I m trying to receive past agent user chat history from my f LiveKit Community #ask-ai

I'm trying to receive past agent/user chat history...

mysterious-computer-23868

07/22/2025, 4:55 PM

I'm trying to receive past agent/user chat history from my front-end, and add this to the initial chat context so that the user can "resume" the conversation. With

AgentSession

, and

aws.realtime.RealtimeModel

as my llm, my code is quite simply and everything is done in entrypoint for now. However the agent doesn't seem to be getting the context from the past conversation, even though I can see it's being received in the metadata. Am I loading the chat context the right way? I'm wondering if I'm using the wrong method, or if it's just getting wiped somehow.

Copy code

async def entrypoint(ctx: agents.JobContext):

  <http://logger.info|logger.info>(f"Connecting to room {ctx.room.name}")
  await ctx.connect()

  participant = await ctx.wait_for_participant()
  <http://logger.info|logger.info>(f"Starting voice assistant for participant {participant.identity}")

  session = AgentSession(
    llm=aws.realtime.RealtimeModel(
      voice="tiffany",
      region="us-east-1",
      temperature=0.7,
      top_p=0.95,
    ),
  )

  await session.start(
    room=ctx.room,
    agent=Kyorla(),
    room_input_options=RoomInputOptions(
      noise_cancellation=noise_cancellation.BVC(),
    ),
    room_output_options=RoomOutputOptions(
      sync_transcription=True,
    ),
  )

  history = json.loads(participant.metadata).get("chat_history", [])
  <http://logger.info|logger.info>(f"HISTORY: {history}")
  history_ctx = session.history.copy()
  for msg in history:
    history_ctx.add_message(role=msg["role"], content=msg["content"])

tall-airline-16756

07/22/2025, 4:55 PM

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

tall-airline-16756

07/22/2025, 4:55 PM

To ensure the agent receives prior chat context and can "resume" conversations, use one of these supported methods: • Preferred: Build a

ChatContext

with your chat history and pass it to the

Agent

constructor before starting the session. This guarantees the LLM receives the full context from the start. • Alternative: If you need to update context after the session has started, use the

update_chat_ctx()

method on the agent to inject prior messages. Directly modifying

session.history

after

session.start()

is not a supported or reliable way to provide context to the LLM. The recommended approaches are:

Copy code

# Build ChatContext from your history
chat_ctx = ChatContext.empty()
for msg in history:
    chat_ctx.add_message(role=msg["role"], content=msg["content"])

# Pass chat_ctx to the Agent
agent = Agent(
    llm=aws.realtime.RealtimeModel(...),
    chat_ctx=chat_ctx
)

await session.start(agent=agent, room=ctx.room)

Or, after starting the session:

Copy code

chat_ctx = ChatContext()
for msg in history:
    chat_ctx.add_message(role=msg["role"], content=msg["content"])
await agent.update_chat_ctx(chat_ctx)

Switching to one of these methods will ensure the LLM receives the intended prior context and can resume conversations as expected. Sources: OpenAI Realtime API integration guide | LiveKit Docs

Open in Slack

Previous Next