microsoft/autogen · error · NotSupportedException

VarObject {response.VarObject} is not supported

Error message

VarObject {response.VarObject} is not supported

What it means

NotSupportedException from the streaming path: the connector requires every ChatCompletionResponse chunk to have object == "chat.completion.chunk". A chunk with a different object value (commonly "chat.completion" when streaming was silently disabled, or an error object) is rejected.

Source

Thrown at dotnet/src/AutoGen.Mistral/Middleware/MistralChatMessageConnector.cs:175

        }
        else if (finishReason == Choice.FinishReasonEnum.ToolCalls)
        {
            var functionContents = choice.Message?.ToolCalls ?? throw new ArgumentNullException("choice.Message.ToolCalls");
            var toolCalls = functionContents.Select(f => new ToolCall(f.Function.Name, f.Function.Arguments) { ToolCallId = f.Id }).ToList();
            return new ToolCallMessage(toolCalls, from: from.Name);
        }
        else
        {
            throw new NotSupportedException($"FinishReason {finishReason} is not supported");
        }
    }

    private IMessage? ProcessChatCompletionResponse(IMessage<ChatCompletionResponse> message, IAgent agent)
    {
        var response = message.Content;
        if (response.VarObject != "chat.completion.chunk")
        {
            throw new NotSupportedException($"VarObject {response.VarObject} is not supported");
        }
        if (response.Choices is null)
        {
            throw new ArgumentNullException("response.Choices");
        }

        if (response.Choices?.Count != 1)
        {
            throw new NotSupportedException("response.Choices.Count != 1");
        }

        var choice = response.Choices[0];
        var delta = choice.Delta;

        // process text message if delta.content is not null
        if (delta?.Content is string content)
        {
            return new TextMessageUpdate(role: Role.Assistant, content, from: agent.Name);

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Verify the raw SSE lines (data: {...}) from the Mistral endpoint and check the "object" field of each chunk.
2. If a proxy buffers the stream, bypass it for Mistral traffic or disable response buffering for text/event-stream.
3. If streaming is genuinely unavailable, use the non-streaming GenerateReplyAsync path instead.
4. Align mocks/tests to emit object="chat.completion.chunk" payloads.

Example fix

// before
if (response.VarObject != "chat.completion.chunk")
{
    throw new NotSupportedException($"VarObject {response.VarObject} is not supported");
}

// after
if (response.VarObject == "chat.completion")
{
    // full completion object where a chunk was expected: fall back to non-streaming processing
    return ProcessChatCompletionResponse(message, agent);
}
if (response.VarObject != "chat.completion.chunk")
{
    throw new NotSupportedException($"VarObject {response.VarObject} is not supported");
}

// verify streaming actually works before relying on it
var probe = await client.StreamingChatCompletionsAsync(testRequest).FirstOrDefaultAsync();
if (probe?.VarObject != "chat.completion.chunk")
{
    // endpoint/proxy does not deliver chunks: use non-streaming calls instead
}

static bool IsChunk(ChatCompletionResponse r) => r?.VarObject == "chat.completion.chunk";

catch (NotSupportedException ex) when (ex.Message.Contains("VarObject"))
{
    // fall back to non-streaming GenerateReplyAsync for this agent
}

When it happens

Trigger: Calling GenerateStreamingReplyAsync on a Mistral agent but receiving non-chunk objects. Typical causes: an intermediary proxy that buffers SSE and returns one full "chat.completion" object; the request's stream flag being overridden; or a non-streaming response fed into the streaming processor in tests.

Common situations: Corporate proxies or gateways (Azure API Management, nginx buffering) collapsing SSE streams; mocking helpers returning full completion objects into streaming code paths; version drift where the object field name/serialization changes.

Related errors

• Failed to deserialize response
• response.Choices.Count != 1
• FinishReason {finishReason} is not supported
• Role {textMessage.Role} is not supported
• Invalid JSON response from server. Check if the MCP route is

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/3ea153e027bc7ea2. Report an issue: GitHub.