Jj. UdZddlZddlZddlZddlZddlZddlZddlZddlZddl Z ddl Z ddl Z ddl Z ddl Z ddlZddlmZddlmZmZmZmZddlmZddlmZddlmZejeZddZdd l m!Z!dd l"m#Z#m$Z$m%Z%m&Z&dd l'm(Z(d Z)d Z*dZ+dZ,dZ-dZ.dZ/dZ0dZ1dZ2dZ3dZ4dZ5e6dhZ7dZ8dZ9dZ:dZ;dZd Z?d!Z@d"ZAd#ZBd$ZCd%ZDd&ZEd'ZFd(ZGd)ZHd*ZId+ZJdZKd,ZLd-ZMd.ZNd'ZOd/ZPd%ZQd0ZRd1ZSd2eTfd3ZUeUZVd4d5d6d7d8d9d7d:d:d4d; ZWeeTeXfeYd<<d4d4d7d7d7d7d=d>d?ZZeeTeXfeYd@<ddAedBe[d2e[fdCZ\d8Z]e]Z^ ddDeeTdEeeeTefd2eXfdFZ_d2eeTeffdGZ`dEeeTefd2eTfdHZae6hdIZbdJZcdKZde6hdLZed4ZfdEeeTefdMeTd2eeTeffdNZgdEeeTefdMeTd2eeTeffdOZhdPeeTefd2e[fdQZidDeTdEeeTefd2eeeTeffdRZjdSeTdTeTdDeTdEeeTefd2eeTf dUZkdDeTd2e[fdVZldEeeTeffdWZmdPeeTefd2enfdXZo ddPeeTefdTeeTd2eTfdYZpdPeeTefd2e[fdZZqd[eTd\eXd2eeTfd]ZrdAeTd^eeTd2eTfd_Zsd[eTd`eeTeTfd2eTfdaZtdbe jud2dfdcZvddeTdeend2e jwfdfZxdgedPeeTefd2efdhZydSeTdTeTdieTdPeeTefdEeeTefd2eTf djZzddEeeeTefd2e[fdkZ{d2e[fdlZ|dmeTd2eeTfdnZ}dSeTdTeTdEeeTefd2eTfdoZ~dSeTdTeTdEeeTefd2eTfdpZdSeTdTeTdEeeTefd2eTfdqZdrZdsZejdtduezdvzduezdwzejxZejdyejxZddAedBe[d2e[fdzZdSeTd2eTfd{ZdSeTdTeTdEeeTefd2eTfd|ZdSeTdTeTdEeeTefd2eTfd}ZdSeTdTeTdEeeTefd2eTfd~ZeQeReSfdedeXdeXdeXd2ef dZdeeTefd2eefdZdeeTefd2eTfdZdeTd2e[fdZdeeTefdeTd2e[fdZdeTd2eTfdZded2eTfdZddSeTdeTd2eTfdZ ddSeTdeeTefdeeTd2eTfdZdSeTdTeTdEeeTefd2eTfdZd2e[fdZd2e[fdZd2eTfdZd2eTfdZdSeTdTeTdEeeTefd2eTfdZiaeeTefeYd<d2e[fdZd2efdZdeTded2eTfdZdSeTdTeTdEeeTefd2eTfdZiaeeTefeYd<dSeTdTeTdEeeTefd2eTfdZ ddSeTdTeeTd2eTfdZd2e[fdZd2eeTeTe[ffdZd2e[fdZejdZejdZejdZejdZejdZejdZejdZejdejxZejdejxZejdZejdZdSeTd2eTfdZ ddejde jde jdeeeTgdffdZedkr-ededdZededee)drdndedee*drdndeded¦rdnděedee+dƦrdndǛede&rdndțededʦrdnd̛ederdndΛede|rdndћedeVe`ZeaeZededdlmZmZddddddڜddedܝdڜdݜdSgdޜdߜZejddededdS)a  Text-to-Speech Tool Module Built-in TTS providers: - Edge TTS (default, free, no API key): Microsoft Edge neural voices - ElevenLabs (premium): High-quality voices, needs ELEVENLABS_API_KEY - OpenAI TTS: Good quality, needs OPENAI_API_KEY - MiniMax TTS: High-quality with voice cloning, needs MINIMAX_API_KEY - Mistral (Voxtral TTS): Multilingual, native Opus, needs MISTRAL_API_KEY - Google Gemini TTS: Controllable, 30 prebuilt voices, needs GEMINI_API_KEY - xAI TTS: Grok voices, uses xAI Grok OAuth credentials or XAI_API_KEY - NeuTTS (local, free, no API key): On-device TTS via neutts - KittenTTS (local, free, no API key): On-device 25MB model - Piper (local, free, no API key): OHF-Voice/piper1-gpl neural VITS, 44 languages Custom command providers: - Users can declare any number of named providers with ``type: command`` under ``tts.providers.`` in ``~/.hermes/config.yaml``. Hermes writes the input text to a temp file and runs the configured shell command, which must produce the audio file at the expected path. See the Local Command section of ``website/docs/user-guide/features/tts.md``. Output formats: - Opus (.ogg) for Telegram voice bubbles (requires ffmpeg for Edge TTS) - MP3 (.mp3) for everything else (CLI, Discord, WhatsApp) Configuration is loaded from ~/.hermes/config.yaml under the 'tts:' key. The user chooses the provider and voice; the model just sends text. Usage: from tools.tts_tool import text_to_speech_tool, check_tts_requirements result = text_to_speech_tool(text="Hello world") N)Path)CallableDictAnyOptional)urljoin)windows_hide_flags)display_hermes_homec~ ddlm}n%#t$rtj||cYSwxYw||}||n|S)aRead env values through the live config module. Tests may monkeypatch and later restore ``hermes_cli.config.get_env_value`` before this module is imported. Resolve the helper at call time so TTS does not keep a stale imported function for the rest of the test process. r) get_env_value)hermes_cli.configr ImportErrorosgetenv)namedefault_get_env_valuevalues 4/home/rurouni/.hermes/hermes-agent/tools/tts_tool.pyr r ;sl(EEEEEEE (((yw'''''( N4 Em77.s  ++)resolve_managed_tool_gateway)managed_nous_tools_enabled%nous_tool_gateway_unavailable_messageprefers_gatewayresolve_openai_audio_api_key)hermes_xai_user_agentc ddlm}|ddn9#t$rYn-t$r!}tt |d}~wwxYwddl}|S)z?Lazy import edge_tts. Returns the module or raises ImportError.rensureztts.edgeFpromptN)tools.lazy_depsrr Exceptionstredge_tts) _lazy_ensureer$s r_import_edge_ttsr'Vs":::::: Z.....      """#a&&!!!"OOO O A  A AA c ddlm}m}|ddn9#t$rYn-t$r!}tt |d}~wwxYwddlm}|S)aLazy import ElevenLabs client. Returns the class or raises ImportError. Calls :func:`tools.lazy_deps.ensure` first so the SDK gets installed on demand if the user picked ElevenLabs as their TTS provider but never ran the post-setup hook (e.g. enabled it by editing config.yaml directly). Raises ``ImportError`` on lazy-install failure so existing callers' error-handling paths keep working. r)FeatureUnavailablerztts.elevenlabsFrN) ElevenLabs)r!r*rrr"r#elevenlabs.clientr+)r*rr&r+s r_import_elevenlabsr-bs">>>>>>>>.....      """#a&&!!!",,,,,, s A AA  Acddlm}|S)zCLazy import OpenAI client. Returns the class or raises ImportError.r)OpenAI)openair/) OpenAIClients r_import_openai_clientr2ws------ c ddlm}|ddn9#t$rYn-t$r!}tt |d}~wwxYwddlm}|S)ajLazy import Mistral client. Returns the class or raises ImportError. Calls :func:`tools.lazy_deps.ensure` first so the ``mistralai`` SDK gets installed on demand if the user picked Mistral as their STT/TTS provider but never ran the post-setup hook (e.g. enabled it by editing config.yaml directly). Mirrors the ElevenLabs lazy-import path. rrz tts.mistralFrN)Mistral)r!rrr"r#mistralai.clientr5)rr&r5s r_import_mistral_clientr7|s"******}U+++++      """#a&&!!!"(((((( Nr(cddl}|S)zJLazy import sounddevice. Returns the module or raises ImportError/OSError.rN) sounddevice)sds r_import_sounddevicer;s Ir3cddlm}|S)z?Lazy import KittenTTS. Returns the class or raises ImportError.r KittenTTS) kittenttsr>r=s r_import_kittenttsr@s###### r3cddlm}|S)atLazy import Piper. Returns the PiperVoice class or raises ImportError. Piper is an optional, fully-local neural TTS engine (Home Assistant / Open Home Foundation). ``pip install piper-tts`` provides cross-platform wheels (Linux / macOS / Windows, x86_64 + ARM64) with embedded espeak-ng. Voice models (.onnx + .onnx.json) are downloaded on first use. r PiperVoice)piperrCrBs r _import_piperrEs!     