*`i/ nddlZddlZddlZddlZddlmZddlmZddlm Z ddl m Z ddl Z ddlZddlmZmZmZmZe rddlmZejeZer*ddlZejZedd eDZnedd d iZerddlZd ejDZ Gd d Z!de"e j#zde"e j#zfdZ$de j#de j#fdZ%de j#de j#de j#fdZ&ede'de'de"de"de'de j#f dZ(dS)N)Enum)cache)Path) TYPE_CHECKING)assert_soundfile_installedassert_soxr_installedis_soundfile_installedis_soxr_installed)RawAudio AudioFormatci|]}||Sr).0 format_names h/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/mistral_common/audio.py r s&e&e&eK{K&e&e&enonec@g|]}|jSr)valuelower)rvs r r's"TTTa!'--//TTTrc6eZdZdejdededdfdZdefdZddZ e de fd Z e dd ed eddfd Ze dded eddfdZe dded eddfdZe dded eddfdZddededefdZe ddZdeddfdZdS) Audio audio_array sampling_rateformatreturnNcX||_||_||_|dS)a Initialize an Audio instance with audio data, sampling rate, and format. Args: audio_array: The audio data as a numpy array. sampling_rate: The sampling rate of the audio in Hz. format: The format of the audio file. N)rrr _check_valid)selfrrrs r__init__zAudio.__init__+s3'*  rcld|jdt|j|jz dd|jjS)NzAudio - sampling_rate=z Hz, duration=z.2fz s, shape=)rlenrshaper"s r__repr__zAudio.__repr__8sQ .T%7 . .D,--0BB K . .%+ . . rc@t|jtjs!Jt tj|jjdksJd|jjt |jtvsJd|jdtdS)Nzself.audio_array.ndim=z self.format= not in EXPECTED_FORMAT_VALUES=) isinstancernpndarraytypendimrrEXPECTED_FORMAT_VALUESr's rr!zAudio._check_valid?s$*BJ77IIbj9I9III7$)))+Fd.>.C+F+F)))"$$${44446h6h6hOe6h6h44444rc:|jjd|jz }|S)z{Calculate the duration of the audio in seconds. Returns: The duration of the audio in seconds. r)rr&r)r"durations rr3zAudio.durationEs!*03d6HHrTurlstrictc, tj|}|t|j|S#tj$r}td||d}~wt$r}td|d|d}~wwxYw)zCreate an Audio instance from a URL. Args: url: The URL of the audio file. strict: Whether to strictly enforce mono audio. Returns: An instance of the Audio class. r5z#Failed to download audio from URL: Nz*Failed to create Audio instance from URL: z .) requestsgetraise_for_statusr from_bytescontentRequestException ValueError Exception)r4r5responsees rfrom_urlzAudio.from_urlPs Z|C((H  % % ' ' '##H$4V#DD D( Q Q QH3HHIIq P Z Z ZQ#QQQRRXY Y Zs$AA BA-- B:BB audio_base64c&ttjd|r|dd} t j|}n"#t $r}td|d}~wwxYwt ||S)aCreate an Audio instance from a base64 encoded string. Args: audio_base64: The base64 encoded audio data. strict: Whether to strictly enforce mono audio. Defaults to True. Returns: An instance of the Audio class. z^data:audio/\w+;base64,,r*zHbase64 decoding failed. Please check the input string is a valid base64.Nr7) rrematchsplitbase64 b64decoder?r>rr;)rCr5 audio_bytesrAs r from_base64zAudio.from_base64ds #$$$ 8. = = 6'--c2215L p *<88KK p p pghhno o p F;;;sA A4A//A4filectt|tr|dr |dd}t |st d|dt|d5}|}dddn #1swxYwYt ||S)zCreate an Audio instance from an audio file. Args: file: Path to the audio file. strict: Whether to strictly enforce mono audio. Defaults to True. Returns: An instance of the Audio class. zfile://Nzfile=z does not existrbr7) rr,str startswithrexistsFileNotFoundErroropenreadrr;)rMr5frKs r from_filezAudio.from_file{s #$$$ dC  T__Y%?%? 8DDzz  "" ?#$=t$=$=$=>> > $   #&&((K # # # # # # # # # # # # # # # F;;;sB((B,/B,rKctj|5}tj|5}|d}|j}|j}dddn #1swxYwYdddn #1swxYwYt|}|j }|j dkr/|rtd|j | d}t|||S)zCreate an Audio instance from bytes. Args: audio_bytes: The audio data as bytes. strict: Whether to strictly enforce mono audio. Defaults to True. Returns: An instance of the Audio class. float32)dtypeNr*zaudio_array.ndim=axis)rrr)ioBytesIOsf SoundFilerV sampleraterr rrr0r>meanr) rKr5 audio_filerWrr audio_format format_enumrs rr;zAudio.from_bytesshZ $ $ ( j)) (Qff9f55 !  