How can I reuse a single thread to preserve chat history in OpenAI's Assistant API?

Below is an explanation of why your chat history isn’t persisting and how to change it.

Issue

Your current code creates a new thread each time you call the helper function (create_thread_and_run). Since each thread represents a separate conversation session, using a new thread for every message means the conversation history isn’t maintained.

Solution

To have the assistant remember previous messages, create a single thread and then add messages to that same thread throughout your interaction. In other words, do not create a new thread for every new user message. Instead, initialize the thread once and pass the same thread object for subsequent calls.

Example

Below is an adjusted snippet where the thread is created once and reused for the conversation:

PYTHONCopied
from openai import OpenAI
import time

ASSISTANT_ID = "asst_xxxxxxxxxxxxxxxx"

client = OpenAI(
    api_key="sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
)

def submit_message(assistant_id, thread, user_message):
    client.beta.threads.messages.create(
        thread_id=thread.id,
        role="user",
        content=user_message
    )
    return client.beta.threads.runs.create(
        thread_id=thread.id,
        assistant_id=assistant_id,
    )

def get_response(thread):
    return client.beta.threads.messages.list(thread_id=thread.id, order="asc")

# Pretty printing helper
def pretty_print(messages):
    print("# Messages")
    for m in messages:
        # Assuming the content is accessed this way
        print(f"{m.role}: {m.content[0].text.value}")
    print()

# Waiting in a loop
def wait_on_run(run, thread):
    while run.status == "queued" or run.status == "in_progress":
        run = client.beta.threads.runs.retrieve(
            thread_id=thread.id,
            run_id=run.id,
        )
        time.sleep(0.5)
    return run

# Create one thread for the entire conversation
thread = client.beta.threads.create()

# Simulate multiple user messages in the same thread
user_inputs = ["Hello!", "How are you today?", "What's the weather like in San Francisco?"]

for user_input in user_inputs:
    run = submit_message(ASSISTANT_ID, thread, user_input)
    run = wait_on_run(run, thread)
    pretty_print(get_response(thread))

Notes

• Reusing the thread: Notice that we create the thread only once (thread = client.beta.threads.create()) and then reuse that thread in every call to submit_message(). This way all messages accumulate in a single conversation session.
• Context management: The API automatically manages the message history and will smartly truncate the context if it ever exceeds the model's context window.

Using this approach, your assistant will remember previous chat messages until you decide to start a new conversation (i.e., create a new thread).