asr": "formosan_xsy", "mt": "賽夏"},
-    "雅美": {"asr": "formosan_tao", "mt": "雅美"}, "邵": {"asr": "formosan_ssf", "mt": "邵"},
-    "噶瑪蘭": {"asr": "formosan_ckv", "mt": "噶瑪蘭"}, "太魯閣": {"asr": "formosan_trv", "mt": "太魯閣"},
-    "撒奇萊雅": {"asr": "formosan_szy", "mt": "撒奇萊雅"}, "賽德克": {"asr": "formosan_sdq", "mt": "賽德克"},
-    "拉阿魯哇": {"asr": "formosan_sxr", "mt": "拉阿魯哇"}, "卡那卡那富": {"asr": "formosan_xnb", "mt": "卡那卡那富"}
+    "阿美": {"mt": "阿美"}, "泰雅": {"mt": "泰雅"}, "排灣": {"mt": "排灣"}, "布農": {"mt": "布農"},
+    "卑南": {"mt": "卑南"}, "魯凱": {"mt": "魯凱"}, "鄒": {"mt": "鄒"}, "賽夏": {"mt": "賽夏"},
+    "雅美": {"mt": "雅美"}, "邵": {"mt": "邵"}, "噶瑪蘭": {"mt": "噶瑪蘭"}, "太魯閣": {"mt": "太魯閣"},
+    "撒奇萊雅": {"mt": "撒奇萊雅"}, "賽德克": {"mt": "賽德克"}, "拉阿魯哇": {"mt": "拉阿魯哇"}, "卡那卡那富": {"mt": "卡那卡那富"}
 }
 
 def get_clean_value(res):
@@ -62,50 +55,31 @@ def format_native_text(text, tribe):
 
 @app.get("/")
 def read_root():
-    return {"status": "ILRDF Avatar Backend is Running Successfully!", "model_in_use": MODEL_NAME}
+    return {"status": "ILRDF Avatar Text Brain is Running Successfully!", "model": MODEL_NAME}
 
 @app.post("/api/chat")
 async def avatar_chat(
     tribe: str = Form(...), 
-    audio: UploadFile = File(...)
+    text: str = Form(...)  # 👈 接收前端秒傳過來的中文文字
 ):
     try:
         config = TRIBE_CONFIG.get(tribe)
         if not config:
             return JSONResponse({"error": f"不支援的族語: {tribe}"}, status_code=400)
 
-        # --- 步驟 A: 處理網頁傳來的錄音檔 ---
-        temp_in = f"static/in_{audio.filename}"
-        with open(temp_in, "wb") as buffer:
-            shutil.copyfileobj(audio.file, buffer)
+        zh_in = text.strip()
 
-        wav_path = f"static/in_converted.wav"
-        audio_segment = AudioSegment.from_file(temp_in)
-        audio_segment.export(wav_path, format="wav")
-
-        # --- 步驟 B: 【100% 複製你 handle_audio 裡的 ASR 程式碼】 ---
-        native_in = asr_client.predict(
-            dialect_id=config["asr"], 
-            audio_data=handle_file(wav_path), 
-            api_name="/automatic_speech_recognition"
-        )
-
-        # --- 步驟 C: 【100% 複製你 handle_audio 裡的 中文翻譯 程式碼】 ---
-        go_code = get_clean_value(mt_client.predict(ethnicity=config["mt"], api_name="/lambda"))
-        zh_in = get_clean_value(mt_client.predict(text=native_in, src_lang=go_code, tgt_lang="zho_Hant", api_name="/translate"))
-        native_in = format_native_text(native_in, tribe)
-
-        # --- 步驟 D: AI 大腦生成對話 (Gemini 3.5 Flash) ---
+        # --- 步驟 A: AI 大腦生成對話 (Gemini 3.5 Flash) ---
         prompt = f"你現在是與我對話的{tribe}族朋友。請用中文聊天。回覆規則：1.口吻生活化親切。2.不要說教解釋。3.字數一定要少（15字內），限一個短句。\n\n使用者說：{zh_in}"
         response = model.generate_content(prompt)
         ai_zh = response.text.strip()
 
-        # --- 步驟 E: 【100% 複製你一般聊天模式裡的 族語翻譯 程式碼】 ---
+        # --- 步驟 B: 翻譯回族語 ---
         back_code = get_clean_value(mt_client.predict(ethnicity=config["mt"], api_name="/lambda_1"))
         ai_native = get_clean_value(mt_client.predict(text=ai_zh, src_lang="zho_Hant", tgt_lang=back_code, api_name="/translate_1"))
         ai_native = format_native_text(ai_native, tribe)
 
-        # --- 步驟 F: 【100% 複製你 handle_text 裡的 TTS 語音合成與路徑搬移程式碼】 ---
+        # --- 步驟 C: 語音合成 (TTS) ---
         speaker = get_clean_value(tts_client.predict(ethnicity=config["mt"], api_name="/lambda"))
         if tribe == "阿美": speaker = "阿美_秀姑巒_女聲1"
         temp_tts = tts_client.predict(ref=speaker, gen_text_input=ai_native, api_name="/default_speaker_tts")
@@ -114,7 +88,7 @@ async def avatar_chat(
         out_filepath = f"static/{unique_str}" 
         shutil.move(temp_tts, out_filepath)
 
-        # --- 步驟 G: 組合網址與清理環境 ---
+        # --- 步驟 D: 組合完整語音網址 ---
         base_url = os.getenv('SPACE_HOST', '') 
         if base_url:
             base_url = f"https://{base_url}"
@@ -123,11 +97,8 @@ async def avatar_chat(
 
         audio_url = f"{base_url}/{out_filepath}"
 
-        os.remove(temp_in)
-        os.remove(wav_path)
-
         return {
-            "user_native_text": native_in,
+            "user_zh_text": zh_in,
             "ai_zh_text": ai_zh,
             "ai_native_text": ai_native,
             "audio_url": audio_url