r3edgezen-US-AriaNeuralpNInz6obpgDQGcFmaJgBeleven_multilingual_v2eleven_flash_v2_5zgpt-4o-mini-ttsz!KittenML/kitten-tts-nano-0.8-int8Jasperzen_US-lessac-mediumalloyzhttps://api.openai.com/v1z speech-02-hdEnglish_expressive_narratorz https://api.minimax.io/v1/t2a_v2zvoxtral-mini-tts-2603z$c69964a6-ab8b-4f8a-9465-ec0925096ec8eveen]Fzhttps://api.x.ai/v1ffffff?g??zgemini-2.5-flash-preview-ttsKorez0https://generativelanguage.googleapis.com/v1betatts_audio_tagsreturnc@ddlm}t|ddS)Nrget_hermes_dirz cache/audio audio_cache)hermes_constantsrZr#rYs r_get_default_output_dirr]s.////// ~~m];; < <1onyestrueenabledT>0noofffalsedisabledF) isinstanceboolintfloatr#striplower)rr normalizeds r _config_boolrs%  }%#u&&E{{%[[]]((** > > >4 @ @ @5 Nr3provider tts_configcl|stS|}|pi}t||t r||ni}|r|dnd}t|t rd}t|tr|dkr|S|dkr[|pidpt}tt|}|r|S|tvr t|S|tvrot||}t|rP|d} t| t rd} t| tr| dkr| StStS)aReturn the input-character cap for *provider*. Resolution order: 1. ``tts..max_text_length`` (user override in config.yaml) 2. ``tts.providers..max_text_length`` for user-declared command providers 3. ElevenLabs model-aware table (keyed on configured ``model_id``) 4. ``PROVIDER_MAX_TEXT_LENGTH`` default 5. ``DEFAULT_COMMAND_TTS_MAX_TEXT_LENGTH`` when the provider is a command-type user provider without an explicit cap 6. ``FALLBACK_MAX_TEXT_LENGTH`` (4000) Non-positive or non-integer overrides fall through to the default so a broken config can't accidentally disable truncation entirely. max_text_lengthNrrcmodel_id)FALLBACK_MAX_TEXT_LENGTHr}r|rxgetdictryrzDEFAULT_ELEVENLABS_MODEL_IDrlr#reBUILTIN_TTS_PROVIDERS_get_named_provider_config_is_command_provider_config#DEFAULT_COMMAND_TTS_MAX_TEXT_LENGTH) rrkeycfgprov_cfgoverridermappednamednamed_overrides r_resolve_max_text_lengthrs& ('' ..   " "C  C *#''#,,==Eswws|||2H2:Dx||-...H(D!!(C  X\\ lN'' 33R7R155c(mm6I6I6K6KLL  M &&&',, '''*344 &u - - 7"YY'899N.$// &!%.#.. &>A3E3E%%6 6 ##r3c  ddlm}|}|diS#t$rtdicYSt $r)}td|dicYd}~Sd}~wwxYw) z Load TTS configuration from ~/.hermes/config.yaml. Returns a dict with provider settings. Falls back to defaults for any missing fields. r) load_configttsz9hermes_cli.config not available, using default TTS configzFailed to load TTS config: %sTexc_infoN)r rrrloggerdebugr"warning)rconfigr&s r_load_tts_configrNs 111111zz%$$$  PQQQ 6DIII s!%(&B BA=7B=Bc|dptS)z%Get the configured TTS provider name.r)rDEFAULT_PROVIDERr}r|)rs r _get_providerras2 NN: & & :*: A A C C I I K KKr3> r_rFrDrbrdr0r`rar?rcxmp3>roggwavflacrct|tsiS||}t|tr|niS)zBReturn a provider config block if it's a dict, else an empty dict.)rxrr)rrsections r_get_provider_sectionrsC j$ ' ' nnT""G $// 777R7r3c t|d}t|tr||nd}t|tr|S|t vrt||}|r|SiS)aReturn the config dict for a user-declared provider. Looks up ``tts.providers.`` first (the canonical location), and falls back to ``tts.`` so users who followed the built-in layout still work. Returns an empty dict when the provider is not declared. providersN)rrxrrr}r)rrrrlegacys rrrs&j+>>I%/ 4%@%@JimmD!!!dG'4   zz||000&z488  M Ir3rcjt|tsdSt|dpd}|r|dkrdS|d}t|to t |S)z;Return True when *config* declares a command-type provider.Ftypecommand)rxrr#rr|r}ry)rptypers rrrs fd # #u  6""(b ) ) / / 1 1 7 7 9 9E )##ujj##G gs # # =W]]__(=(==r3c|sdS|}|tvrdSt||}t |r|SdS)zReturn the provider config if *provider* resolves to a command type. Built-in provider names are rejected (they have native handlers). Returns None when the name is a built-in, unknown, or not a command type. N)r}r|rrr)rrrrs r _resolve_command_provider_configrsf t ..   " "C ###t ' C 8 8F"6** 4r3text output_pathc |sdS|}|tvrdStt ||rdS ddlm}ddlm}|||}||d||}n3#t$r&}t d|Yd}~dSd}~wwxYw|dSt|tr|dnd} t|tr|d nd} t|tr|d nd} t|tr|d tnt} td ||||t| t$r| r| ndt| t$r| r| ndt| t&t(frt)| nd| r!t%| nd } t| t$r| r| n|S)u Route the call to a plugin-registered TTS provider, or return None. Returns the path to the written audio file on dispatch, or ``None`` to fall through to the next resolution layer (built-in dispatch or Edge TTS default). Resolution invariants enforced here (matches issue #30398): 1. Built-in provider names short-circuit — never reach the plugin registry. The caller is responsible for the elif chain that handles ``edge``/``openai``/etc.; this function explicitly rejects those names defensively. 2. Command-type providers declared under ``tts.providers.: type: command`` (PR #17843) win over a plugin with the same name. The caller passes us only when its own command-provider check returned None — we re-verify here so a refactor of the caller can't silently break the invariant. 3. Plugin dispatch fires only when ``provider`` matches a registered :class:`TTSProvider` whose ``name`` equals the configured value. Unknown names return None (caller falls through to Edge default). Plugin exceptions are caught and re-raised — the outer ``text_to_speech_tool`` try/except converts them to the standard error envelope, matching how command-provider failures surface. Nr get_provider)_ensure_plugins_discoveredT)forcez2tts plugin dispatch skipped (discovery failed): %svoicemodelspeed output_formatz2Generating speech with plugin TTS provider '%s'...r)rrrformat)r}r|rrragent.tts_registryrhermes_cli.pluginsrr"rrrxrr!DEFAULT_COMMAND_TTS_OUTPUT_FORMATinfo synthesizer#rzr{)rrrrrrrplugin_providerexcrrrfmtwrittens r_dispatch_to_plugin_providerrs> t ..   " "C ###t##=j##N#NOOt333333AAAAAA""$$$&,s++  " ' &T 2 2 2 2*l3//O  I3OOOtttttt (2*d'C'C MJNN7 # # #E'1*d'C'C MJNN7 # # #E'1*d'C'C MJNN7 # # #E j$ ' ' / (IJJJ .  KK>Io2,,..  c dC  TZZ\\9N%N%N*3// Ci  r3c|d|dt} t|}n+#ttf$rttcYSwxYw|dkrttS|S)z5Return timeout in seconds, falling back when invalid.timeouttimeout_secondsr)r#DEFAULT_COMMAND_TTS_TIMEOUT_SECONDSr{ TypeError ValueError)rrawrs r_get_command_tts_timeoutrXs **Y +<>a b b c cC:c  z ":::899999: zz8999 LsA%A)(A)c|rVt|jd}|t vr|S|dp|dpt}t|d}|t vr|ntS)z6Return the validated output format (mp3/wav/ogg/flac)..rr) rsuffixr}r|lstripCOMMAND_TTS_OUTPUT_FORMATSrrr#)rrrrrs r_get_command_tts_output_formatrds k"")//117799@@EE / / /M 8 - ::o & & - , c((..   " " ) )# . .C333339ZZr3c|dd}t|tr(|dvSt |S)zEReturn True only when the user explicitly opted in to voice delivery.rF>rnrorprq)rrxr#r|r}ry)rrs r _is_command_tts_voice_compatiblervsU JJ)5 1 1E%C{{}}""$$(BBB ;;r3command_templatepositioncd}d}d}||kr\||}|dkr |dkrd}n:|dkr|rd}n/|dkrd}n&|dkrd}n|dkrd}n|dkrd}n |dkr|dz }|dz }||k\|S) zReturn the shell quote character active right before *position*. Returns ``"'"`` / ``'"'`` when inside a single- / double-quoted region of the template, ``None`` for bare context. NFr'"\TrU)rrquoteescapedichars r_shell_quote_contextr~s  EG A h,," C<<s{{ c\\  S[[EE S[[EE T\\ FA Q% h,,& Lr3 quote_contextc\|dkr|ddS|dkrR|ddddddd d Stjd krtj|gSt j|S) zGQuote a placeholder value for its position in a shell command template.rz'\''rrz\\z\"$z\$`z\`nt)replacerr subprocess list2cmdlineshlexr)rrs r_quote_command_tts_placeholderrs}}S'***  WT6 " " WS% WS% WS%   w$&w/// ;u  r3 placeholderscddD}tjd|d|d}gdtjtdtffd }||}|d d d d }D]\}}|||}|S) z@Replace supported placeholders while preserving ``{{`` / ``}}``.|c3>K|]}tj|VdSN)reescape).0rs r z/_render_command_tts_template..s*>>RYt__>>>>>>r3z(?z)\}\}|\{(?Pz)\})matchrWc |dp|d}dtd}|t|t |f|S)Ndoublesingle__HERMES_TTS_PLACEHOLDER___)grouplenappendrrstart)rrtokenrr replacementss r replace_matchz3_render_command_tts_template..replace_matchs{{8$$= H(=(=AC ,=,=AAA  *T"$%5u{{}}EE       r3z{{{z}}})joinrcompileMatchr#subr) rrnamespatternr renderedr rr s `` @r_render_command_tts_templaters HH>>>>> > >EjPuPPEPPPG+-L RXc] s        {{=*:;;Hc**224==H$22 u##E511 Or3procc |dStjdkrv tjddddt |jgtjtjdtjn$#t$r| YnwxYwdSd dl } | |j}| d D]'} | #|j$rY$wxYw| n0#|j$rYdSt$r| YnwxYw |d dS#tj$rYnwxYw | |j}| d D]'} | #|j$rY$wxYw| dS#|j$rYdSt$r| YdSwxYw)zCBest-effort termination of a shell process and all of its children.Nrtaskkillz/Fz/Tz/PID)stdoutstderrrstdinrT) recursiverVr)pollrrrrunr#pidDEVNULLr"killpsutilProcesschildren terminate NoSuchProcesswaitTimeoutExpired)rr$parentchilds r#_terminate_command_tts_process_treer-s yy{{ w$  NT4TX?!)!) (          IIKKKKK MMM ))__t_44  E !!!!'          !   $      ))__t_44  E  '        sAA77BB"2DC*)D* C74D6C77D D<D<;D<EE*)E*.2G!F65G6 GGGG H )H  H rrcdtjtjdd}tjdkrt tdd|d<nd|d<tj|fi|dtji} || \}}n#tj$rz}t| |d \}}n2#t$r%t |d d }t |d d }YnwxYwtj|||||d }~wwxYw|j rtj |j |||tj ||j ||S)zGRun a command-provider shell command with process-tree timeout cleanup.T)shellrrrrCREATE_NEW_PROCESS_GROUPr creationflagsstart_new_sessionrrrUoutputNr)r3r)rPIPErrgetattrPopenr" communicater*r-r" returncodeCalledProcessErrorCompletedProcess)rr popen_kwargsrrrrs r_run_command_ttsr<s// $$L  w$(/ D ?,C.+D -C..D  DpathcPt|}|d|S)zKReturn an output path whose extension matches the provider's output_format.r)r with_suffix)r=rrs r#_configured_command_tts_output_pathr@(s) ( 0 0C   III & &&r3 provider_namec nt|dpd}|std|dt |}|jdd|r| t|}t|t|}|d|dd} tj 5} t | dz } | |d t| t| t||t|d dt|d dt| d } t|| } t!| |n#t"j$r}t'd|d|dd|d}~wt"j$r}g}|jr/|d|j|jr/|d|jd|pd}t'd|d|jd||d}~wwxYw dddn #1swxYwY|r|jdkrt'd|d|t|S)a Generate speech by running a user-configured shell command. Returns the absolute path of the audio file the command wrote. Raises ``ValueError`` when the provider config is invalid, and ``RuntimeError`` for timeouts / non-zero exits / empty output. rrztts.providers.z.command is not configuredTparentsexist_okrz input.txtutf-8encodingrr) input_path text_pathrrrrrzTTS provider 'z' timed out after gsNzstderr: zstdout: z; zno command outputz' exited with code z: rz' produced no output at )r#rr|rr expanduserr+mkdirexistsunlinkrrtempfileTemporaryDirectory write_textrr<rr* RuntimeErrorr9rrrrr8statst_size)rrrArrrr3rrrtmpdirrJrrr detail_partsdetails r_generate_command_ttsrZ.s6::i006B77==??   F] F F F   +   ) ) + +F Mt444 }} &v..G263v;;GGM JJw w ; ; < ..%+..   /!B ==?? fkkmm3q88 L] L LF L L    v;;sDB:KG&%K&K5H  KB#KKKKKcT|t}t|D]\}}dSdS)z=Return True when any command-type TTS provider is configured.NTF)rr)r_name_cfgs r_has_any_command_tts_providerr^rs:%'' .z:: ttt 5r3c.tjdduS)z+Check if ffmpeg is available on the system.ffmpegN)shutilwhichrr3r _has_ffmpegrc~s < ! ! --r3mp3_pathctsdS|ddddz} tjdd|dd d d d d dd|dg ddtjt }|jdkrEtd|j|j dddddStj |r%tj |dkr|Sn#tj$rtdYnXt $rtdYn3t"$r'}td|dYd}~nd}~wwxYwdS)z Convert an MP3 file to OGG Opus format for Telegram voice bubbles. Args: mp3_path: Path to the input MP3 file. Returns: Path to the .ogg file, or None if conversion fails. NrrUr.oggr`-i-acodeclibopus-acrn-b:a64k-vbrru-yTcapture_outputrrr1z0ffmpeg conversion failed with return code %d: %srFignoreerrorsz)ffmpeg OGG conversion timed out after 30szffmpeg not found in PATHz ffmpeg OGG conversion failed: %sr)rcrsplitrr r"r r8rrrdecoderr=rOgetsizer*FileNotFoundErrorr")rdogg_pathresultr&s r_convert_to_opusr|s ==tsA&&q)F2HM tXy) Cx G$,..       ! ! NNM +V]-A-A'RZ-A-[-[\`]`\`-a c c c4 7>>( # # (A(AA(E(EO  $DDDBCCCCC 333122222 MMM91tLLLLLLLLM 4s+B D=AD)F-#F FE==Fc Kt}|di}|dt}t|d|dd}d|i}|dkrt |dz dz}|dd|d<|j|fi|} | |d {V|S) z Generate audio using Edge TTS. Args: text: Text to convert. output_path: Where to save the MP3 file. tts_config: TTS config dict. Returns: Path to the saved audio file. rFrrrRdz+d%rateN)r'rDEFAULT_EDGE_VOICEr{round Communicatesave) rrr _edge_tts edge_configrrkwargspctr7s r_generate_edge_ttsrs!""I..,,K OOG%7 8 8E +//':>>'3+G+GHH I IEu F ||US[C'((v')'7777K   ; ' '''''''' r3c tdpd}|std|di}|dt}|dt}|drd}nd }t }|| } | j|||| } t|d 5} | D]} | |  d d d n #1swxYwY|S)z Generate audio using ElevenLabs. Args: text: Text to convert. output_path: Where to save the audio file. tts_config: TTS config dict. Returns: Path to the saved audio file. ELEVENLABS_API_KEYrz=ELEVENLABS_API_KEY not set. Get one at https://elevenlabs.io/rcvoice_idrrf opus_48000_64 mp3_44100_128api_keyrrrrwbN) r rrDEFAULT_ELEVENLABS_VOICE_IDrendswithr-text_to_speechconvertopenwrite) rrrr el_configrrrr+clientaudio_generatorfchunks r_generate_elevenlabsrse1228bG ZXYYY|R00I}}Z)DEEH}}Z)DEEHF##(' ' #%%J Z ( ( (F+33 # 4O k4 A$  E GGENNNN  sC99C=C=c t\}}}|di}|dt}|dt}|d} | r| }t |d|dd} |r4| s2|t vr)t d|t|t}|drd } nd } t} | || } |||| d ttj id }| dkr!tdtd| |d<| jjjdi|}|||t'| dd}t)|r |SS#t'| dd}t)|r |wwxYw)z Generate audio using OpenAI TTS. Args: text: Text to convert. output_path: Where to save the audio file. tts_config: TTS config dict. Returns: Path to the saved audio file. r0rrbase_urlrrRzTTS: managed OpenAI audio gateway does not support model %r; falling back to %s. Set VOICE_TOOLS_OPENAI_KEY or OPENAI_API_KEY to use %r directly.rfopusr)rrzx-idempotency-key)rrinputresponse_format extra_headersg?g@closeNr)#_resolve_openai_audio_client_configrDEFAULT_OPENAI_MODELDEFAULT_OPENAI_VOICEr{MANAGED_OPENAI_TTS_MODELSrrrr2r#uuiduuid4maxminaudiospeechcreatestream_to_filer5callable)rrrrr is_managed oai_configrrcustom_base_urlrrr1r create_kwargsresponsers r_generate_openai_ttsrs%H$I$I!GXz"--J NN7$8 9 9E NN7$8 9 9E nnZ00O#" *..*..#*F*FGG H HE %/%e;T.T.T " '    %F##  (**L \'H = = =F.13tz||3D3DE    C<<%(s3%?%?M' "-6<&->> >> ,,,.. E??  EGGGG .. E??  EGGGG s A)flagsu^(.{12,120}?[.!?…])\s+(?=\S)c$t||S)Nr)r)rrs r_xai_bool_configr_s w / / //r3c|}|s|S|}tjdd|}tjdd|}t|st d|d}tjdd|}t|r|Sd t}d t}d |zd z|zd z} d dl m }|dd|ddd|dgd}t|}tj d|tj } | r'| d}|p|S#t$r'} t d| |cYd} ~ Sd} ~ wwxYw)aAdd xAI speech tags for more natural voice-mode replies. First applies a conservative local transform (inserts [pause] between paragraphs and after the first sentence). Then, if the result contains no explicit user/model speech tags, asks the configured auxiliary model to rewrite the transcript with a richer set of xAI-supported tags (laughs, sighs, whispers, soft/loud, slow/fast, etc.) so the voice output sounds more expressive. Falls back to the local result on any auxiliary-model failure. z\n\s*\n+z [pause] z\s*\n\s* z \1 [pause] rU)countz\s{2,}z, zYou rewrite transcripts for the xAI /v1/tts endpoint by inserting expressive speech tags. Valid inline tags (use as `[tag]`): z1. Valid wrapping tags (use as `[tag]...[/tag]`): u. Rules: - Preserve the spoken words, order, and meaning. - Do not add new spoken sentences or remove existing spoken words. - Use inline `[tag]` for short modifiers (laughs, sighs, pause, etc.). - Use wrapping `[tag]...[/tag]` for sustained effects (whisper, soft, slow, fast, loud, etc.). - Do not use angle-bracket tags like `...` — xAI uses BBCode-style closing tags with `[/tag]`. - Do not use SSML. - Do not explain or comment. - Return only the tagged TTS script.rcall_llmrTsystemrolecontentuserzTRANSCRIPT TO TAG: rQtaskmessages temperature$```(?:[A-Za-z0-9_-]+)?\s*(.*?)\s*```rz?xAI TTS audio tag rewrite failed; using locally-tagged text: %sN)r|rr_XAI_SPEECH_TAG_REsearch_XAI_FIRST_SENTENCE_REr_XAI_INLINE_SPEECH_TAGS_XAI_WRAPPING_SPEECH_TAGSagent.auxiliary_clientr"_extract_auxiliary_message_content fullmatchDOTALLrr"rr) rcleanlocalinlinewrapping system_promptrrtaggedfencers r_apply_xai_auto_speech_tagsrcs JJLLE   E F; U 3 3E F;U + +E  $ $U + +K&**>5*JJ F9c5 ) ) / / 1 1E  '' YY. / /Fyy233H /17 8;: :=E E H/ /3333338!!m<<,J5,J,JKK    4H==CCEE DfTVT]^^^  ,[[^^))++F  VX[\\\ sB F GF?9G?Gcddl}ddlm}|}t|dpd}|st d|di}t|dtpt}t|d tpt} t|d t} t|d t} t|d |d t} |d|d} | 0| dkr* t| } n#tt f$rd} YnwxYw| (t!t"t%t&| } |d|d}|0|dkr* t|}n#tt f$rd}YnwxYw|t!dt%d|}| rt)|}t|dp*|dpt+dpt,d}|drdnd}||| d}|dks| tks|dkr(| tkrd|i}| r| |d <|dkr| r| |d <||d<| | t2kr| |d<||t4kr||d<||dd|dt9d|d}|t=|d 5}||j dddn #1swxYwY|S)!z Generate audio using xAI TTS. xAI exposes a dedicated /v1/tts endpoint instead of the OpenAI audio.speech API shape, so this is implemented as a separate backend. rNresolve_xai_http_credentialsrrzSNo xAI credentials found. Configure xAI OAuth in `hermes model` or set XAI_API_KEY.r_rlanguage sample_ratebit_rateauto_speech_tags speech_tagsroptimize_streaming_latencyrVr XAI_BASE_URL/.wavrr)rrrcodecrz/ttsBearer application/json) Authorization Content-Typez User-Agent<)headersjsonrr)!requeststools.xai_httprr#rr|rDEFAULT_XAI_VOICE_IDDEFAULT_XAI_LANGUAGErzDEFAULT_XAI_SAMPLE_RATEDEFAULT_XAI_BIT_RATErDEFAULT_XAI_AUTO_SPEECH_TAGSr{rrDEFAULT_XAI_SPEED_MINrDEFAULT_XAI_SPEED_MAXrr DEFAULT_XAI_BASE_URLrstriprDEFAULT_XAI_SPEED_DEFAULT.DEFAULT_XAI_OPTIMIZE_STREAMING_LATENCY_DEFAULTpostrraise_for_statusrrr)rrrrrcredsr xai_configrrrrrrrrrpayloadrrrs r_generate_xai_ttsrsOOO;;;;;; ( ( * *E%))I&&,"--3355G pnooor**J:>>*.BCCDDJJLLdPdH:>>*.BCCDDJJLLdPdHjnn]4KLLMMK:>>*.BCCDDH'):>>-+H+HII$ NN7JNN7$;$; < >h>(0M* %#0  U&???  #. &*X X X0J,-}} 0w00./11    H  k4 "A  !!!""""""""""""""" s62GGG7IIIO??PPc ddl}tdpd}|std|di}|dt}|dt }|d t }|d d } |d d } |d d} |dd} |dd} |dd}t|dpdp"tdpd}|rd|vrd|vrdnd}||d|}dd|d}d|v}|r|||| | | | d| |ddd d!}n|||d"}| |||d#$}|r| | }|d%i}|d&d'}|dkr+|d(d)}td*|d+||d,id-d}|std.t|}t|d/5}||dddn #1swxYwY|S|jd0d}d1|vrDt|d/5}||jdddn #1swxYwY|S | }|d%i}|d&d'}|dkr+|d(d)}td*|d+|nJ#t&$r=| td2|d3t)|jd4wxYwtd5)6a Generate audio using MiniMax TTS API. Supports two endpoints: - v1/text_to_speech: simple payload, returns raw audio (Content-Type: audio/mpeg) - v1/t2a_v2: nested voice_setting/audio_setting, returns JSON with hex-encoded audio Args: text: Text to convert (max 10,000 characters). output_path: Where to save the audio file. tts_config: TTS config dict. Returns: Path to the saved audio file. rNMINIMAX_API_KEYrz@MINIMAX_API_KEY not set. Get one at https://platform.minimax.io/r`rrrrrRvolpitchemotionneutralrr^bitraterPgroup_idMINIMAX_GROUP_IDzGroupId=?&rr)rrt2a_v2)rrrrrrrU)rrrchannel)rr voice_setting audio_setting)rrrr)rrr base_resp status_code status_msg unknown errorzMiniMax TTS API error (code ): datarz%MiniMax TTS returned empty audio datarrzaudio/z.MiniMax TTS returned unexpected Content-Type 'z' (z bytes)z"MiniMax TTS returned no audio data)rr rrDEFAULT_MINIMAX_MODELDEFAULT_MINIMAX_VOICE_IDDEFAULT_MINIMAX_BASE_URLr#r|rrrrTbytesfromhexrrrrr"r)rrrrr mm_configrrrrrrrrrrsepr is_t2a_v2rrr{r%r&r( hex_audio audio_bytesr content_types r_generate_minimax_ttsr7s OOO.