microsoft/autogen · error · NotSupportedException

response.Choices.Count != 1

Error message

response.Choices.Count != 1

What it means

After the null check, MistralChatMessageConnector.PostProcessMessage insists that Choices.Count equals 1 — the connector maps exactly one choice to one AutoGen message and cannot represent n>1 or n=0 completions, so any other count throws NotSupportedException.

Source

Thrown at dotnet/src/AutoGen.Mistral/Middleware/MistralChatMessageConnector.cs:148

                    ToolCallMessage toolCallMessage when (toolCallMessage.From is null || toolCallMessage.From == agent.Name) => ProcessToolCallMessage(toolCallMessage, agent),
                    ToolCallResultMessage toolCallResultMessage => ProcessToolCallResultMessage(toolCallResultMessage, agent),
                    AggregateMessage<ToolCallMessage, ToolCallResultMessage> aggregateMessage => ProcessFunctionCallMiddlewareMessage(aggregateMessage, agent), // message type support for functioncall middleware
                    _ => [m],
                };
            }
        });
    }

    private IMessage PostProcessMessage(ChatCompletionResponse response, IAgent from)
    {
        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 finishReason = choice.FinishReason ?? throw new ArgumentNullException("choice.FinishReason");

        if (finishReason == Choice.FinishReasonEnum.Stop || finishReason == Choice.FinishReasonEnum.Length)
        {
            return new TextMessage(Role.Assistant, choice.Message?.Content ?? throw new ArgumentNullException("choice.Message.Content"), from: from.Name);
        }
        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");

View on GitHub (pinned to 027ecf0a37)

Solutions

  1. Do not request multiple completions (n > 1) when using this connector — it only supports a single choice.
  2. Log the raw response; if Choices is empty, investigate the blocking/error reason in the payload (content filter, invalid prompt).
  3. Catch NotSupportedException at the orchestration layer and retry with a simplified prompt if the emptiness is transient.

Example fix

// before
var request = new ChatCompletionRequest { Model = "mistral-large-latest", Messages = msgs, N = 3 };

// after
var request = new ChatCompletionRequest { Model = "mistral-large-latest", Messages = msgs }; // n defaults to 1
Defensive patterns

Strategy: validation

Validate before calling

if (response.Choices is { Count: 1 })
{
    // safe to map
}
else
{
    // log raw response; likely blocked generation or n>1 config
}

Try / catch

try { var reply = await mistralAgent.SendAsync(msg); } catch (NotSupportedException ex) when (ex.Message.Contains("Choices.Count != 1")) { /* retry with simplified prompt or report */ }

Prevention

When it happens

Trigger: A completion response with zero choices (blocked/empty generation) or multiple choices (if n > 1 was requested or the API returned extra candidates) reaching PostProcessMessage.

Common situations: Requesting n > 1 completions from a Mistral model; safety-filtered or empty responses with an empty choices array; API error payloads that deserialize to a Choices collection of size 0.

Related errors

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/715b0c2a2a941beb. Report an issue: GitHub.