initial input layout fix

use also file mode to send audio from mic to whisper

remove audio debug zone

make whisper pipeline read file instead of numpy array

add chunk_length_s=30 to asr pipeline

add debug on processed audio

add audio file input

add audio preprocessing to make sure asr input is single channel audio

fix type error

fix the issue microphone dosen't release

re-enable zero gpu

import spaces at first

debug zero gpu

enable zero gpu

Update README.md

drop deoendency on llama.cpp

Update README.md

switch to gradio

add sentencepiece

update outlines to latest

use transformers library for asr and llm inference, too.

switch to 3b llm

quote messgae to classify in prompt text

secure get_tts_message func

disable speculative decoding as it harms precision in classification

add pypinyin

add missing pkgs

fix numpy error

fix requirements

llama.cpp optimization: speculative dec + use cpu optmized build

add llama-cpp-python again

specify docopt by wheel link

specify jieba wheel link

update requirements

addnecessary modules

specify llama cpp pyhton ver

specify jieba ver

add jieba to requirements

downgrade llamacpppython again

downgrade llamacpppython again

update requirements

downgrade llamacpp python version

fix tts and improve rrs-related ergonimicity

try qwen2.5 7b q2

try deepseek 3b

add intent def in chinese too

adjust prompt

downgrade outlines to 0.0.36 to avoid issue 820

add missing pkg

fix numpy error