//52G ][\\\y"--I MM'#8 9 9E}}Z)ABBH}}Z)ABBH MM'3 ' 'E --s # #C MM'1 % %EmmIy11G-- u55KmmIv..G IMM* % % +,,2244 = , - - 3 : : < < 8Jh..H__cc#777X77+,7,,G H$I $"  +"   (    }}XGWb}QQH,A!!###JJ{B// mmM266 !  "|_EEJZkZZjZZ[[ [JJvr**..w;;  HFGG GmmI.. +t $ $ ! GGK  ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! '++NB?? | # #k4(( *A())) * * * * * * * * * * * * * * *  ]]__F ;33I#-- r::Ka&]]<II "#^+#^#^R\#^#^___      % % ' ' '333())333  ?@@@s2/KKK L44L8;L8A1N44AO;ctdpd}|std|di}|dt}|dpt}|drd}n2|d rd }n|d rd }nd }t } ||5} | jj ||||} tj | j } dddn #1swxYwYn^#t$rt$rG} td| dt!dt#| j| d} ~ wwxYwt'|d5} | | dddn #1swxYwY|S)zGenerate audio using Mistral Voxtral TTS API. The API returns base64-encoded audio; this function decodes it and writes the raw bytes to *output_path*. Supports native Opus output for Telegram voice bubbles. MISTRAL_API_KEYrz?MISTRAL_API_KEY not set. Get one at https://console.mistral.ai/rarrrfrrrz.flacrrr)rrrrNzMistral TTS failed: %sTrzMistral TTS failed: r)r rrDEFAULT_MISTRAL_TTS_MODELDEFAULT_MISTRAL_TTS_VOICE_IDrr7rrcompletebase64 b64decode audio_datar"rerrorrTr__name__rr)rrrr mi_configrrrr5rrr5r&rs r_generate_mistral_ttsrCs.//52G \Z[[[y"--I MM'#< = =E}}Z((H,HHF##     f % %    g & &  $&&G M WW % % % @|*33! / 4H !*8+>??K @ @ @ @ @ @ @ @ @ @ @ @ @ @ @  MMM -q4 @@@D$q''2BDDEE1LM k4 A   sO D*=D D*D""D*%D"&D**F>AFFF::F>F> pcm_bytesrchannels sample_widthc @ddl}||z|z}||z}t|}|dddd|||||dz }|dd |} d t|zt| z|z} |d d | d } | |z| z|zS)aWrap raw signed-little-endian PCM with a standard WAV RIFF header. Gemini TTS returns audio/L16;codec=pcm;rate=24000 -- raw PCM samples with no container. We add a minimal WAV header so the file is playable and ffmpeg can re-encode it to MP3/Opus downstream. rNz <4sIHHIIHHsfmt rUz<4sIsdataz<4sI4ssRIFFsWAVE)structrpack) rDrrErFrK byte_rate block_align data_size fmt_chunkdata_chunk_header riff_size riff_headers r_wrap_pcm_as_wavrTsMMMh&5I\)KII   q  I FGY??C NN"S):%;%;;iGI++hGDDK  "%6 6 BBr3 gemini_configc|d}t|tr|sdStj|}t|}| s; ddl m }||z }n&#t$rtj |z }YnwxYw|S)z7Return the configured persona prompt file path, if any.persona_prompt_fileNr)get_hermes_home)rrxr#r|rr= expandvarsrrM is_absoluter\rXr"cwd)rUrexpandedr=rXs r#_resolve_gemini_persona_prompt_pathr]s   1 2 2C c3  syy{{tw!!#))++..H >> $ $ & &D     % % 8 8 8 8 8 8"?$$t+DD % % %8::$DDD % Ks(B<< CCct|}|dS |dS#ttf$r'}t d||Yd}~dSd}~wwxYw)zERead the Gemini persona prompt file, failing soft on config mistakes.NrrFrGz4Gemini TTS persona prompt file unavailable at %s: %s)r] read_textr|OSErrorUnicodeDecodeErrorrr)rUr=rs r_read_gemini_persona_promptrbs .} = =D |r~~w~//55777 ' ( B     rrrrr s'=A5A00A5rc|pdddd}d|vod|vS)zIReturn True for Gemini TTS models known to support expressive audio tags.rrrUr'z gemini-3.1r)r|r}rv)rr~s r!_gemini_model_supports_audio_tagsrdsM+2$$&&,,..55c1==bAJ : % =%:*==r3c|d}t|tr|d}t|t}|sdSt |st d|dSdS)N audio_tagsrrrFzrGemini TTS audio_tags enabled, but model %s is not known to support Gemini audio tags; skipping hidden tag rewriteT)rrxrrDEFAULT_GEMINI_AUDIO_TAGSrdrr)rUrrrrs r_gemini_audio_tags_enabledrhs   L ) )C#t!ggi  3(ABBBG u ,U 3 3 =    u 4r3rc|pd}tjd|tj}|r'|d}|S)NrrrrU)r|rrrr)rrrs r_clean_gemini_audio_tag_rewriterj&sX ] ! ! # #E L@%ry Y Y YE ' A$$&& Lr3rc |jd}t|dd}t|tr$t |dpdSt t|ddpdS#t $rYdSwxYw)Nrmessagerr)choicesr5rxrr#rr")rchoicerls rrr.s!!$&)T22 gt $ $ 5w{{9--344 477Ir228b999 rrsAA9A99 BBrpersona_promptc^|}|s|Sd}|pd}d|d|} ddlm}|td|dd |dgd }t t |}|p|S#t $r'} td | |cYd } ~ Sd } ~ wwxYw)z?Use the configured auxiliary model to insert Gemini audio tags.aYou rewrite transcripts for Gemini 3.1 Flash TTS by inserting expressive audio tags. Audio tags are inline square-bracket modifiers such as [whispers], [excitedly], [very slow], [sarcastically], [laughs], [sighs], or [gasp]. There is no fixed allowlist. Use creative freeform tags generously but naturally to control tone, pace, emotional vibe, emphasis, section-level delivery, and non-verbal sounds. Use English audio tags even when the spoken transcript is not English. Rules: - Preserve the spoken words, order, and meaning. - Do not add new spoken sentences or remove existing spoken words. - Use square brackets for every audio tag. - Do not use SSML or XML tags. - Do not explain or comment. - Return only the tagged TTS script.z(none)zPERSONA AND DIRECTOR CONTEXT: z TRANSCRIPT TO TAG: rrrrrrQrz>& ! ! 5[[V44F,,F,,2244 4 4 4 5 M MN M M M M S S U UUr3cx ddl}tdptdpd}|std|di}t |t r|ni}t|dtpt}t|d tpt}t|d ptd pt d } t|} |} t||rt|| } t| || } t!d|} t#| | kr3t$dt#| | | d| } dd| igigdgddd|iiidd}| d|d}||d|iddi|d}|jdkr |di}|d p|jdd!}n#t0$r|jdd!}YnwxYwt3d"|jd#| |}|d$dd%d}t5d&|Dd}|t3d'|d(p|d)pi}|d*d}n2#t6t8t:f$r}t3d+||d}~wwxYw|st3d,t=j|}tA|}|!"d-r?tG|d.5}|$|dddn #1swxYwY|StKj&d-d/05}|$||j'}dddn #1swxYwY tQj)d1}|r|!"d2r|d3|d4d5d6d7d8d9d:d;dd?tTj,t[@} | j.dkr6| j/0dAdBCdd!}!t3dD|!n0t$dE|tQj1|| tej3|n:#th$rYn.wxYw# tej3|w#th$rYwwxYwxYw|S)FaGenerate audio using Google Gemini TTS. Gemini's generateContent endpoint with responseModalities=["AUDIO"] returns raw 24kHz mono 16-bit PCM (L16) as base64. We wrap it with a WAV RIFF header to produce a playable file, then ffmpeg-convert to MP3 / Opus if the caller requested those formats (same pattern as NeuTTS). Args: text: Text to convert (prompt-style; supports inline direction like "Say cheerfully:" and audio tags like [whispers]). output_path: Where to save the audio file (.wav, .mp3, or .ogg). tts_config: TTS config dict. Returns: Path to the saved audio file. rNGEMINI_API_KEYGOOGLE_API_KEYrzIGEMINI_API_KEY not set. Get one at https://aistudio.google.com/app/apikeyrbrrrGEMINI_BASE_URLr)roz@Gemini TTS composed prompt too long (%d chars), truncating to %dpartsrAUDIO voiceConfigprebuiltVoiceConfig voiceName)responseModalities speechConfig)contentsgenerationConfigz/models/z:generateContentrrrr)paramsrrrrur@rl,zGemini TTS API error (HTTP r* candidatesrc3*K|]}d|vsd|v |VdS) inlineData inline_dataNr)rps rrz'_generate_gemini_tts..s7WW|q/@/@MUVDVDV1DVDVDVDVWWr3z+Gemini TTS response contained no audio datarrr+z#Gemini TTS response was malformed: z$Gemini TTS returned empty audio datarrFrdeleter`rfrgrhrirjrnrkrlrmrurn -loglevelTrorprFrrrszffmpeg conversion failed: zEffmpeg not found; writing raw WAV to %s (extension may be misleading))5rr r|rrrxrr#DEFAULT_GEMINI_TTS_MODELDEFAULT_GEMINI_TTS_VOICEDEFAULT_GEMINI_TTS_BASE_URLr rbrhrur{rrrrrr&rrr"rTnextKeyError IndexErrorrr=r>rTr}rrrrQNamedTemporaryFilerrarbrr r"r r8rrwcopyfilerremover`)"rrrrrraw_gemini_configrUrrrro tts_script prompt_textmax_lenrendpointrerrrYr+r audio_partr audio_b64r&rD wav_bytesrtmpwav_pathr`cmdr{rs" r_generate_gemini_ttsrs"OOO-..W-@P2Q2QWUW^^``G   W   #x44)34Et)L)LT%%RTM  !!'