What is the current way to export an thread and preload a chat with a team in v0.4

[bih]

I'm fairly new to the AutoGen project and we're exploring it for our use case, but as I was prototyping it in a notebook I was wondering what is the way for us to be able to export a chat and preload it in a team. For example:

team = SelectorGroupChat(
    participants=[customer_service_agent, manager_agent, finance_agent],
    allow_repeated_speaker=False,
    model_client=OpenAIChatCompletionClient(model="gpt-4o"),
)

I can do this to get the conversation history

conversation = await team.run(task="Hello I would like a refund!")
print(conversation.messages)

But I would love to be able to export it as JSON

import json

conversation = await team.run(task="Hello I would like a refund!")
print(json.dumps(conversation.messages))

And equally I would love to be able to import the conversation history JSON so anything ran against in team.run()

conversation_history = json.loads(conversation_history_json)
conversation = await team.run(conversation=conversation_history, task="Hi world!")

Also, is it possible to message the agents with nothing (ideally as a specific flag)? We would love to be able to have agents give a meaningful intro message, but this isn't critical - we can build workarounds for this.

Note that this is for the v0.4 release!

[victordibia]

Hi @bih ,

We are currently working a memory interface that should help with this for the high level AgentChat API.
These capabilities are being tracked in issue #4039 and #4100 .
Stay tuned.

[bih]

Sounds good, thanks for the quick response!