- Prepare the environment. Note that parameters in
grpc_audio_client.py
are adjusted toquartznet15x5
model../fetch_models.sh # CV models; ASR ─ you load manually (onnx) pip install requests aiogram # telegram bot API dependencies
- Create a Telegram bot, obtain its token (all can be done via BotFather).
- Use
./start.sh
to launch triton-server andfastapi
"dummy client".
You can also check the script options with./start.sh -h
command. - Pass the token to the script launching the telegram bot
python voice_transcription_bot.py -t YOUR_API_TOKEN
- triton-inference-server
─ some files from there (
grpc_image_client.py
andfetch_models.sh
)
are adjusted to the use in this repository.