§ øžyjpHãó–—UdZddlmZddlZddlZddlZddlmZmZddl m Z mZmZm Z mZddlmZddlmZmZmZeje¦«ZdZd=d „ZdZdZdaded<d>d„Zd?d„Zej d¦«Z!ej dej"¬¦«Z#Gd„d¦«Z$Gd„de¦«Z%iZ&ded<d@d!„Z'dAd%„Z(gd&¢Z)d'ed(< dBdCd+„Z*e'd,¦«Gd-„d.e%¦«¦«Z+dDd/„Z,e'd0¦«Gd1„d2e%¦«¦«Z-dEd6„Z.e'd7¦«Gd8„d9e%¦«¦«Z/e'd:¦«Gd;„d.streaming`), and resolver come free. é)ÚannotationsN)ÚABCÚabstractmethod)ÚCallableÚDictÚIteratorÚListÚOptional)Úresolve_openai_audio_api_key)Ú _get_providerÚ_load_tts_configÚ get_env_valueiÚenv_varÚstrÚprovider_idÚreturncón— ddlm}|||¦«pdS#t$rt|¦«pdcYSwxYw)uProvider secret lookup: config > env/.env > credential pool. Thin, monkeypatchable seam over ``tools.tts_tool._resolve_provider_key`` (which delegates to ``resolve_provider_secret``). ALL streaming-provider key lookups go through here â€” never bare ``get_env_value``. r)Ú_resolve_provider_keyÚ)Útools.tts_toolrÚ Exceptionr)rrrs ú9/home/ragecks/.hermes/hermes-agent/tools/tts_streaming.pyÚ_resolve_keyr*se€ð,Ø8Ð8Ð8Ð8Ð8Ð8à$Ð$ W¨kÑ:Ô:Ð@¸bÐ@øÝð,ð,ð,Ý˜WÑ%Ô%Ð+¨Ð+Ð+Ð+ð,øøøs‚–4³4zK[Note: the user interrupted your previous spoken reply before it finished.]g^@zOptional[float]Ú_interrupted_atÚNonecó,—tj¦«adS©N)ÚtimeÚ monotonicr©órÚmark_speech_interruptedr"Is€å”nÑ&Ô&€O€O€Or!Úboolcó\—tdc}a|duotj¦«|z tkS)z9Pop the latch; True when a barge happened within the TTL.N)rrrÚ_INTERRUPT_TTL_S)Úats rÚtake_speech_interruptedr'Ns2€õ*¨4Ð€BˆØ Tˆ>ÐFdœnÑ.Ô.°Ñ3Õ6FÒFÐFr!z(?<=[.!?])(?:\s|\n)|(?:\n\n)z].*?)Úflagscó,—eZdZdZd dd„Zdd „Zdd„ZdS)ÚSentenceChunkeraŸIncremental sentence cutter for LLM token deltas. Shared by the speaker pipeline (`stream_tts_to_speaker`) and the speak-stream WebSocket so every surface cuts speech identically. Strips ```` blocks (even split across deltas) and merges fragments shorter than *min_len* into the following sentence, so "Ha!" rides along with the sentence after it instead of stalling as a tiny clip. éÚmin_lenÚintcó"—||_d|_dS)Nr)r,Úbuf)Úselfr,s rÚ__init__zSentenceChunker.