when agent disconnect the room and enter the room again its start with the new clear new session.history

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

what is the json structure of chat_ctx i should push to the session becuse when add the data to session the agent still start as there is not injected data for item_data in history_result["history"]["items"]: # Convert dict to ChatContext item format chat_item = type('ChatItem', (), item_data)() saved_ctx.items.append(chat_item) # Inject the history into the main agent await main_agent.update_chat_ctx(saved_ctx)

Thanks — I’m responding to your question about the JSON structure for

chat_ctx

and why your current reconstruction isn’t injecting history. Quick answer:

chat_ctx

must be a ChatContext containing properly typed messages (not plain dynamic objects). Each message should have a

role

and

content

(content is a list). Use the ChatContext builder functions (e.g.,

add_message

) or create proper ChatMessage objects instead of using

type(...)

to mimic objects. Minimal JSON shape (conceptual):

{
  "messages": [
    {
      "type": "message",
      "role": "user",
      "content": [
        "Hello, how are you?"
      ]
    },
    {
      "type": "message",
      "role": "assistant",
      "content": [
        "I am fine, thanks."
      ]
    }
  ]
}

Correct Python approach (recommended):

from livekit.agents.llm import ChatContext

# rebuild chat context from stored history_result
saved_ctx = ChatContext()
for item in history_result["history"]["items"]:
    if item.get("type") == "message":
        # content must be a list[str] or list[Content] objects
        saved_ctx.add_message(role=item["role"], content=item["content"])
    # handle function/tool calls here if present

await main_agent.update_chat_ctx(saved_ctx)

Key checks if it still doesn't work: • Ensure each message's

content

is a list (e.g.,

["text"]

). • Verify

item["type"]

is

"message"

(or handle other types explicitly). • After

update_chat_ctx

, inspect

main_agent.chat_ctx.messages

to confirm messages were added. • Make sure any

on_enter

or startup hooks don’t overwrite the injected context after you update it. If you want, tell me: • which SDK/language you’re using (Python or JS), and • an example of one saved item from

history_result["history"]["items"]

(redact private text) — and I’ll give a precise snippet that matches that shape. Sources: Pipeline nodes and hooks | LiveKit docs | External data and RAG | LiveKit docs