Upload 4 files

app.py

@@ -0,0 +1,45 @@
+from fastapi import FastAPI, Form
+from fastapi.middleware.cors import CORSMiddleware
+from translator import standard_translator, smart_auto_translator
+from utils import source_tts_handler, target_tts_handler
+
+app = FastAPI(title="Clean Translator Backend", version="2.0")
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.get("/")
+async def root():
+    return {"status": "ok", "message": "Backend running (In-Memory Mode) 🚀"}
+
+# --- TRANSLATION ENDPOINTS (Unchanged) ---
+@app.post("/translate/manual")
+async def manual_translate(text: str = Form(...), src_lang: str = Form(...), tgt_lang: str = Form(...)):
+    translated = standard_translator(text, src_lang, tgt_lang)
+    return {"mode": "manual", "source_text": text, "translated_text": translated}
+
+@app.post("/translate/auto")
+async def auto_translate(text: str = Form(...), target_lang: str = Form(...)):
+    result = smart_auto_translator(text, target_lang)
+    return result
+
+# --- TTS ENDPOINTS (Updated for Base64) ---
+
+@app.post("/tts/source")
+async def source_tts(text: str = Form(...)):
+    b64_audio = await source_tts_handler(text)
+    if not b64_audio:
+        return {"error": "Could not generate audio"}
+    return {"audio_base64": b64_audio}
+
+@app.post("/tts/target")
+async def target_tts(translated_text: str = Form(...), target_lang: str = Form(...)):
+    b64_audio = await target_tts_handler(translated_text, target_lang)
+    if not b64_audio:
+        return {"error": "Could not generate audio"}
+    return {"audio_base64": b64_audio}

requirements.txt

@@ -0,0 +1,11 @@
+langdetect
+transformers
+torch
+sentencepiece
+groq
+edge-tts
+dotenv
+uuid
+fastapi 
+uvicorn
+python-multipart