+CDD E E K K M M iQiE  !!'+CDD E E K K M M iQiE*%% ' * + + ' & eggffSkk  1??NJ!-77Y3DXXX ,%K 'x<8=#+>FF ) ) )]4C4(FFF ) K(*> K K6 K K    M}}\"1%i09WWeWWWY]^^  LMM M --T 1N1NTRTJJvr** j) ,MMMDDDEE1LM CABBB ++I ++I##F++ +t $ $  GGI                    $F5 A A AS )8h''  3  ""++F33 XD(y%E65+w tXt[';W^CbPZPbsEsGsGHHHF A%%--gh-GGM"#H#H#HIII& NNW    OHk 2 2 2  Ih        D   Ih        D  s7A JJ J >BMN 1NN 7PP P<Q%%Q),Q)1C9V+V V  V V7V'&V7' V41V73V44V7cf ddl}|jdduS#t$rYdSwxYw)z=Check if the neutts engine is importable (installed locally).rNrdFimportlib.utilutil find_specr" importlibs r_check_neutts_availabler sP~''11== uu " 00cf ddl}|jdduS#t$rYdSwxYw)z@Check if the kittentts engine is importable (installed locally).rNr?Frrs r_check_kittentts_availabler)sP~'' 44D@@ uurcZtttjdz dz S)z9Return path to the bundled default voice reference audio.neutts_sampleszjo.wavr#r__file__r+rr3r_default_neutts_ref_audior2$ tH~~$'77(B C CCr3cZtttjdz dz S)z>Return path to the bundled default voice reference transcript.rzjo.txtrrr3r_default_neutts_ref_textr7rr3c8ddl}|di}|ddp t}|ddp t}|dd}|d d }|} |d s|d d dd z} t ttj dz } |j | d|d| d|d|d|d|g} tj | dddtj } | jdkrk| j} d| D}t%dt'd|pd| |krxt+jd}|rM|d| ddd |g}tj |dd!tj t/"t1j| nt1j| ||S)#aGenerate speech using the local NeuTTS engine. Runs synthesis in a subprocess via tools/neutts_synth.py to keep the ~500MB model in a separate process that exits after synthesis. Outputs WAV; the caller handles conversion for Telegram if needed. rNrd ref_audiorref_textrzneuphonic/neutts-air-q4-ggufdevicecpurrrUzneutts_synth.pyz--textz--outz --ref-audioz --ref-textz--modelz--deviceTrrqrrrc<g|]}|d|S)zOK:) startswith)rls r z$_generate_neutts..`s)QQQQQ\\%=P=PQqQQQr3zNeuTTS synthesis failed: r)r`rgrnrr@rocheckrrr1)sysrrrrrvr#rrr+ executablerr r"r8rr| splitlinesrTchrrrarbr rrrename)rrrr neutts_configrrrrr synth_scriptrr{r error_linesr`conv_cmds r_generate_neuttsr<s*JJJNN8R00M!!+r22Q6O6Q6QI  R00N4L4N4NH   g'E F FE   x / /FH    ' ':%%c1--a069tH~~,/@@AAL  $yh5F C^C4T^Tf g g gF A$$&&QQ&"3"3"5"5QQQ es2ww||K7P7P7cTceefff;h''  -hk7KXH N84:CUeweyey z z z z Ih     Ih , , , r3_piper_voice_cachecf ddl}|jdduS#t$rYdSwxYw)z2Check whether the piper-tts package is importable.rNrDFrrs r_check_piper_availabler{sP~''00<< uurcrddlm}t|dd}|dd|S)zReturn the directory where Hermes caches Piper voice models. Resolves to ``~/.hermes/cache/piper-voices/`` under the active HERMES_HOME so voice downloads follow profile boundaries. rrYzcache/piper-voicespiper_voices_cacheTrC)r\rZrrN)rZroots r_get_piper_voices_dirrsL 0///// 35IJJ K KDJJtdJ+++ Kr3r download_dirc f|st}t|}|jdkr#|rt |S||dz }|r)||dz rt |Sddl}t d|| tj |j dd|dt |gd d d tj }n+#tj$r}td |d |d}~wwxYw|jdkr:|jpdpd}td|d|dd|std|dt |S)a}Resolve *voice* (a model name or path) to a concrete .onnx file path. Accepts any of: - Absolute / expanded path to an .onnx file the user already has - A voice *name* like ``en_US-lessac-medium`` (downloads to ``download_dir`` on first use via ``python -m piper.download_voices``) Raises RuntimeError if the model can't be located or downloaded. z.onnxz .onnx.jsonrNz0[Piper] Downloading voice '%s' to %s (first use)z-mzpiper.download_voicesz--download-dirTrrz/Piper voice download timed out after 300s for 'rrzno stderr outputz!Piper voice download failed for 'z': iz#Piper voice download completed but uh is missing — check voice name (see: https://github.com/OHF-Voice/piper1-gpl/blob/main/docs/VOICES.md))DEFAULT_PIPER_VOICErrMrr}rOr#rrrrr rr"r*rTr8rr|)rr candidatecached_sysr{rrs r_resolve_piper_voice_pathrs $#U &&((I7**y/?/?/A/A*9~~uOOO +F }}Le+?+?+??GGII6{{ KKBE<XXX  _d$;U s<00 2dC$      $ Fe F F F    A-%2,,..D2D H H H&#, H H    ==??  )& ) ) )   v;;s;D D5D00D5c t}ddl}t|tr|dinidpt }t dp t}| ddtdd }t||}d d} t| tst| tsd} n| } |d |} | tvrTtd |||| t| <tdt| } d} t#fddD}|r ddlm}|t)ddt)ddt)ddt)ddtdd| } n*#t*$rtdYnwxYw|}|ds|ddddz}||d5}| | ||| n| ||dddn #1swxYwY||krt7jd }|r^|d!|d"d#d$|g}t;j|dd%t:jtA& tCj"|n%#tF$rYnwxYwtCj$|||S)'zGenerate speech using the local Piper engine. Loads the voice model once per process (cached by absolute path) and writes a WAV file. Caller is responsible for converting to MP3/Opus via ffmpeg when a different output format is required. rNrDr voices_dirTrCuse_cudaF speaker_idz::cuda=z[Piper] Loading voice: %s)rz[Piper] Voice loadedc3 K|]}|vV dSrr)rk piper_configs rrz&_generate_piper_tts..s<    \      r3) length_scale noise_scale noise_w_scalevolumenormalize_audior)SynthesisConfigrrRrgMbX?rg?rruZ[Piper] SynthesisConfig not available in this piper-tts version — advanced knobs ignoredrrrUr) syn_configr`rgrnrr@ror)%rEwaverxrrrrrrMrNryrrzrrrloadanyrDrr{rrrrvrsynthesize_wavrarbrr r"r rrr`r)rrrrCr voice_namerr model_path _raw_speakerr cache_keyrr has_advancedrrwav_filer`rrs @r_generate_piper_ttsrsSJKKK2>'2...TVL!!'**A.AJ ((66Q:O:Q:QRR]]__Ltd333L$$Z7788H*:|DDJ  ##L!44L,%%"Z c-J-J" ! 00h00I*** /<<<(2 X(V(V9% *+++ y )E J          L  - - - - - -("<#3#3NC#H#HII!,"2"2=%"H"HII#L$4$4_c$J$JKK\--h<<== $\%5%56G%N%N O O% JJ    NN5      H    ' ':%%c1--a069 8T " "1h  !  xJ G G G G  x 0 0 0 111111111111111;h''  -hk7KXH N84:CUeweyey z z z z  (####     Ih , , , s7?B kt_config model_namerrrrrsfrr`rs r_generate_kittenttsr/s"##I{B//Iw(?@@J MM'#: ; ;E MM'3 ' 'E|T22J/// 3Z@@@-6Yz-B-Bz* ;<<< ": .E NN4uEjN Q QEH    ' ':%%c1--a069HHXue$$$;h''  -hk7KXH N84:CUeweyey z z z z Ih     Ih , , , r3cDrstddStt}t |}t |}t |kr4td|t |d|ddl m }|dd  }|d k}|rbdd l m }||rtjdd |d ddSt!