StandIn

pip install standin-sdk

npm install @komaa/standin-sdk

​

Start with your framework

​

How it fits together

​

The seam

from standin import CallServer, CallSession


class EchoHandler:
    async def on_start(self, session: CallSession) -> None:
        self._call = session

    async def on_caller_audio(self, pcm: bytes) -> None:
        await self._call.send_audio(pcm)


server = CallServer(handler_factory=EchoHandler)
await server.start()

import { CallServer, type CallSession } from "@komaa/standin-sdk";

class EchoHandler {
  #call!: CallSession;
  async onStart(session: CallSession) {
    this.#call = session;
  }
  async onCallerAudio(pcm: Buffer) {
    await this.#call.sendAudio(pcm);
  }
}

const server = new CallServer({ handlerFactory: () => new EchoHandler() });
await server.start();

​

What your agent can do on the call

• Two dialogue modes, both in the SDK. One realtime speech-to-speech provider doing its own turn-taking, or speech to text, then your agent, then speech, with the turn-taking, the pacing and the barge-in already written. Dialogue modes.
• Chat is a lane of its own. A pasted screenshot, a dragged-in file and a voice note land as one turn your agent can answer, voice notes transcribed by the engine you supply, and a reply can carry a picture back. Chat.
• The expensive and the sensitive are off until you switch them on. Outbound calling refuses everyone until you list who the agent may ring, then caps it at six an hour. A local file is refused until you name a root it may come from. No history of the caller’s screen is kept unless the call is being recorded. And who may call the agent is a check you write, because the handshake proves the call came from StandIn and nothing more. Security.

​

Two lanes

​

Next