Saidwire plugin for LiveKit Agents

Support for Saidwire real-time speech-to-text in LiveKit Agents.

The plugin streams the user's audio to Saidwire over a WebSocket and emits interim and final transcripts. When the agent session's VAD or turn detector ends the user's turn, it flushes the stream; the plugin sends Finalize, and the final transcript comes straight back.

Saidwire is English-only. Pricing is $0.30 per audio hour, billed per second of audio sent. This plugin is built and maintained by Saidwire, the company that runs the service.

Tested with livekit-agents 1.8.10.

Installation

pip install livekit-plugins-saidwire

Pre-requisites

You'll need an API key from Saidwire (saidwire.com/dev). It can be set as an environment variable: SAIDWIRE_API_KEY.

Usage

from livekit.agents import AgentSession
from livekit.plugins import saidwire, silero

session = AgentSession(
    stt=saidwire.STT(),  # reads SAIDWIRE_API_KEY
    vad=silero.VAD.load(),
    llm=...,
    tts=...,
)

Already have an agent? Replace your current stt= with saidwire.STT(). Nothing else changes.

A complete agent is in examples/agent.py (its LLM and TTS go through LiveKit Inference, so only LiveKit credentials are needed for them):

pip install "livekit-agents[silero]" livekit-plugins-saidwire python-dotenv
export SAIDWIRE_API_KEY=...
export LIVEKIT_URL=... LIVEKIT_API_KEY=... LIVEKIT_API_SECRET=...
python examples/agent.py console      # talk to it in your terminal

Options

stt = saidwire.STT(
    api_key="...",  # default: $SAIDWIRE_API_KEY
    model="saidwire-en-2026-10-30",
    language="en",  # en, en-US, en-GB, ... (English only)
    sample_rate=16000,  # 8000 for telephony audio
    interim_results=True,
    punctuate=True,
    endpointing_ms=None,  # server default; 0 = leave turn-taking to the session
    utterance_end_ms=None,  # e.g. 1000: END_OF_SPEECH after 1 s of silence
    vad_events=False,  # True: START_OF_SPEECH from the server
    early_words=False,  # True: "early_word" events (below)
    warmup_timeout=300.0,  # seconds to wait while the service starts up
)

stt.update_options(...) changes options at runtime; open streams reconnect with them. recognize() sends pre-recorded audio to POST /v1/listen.

Early words

stt = saidwire.STT(early_words=True)


@stt.on("early_word")
def on_early_word(word: dict) -> None:
    # {"word": "cancel", "punctuated_word": "cancel.", "start": 6.84, "end": 7.29, ...}
    # Sent as soon as the word is committed, usually before the final that contains it.
    # Useful for keyword triggers ("stop", "cancel"). The final transcript is authoritative.
    ...

Behavior

• The session's flush at end of turn sends Finalize; the answer is a final transcript (empty if everything was already final), so the session never waits on a timeout.
• While audio is not flowing, a KeepAlive message holds the stream open. Silence would be billed as audio; KeepAlive is free.
• Dropped connections are retried per APIConnectOptions. A bad or revoked key, or used-up credit, fails at once with a non-retryable APIStatusError.
• stream.request_id is the id of the current connection; include it when you contact support.

Protocol

Saidwire's streaming API uses the Deepgram-compatible /v1/listen protocol. API reference: saidwire.com/docs/api.

Development

pip install -e . pytest pytest-asyncio ruff mypy numpy
pytest                     # runs against a fake server on 127.0.0.1; no key, no network
ruff check . && ruff format --check . && mypy livekit

Support

Open an issue on the source repository, or email support@saidwire.com.