CoolFace
Apppublic

Spock24/nav-willow

sourceHugging Faceupdated 3mo agoView on Hugging Face
0likes
App README

Nav Willow STT

Self-hosted speech-to-text endpoint for Nav. This Space exposes a Willow/WIS-style ASR endpoint:

text
POST /api/asr
multipart field: audio_file

It also exposes an OpenAI-style compatibility endpoint:

text
POST /audio/transcriptions
multipart field: file

After deploying this folder as a Hugging Face Docker Space, set Nav backend/.env:

env
STT_PROVIDER=willow
WILLOW_API_URL=https://YOUR-USERNAME-nav-willow.hf.space
WILLOW_ENDPOINT_STYLE=wis
WILLOW_LANGUAGE=en
WILLOW_MODEL=small
HF_TOKEN=hf_...   # only if the Space is private

Health check:

bash
curl https://YOUR-USERNAME-nav-willow.hf.space/health

Transcription test:

bash
curl -X POST "https://YOUR-USERNAME-nav-willow.hf.space/api/asr" \
  -H "Authorization: Bearer $HF_TOKEN" \
  -F "audio_file=@test.webm" \
  -F "model=small"

Notes:

  • —tiny and base are faster on CPU.
  • —small is a better accuracy/speed compromise.
  • —medium and above are best on GPU Spaces.