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!

[victordibia]

Hi @bih ,

We have the PR for loading/saving state merged into main (should be available in the next dev release).

Documentation page is here

Save state ..

# Define a team.
assistant_agent = AssistantAgent(
    name="assistant_agent",
    system_message="You are a helpful assistant",
    model_client=OpenAIChatCompletionClient(
        model="gpt-4o-2024-08-06",
    ),
)
agent_team = RoundRobinGroupChat([assistant_agent], termination_condition=MaxMessageTermination(max_messages=2))

# Run the team and stream messages to the console.
stream = agent_team.run_stream(task="Write a beautiful poem 3-line about lake tangayika")

# Use asyncio.run(...) when running in a script.
await Console(stream)

# Save the state of the agent team.
team_state = await agent_team.save_state()

load team state

print(team_state)

# Load team state.
await agent_team.load_state(team_state)
stream = agent_team.run_stream(task="What was the last line of the poem you wrote?")
await Console(stream)

Let us know if it works for you.

[zzzzxcshuijiao]

If load_state work in anther context, how can i get the agent_team instance? In the same context, i don't need get team instance by load_state method

[victordibia]

Does this work for you / answer your question?

• Run your team
• save_state ..
• persist to json file ...

In new context

• create your team
• load state (from file on disc)
• proceed ...

import json


## save state to disk 
with open("coding/team_state.json", "w") as f:
    json.dump(team_state, f)

## load state from disk
with open("coding/team_state.json", "r") as f:
    team_state = json.load(f)

## create team
new_agent_team = RoundRobinGroupChat([assistant_agent], termination_condition=MaxMessageTermination(max_messages=2)) 

## load state
await new_agent_team.load_state(team_state)
stream = new_agent_team.run_stream(task="What was the last line of the poem you wrote?") 
await Console(stream)