@acpfx/stt-deepgram
v0.2.6
Published
Speech-to-text via Deepgram streaming API. Streams partial transcriptions in real time with configurable VAD and endpointing.
Readme
@acpfx/stt-deepgram
Speech-to-text via Deepgram streaming API. Streams partial transcriptions in real time with configurable VAD and endpointing.
Usage
This package is a pipeline node for @acpfx/cli. See the CLI package for installation and usage.
Requires a DEEPGRAM_API_KEY environment variable.
Manifest
- Consumes:
audio.chunk - Emits:
speech.partial,speech.final,speech.pause,lifecycle.ready,lifecycle.done,control.error
Settings
| Name | Type | Default | Description |
|------|------|---------|-------------|
| language | string | en | Language code |
| model | string | nova-3 | Deepgram model name |
| utteranceEndMs | number | 1000 | Silence ms before utterance end |
| endpointing | number | 300 | VAD endpointing threshold in ms |
| apiKey | string | | Overrides DEEPGRAM_API_KEY env var |
Pipeline Example
nodes:
stt:
use: "@acpfx/stt-deepgram"
settings: { language: en, model: nova-3 }
outputs: [bridge]
env:
DEEPGRAM_API_KEY: ${DEEPGRAM_API_KEY}External Links
- Deepgram -- Speech AI platform
- Deepgram Developer Docs -- API reference and guides
License
ISC