|} |t%| |} nt&jd} t!t,} | dd|t1|} | d| d| z } n|r|dvr | d| dz } n | d| dz } | jddt5|  |0td|t9||n`|t:vrt=|x} | n>|dkrf t?n)#t@$rtjddddcYSwxYwtdtCn|dkrf tEn)#t@$rtjddddcYSwxYwtdtGnf|dkr-td tIn3|d!kr-td"tKn|d#krf tMn)#t@$rtjdd$ddcYSwxYwtd%tOn|d&kr-td'tQna|d(krTtSstjdd)ddStd*tUn|d+krf tWn)#t@$rtjdd,ddcYSwxYwtd-tYn|d.krf t[n)#t@$rtjdd/ddcYSwxYwtd0t]n/d} t_n#t@$rd}YnwxYw|rtd1 ddl0}|j12d235}|3fd44d56dddn #1swxYwYn#tj$r&tmj7tqYnYwxYwtSr.td7d(}tUntjdd8ddStrj:;r#trj:<dkrtjdd9|d:ddSd}|Mt{|r=>dst}|r|>d}n|t:vrOt|}|r=>dst}|r|>d}nL|r/|d;vr+>dst}|r|d}n|dvr|o>d}trj:<}td<|d=|d>}|rd?|}tjd|||d@dS#t$r>}dA|dB|}tBdC|t|dcYd}~Sd}~wt$r@}dD|dB|}tBdC|dEt|dcYd}~Sd}~wt$r@}dF|dB|}tBdC|dEt|dcYd}~Sd}~wwxYw)Gac Convert text to speech audio. Reads provider/voice config from ~/.hermes/config.yaml (tts: section). The model sends text; the user configures voice and provider. On messaging platforms, the returned MEDIA: tag is intercepted by the send pipeline and delivered as a native voice message. In CLI mode, the file is saved to ~/voice-memos/. Args: text: The text to convert to speech. output_path: Optional custom save path. Defaults to ~/voice-memos/.mp3 Returns: str: JSON result with success, file_path, and optionally MEDIA tag. zText is requiredF)successz>TTS text too long for provider %s (%d chars), truncating to %dNr)get_session_envHERMES_SESSION_PLATFORMrtelegram)has_traversal_componentz/output_path contains '..' traversal component: zM. Use an absolute path or one relative to the current directory without '..'.)rr@) ensure_asciiz %Y%m%d_%H%M%STrCtts_r>rbr0rarcrfz.mp3z3Generating speech with command TTS provider '%s'...rcz`ElevenLabs provider selected but 'elevenlabs' package not installed. Run: pip install elevenlabsz$Generating speech with ElevenLabs...r0zz%text_to_speech_tool..C sGK0B4S]0^0^$_$_r3rrz9Edge TTS not available, falling back to NeuTTS (local)...zhNo TTS provider available. Install edge-tts (pip install edge-tts) or set up NeuTTS for local synthesis.z-TTS generation produced no output (provider: )>r_rFrDrdr`r?z,TTS audio saved: %s (%s bytes, provider: %s),zMEDIA:z[[audio_as_voice]] )r file_path media_tagrrzTTS configuration error (r*z%szTTS dependency missing (rzTTS generation failed ()Er| tool_errorrrrrrrrgateway.session_contextrr}tools.path_securityr rdumpsrrMr@datetimenowstrftimeDEFAULT_OUTPUT_DIRrNrr+r#rrZrrr-rrr2rr7rr7rCrrrr@rrErr'concurrent.futuresfuturesThreadPoolExecutorsubmitr{rTrr rrr=rOrxrrr|rrr@ryr")rrrcommand_provider_configrrplatform want_opusr r timestampout_dirr _plugin_pathedge_available concurrentpoolr opus_pathplugin_voice_compatible file_sizerr& error_msgrrs` @@rtext_to_speech_toolr/isU * =tzz||=,e<<<<!##JZ((H ?xTT'x< ? ? ? x < < < <  ! ! '%'''' ' ' 'z$[##!&'''''' ' KK> ? ? ? x < < < <  " " KK? @ @ @ !$* = = = =    KK; < < < dHj 9 9 9 9  " " '&(((( ' ' 'z$H##!& '''''' ' KKG H H H !$* = = = =  ! ! KKE F F F x < < < <  ! !*,, 'z$F##!& '''' KKB C C C T8Z 8 8 8 8  $ $ '!#### ' ' 'z$J## !& '''''' ' KKL M M M h ; ; ; ;   ' ' ' 'z$5## !& '''''' ' KKA B B B h ; ; ; ;"N ' """" ' ' '!& ' ' @AAAP----#+>>1>MM-QU ______ &&,,,---------------$PPPK 24: N NOOOOOP(** ' WXXX# x<<<<z$E##!& ''''w~~h'' #27??8+D+D+I+I: TTTT"### #! " .00GHH =((00- 0 : :I -#,#+#4#4V#<#< 2 2 2 'K8&T&T #& =((00- 0 : :I -#,#+#4#4V#<#<  GVVV%%f--W)22I ($#' F F F(FX->->v-F-F GOOH--  BHQZN^N^`hiii(X''  ;:y::Iz!" 0      444@@@Q@@  T9%%%)U333333333 444?x??A??  T9t 444)U333333333 444>h>>1>>  T9t 444)U333333333 4sz;AaI%$a%#J a J  6aKa#K74a6K77BaN#"a##O aO  Ba3a Ra#S?aS6a9Ta#T.+a-T..2a!U0/a0 U?<a>U??aX>/W9- X9W==XW=Xa-X52a4X55Aa Aa,Ea d3b d d5c d d5dddctrdS tdS#t$rYnwxYw tt drdSn#t$rYnwxYw t t rdSn#t$rYnwxYwt drdS ddlm}| drdSn#t$rYnwxYwt dst drdS tt d rdSn#t$rYnwxYwtrdStrdStrdSd S) a, Check if at least one TTS provider is available. Edge TTS needs no API key and is the default, so if the package is installed, TTS is available. A user-declared command provider also satisfies the requirement. Returns: bool: True if at least one provider can work. Trrrrrr}r~r9F)r^r'rr-r r2_has_openai_audio_backendrrrr"r7rrrrs rcheck_tts_requirementsr2 s%&&t t        - . . 4         $ & & 4       &''t ?????? ' ' ) ) - -i 8 8 4       %&&-8H*I*It    * + + 4         t!##tt 5sS" //A A A $B BB%#C CC<D D)(D)cbt}|rtds |tdfStd}|Cd}t stdr|dt dzz }t ||jt|j ddd d fS) amReturn ``(api_key, base_url, is_managed)`` for the OpenAI audio client. ``is_managed`` is True when the config resolves to the Nous managed audio gateway (a restricted proxy), so callers can coerce the request to what the gateway supports. When ``tts.use_gateway`` is set the gateway is preferred even if direct OpenAI credentials are present. rF openai-audioNz8Neither VOICE_TOOLS_OPENAI_KEY nor OPENAI_API_KEY is setz. zmanaged OpenAI audio for TTSrv1T) rrDEFAULT_OPENAI_BASE_URLrrrrnous_user_tokenrgateway_originr )direct_api_keymanaged_gatewayrls rrr s233N>oe44>6==2>BBOL % ' ' ?5+A+A  72 G !!! '?188==@@@$GG  r3cVttptdS)zPReturn True when OpenAI audio can use direct credentials or the managed gateway.r4)ryrrrr3rr1r1 s% ,..^2N~2^2^ _ __r3z(?<=[.!?])(?:\s|\n)|(?:\n\n)z```[\s\S]*?```z\[([^\]]+)\]\([^)]+\)z https?://\S+z \*\*(.+?)\*\*z \*(.+?)\*z`(.+?)`z^#+\s*z ^\s*[-*]\s+z---+z\n{3,}cFtd|}td|}td|}td|}t d|}t d|}td|}td|}td|}td|}| S)z:Remove markdown formatting that shouldn't be spoken aloud.rz\1rrw) _MD_CODE_BLOCKr_MD_LINK_MD_URL_MD_BOLD _MD_ITALIC_MD_INLINE_CODE _MD_HEADER _MD_LIST_ITEM_MD_HR _MD_EXCESS_NLr|)rs r_strip_markdown_for_ttsrG s   c4 ( (D <<t $ $D ;;r4 D <<t $ $D >>% & &D   ud + +D >>"d # #D   R & &D ::b$  D   VT * *D ::<<r3 text_queue stop_eventtts_done_eventdisplay_callbackc | ddttt}|di}|d|d|dt di|di|diit dpd}|stdn t}|| n*#t$rtd YnwxYw t}| d d d  nj#ttf$r'} td| dYd} ~ n7d} ~ wt $r'} td| dYd} ~ nd} ~ wwxYwd} d} d} d} gt#jdt"j}dt(ff d }ds; || }n5#t,j$r#t1| | kr || d} Y^wxYw|6|d| } | r || n| |z } |d| } d| vrd| vr t6| }|n_|}| d|}| |d} t1|| kr|| z} n |||; |n#t,j$rYnwxYw,n2#t $r%} td| Yd} ~ nd} ~ wwxYw:  n#t $rYnwxYw|!dS#:  n#t $rYnwxYw|!wxYw)aConsume text deltas from *text_queue*, buffer them into sentences, and stream each sentence through ElevenLabs TTS to the speaker in real-time. Protocol: * The producer puts ``str`` deltas onto *text_queue*. * A ``None`` sentinel signals end-of-text (flush remaining buffer). * *stop_event* can be set to abort early (e.g. user interrupt). * *tts_done_event* is **set** in the ``finally`` block so callers waiting on it (continuous voice mode) know playback is finished. Nrcrstreaming_model_idrrrz8ELEVENLABS_API_KEY not set; streaming TTS audio disabledrz8elevenlabs package not installed; streaming TTS disabledrOrUint16) sampleraterEdtypezsounddevice not available: %sz#sounddevice OutputStream failed: %sr~g?z].