__init__cs€ØˆŒØˆŒˆˆr!Údeltarrú List[str]cóH—t d|j|z¦«|_d|jvrd|jvrgSg}d}t |j|¦«x}r¿|jd| ¦«…}t | ¦«¦«|jkr| ¦«}Œ‚| |¦«|j| ¦«d…|_d}t |j|¦«x}°¿|S)zBAbsorb *delta*; return every complete sentence now ready to speak.rzrN) Ú_THINK_BLOCK_REÚsubr/ÚSENTENCE_BOUNDARY_REÚsearchÚendÚlenÚstripr,Úappend)r0r2ÚoutÚstartÚmÚheads rÚfeedzSentenceChunker.feedgs€å"×&Ò& r¨4¬8°eÑ+;Ñ<Ô<ˆŒØt”xÐÐ J°d´hÐ$>Ð$>ØˆIØˆØˆÝ'×.Ò.¨t¬x¸Ñ?Ô?Ð?ˆað Ø”8˜I˜aŸeše™gœg˜IÔ&ˆDÝ4—:’:‘<”<Ñ Ô 4¤<Ò/Ð/ØŸš™œØØJŠJtÑÔÐØ”x §¢¡¤ Ô)ˆDŒHØˆEõ(×.Ò.¨t¬x¸Ñ?Ô?Ð?ˆað ðˆ r!có‚—t d|j¦« ¦«}d|_|r|gngS)z0Drain the tail (end-of-text or long-idle flush).r)r5r6r/r;)r0Útails rÚflushzSentenceChunker.flushxs=€å×"Ò" 2 t¤xÑ0Ô0×6Ò6Ñ8Ô8ˆØˆŒØÐ%ˆvˆv 2Ð%r!N)r+)r,r-)r2rrr3)rr3)Ú__name__Ú __module__Ú__qualname__Ú__doc__r1rArDr r!rr*r*Ys_€€€€€ðððððððððððð"&ð&ð&ð&ð&ð&r!r*có†—eZdZUdZdZded<dZded<dZded<dd„Ze e dd„¦«¦«Ze dd„¦«ZdS)ÚStreamingTTSProviderzDYields raw int16, little-endian, mono PCM chunks at ``sample_rate``.éÀ]r-Úsample_rateéÚchannelséÚsample_widthÚ tts_configrÚsectioncó"—||_||_dSr)rQrR)r0rQrRs rr1zStreamingTTSProvider.__init__Šs€Ø$ˆŒØˆŒˆˆr!rr#có—dS)z?True when this provider's credentials/SDK are usable right now.Nr r r!rÚ availablezStreamingTTSProvider.availableŽó€€€r!ÚtextrúIterator[bytes]có—dS)z>Yield PCM chunks for ``text``. Raise on failure (caller logs).Nr ©r0rWs rÚstreamzStreamingTTSProvider.stream“rVr!N)rQrrRr©rr#©rWrrrX) rErFrGrHrLÚ__annotations__rNrPr1ÚstaticmethodrrUr[r r!rrJrJƒs·€€€€€€ØNÐNà€KÐÐÐÑØ€HÐÐÐÑØ€LÐÐÐÑðððððØðNðNðNñ„^ñ„\ðNððMðMðMñ„^ðMðMðMr!rJz%Dict[str, type[StreamingTTSProvider]]Ú _REGISTRYÚnameúBCallable[[type[StreamingTTSProvider]], type[StreamingTTSProvider]]có‡—dˆfd„}|S)NÚclsútype[StreamingTTSProvider]rcó•—|t‰<|Sr)r`)rdras €rÚ_wrapzregister.._wrapœsø€Ø $‰Øˆ r!)rdrerrer )rargs` rÚregisterrh›s(ø€ððððððð€Lr!rQrúOptional[StreamingTTSProvider]có—t |¦«}|| ¦«sdS ||| |¦«pi¦«S#t$r'}t d||¦«Yd}~dSd}~wwxYw)zCConstruct the registered streamer *name* if it's usable, else None.Nz%streaming provider %s init failed: %s)r`ÚgetrUrÚloggerÚdebug)rarQrdÚexcs rÚ_try_instantiatero£s”€å -Š-˜Ñ Ô €CØ €{˜#Ÿ-š-™/œ/€{ØˆtðØˆs:˜zŸ~š~¨dÑ3Ô3Ð9°rÑ:Ô:Ð:øÝðððÝŠÐ<¸dÀCÑHÔHÐHØˆtˆtˆtˆtˆtøøøøðøøøs´ AÁ BÁBÂB)Ú elevenlabsÚgeminiÚopenaiÚxair3Ú_PROVIDER_PRIORITYÚ preferredú Optional[str]có¾—| d¦«pi}t| d¦«pd¦« ¦« ¦«}|dkr"tD]}t||¦«}||cSŒdS|rt||¦«S|pt |¦« ¦« ¦«}t||¦«S)uReturn a ready streamer for the *configured* provider, else ``None``. Resolution order: 1. ``tts.streaming.provider`` (config knob) when set: * a provider name pins that exact streamer (or ``None`` if unusable); * ``auto`` walks the priority list (``elevenlabs â†’ gemini â†’ openai â†’ xai``) and returns the first usable streamer â€” an explicit opt-in to "give me the best chunked voice available". 2. Otherwise the *configured* TTS provider (or ``preferred`` override). ``None`` means "no chunked API for this provider" â€” the dispatcher then speaks per-sentence via the sync path, preserving the user's chosen voice. We never silently swap to a different provider just to get streaming. Ú streamingÚproviderrÚautoN)rkrÚlowerr;rtror)rQruÚ streaming_cfgÚpinnedraÚinsts rÚresolve_streaming_providerr¶sæ€ð&—N’N ;Ñ/Ô/Ð5°2€MÝ ×"Ò" :Ñ.Ô.Ð4°"Ñ 5Ô 5× ;Ò ;Ñ =Ô =× CÒ CÑ EÔ E€FØ ÒÐÝ&ð ð ˆDÝ# D¨*Ñ5Ô5ˆDØÐØð àˆtØ ð4Ý ¨ Ñ3Ô3Ð3àÐ2 zÑ2Ô2×9Ò9Ñ;Ô;×AÒAÑCÔC€DÝ˜D *Ñ-Ô-Ð-r!rpcó6—eZdZdZdZedd„¦«Zdd „Zd S) ÚElevenLabsStreameruDElevenLabs chunked HTTP â†’ pcm_24000 (the original reference path).rKrr#có<—ttdd¦«¦«S)NÚELEVENLABS_API_KEYrp©r#rr r!rrUzElevenLabsStreamer.availableâs€å•LÐ!5°|ÑDÔDÑEÔEÐEr!rWrrXc#ódK—ddlm}m}m}m}|¦«ddtdd¦«i||j¦«¤Ž}|j d|¦«}|j d|j d|¦«¦«}|j |||d ¬ ¦«Ed{V—†dS)Nr)Ú%DEFAULT_ELEVENLABS_STREAMING_MODEL_IDÚDEFAULT_ELEVENLABS_VOICE_IDÚ_elevenlabs_environment_kwargsÚ_import_elevenlabsÚapi_keyrƒrpÚvoice_idÚstreaming_model_idÚmodel_idÚ pcm_24000)rWr‹rÚ output_formatr ) rr†r‡rˆr‰rrRrkÚtext_to_speechÚconvert) r0rWr†r‡rˆr‰Úclientr‹rs rr[zElevenLabsStreamer.streamæs,èè€ð ð ð ð ð ð ð ð ð ð ð ð ð&Ð#Ð#Ñ%Ô%ð ð Ý Ð!5°|ÑDÔDð à,Ð,¨T¬\Ñ:Ô:ð ð ˆð”<×#Ò# JÐ0KÑLÔLˆØ”<×#Ò#Ø ØŒL×Ò˜ZÐ)NÑOÔOñ ô ˆðÔ(×0Ò0ØØØØ%ð 1ñ ô ð ð ð ð ð ð ð ð ð r!Nr\r]©rErFrGrHrLr_rUr[r r!rrrÜsU€€€€€àNÐNà€KàðFðFðFñ„\ðFð ð ð ð ð ð r!rcóœ— t¦« d¦«pi}n#t$rYdSwxYw| d¦«pdS)z@Return ``tts.openai.api_key`` from config.yaml, or empty string.rrrrŠ)r rkr)Ú openai_cfgs rÚ_openai_config_api_keyr–ÿs`€ðÝ&Ñ(Ô(×,Ò,¨XÑ6Ô6Ð<¸"ˆ ˆ øÝðððØˆrˆrðøøøà>Š>˜)Ñ$Ô$Ð*¨Ð*s‚#&¦ 4³4rrcó6—eZdZdZdZedd„¦«Zdd „Zd S) ÚOpenAIStreamerz?OpenAI speech with ``response_format=pcm`` (24 kHz mono int16).rKrr#cóT—tt¦«p t¦«¦«Sr)r#r–rr r!rrUzOpenAIStreamer.availables#€åÕ*Ñ,Ô,ÐNÕ0LÑ0NÔ0NÑOÔOÐOr!rWrrXc#óK—ddlm}||j d¦«p t ¦«|j d¦«ptd¦«pd¬¦«}|j dd¦«}|j d d ¦«}|jjj |||d¬¦«5}t| ¦«d ¦«Ed{V—†ddd¦«dS#1swxYwYdS)Nr)ÚOpenAIrŠÚbase_urlÚOPENAI_BASE_URL)rŠrœÚmodelzgpt-4o-mini-ttsÚvoiceÚalloyÚpcm)ržrŸÚinputÚresponse_formatzOpenAI streaming TTS)rrr›rRrkrrÚaudioÚspeechÚwith_streaming_responseÚcreateÚ_cappedÚ iter_bytes)r0rWr›r’ržrŸÚresponses rr[zOpenAIStreamer.streams€èè€Ø!Ð!Ð!Ð!Ð!Ð!àØ”\×%Ò% iÑ0Ô0ÐRÕ4PÑ4RÔ4Rà”× Ò Ñ,Ô,ðÝ Ð!2Ñ3Ô3ðàð ñ ô ˆð”× Ò Ð*;Ñ<Ô<ˆØ”× Ò ¨'Ñ2Ô2ˆØ Œ\Ô Ô 8× ?Ò ?ØØØØ!ð @ñ ô ð Nð Ý˜x×2Ò2Ñ4Ô4Ð6LÑMÔMÐMÐMÐMÐMÐMÐMÐMð Nð Nð Nñ Nô Nð Nð Nð Nð Nð Nð Nð Nøøøð Nð Nð Nð Nð Nð NsÃ)C9Ã9C=ÄC=Nr\r]r“r r!rr˜r˜s[€€€€€àIÐIà€KàðPðPðPñ„\ðPðNðNðNðNðNðNr!r˜ÚchunksrXÚlabelc#ó¢K—d}|D]G}|t|¦«z }|tkr$t d|t¦«dS|V—ŒHdS)uPass chunks through, aborting past the 16 MiB per-sentence cap. The streaming mirror of ``_read_tts_response_bytes``'s bounded-body invariant: one sentence of PCM should never approach the cap, so exceeding it means a runaway/hostile upstream â€” stop pulling. rz1%s exceeded %d bytes for one sentence; truncatingN)r:Ú_STREAM_SENTENCE_BYTE_CAPrlÚwarning)r«r¬ÚtotalÚchunks