How can I retrieve and display an assistant's response using OpenAI's thread run API?

Below is an explanation of what the provided code snippet does:

Overview

The code demonstrates how to interact with the OpenAI API to:

• Retrieve an existing assistant.
• Create a new thread (conversation).
• Send a user message (requesting help to solve a math equation).
• Start a run (execution) with the retrieved assistant.
• Poll for the run's status until it's completed.
• List and filter the messages to display only those from the assistant.

Step-by-Step Explanation

1. Imports and API Key Setup

   The code imports required modules and sets up the OpenAI API key from an environment variable:

   PYTHONCopied
   from openai import OpenAI
   import os
   import time
   global client
   global messages
   os.environ["OPENAI_API_KEY"] = "{key}"
   client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

2. Retrieve the Assistant

   It retrieves an assistant by its unique identifier:

   PYTHONCopied
   assistant = client.beta.assistants.retrieve("{assistant id}")
   print("Assistant Located")

3. Creating a New Conversation (Thread)

   A new thread is created to represent the conversation:

   PYTHONCopied
   thread = client.beta.threads.create()
   print("Thread  Created")

4. Posting a User Message

   A message is added to the thread with the role "user". In this case, the user is asking for help to solve the equation 3x + 11 = 14:

   PYTHONCopied
   message = client.beta.threads.messages.create(
       thread_id=thread.id,
       role="user",
       content="I need to solve the equation `3x + 11 = 14`. Can you help me?"
   )
   print("Thread Ready")

5. Starting the Run with the Assistant

   A run is started, linking the thread with the assistant. This action triggers the assistant to process the conversation and generate a response:

   PYTHONCopied
   run = client.beta.threads.runs.create(
     thread_id=thread.id,
     assistant_id=assistant.id
   )
   print("Assistant Loaded")
   print("Run Started - Please Wait")

6. Polling for Run Completion

   The code enters a loop where it checks the status of the run every 10 seconds. Once the run status is "completed", it fetches all messages from the thread:

   PYTHONCopied
   while True:
     time.sleep(10)
   
     run_status = client.beta.threads.runs.retrieve(
       thread_id=thread.id,
       run_id=run.id
     )
   
     if run_status.status == "completed":
       print("Run is Completed")
       messages = client.beta.threads.messages.list(
         thread_id=thread.id
       )
       # Filter messages for Assistant role
       assistant_messages = [msg for msg in messages if msg.role == 'assistant']
       print(assistant_messages)
       break
     else:
       print("Run is in progress - Please Wait")
       continue

   This loop ensures that the program waits until the assistant has finished processing the request before printing the assistant's messages.

Summary

The code effectively demonstrates how to use the OpenAI API to create an interactive session where:

• An assistant (configured, for example, as a personal math tutor) is recalled.
• A conversation is initiated with a user request.
• The assistant processes the request by leveraging its tools (such as executing Python code to solve the equation).
• The run is monitored until the assistant finishes processing, then the assistant's responses are printed.

This example is a practical demonstration of how to build and manage conversational interactions with the OpenAI API using threads and runs.