Add local LLM voice model transcription - #1

Open

pdufour wants to merge 17 commits into

mainfrom

paul.dufour/test-local-trans

Commits on Mar 15, 2026

wip: rip out openai-only transcription, wire up transcribe-rs with parakeet onnx backend
pdufour
committed
cdc85f4
download missing parakeet model files from huggingface on first use
pdufour
committed
50f9cee
add /voicemodel slash command, voice model popup, bring back openai transcription path
pdufour
committed
0c379cb
restore chat_composer.rs voice keybinding block that got nuked
pdufour
committed
f338a8e
switch model download from curl to hf-hub api, add tracing, kill /tmp debug log
pdufour
committed
7e29b12
clean up voice model validation, replace alias map with allowed-id list, trim chatwidget
pdufour
committed
61b409b
drop normalize_selected_model and validate wrapper, inline validation into try_set
pdufour
committed
a4849dc
add 24k to 16k wav resampling for local models, simplify voice popup and set_voice_model
pdufour
committed
d633852
revert .gitignore additions for model artifacts and sample.ogg
pdufour
committed
c0cd0c4
restore openai transcribe duration_seconds param and 24k sample rate test expectations from main
pdufour
committed
b7b0c33
fix normalize_chatgpt_base_url: use parameter name base_url instead of input
pdufour
committed
185279f
voice: repo-to-class map, generic SpeechModel loading, voice_model_picker_options()
pdufour
committed
18a084d

Commits on Mar 16, 2026

voice: default to OpenAI when no /voicemodel selected
pdufour
committed
7cf9b15
Re-enable features
pdufour
committed
d88d05f
Disable voice-model when voice-input feature is off
pdufour
committed
a333f38
Add tests
pdufour
committed
8971a82
terminal clear after download finished
pdufour
committed
90af45e