Error "Thread ... already has an active run" when using assistants with function calling and streaming

[rstropek]

Confirm this is not an issue with the OpenAI Python Library

• This is not an issue with the OpenAI Python Library

Confirm this is not an issue with the underlying OpenAI API

• This is not an issue with the OpenAI API

Confirm this is not an issue with Azure OpenAI

• This is not an issue with Azure OpenAI

Describe the bug

When using the C# OpenAI SDK with assistants, function calling, and streaming, an error is thrown when calling CreateRunStreamingAsync. I described the issue already in https://community.openai.com/t/error-400-already-has-an-active-run/930753/6?u=rainer1, but got no feedback. Other users are reporting this problem, too.

To Reproduce

I copied the corresponding example from this repository and ran it with GPT-4o, GPT-4o-2024-08-06, and GPT-4-turbo. All models show the same behavior.

Code snippets

Here is the exact code that I used (it is an exact copy of the example from the OpenAI repo, except it is no longer a unit test):

using System.ClientModel;
using System.ClientModel.Primitives;
using System.Text.Json;
using dotenv.net;
using OpenAI.Assistants;
#pragma warning disable OPENAI001

var env = DotEnv.Read(options: new DotEnvOptions(probeForEnv: true, probeLevelsToSearch: 7));

// This example parallels the content at the following location:
// https://platform.openai.com/docs/assistants/tools/function-calling/function-calling-beta
#region Step 1 - Define Functions

// First, define the functions that the assistant will use in its defined tools.

FunctionToolDefinition getTemperatureTool = new()
{
    FunctionName = "get_current_temperature",
    Description = "Gets the current temperature at a specific location.",
    Parameters = BinaryData.FromString("""
    {
        "type": "object",
        "properties": {
        "location": {
            "type": "string",
            "description": "The city and state, e.g., San Francisco, CA"
        },
        "unit": {
            "type": "string",
            "enum": ["Celsius", "Fahrenheit"],
            "description": "The temperature unit to use. Infer this from the user's location."
        }
        }
    }
    """),
};

FunctionToolDefinition getRainProbabilityTool = new()
{
    FunctionName = "get_current_rain_probability",
    Description = "Gets the current forecasted probability of rain at a specific location,"
        + " represented as a percent chance in the range of 0 to 100.",
    Parameters = BinaryData.FromString("""
    {
        "type": "object",
        "properties": {
        "location": {
            "type": "string",
            "description": "The city and state, e.g., San Francisco, CA"
        }
        },
        "required": ["location"]
    }
    """),
};

#endregion

// Assistants is a beta API and subject to change; acknowledge its experimental status by suppressing the matching warning.
AssistantClient client = new(env["OPENAI_KEY"]!);

#region Create a new assistant with function tools
// Create an assistant that can call the function tools.
AssistantCreationOptions assistantOptions = new()
{
    Name = "Example: Function Calling",
    Instructions =
        "Don't make assumptions about what values to plug into functions."
        + " Ask for clarification if a user request is ambiguous.",
    Tools = { getTemperatureTool, getRainProbabilityTool },
};

Assistant assistant = await client.CreateAssistantAsync(env["OPENAI_MODEL"]!, assistantOptions);
#endregion

#region Step 2 - Create a thread and add messages
AssistantThread thread = await client.CreateThreadAsync();
ThreadMessage message = await client.CreateMessageAsync(
    thread,
    MessageRole.User,
    [
        "What's the weather in San Francisco today and the likelihood it'll rain?"
    ]);
#endregion

#region Step 3 - Initiate a streaming run
AsyncCollectionResult<StreamingUpdate> asyncUpdates
    = client.CreateRunStreamingAsync(thread, assistant);

ThreadRun? currentRun = null;
do
{
    currentRun = null;
    List<ToolOutput> outputsToSubmit = [];
    await foreach (StreamingUpdate update in asyncUpdates)
    {
        if (update is RunUpdate runUpdate)
        {
            currentRun = runUpdate;
        }
        else if (update is RequiredActionUpdate requiredActionUpdate)
        {
            if (requiredActionUpdate.FunctionName == getTemperatureTool.FunctionName)
            {
                outputsToSubmit.Add(new ToolOutput(requiredActionUpdate.ToolCallId, "57"));
            }
            else if (requiredActionUpdate.FunctionName == getRainProbabilityTool.FunctionName)
            {
                outputsToSubmit.Add(new ToolOutput(requiredActionUpdate.ToolCallId, "25%"));
            }
        }
        else if (update is MessageContentUpdate contentUpdate)
        {
            Console.Write(contentUpdate.Text);
        }
    }
    if (outputsToSubmit.Count > 0)
    {
        asyncUpdates = client.SubmitToolOutputsToRunStreamingAsync(currentRun, outputsToSubmit);
    }
}
while (currentRun?.Status.IsTerminal == false);

#endregion

// Optionally, delete the resources for tidiness if no longer needed.
RequestOptions noThrowOptions = new() { ErrorOptions = ClientErrorBehaviors.NoThrow };
_ = await client.DeleteThreadAsync(thread.Id, noThrowOptions);
_ = await client.DeleteAssistantAsync(assistant.Id, noThrowOptions);

OS

Ubuntu

.NET version

8

Library version

2.0.0-beta.11

[winston0410]

Encountering the same issue, even using the example here
https://github.com/openai/openai-dotnet/blob/7a8bc8bebc831cf95707220e38a56b14297adf33/examples/Assistants/Example02b_FunctionCallingStreaming.cs

@rstropek Have you found a solution to this yet?

[rstropek]

Encountering the same issue, even using the example here https://github.com/openai/openai-dotnet/blob/7a8bc8bebc831cf95707220e38a56b14297adf33/examples/Assistants/Example02b_FunctionCallingStreaming.cs

@rstropek Have you found a solution to this yet?

No, not yet. However, at https://community.openai.com/t/error-400-already-has-an-active-run/930753/10?u=rainer1, some people posted workarounds and/or fixes that worked for them. Maybe you are lucky.

[winston0410]

At the moment my workaround is not to use StreamingAsync but just to use Async API, which worked but obviously not ideal for a chat application.

[joseharriaga]

Thank you for reaching out, @rstropek ! We identified a bug in our example and merged a fix. Could you try the latest version and see if it works for you?
🔗 https://github.com/openai/openai-dotnet/blob/main/examples/Assistants/Example02b_FunctionCallingStreaming.cs