translator.py

@@ -0,0 +1,229 @@
+# translator.py
+
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+import torch
+from langdetect import detect
+
+MODEL_NAME = "facebook/nllb-200-distilled-600M"
+
+# ===============================
+# LANGUAGE MAP
+# ===============================
+LANGUAGE_MAP = {
+    # ===== EXISTING =====
+    "english": "eng_Latn", "french": "fra_Latn", "german": "deu_Latn", "spanish": "spa_Latn",
+    "hindi": "hin_Deva", "marathi": "mar_Deva", "tamil": "tam_Taml", "telugu": "tel_Telu",
+    "kannada": "kan_Knda", "bengali": "ben_Beng", "urdu": "urd_Arab", "arabic": "arb_Arab",
+    "persian": "pes_Arab", "japanese": "jpn_Jpan", "chinese": "zho_Hans", "korean": "kor_Hang",
+    "russian": "rus_Cyrl", "italian": "ita_Latn", "portuguese": "por_Latn", "dutch": "nld_Latn",
+    "swedish": "swe_Latn", "norwegian": "nob_Latn", "danish": "dan_Latn", "finnish": "fin_Latn",
+    "polish": "pol_Latn", "czech": "ces_Latn", "slovak": "slk_Latn", "hungarian": "hun_Latn",
+    "romanian": "ron_Latn", "bulgarian": "bul_Cyrl", "ukrainian": "ukr_Cyrl", "greek": "ell_Grek",
+    "gujarati": "guj_Gujr", "punjabi": "pan_Guru", "malayalam": "mal_Mlym",
+    "thai": "tha_Thai", "vietnamese": "vie_Latn", "indonesian": "ind_Latn",
+    "turkish": "tur_Latn", "hebrew": "heb_Hebr",
+
+    # ===== KURDISH (IMPORTANT) =====
+    "kurdish_kurmanji": "kmr_Latn",   # Northern Kurdish (Latin)
+    "kurdish_sorani": "ckb_Arab",     # Central Kurdish (Arabic)
+
+    # ===== SOUTH ASIAN =====
+    "nepali": "npi_Deva",
+    "sinhala": "sin_Sinh",
+    "odia": "ory_Orya",
+    "assamese": "asm_Beng",
+    "maithili": "mai_Deva",
+    "santali": "sat_Olck",
+
+    # ===== SOUTHEAST ASIA =====
+    "malay": "zsm_Latn",
+    "filipino": "tgl_Latn",
+    "khmer": "khm_Khmr",
+    "lao": "lao_Laoo",
+    "burmese": "mya_Mymr",
+
+    # ===== EAST ASIA =====
+    "traditional_chinese": "zho_Hant",
+    "mongolian": "mon_Cyrl",
+
+    # ===== CENTRAL ASIA =====
+    "kazakh": "kaz_Cyrl",
+    "uzbek": "uzn_Latn",
+    "tajik": "tgk_Cyrl",
+    "kyrgyz": "kir_Cyrl",
+    "turkmen": "tuk_Latn",
+
+    # ===== MIDDLE EAST =====
+    "pashto": "pbt_Arab",
+    "sindhi": "snd_Arab",
+
+    # ===== AFRICAN =====
+    "swahili": "swh_Latn",
+    "amharic": "amh_Ethi",
+    "yoruba": "yor_Latn",
+    "igbo": "ibo_Latn",
+    "hausa": "hau_Latn",
+    "zulu": "zul_Latn",
+    "xhosa": "xho_Latn",
+    "somali": "som_Latn",
+    "afrikaans": "afr_Latn",
+
+    # ===== EUROPE EXTRA =====
+    "estonian": "est_Latn",
+    "latvian": "lav_Latn",
+    "lithuanian": "lit_Latn",
+    "icelandic": "isl_Latn",
+    "irish": "gle_Latn",
+    "welsh": "cym_Latn",
+    "albanian": "sqi_Latn",
+    "serbian": "srp_Cyrl",
+    "croatian": "hrv_Latn",
+    "slovenian": "slv_Latn",
+
+    # ===== OTHERS =====
+    "latin": "lat_Latn",
+    "esperanto": "epo_Latn"
+}
+
+ISO_TO_LANGUAGE_KEY = {
+    # ===== EXISTING =====
+    "en": "english", "fr": "french", "de": "german", "es": "spanish", "hi": "hindi",
+    "mr": "marathi", "ta": "tamil", "te": "telugu", "kn": "kannada", "bn": "bengali",
+    "ur": "urdu", "ar": "arabic", "fa": "persian", "ja": "japanese", "zh": "chinese",
+    "ko": "korean", "ru": "russian", "it": "italian", "pt": "portuguese", "nl": "dutch",
+    "sv": "swedish", "no": "norwegian", "da": "danish", "fi": "finnish", "pl": "polish",
+    "cs": "czech", "sk": "slovak", "hu": "hungarian", "ro": "romanian", "bg": "bulgarian",
