Conversational Assistant API

API
1

Hi all. I have been working on building Assistants manually for awhile now. I am new to coding but GPT has helped greatly in showing me how to build. One thing I am stuck on at the moment is trying to run an assistant, maintain memory, and carry a conversation with the data/potential files as a user would.

Currently I am stuck at a spot where the data gets loaded and I am able to ask one question but it quits after that. Would anyone know how to keep this going? Id like to basically continue the conversation every time I run it. Eventually also share the tool with others via streamlit or something along those lines.

Any ideas are welcome! Thank you.

2

What do you mean by “it quits after that”?
Something to keep in mind is assistant API is a bit different than the chat completion API, in that you create a run with a thread and an assistant. Its the thread object that keeps the messages, not the assistant. So if you wanted a multi-turn conversation, you would simply grab the thread and create a new message on it:

message = client.beta.threads.messages.create(
    thread_id=thread.id, role="user", content="Here is some new info"
)

Then you simply run the thread with an assistant:

# Execute our run
run = client.beta.threads.runs.create(
    thread_id=thread.id,
    assistant_id=assistant.id,
)

And grab the messages again:

messages = client.beta.threads.messages.list(
    thread_id=thread.id, order="asc", after=message.id
)

Something that confused me at first was the use of .create() as the method call, made me think I was resetting everything, but you are creating a new message on the thread, not a new thread.

1 Like
3

Quits as it the program would just end after answering my query. Either way, thank you for sharing some insights - this is definitely helpful and I will look to implement the recurring messages within my thread!

1 Like