rr¨r¨(ssèè€ð €EØððˆØ •U‘”ÑˆØÕ,Ò,Ð,ÝNŠNÐNØ Õ";ñ =ô =ð =àˆFˆFØˆˆˆˆð ðr!rqcó6—eZdZdZdZedd„¦«Zdd „Zd S) ÚGeminiStreameru(Gemini ``streamGenerateContent?alt=sse`` â†’ base64 PCM chunks (24 kHz). Salvaged from PR #47588 (@Cdddo) and rebased onto the post-campaign infrastructure: credentials via the provider-secret resolver, requests (not httpx) with a bounded streamed body, and main's provider ABC. rKrr#có\—ttdd¦«ptdd¦«¦«S)NÚGEMINI_API_KEYrqÚGOOGLE_API_KEYr„r r!rrUzGeminiStreamer.availableDs5€åÝÐ)¨8Ñ4Ô4ð 8ÝÐ,¨hÑ7Ô7ñ ô ð r!rWrrXc#ó¸‡ ‡ ‡‡‡ ‡K—ddlŠddlŠ ddlŠ ddlm}m}m}tdd¦«ptdd¦«Š t|j d|¦«¦« ¦«p|}t|j d|¦«¦« ¦«p|}t|j d¦«ptd ¦«p|¦« ¦« d ¦«}dd|igigd gddd|iiidœdœŠ|›d|›dŠdˆ ˆ ˆˆˆ ˆfd„}t|¦«d¦«Ed{V—†dS)Nr)ÚDEFAULT_GEMINI_TTS_BASE_URLÚDEFAULT_GEMINI_TTS_MODELÚDEFAULT_GEMINI_TTS_VOICErµrqr¶ržrŸrœÚGEMINI_BASE_URLú/ÚpartsrWÚAUDIOÚvoiceConfigÚprebuiltVoiceConfigÚ voiceName)ÚresponseModalitiesÚspeechConfig)ÚcontentsÚgenerationConfigz/models/z:streamGenerateContentrrXc3ó•K—‰ ‰ d‰ dœ‰dd¬¦«5}| ¦«| d¬¦«D]}|r| d¦«sŒ ‰ |td¦«d…¦«}|dd d d}n##ttttf$rYŒ€wxYw|D]˜}| d¦«p| d ¦«pi}| dd¦«}|sŒG ‰ |¦«V—Œ`#ttf$r%}t d|¦«Yd}~Œ‘d}~wwxYwŒ! ddd¦«dS#1swxYwYdS)NÚsse)ÚaltÚkeyé<T)ÚparamsÚjsonÚtimeoutr[)Údecode_unicodezdata: Ú candidatesrÚcontentr½Ú inlineDataÚinline_dataÚdatarz Gemini SSE: bad base64 audio: %s)ÚpostÚraise_for_statusÚ iter_linesÚ startswithÚloadsr:Ú ValueErrorÚKeyErrorÚ IndexErrorÚ TypeErrorrkÚ b64decoderlr¯)rªÚlineÚeventr½ÚpartÚinlineÚb64rnÚ_jsonrŠÚbase64ÚpayloadÚrequestsÚurls €€€€€€rÚ_sse_chunksz*GeminiStreamer.stream.._sse_chunksrs;øèè€Ø—’ØØ$¨WÐ5Ð5ØØØðñôð TðØ×)Ò)Ñ+Ô+Ð+Ø$×/Ò/¸tÐ/ÑDÔDðTñTDØð! t§¢°xÑ'@Ô'@ð!Ø ð!Ø %§¢¨Dµ°X±´°°Ô,@Ñ AÔ A˜Ø % lÔ 3°AÔ 6°yÔ AÀ'Ô J˜˜øÝ&µ*½iÐHð!ð!ð!Ø ˜ð!øøøà %ðTðT˜Ø!%§¢¨,Ñ!7Ô!7Ð!X¸4¿8º8ÀMÑ;RÔ;RÐ!XÐVX˜Ø$Ÿjšj¨°Ñ4Ô4˜Ø"ð%Ø$ðTØ"(×"2Ò"2°3Ñ"7Ô"7Ð7Ð7Ð7Ð7øÝ *IÐ6ðTðTðTÝ"ŸNšNÐ+MÈsÑSÔSÐSÐSÐSÐSÐSÐSøøøøðTøøøñTðTð Tð Tð Tñ Tô Tð Tð Tð Tð Tð Tð Tð Tøøøð Tð Tð Tð Tð Tð Tsb¡AE;Á(AB-Â,E;Â-C Ã E;ÃC Ã AE;ÄD2Ä1E;Ä2E( ÅE# ÅE;Å#E( Å(E;Å;E?