+    "uk": "ukrainian", "el": "greek", "gu": "gujarati", "pa": "punjabi", "ml": "malayalam",
+    "th": "thai", "vi": "vietnamese", "id": "indonesian", "tr": "turkish", "he": "hebrew",
+
+    # ===== KURDISH =====
+    "ku": "kurdish_kurmanji",   # default kurdish
+    "ckb": "kurdish_sorani",
+
+    # ===== SOUTH ASIAN =====
+    "ne": "nepali",
+    "si": "sinhala",
+    "or": "odia",
+    "as": "assamese",
+    "mai": "maithili",
+    "sat": "santali",
+
+    # ===== SOUTHEAST ASIA =====
+    "ms": "malay",
+    "tl": "filipino",
+    "km": "khmer",
+    "lo": "lao",
+    "my": "burmese",
+
+    # ===== CENTRAL ASIA =====
+    "kk": "kazakh",
+    "uz": "uzbek",
+    "tg": "tajik",
+    "ky": "kyrgyz",
+    "tk": "turkmen",
+
+    # ===== MIDDLE EAST =====
+    "ps": "pashto",
+    "sd": "sindhi",
+
+    # ===== AFRICAN =====
+    "sw": "swahili",
+    "am": "amharic",
+    "yo": "yoruba",
+    "ig": "igbo",
+    "ha": "hausa",
+    "zu": "zulu",
+    "xh": "xhosa",
+    "so": "somali",
+    "af": "afrikaans",
+
+    # ===== EUROPE EXTRA =====
+    "et": "estonian",
+    "lv": "latvian",
+    "lt": "lithuanian",
+    "is": "icelandic",
+    "ga": "irish",
+    "cy": "welsh",
+    "sq": "albanian",
+    "sr": "serbian",
+    "hr": "croatian",
+    "sl": "slovenian",
+
+    # ===== OTHERS =====
+    "la": "latin",
+    "eo": "esperanto"
+}
+
+
+# ===============================
+# LOAD MODEL ONCE
+# ===============================
+print("Loading translation model...")
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
+print("Model loaded successfully.")
+
+# ===============================
+# HELPERS
+# ===============================
+def get_nllb_code(lang: str):
+    lang = lang.lower().strip()
+
+    if lang in LANGUAGE_MAP:
+        return LANGUAGE_MAP[lang]
+
+    if lang in ISO_TO_LANGUAGE_KEY:
+        key = ISO_TO_LANGUAGE_KEY[lang]
+        return LANGUAGE_MAP[key]
+
+    return "eng_Latn"  # fallback
+
+
+# ===============================
+# MANUAL TRANSLATOR
+# ===============================
+def standard_translator(text: str, src_lang: str, tgt_lang: str) -> str:
+    src_code = get_nllb_code(src_lang)
+    tgt_code = get_nllb_code(tgt_lang)
+
+    tokenizer.src_lang = src_code
+    inputs = tokenizer(text, return_tensors="pt")
+
+    with torch.no_grad():
+        output = model.generate(
+            **inputs,
+            forced_bos_token_id=tokenizer.convert_tokens_to_ids(tgt_code),
+            max_length=512
+        )
+
+    return tokenizer.decode(output[0], skip_special_tokens=True)
+
+
+# ===============================
+# AUTO TRANSLATOR
+# ===============================
+def smart_auto_translator(text: str, target_lang: str):
+    detected_iso = detect(text)
+    detected_lang_key = ISO_TO_LANGUAGE_KEY.get(detected_iso, detected_iso)
+
+    src_code = get_nllb_code(detected_iso)
+    tgt_code = get_nllb_code(target_lang)
+
+    tokenizer.src_lang = src_code
+    inputs = tokenizer(text, return_tensors="pt")
+
+    with torch.no_grad():
+        output = model.generate(
+            **inputs,
+            forced_bos_token_id=tokenizer.convert_tokens_to_ids(tgt_code),
+            max_length=512
+        )
+
+    translated_text = tokenizer.decode(output[0], skip_special_tokens=True)
+
+    return {
+        "translated_text": translated_text,
+        "detected_iso": detected_iso,
+        "detected_language": detected_lang_key
+    }
+

