In New foundry agent , How to check whether conversation id is exist or not using python code without api

Hello Raghav Mittal,

Greetings! Thanks for raising this question in Q&A forum.

First, an important clarification on the architecture of the New Microsoft Foundry Agent. In the new Foundry experience, conversations are managed using a conversation ID (created via openai.conversations.create()), not the classic thread ID model used in the older Assistants API. The conversation ID is what you need to track and validate across turns.

There is no dedicated "does this conversation exist?" method in the SDK, so the supported approach is to attempt to retrieve the conversation's messages and catch the exception if it does not exist. This is the standard try/except pattern and it does use the SDK, which itself wraps the underlying API, but it requires no direct raw API call or HTTP request from your code.

Here is the complete working approach using the azure-ai-projects package (version 2.x):

Step 1: Install the required packages.

pip install azure-ai-projects azure-identity

Step 2: Use try/except to check if a conversation ID exists.

import os
from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential
from azure.core.exceptions import HttpResponseError

PROJECT_ENDPOINT = os.environ["FOUNDRY_PROJECT_ENDPOINT"]

project = AIProjectClient(
    endpoint=PROJECT_ENDPOINT,
    credential=DefaultAzureCredential(),
)

openai_client = project.get_openai_client()

def conversation_exists(conversation_id: str) -> bool:
    try:
        # Attempt to retrieve items (messages) from the conversation
        items = openai_client.conversations.items.list(
            conversation_id=conversation_id
        )
        # If this succeeds without error, the conversation exists
        _ = list(items)
        return True
    except HttpResponseError as e:
        if e.status_code == 404:
            return False
        # Re-raise for unexpected errors
        raise

# Example usage
conversation_id_to_check = "conv_abc123yourIDhere"

if conversation_exists(conversation_id_to_check):
    print(f"Conversation {conversation_id_to_check} exists.")
else:
    print(f"Conversation {conversation_id_to_check} does not exist.")

Step 3: Best practice - persist the conversation ID yourself.

The recommended pattern for multi-turn agent applications is to store the conversation ID when you first create it, and reuse it for subsequent turns. This way you avoid needing to check existence at all, because you only attempt to reuse IDs that your own code created.

# First turn: create and persist
conversation = openai_client.conversations.create()
saved_conversation_id = conversation.id  # Save this to your database or session store
print(f"Conversation created with ID: {saved_conversation_id}")

# Subsequent turn: reuse the persisted ID
response = openai_client.responses.create(
    conversation=saved_conversation_id,
    extra_body={"agent_reference": {"name": "your_agent_name", "type": "agent_reference"}},
    input="Your next message here",
)
print(response.output_text)

Important note on "without API": The Foundry Agent Service is entirely API-driven. The SDK is the supported way to interact with it and uses the underlying APIs internally, but it abstracts away the raw HTTP calls so you do not need to write REST requests manually. There is no local file, database, or SDK property you can query to check conversation existence without going through the SDK or API, because all conversation state is stored on the Foundry service backend.

If this answer helps you kindly accept the answer which will help others who have similar questions.

Best Regards,

Jerald Felix.