ÆE?zGemini streaming TTS)rrX)rärÌrærr¸r¹rºrrrRrkr;rÚrstripr¨)r0rWr¸r¹rºržrŸrœrèrãrŠrärårærçs @@@@@@rr[zGeminiStreamer.streamKsøøøøøøèè€Øˆ ˆ ˆ ØÐÐÐàˆˆˆð ð ð ð ð ð ð ð ð ð õ Ð)¨8Ñ4Ô4ð 8ÝÐ,¨hÑ7Ô7ð õD”L×$Ò$ WÐ.FÑGÔGÑHÔH×NÒNÑPÔPÐlÐTlˆÝD”L×$Ò$ WÐ.FÑGÔGÑHÔH×NÒNÑPÔPÐlÐTlˆÝØŒL×Ò˜ZÑ(Ô(ð +ÝÐ.Ñ/Ô/ð +à*ñ ô ÷Š%‰'Œ'—&’&˜‘+”+ð ð" V¨T NÐ#3Ð4Ð5à'. ià!Ø-°¸UÐ/Cð$ð!ð!ð!ð ð ˆðÐ@Ð@ 5Ð@Ð@Ð@ˆð Tð Tð Tð Tð Tð Tð Tð Tð Tð Tð Tõ6˜;˜;™=œ=Ð*@ÑAÔAÐAÐAÐAÐAÐAÐAÐAÐAÐAr!Nr\r]r“r r!rr³r³9sc€€€€€ððð€Kàð ð ð ñ„\ð ðBBðBBðBBðBBðBBðBBr!r³rscóN—eZdZdZdZedd„¦«Zdd „Zdd„Zdd„Z dd „Z dS)ÚXAIStreameru¸xAI WebSocket TTS â†’ binary PCM frames (24 kHz mono int16). Salvaged from PR #47588 (@Cdddo): xAI's chunked TTS API is WebSocket-only (``wss://api.x.ai/v1/tts``). Credentials route through ``resolve_xai_http_credentials`` (OAuth or XAI_API_KEY), same as the sync ``_generate_xai_tts`` path. The async WS loop is bridged to the sync iterator contract via ``_collect_async`` â€” the seam unit tests monkeypatch. rKrr#cóÌ— ddlm}|¦«}tt| d¦«pd¦« ¦«¦«S#t$rYdSwxYw)Nr©Úresolve_xai_http_credentialsrŠrF)Útools.xai_httprîr#rrkr;r)rîÚcredss rrUzXAIStreamer.availabležs|€ð ØCÐCÐCÐCÐCÐCà0Ð0Ñ2Ô2ˆEÝ˜EŸIšI iÑ0Ô0Ð6°BÑ7Ô7×=Ò=Ñ?Ô?Ñ@Ô@Ð@øÝð ð ð Ø55ð øøøs‚AAÁ A#Á"A#rWrrXc#óvK—tt| |¦«¦«d¦«Ed{V—†dS)NzxAI streaming TTS)r¨ÚiterÚ_collect_asyncrZs rr[zXAIStreamer.stream¨sEèè€Ý4 × 3Ò 3°DÑ 9Ô 9Ñ:Ô:Ðr sŒððððð*#Ð"Ð"Ð"Ð"Ð"à€€€Ø € € € Ø€€€Ø#Ð#Ð#Ð#Ð#Ð#Ð#Ð#Ø;Ð;Ð;Ð;Ð;Ð;Ð;Ð;Ð;Ð;Ð;Ð;Ð;Ð;àCÐCÐCÐCÐCÐCØIÐIÐIÐIÐIÐIÐIÐIÐIÐIà ˆÔ ˜8Ñ $Ô $€ð-Ðð,ð,ð,ð,ð2RððÐØ#'€Ð'Ð'Ð'Ñ'ð'ð'ð'ð'ð GðGðGðGð"r”zÐ"AÑBÔBÐØ"”*Ð6¸b¼iÐHÑHÔH€ð#&ð#&ð#&ð#&ð#&ñ#&ô#&ð#&ðTMðMðMðMðM˜3ñMôMðMð*46€ Ð5Ð5Ð5Ñ5ððððð ð ð ð ð !JÐ IÐ IÐÐIÐIÐIÑIð $ð.ð.ð.ð.ð.ðL €ˆ,ÑÔð ð ð ð ð Ð-ñ ô ñÔð ðD+ð+ð+ð+ð €ˆ(ÑÔðNðNðNðNðNÐ)ñNôNñÔðNð>ðððð" €ˆ(ÑÔðSBðSBðSBðSBðSBÐ)ñSBôSBñÔðSBðl €ˆ%„ðWðWðWðWðWÐ&ñWôWñ„ðWðWðWr!