*?rsentencecH rdSt|}|sdS|d} D]0}|d|krdS1 |  | dSt |kr |d} j| d}h|D]a}rnLddl }| ||j } | ddbdSdS |dS#t$r&}td |Yd}~dSd}~wwxYw) z6Display sentence and optionally generate + play audio.Nz.!, pcm_24000rr)rPr'rUz!Streaming TTS sentence failed: %s)is_setrGr|r}r rrrrnumpy frombufferrNrreshaper"rr)rRcleaned cleaned_lowerprev audio_iterr_np audio_arrayr_play_via_tempfile_spoken_sentencesrrKr output_streamrIstream_max_lenrs r_speak_sentencez.stream_tts_to_speaker.._speak_sentencen s  "" -h77==??G #MMOO22599M)  ::<<&&u-->>FF?  $ $W - - -+  ***~7||n,,!/>/2 I#2:: %%"- ; !,!+HH%,,.."!E++++&)nnU#)n&L&L %++K,?,?A,F,FGGGG HH!E '&z:>>>>> I I IBCHHHHHHHHH IsBE1# E11 F!;FF!cd} ddl}tjdd}|j}||d5}|d|d|d |D]-}|rn| |.dddn #1swxYwYdd l m }||n2#t$r%}td |Yd}~nd}~wwxYw|r( tj|dS#t"$rYdSwxYwdS#|r& tj|w#t"$rYwwxYwwxYw) z0Write PCM chunks to a temp WAV file and play it.NrrFrrrUrVrO)play_audio_filez!Temp-file TTS fallback failed: %s)rrQrrr setnchannels setsampwidth setframeraterU writeframestools.voice_moderer"rrrrPr`) r\stop_evttmp_pathrrwfrrers rr_z1stream_tts_to_speaker.._play_via_tempfile sH  1NNN8YYsD)).ROOA&&&OOA&&&OOE***!+..#??,,"!Eu----...............=<<<<<)))) I I IBCHHHHHHHH I (+++++"8 (++++"s7CA0B7+ C7B;;C>B;?CD2 DC>9D2>DD2 D D-,D-2E6E  E EEEErzTz Streaming TTS pipeline error: %s)"clearr%DEFAULT_ELEVENLABS_STREAMING_MODEL_IDrrrr rrr-rr; OutputStreamrr`rr"rrrr#rUqueueEmptyrrr|_SENTENCE_BOUNDARY_RErend get_nowaitstoprset)rHrIrJrKrrrr+r:r sentence_bufmin_sentence_lenlong_flush_len queue_timeout_think_block_rercdeltamend_posrRr_r`rrrarbrs ` ` @@@@@@@rstream_tts_to_speakerr# s"{ .8%'' NN<44 ==X66==!5!*z8!D!DFF 2  Mz M<)LI)Lz8)L)L M M   !!566<" ) NNU V V V V [/11 #G444 [ [ [YZZZZZ [ ! ),..B$&OO#(1G%4%%M"''))))#W-)))LL!@#FFF$(MMMMMM )))NN#H#NNN$(MMMMMM)  ')*%>biPPP( Ic( I( I( I( I( I( I( I( I( I( I( I( I( I( IT   4##%%* * "}==;   |$$~55#OL111#%L  }.222|DD %%''2#OL111 E !L +..r<@@L<''Jl,J,J *)00>>9%%'''1+GHH5 x~~''((+;;;#+l#:L))) *?##%%* *Z  %%'''';      @@@93????????@  $ ""$$$##%%%%      $ ""$$$##%%%%    sCN C;:N ;$D"N !D""N (:E#"N #G 4FN  G #GN G  AN %H<;N =Q#> Q Q# Q  Q#__main__u🔊 Text-to-Speech Tool Modulez2==================================================c> |dS#t$rYdSwxYw)NTF)r)importerlabels r_checkr s9  HJJJ4   55 s  z Provider availability:z Edge TTS: installedz$not installed (pip install edge-tts)z ElevenLabs: elz¬ installed (pip install elevenlabs)z API Key: rrwznot setz OpenAI: oaiz not installedz2not set (VOICE_TOOLS_OPENAI_KEY or OPENAI_API_KEY)z MiniMax: rz API key setznot set (MINIMAX_API_KEY)z Piper: z%not installed (pip install piper-tts)z ffmpeg: u ✅ foundu(❌ not found (needed for Telegram Opus)z Output dir: z Configured provider: )registryrraConvert text to speech audio. Returns a MEDIA: path that the platform delivers as native audio. Compatible providers render as a voice bubble on Telegram; otherwise audio is sent as a regular attachment. In CLI mode, saves to ~/voice-memos/. Voice and provider are user-configured (built-in providers like edge/openai or custom command providers under tts.providers.), not model-selected.objectstringzThe text to convert to speech. Provider-specific character caps apply and are enforced automatically (OpenAI 4096, xAI 15000, MiniMax 10000, ElevenLabs 5k-40k depending on model); over-long input is truncated.)r descriptionz9Optional custom file path to save the audio. Defaults to z/audio_cache/.mp3rr)r propertiesrequired)rr parametersrc rt|dd|dS)Nrrrr)r/r)argskws rrr1 s62 XXfb ! !HH]++ - - -r3u🔊)rtoolsetschemahandlercheck_fnemojir)F)r)__doc__rr=rrloggingrrqrrrarrQ threadingrpathlibrtypingrrrr urllib.parserhermes_cli._subprocess_compatr r\r getLoggerrArr tools.managed_tool_gatewayrtools.tool_backend_helpersrrrrrrr'r-r2r7r;r@rErrrrror frozensetrrrrrr6r,r-r.r:r;rrrrr r r r rrrrrrgrqGEMINI_TTS_SAMPLE_RATEGEMINI_TTS_CHANNELSGEMINI_TTS_SAMPLE_WIDTHr#r]rrerz__annotations__rlryrrMAX_TEXT_LENGTHrrrrrrrrrrrrrrrr{rrrrrrr6r-r:r<r@rZr^rcr|rrrrrrrrxrrrrrrr7rCr/rTr]rbrdrhrjrrur{rrrrrrrrrrrrrr/r2tuplerr1rsr=r>r?r@rArB MULTILINErCrDrErFrGQueueEventrprintrrrtools.registryrr TTS_SCHEMAregisterrr3rrs!!!F    000000000000 <<<<<<000000  8 $ $ / / / /DCCCCC 100000   * $    '46(;%( &I'8&9::="+5&8=3E$,23.9!P! 0=====-,..    , ,$sCx.   ##"" 4 4 $sCx.   dt" + ,05$5$sm5$c3h(5$ 5$5$5$5$v$sCx.&Ld38nLLLLLH" # # #  '*#$)!&Y'D'D'DEE&*#8d38n8C8DcN8888S#X  #s(^0>S#X>4>>>>S#Xd38n*XL XLXLXLS#X XL c] XLXLXLXLv344 S#X     T#s(^     "&[[ cN[#[ [[[[$T#s(^3#(3->#hsmPS"sCx. <1j.>141111h%Qc%QE%Qj6Q%Q%Q%Q%QP'd'DcN't'''' A AAA cN A S#X A  AAAAHhtCH~.FRV.T.... "s"x}""""P3Sd3PS8nY\<(s(($sCx.(UX((((\<s<<$sCx.<UX<<<<D  RZ 0111MACHHMfDgDggjpp -$$ERYWWW00C0$040000AcAcAAAAHjCjcjtCH~jRUjjjj`~A~A#~A4S>~AVY~A~A~A~AH++#+4S>+VY++++f.'/ CCCCC C  CCCCDtCH~(SW." tCH~ #     >S>T>>>> d38nST"SS---c-3----f%)VV VS>VSMV VVVV>QsQQ$sCx.QUXQQQQpDD3DDDD D#DDDD 232S2d38n2QT2222x&(DcN''' t    2S222222j_c__c3h_TW____N*,S#X+++4c44c3h4TW4444x"&u4u4 u4#u4 u4u4u4u4v 66666rU3T>-B>`4````# #BCC-.. 2:. / / "*_ % % 2:& ' ' RZ % % "*Z(( RZ  6 6 6  >>>> G   9%%  # #    &9= NN NNONxt 45 NNNNh z E +,,, E(OOO E $%%% E x&&1A6*J*Jv;;Pv x xyyy E z&&1CT*J*Jx;;Px z z{{{ E XMM2F$G$GV55Y X XYYY E e&&1F*N*Nc;;Tc e efff E o0022 l558l o o E oMM:K,L,Lm==Rm o oppp E q*@*@*B*Bo;;Ho q qrrr E i++--g;;=g i ijjj E 1/ 1 1222    F}V$$H E .H . ./// 0/////// `! s ! O[n[n[p[p O O O   H   &    - -$       r3