Runnable demos that exercise the v0.1 API surface end-to-end.
docker compose up -d
chmod +x examples/curl_voice_note.sh
examples/curl_voice_note.shThe first docker compose up pulls the faster-whisper-server image and
downloads the whisper model on first transcription (~2 min). Subsequent
runs are fast.
| Path | What it shows |
|---|---|
curl_voice_note.sh |
Multipart upload -> poll -> transcript. Fastest sanity check. |
curl_call_recording.sh |
Deepgram provider with native speaker diarization. |
chatwoot_webhook_bridge/ |
Sinatra app wiring scribed into Chatwoot's message_created webhook. |
samples/CREDITS.md |
Sample audio attribution and re-download instructions. |
openai_compatible(default) -- works with the bundledwhispersidecar, zero external credentials needed.deepgram-- setDEEPGRAM_API_KEYin.env. Native diarization, fast.openai-- setOPENAI_API_KEYin.env. 25 MB file cap.whisper_cpp-- fully local, slowest, no network egress.