@alfe.ai/openclaw-voice
v0.2.20
Published
OpenClaw tools for Alfe one-shot text-to-speech and speech-to-text
Readme
@alfe.ai/openclaw-voice
OpenClaw tools for Alfe's agent-authenticated one-shot voice endpoints.
The plugin registers two tools:
voice_ttsconverts 1–5000 characters to a new workspace-relative WAV or raw PCM file.voice_stttranscribes a workspace-relative WAV or raw mono 16-bit PCM file of at most 10 MiB.
Both tools resolve the agent API key and voice-service URL from
@alfe.ai/config, call through AgentApiClient, and leave billing to the
server. The plugin has no provider credential fields, gateway RPC, daemon
connection, or channel-specific call controls.
File arguments are treated as model-controlled input: real paths must stay inside the current workspace, input files must be regular files within the size cap, and TTS creates output exclusively without replacing existing files.
Channel voice sessions, calls, barge-in, and streaming audio live in the voice service and channel adapters; they are not part of this tool-only package.
Development
pnpm --filter @alfe.ai/openclaw-voice test
pnpm --filter @alfe.ai/openclaw-voice lint
pnpm --filter @alfe.ai/openclaw-voice buildSee DEVELOPING.md for the file-I/O and runtime contracts.