x  ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (",// "((**  q  7 !7K$4!7!7888)..A.66 MRXYYYYs4A2%A A2A A2"A #A22A69A6Fprefixct|tvsJd|dttj5}t j||j|j|| dtj |  d}dddn #1swxYwY|rd|d|}|S) a Convert the audio data to a base64 encoded string. Args: format: The format to encode the audio in. prefix: Whether to add a data prefix to the base64 encoded string. Returns: The base64 encoded audio data. zformat=r+)rrzutf-8Nz data:audio/z;base64,)rr1r^r_r`writerrupperseekrI b64encoderVdecoder)r"rrgrd base64_strs r to_base64zAudio.to_base64s9 #$$$////1^F1^1^E[1^1^/// Z\\ MZ HZ!143Efllnn ] ] ] ] OOA   )*//*;*;<<CCGLLJ M M M M M M M M M M M M M M M  LKv||~~KKzKKJsBCCCaudior c.t|jtrt|jSt|jt rt|jStdt|j)zCreate an Audio instance from a RawAudio object. Args: audio: The RawAudio object containing audio data. Returns: An instance of the Audio class. zUnsupported audio data type: ) r,databytesrr;rQrLr>r/)rps rfrom_raw_audiozAudio.from_raw_audios{ ej% ( ( Q##EJ// /  C ( ( Q$$UZ00 0OT%*=M=MOOPP Prnew_sampling_ratec|j|krdSttj|j|j|d|_||_dS)zResample audio data to a new sampling rate. Args: new_sampling_rate: The new sampling rate to resample the audio to. NHQ)quality)rrsoxrresampler)r"rus rrzzAudio.resamplesU  !2 2 2 F=)94;MO`jnooo.r)rN)T)F)rpr rr)__name__ __module__ __qualname__r-r.intrQr#r(r!propertyfloatr3 staticmethodboolrBrLrXrsr;rortrzrrrrr*s  BJ s C TX     #    iiii %XZZcZ4Z7ZZZ\Z&<<#/K$44Kqqq!""u ABB/I j!bji6X6XYYK rnum_frequency_bins num_mel_bins min_frequency max_frequencyrc6|dkrtd|d||krtd|d|t|}t|}tj|||dz}t |}tjd|dz|} t | |} d|d|dz|d|z z } | tj| dz} | d d krtd |d |d | S)aVCreate a Mel filter bank matrix for converting frequency bins to the Mel scale. This function generates a filter bank matrix that can be used to transform a spectrum represented in frequency bins to the Mel scale. The Mel scale is a perceptual scale of pitches judged by listeners to be equal in distance from one another. Args: num_frequency_bins: The number of frequency bins in the input spectrum. num_mel_bins: The number of desired Mel bins in the output. min_frequency: The minimum frequency (in Hz) to consider. max_frequency: The maximum frequency (in Hz) to consider. sampling_rate: The sampling rate of the audio signal. Returns: A filter bank matrix of shape (num_mel_bins, num_frequency_bins) that can be used to project frequency bin energies onto Mel bins. rzRequire num_frequency_bins: z >= 2zRequire min_frequency: z <= max_frequency: rg@Nr\gzNAt least one mel filter has all zero values. The value for `num_mel_filters` (z?) may be set too high. Or, the value for `num_frequency_bins` (z) may be set too low.) r>rr-linspacerrrmaxany) rrrrrmel_minmel_max mel_freqsrr mel_filtersenorms rmel_filter_bankr&sf2AQ8JQQQRRR}$$d=ddUbddeee=))G=))G GWlQ.>??I **L A}13EFFI0LIIK <L1$4 45 ]l]8SS TE2>%+++KQ3&++--  :0< : :#  : : :    r))rIr^loggingrFenumr functoolsrpathlibrtypingrnumpyr-r8mistral_common.importsrrr r &mistral_common.protocol.instruct.chunkr getLoggerr{logger soundfiler`available_formatsr ry __members__valuesr1rrr.rrrr~rrrrrs   @??????  8 $ $ 8-,..$}&e&eSd&e&e&effKK$}vv&677KKKKTT;3J3Q3Q3S3STTTz/z/z/z/z/z/z/z/zurz)ebj.@0rzbj&bj WYWa(5555 5  5 Z 555555r