How to reuse the same assistant for multiple instances instead of creating a separate one for each instance

API
1

Using the Python OpenAI SDK, I am developing a personal assistant. The issue I am facing is that every time I launch my Streamlit app (new sessions), it creates a new assistant. I don’t mind launching a new thread, but I believe recreating the same assistant for the same app for every new session is inefficient. Is there a way to list all available assistants and select one instead of creating a new one each time?

2 Likes
2

You need to use the listAssistants endpoint

3

If i am correct listassistant only returns the list of assistants that are already created.
Any chance you could provide a code snippet how we meant to use it.
Goal, avoided creating new assistants each time and just call the already existing one (e.g. by using assistant id)

Thanks

4

You could use the assistant ID from your existing assistant, to then initate the assistant with
assistant = client.beta.assistants.retrieve("{your_assistant_id}")

Then proceed as normally!

1 Like
5

Here are some python code i wrote that helps with the issue:


def list_assistants():

    assistant_object = client.beta.assistants.list()
    return assistant_object

def delete_assistant(assistant_id):
    """Delete an assistant by ID."""
    delete_url = f"{BASE_URL}/{assistant_id}"
    response = requests.delete(delete_url, headers=HEADERS)
    if response.status_code == 200:
        print(f"Deleted assistant with ID: {assistant_id}")
    else:
        print(f"Failed to delete assistant with ID: {assistant_id}. Status Code: {response.status_code}")

def delete_all_assistants():
    """Delete all assistants."""
    a_list = list_assistants()
    assitant_obj_list = a_list.data
    for i in range(len(assitant_obj_list)):
        delete_assistant(assitant_obj_list[i].id)

def select_assistant(assistant_id):
    # Use the 'beta.assistants' attribute, not 'Assistant'
    assistant = client.beta.assistants.retrieve(assistant_id)
    return assistant.id

def create_assistant(name, instructions, tools, model):
    assistant = client.beta.assistants.create(
        name=name,
        instructions=instructions,
        tools=tools,
        model=model
    )
    return assistant.id  # Return the assistant ID



def get_assistant_by_id(assistant_id):
    assistant = client.beta.assistants.retrieve(assistant_id)
    return assistant.id


def create_thread():
    
    thread = client.beta.threads.create()
    return thread


def select_assistant(assistant_id):
    return get_assistant_by_id(assistant_id)
3 Likes
6

In Langroid I store the assistant, thread ids in a redis cache, based on a hash from user + machine + org. This way I can retrieve these ids across sessions and re-use the same assistant and thread (or not) depending on user’s config setup. You can see code here for example:

2 Likes
7

Incredibly helpful, thank you!

8

Thanks a lot.
I did some tinkering on the code and it is working!
Got back my assistant!

1 Like