Hi, I would like to ask you about this error and how to avoid it:

node_modules/@livekit/agents/src/stream/deferred_stream.ts:52
throw new Error('Stream source already set');

(note: it worked with 0.x version of node.js) Steps to reproduce: 1. use

multi_agent.ts

(from agents-js/examples/src/, main branch) 2. use this session: const session = new voice.AgentSession({llm: new openai.realtime.RealtimeModel(),userData: userdata}); 3. in session.start, add

inputOptions.closeOnDisconnect

as

false

4. remove whole

prewarm

step in

defineAgent

5. create node.js endpoint, ask for a new token:

const at = new AccessToken(process.env.LIVEKIT_API_KEY, process.env.LIVEKIT_API_SECRET, {
      identity: userId,
    });
    at.addGrant({ roomJoin: true, room: `room-of-${userId}` });
    const roomConfig = new RoomConfiguration({
      agents: [
        new RoomAgentDispatch({
          agentName: "my-agent",
        }),
      ],
      departureTimeout: 180,
      emptyTimeout: 5,
    });
    at.roomConfig = roomConfig;
const token = await at.toJwt();
    return res.json({ token });

6. on front end, run LiveKitRoom from "@livekit/components-react" like this:

<LiveKitRoom token={liveKitToken} serverUrl={process.env.LIVEKIT_URL} audio={true}>
          <RoomAudioRenderer />
        </LiveKitRoom>

7. fetch token, save it into react state, use it in LiveKitRoom 8. agent starts, you can talk, etc. 9. disconnect the user and let him to connect back (refresh the browser), that means we use same userId, same identity, same room name, but fresh token 10. server error 'Stream source already set'

Thanks for reaching out, Pavel. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

how to do it, do you have a code snippet ? I do not work with setSource or DeferredReadableStream in my code

I understand that encountering the error "Stream source already set" can be a bit frustrating, especially when you're not directly working with

setSource

or

DeferredReadableStream

. This usually happens because of some internal agent processes, like when the agent is reinitializing or when stream objects are being reused. Let's walk through some steps to help you resolve this: 1. First, try to identify which part of your setup might be causing the issue. For instance, methods such as

AgentActivity.attachAudioInput()

or

SpeechStream.updateInputStream()

internally call

setSource

. If these are reused without proper cleanup, it can lead to this error. 2. Make sure to clean up properly before reusing streams. Here are a few tips: ◦ Before attaching a new audio stream in

AgentActivity

, call

detachAudioInput()

. ◦ In

SpeechStream

, use

detachInputStream()

before you call

updateInputStream()

again. ◦ Consider using an

AbortController

in

SynthesizeStream

to help with cleanup. 1. Take a look at how your agent session restarts. Ensuring that streams and components are properly torn down before starting again can make a big difference. 2. It might also help to enable logging so you can trace exactly when streams are set or detached. This can give you clues about which internal component might be causing the problem. If you’d like, feel free to share more details about the specific agent or stream-related components you’re working with. I’d be happy to offer more tailored advice or code examples to help you out! Sources: agents/src/voice/room_io/_input.ts | livekit-agents/livekit/agents/voice/agent_session.py