o Si#@sUdZddlmZddlZddlZddlZddlZddlZddl Z ddl Z ddl m Z m Z mZddlZddlZddlZejejdedZdZdZd Zd d d d dddd dddd dddd dddd dddd dddd ddd d d!d"d#d d$d%d&d d'd(d)d d*d+d,d d-d.dd d/d0dd d1d2dd d3Zd4ed5<eeZd6ed7<d8Z d9ed:<dQd=d>Z!dRdAdBZ"dSdHdIZ#dTdKdLZ$e fdUdOdPZ%dS)Vu response_iq_tts.py ────────────────────────────────────────────────────────────────────────────── Text-to-Speech module for Response IQ powered by Microsoft Edge Neural TTS. Architecture ──────────── • 100% Serverless/Cloud-based CPU engine (0 VRAM used). • Zero local transformers models required. • Spawns a dedicated ThreadPoolExecutor background pipeline to prevent FastAPI uvloop event-loop blocking. • Decodes cloud-generated MP3 streams to native PCM WAV formats via PyTorch audio decoders seamlessly. Supported languages (14) english · hinglish · hindi · marathi · punjabi · tamil · telugu · kannada gujarati · bengali · odia · assamese · malayalam · urdu Version: 5.0.0 (Edge Neural Migration) ────────────────────────────────────────────────────────────────────────────── ) annotationsN)DictListTuple)levelresponse_iq_ttsi"VWAVzMicrosoft Edge Azure Neural TTSEnglishzen-INzen-IN-NeerjaNeural) display_namebcp47_lang_tagvoiceHinglishzhi-INzhi-IN-SwaraNeuralHindiBengalizbn-INzbn-IN-TanishaaNeuralUrduzur-INzur-IN-GulNeuralMarathizmr-INzmr-IN-AarohiNeuralGujaratizgu-INzgu-IN-DhwaniNeuralPunjabizpa-INzpa-IN-OjasNeuralTamilzta-INzta-IN-PallaviNeuralTeluguzte-INzte-IN-ShrutiNeuralKannadazkn-INzkn-IN-SapnaNeural Malayalamzml-INzml-IN-SobhanaNeuralOdiazor-INAssamesezas-INMaithilizmai-IN)englishhinglishhindibengaliurdumarathigujaratipunjabitamiltelugukannada malayalamodiaassamesemaithilizDict[str, Dict[str, str]]TTS_LANGUAGE_CONFIGz List[str]TTS_SUPPORTED_LANGUAGESrstrTTS_DEFAULT_LANGUAGEtextreturncCs0td|}tdd|}tdd|}|S)z>NFC normalisation and removal of invisible control characters.NFCz [\x00-\x08\x0B\x0C\x0E-\x1F\x7F]z[ \t]+ ) unicodedata normalizeresubstrip)r.r8@/var/www/eduai.edurigo.com/responseiq/staging/response_iq_tts.py_normalise_textOs r:r Tuple[np.ndarray, int]c szddlWn tytdwdfdd }z"tjjdd }||jd d }Wdn1s7wYWntyP}ztd |d}~wwzt j t |d d\}}| }||fWSty}}ztd|d}~ww)z Synthesises high-quality text using Microsoft Azure Neural voices via edge-tts. Executes entirely in a separate thread to cleanly bypass FastAPI uvloop. rNzJThe 'edge-tts' package is missing. Please install it: pip install edge-ttsr/bytescsdfdd }t|S)Nr/r<csN}t}|2z3dHW}|ddkr!||dq6t|S)Ntypeaudiodata) Communicate bytearraystreamextendr<) communicate audio_datachunkedge_ttsr.r r8r9 _generatehs  z?_synthesise_edge_tts.._run_async_tts.._generater/r<)asynciorun)rIrGr8r9_run_async_ttsgs z,_synthesise_edge_tts.._run_async_tts) max_workers<)timeoutz)Edge TTS Cloud Engine generation failed: mp3)formatz(Failed to decode MP3 stream from cloud: rJ)rH ImportError RuntimeError concurrentfuturesThreadPoolExecutorsubmitresult Exception torchaudioloadioBytesIOsqueezecpunumpy) r.r rMexecutor mp3_bytesexcwaveform sample_rateaudio_npr8rGr9_synthesise_edge_ttsZs2    rir> np.ndarraysrc_rateint target_ratecCsH||kr|Sz ddl}|j|||dWSty#td|YSw)Nr)orig_sr target_sru5librosa not installed — skipping fallback resample.)librosaresamplerTloggerwarning)r>rkrmrpr8r8r9_resample_if_neededs  rtrgcCs:t}tj|||ddd|dt|dS)NrPCM_16)rSsubtyperzutf-8) r^r_sfwriteseekbase64 b64encodereaddecode)r>rgbufr8r8r9_audio_to_base64_wavs rlanguagerc Cs*|pt}|tvrdd|dtiS|r|s ddiSt|}t|}|d}td|d|dd |d t|z3t ||\}}t ||t }t |t }t t|t d }td |d t|||t |t|dWSty} ztd| dt| iWYd} ~ Sd} ~ ww)u` Convert text to human-like speech using Microsoft Edge Neural Cloud APIs. Returns (success) ───────────────── { "audio_tts" : "", "language" : "hindi", "sample_rate" : 22050, "duration_sec": 4.83, "char_count" : 232, } errorzUnsupported language 'z'. Valid values: z>Text payload is empty. Provide a non-empty 'predefinedAnswer'.r zTTS Cloud Request | language=z (r z ) | voice=z | chars=z%TTS Cloud Output complete | duration=zs | base64_len=) audio_ttsrrg duration_sec char_countzTTS Cloud Pipeline failed: N)r-r7lowerr*r+r:rrinfolenrirtTARGET_SAMPLE_RATErroundr[rr,) r.rlang_cfgr rhraw_sample_rate final_audio audio_b64rrer8r8r9 generate_ttssP   r)r.r,r/r,)r.r,r r,r/r;)r>rjrkrlrmrlr/rj)r>rjrgrlr/r,)r.r,rr,r/r)&__doc__ __future__rrKrzconcurrent.futuresrVr^loggingr5r3typingrrrrbnp soundfilerwr\ basicConfigINFO getLoggerrrr OUTPUT_FORMAT TTS_MODEL_IDr*__annotations__sortedkeysr+r-r:rirtrrr8r8r8r9sR                    *