utils.py

@@ -0,0 +1,184 @@
+# utils.py
+
+
+# ===============================
+# VOICE MAP
+# ===============================
+VOICE_MAP = {
+    # ===== EXISTING (UNCHANGED) =====
+    "english": "en-US-AriaNeural",
+    "french": "fr-FR-DeniseNeural",
+    "german": "de-DE-KatjaNeural",
+    "spanish": "es-ES-ElviraNeural",
+    "hindi": "hi-IN-SwaraNeural",
+    "arabic": "ar-SA-ZariyahNeural",
+    "japanese": "ja-JP-NanamiNeural",
+    "korean": "ko-KR-SunHiNeural",
+    "chinese": "zh-CN-XiaoxiaoNeural",
+    "russian": "ru-RU-SvetlanaNeural",
+    "marathi": "mr-IN-AarohiNeural",
+    "tamil": "ta-IN-PallaviNeural",
+    "telugu": "te-IN-ShrutiNeural",
+    "kannada": "kn-IN-SapnaNeural",
+    "bengali": "bn-IN-TanishaaNeural",
+    "urdu": "ur-PK-UzmaNeural",
+    "gujarati": "gu-IN-DhwaniNeural",
+    "punjabi": "pa-IN-GurpreetNeural",
+    "malayalam": "ml-IN-SobhanaNeural",
+    "italian": "it-IT-ElsaNeural",
+    "portuguese": "pt-PT-RaquelNeural",
+    "dutch": "nl-NL-ColetteNeural",
+    "swedish": "sv-SE-SofieNeural",
+    "norwegian": "nb-NO-IselinNeural",
+    "danish": "da-DK-ChristelNeural",
+    "finnish": "fi-FI-NooraNeural",
+    "polish": "pl-PL-ZofiaNeural",
+    "czech": "cs-CZ-VlastaNeural",
+    "slovak": "sk-SK-ViktoriaNeural",
+    "hungarian": "hu-HU-NoemiNeural",
+    "romanian": "ro-RO-AlinaNeural",
+    "bulgarian": "bg-BG-KalinaNeural",
+    "ukrainian": "uk-UA-PolinaNeural",
+    "greek": "el-GR-AthinaNeural",
+    "thai": "th-TH-PremwadeeNeural",
+    "vietnamese": "vi-VN-HoaiMyNeural",
+    "indonesian": "id-ID-GadisNeural",
+    "turkish": "tr-TR-EmelNeural",
+    "hebrew": "he-IL-HilaNeural",
+
+    # ===== NEW – Edge TTS Supported =====
+
+    # South Asia
+    "nepali": "ne-NP-HemkalaNeural",
+    "sinhala": "si-LK-ThiliniNeural",
+
+    # Southeast Asia
+    "malay": "ms-MY-YasminNeural",
+    "filipino": "fil-PH-BlessicaNeural",
+    "khmer": "km-KH-SreymomNeural",
+    "lao": "lo-LA-KeomanyNeural",
+    "burmese": "my-MM-NilarNeural",
+
+    # Central Asia
+    "kazakh": "kk-KZ-AigulNeural",
+    "uzbek": "uz-UZ-MadinaNeural",
+
+    # Africa
+    "swahili": "sw-KE-ZuriNeural",
+    "amharic": "am-ET-MekdesNeural",
+    "zulu": "zu-ZA-ThandoNeural",
+    "xhosa": "xh-ZA-NolwaziNeural",
+    "afrikaans": "af-ZA-AdriNeural",
+
+    # Middle East
+    "azerbaijani": "az-AZ-BanuNeural",
+    "persian": "fa-IR-DilaraNeural",
+
+    # Europe Extra
+    "estonian": "et-EE-AnuNeural",
+    "latvian": "lv-LV-NeveraNeural",
+    "lithuanian": "lt-LT-OnaNeural",
+    "icelandic": "is-IS-GudrunNeural",
+    "irish": "ga-IE-OrlaNeural",
+    "welsh": "cy-GB-NiaNeural",
+    "albanian": "sq-AL-AnilaNeural",
+    "serbian": "sr-RS-SophieNeural",
+    "croatian": "hr-HR-GabrijelaNeural",
+    "slovenian": "sl-SI-PetraNeural"
+}
+ISO_TO_LANGUAGE_KEY = {
+    # ===== EXISTING =====
+    "en": "english", "fr": "french", "de": "german", "es": "spanish", "hi": "hindi",
+    "mr": "marathi", "ta": "tamil", "te": "telugu", "kn": "kannada", "bn": "bengali",
+    "ur": "urdu", "ar": "arabic", "fa": "persian", "ja": "japanese", "zh": "chinese",
+    "ko": "korean", "ru": "russian", "it": "italian", "pt": "portuguese", "nl": "dutch",
+    "sv": "swedish", "no": "norwegian", "da": "danish", "fi": "finnish", "pl": "polish",
+    "cs": "czech", "sk": "slovak", "hu": "hungarian", "ro": "romanian", "bg": "bulgarian",
+    "uk": "ukrainian", "el": "greek", "gu": "gujarati", "pa": "punjabi", "ml": "malayalam",
+    "th": "thai", "vi": "vietnamese", "id": "indonesian", "tr": "turkish", "he": "hebrew",
+
+    # ===== NEW =====
+    "ne": "nepali",
+    "si": "sinhala",
+    "ms": "malay",
+    "tl": "filipino",
+    "km": "khmer",
+    "lo": "lao",
+    "my": "burmese",
+    "kk": "kazakh",
+    "uz": "uzbek",
+    "sw": "swahili",
+    "am": "amharic",
+    "zu": "zulu",
+    "xh": "xhosa",
+    "af": "afrikaans",
+    "az": "azerbaijani",
+    "et": "estonian",
+    "lv": "latvian",
+    "lt": "lithuanian",
+    "is": "icelandic",
+    "ga": "irish",
+    "cy": "welsh",
+    "sq": "albanian",
+    "sr": "serbian",
+    "hr": "croatian",
+    "sl": "slovenian"
+}
+
+
+import os
+import time
+from uuid import uuid4
+import edge_tts
+from langdetect import detect, LangDetectException
+from groq import Groq
+from dotenv import load_dotenv
+import base64
+
+load_dotenv()
+client = Groq()
+
+# ===============================
+# CONFIGURATION
+# ===============================
+STATIC_DIR = "static"
+DELETE_AFTER_SECONDS = 300  # Delete audio files older than 5 minutes
+
+# ===============================
+# MEMORY TTS (Base64)
+# ===============================
+async def TTS(text: str, voice: str):
+    """Generates audio in memory and returns Base64 string"""
+    communicate = edge_tts.Communicate(text, voice)
+    audio_data = b""
+
+    # Stream audio bytes into memory
+    async for chunk in communicate.stream():
+        if chunk["type"] == "audio":
+            audio_data += chunk["data"]
+    
+    # Encode to Base64 string
+    b64_string = base64.b64encode(audio_data).decode('utf-8')
+    return b64_string
+
+# ===============================
+# HANDLERS
+# ===============================
+async def source_tts_handler(text: str):
+    if not text or not text.strip():
+        return None
+    try:
+        iso = detect(text)
+        lang_key = ISO_TO_LANGUAGE_KEY.get(iso, "english")
+        voice = VOICE_MAP.get(lang_key, "en-US-AriaNeural")
+    except LangDetectException:
+        voice = "en-US-AriaNeural"
+
+    return await TTS(text, voice)
+
+async def target_tts_handler(text: str, target_lang: str):
+    if not text:
+        return None
+    clean_lang = target_lang.lower().strip()
+    voice = VOICE_MAP.get(clean_lang, "en-US-AriaNeural")
+    return await TTS(text, voice)