microsoft/semantic-kernel · error · InvalidOperationException

Failed to get a response from the chat completion service.

Error message

Failed to get a response from the chat completion service.

What it means

Thrown by the chatbot process step when IChatCompletionService.GetChatMessageContentAsync returns a null ChatMessageContent. The Semantic Kernel chat completion contract normally returns a non-null message, so a null result indicates the underlying service failed to produce any content (model error, empty deployment, transport failure swallowed).

Source

Thrown at dotnet/samples/GettingStartedWithProcesses/Step01/Step01_Processes.cs:170

            return ValueTask.CompletedTask;
        }

        /// <summary>
        /// Generates a response from the chat completion service.
        /// </summary>
        /// <param name="context">The context for the current step and process. <see cref="KernelProcessStepContext"/></param>
        /// <param name="userMessage">The user message from a previous step.</param>
        /// <param name="_kernel">A <see cref="Kernel"/> instance.</param>
        /// <returns></returns>
        [KernelFunction(ProcessFunctions.GetChatResponse)]
        public async Task GetChatResponseAsync(KernelProcessStepContext context, string userMessage, Kernel _kernel)
        {
            _state!.ChatMessages.Add(new(AuthorRole.User, userMessage));
            IChatCompletionService chatService = _kernel.Services.GetRequiredService<IChatCompletionService>();
            ChatMessageContent response = await chatService.GetChatMessageContentAsync(_state.ChatMessages).ConfigureAwait(false);
            if (response == null)
            {
                throw new InvalidOperationException("Failed to get a response from the chat completion service.");
            }

            System.Console.ForegroundColor = ConsoleColor.Yellow;
            System.Console.WriteLine($"ASSISTANT: {response.Content}");
            System.Console.ResetColor();

            // Update state with the response
            _state.ChatMessages.Add(response);

            // emit event: assistantResponse
            await context.EmitEventAsync(new KernelProcessEvent { Id = ChatBotEvents.AssistantResponseGenerated, Data = response });
        }
    }

    /// <summary>
    /// The state object for the <see cref="ChatBotResponseStep"/>.
    /// </summary>
    private sealed class ChatBotState

View on GitHub (pinned to c028a0c7dc)

Solutions

  1. Verify an IChatCompletionService is registered on the kernel and the deployment/model id is correct.
  2. Inspect _state.ChatMessages passed in to confirm it is non-empty and well-formed.
  3. Upgrade or align the Semantic Kernel connector packages so GetChatMessageContentAsync throws on failure rather than returning null.
  4. Wrap the call to surface the underlying HTTP error instead of a generic message.

Example fix

// before
ChatMessageContent response = await chatService.GetChatMessageContentAsync(_state.ChatMessages).ConfigureAwait(false);
if (response == null) { throw new InvalidOperationException("Failed to get a response from the chat completion service."); }
// after - also log the request shape and inner cause
ChatMessageContent? response;
try { response = await chatService.GetChatMessageContentAsync(_state.ChatMessages).ConfigureAwait(false); }
catch (Exception ex) { throw new InvalidOperationException("Chat completion call failed.", ex); }
if (response is null) throw new InvalidOperationException($"No content returned for {_state.ChatMessages.Count} messages.");
Defensive patterns

Strategy: try-catch

Validate before calling

var chatService = kernel.Services.GetService<IChatCompletionService>();
if (chatService is null) throw new InvalidOperationException("No IChatCompletionService registered.");
if (_state.ChatMessages.Count == 0) _state.ChatMessages.Add(new(AuthorRole.User, "hello"));

Type guard

bool HasChatCompletion(Kernel k) => k.Services.GetService<IChatCompletionService>() is not null;

Try / catch

ChatMessageContent response;
try { response = await chatService.GetChatMessageContentAsync(_state.ChatMessages); }
catch (HttpRequestException ex) { /* surface real HTTP error */ throw; }
if (response is null) { /* log + fallback or retry */ }

Prevention

When it happens

Trigger: Invoking the GetChatResponseAsync kernel function in Step01_Processes when the registered IChatCompletionService returns null; e.g. misconfigured model/deployment id, empty chat history, or a backend that returned an empty body.

Common situations: Wrong deployment/model id for Azure OpenAI; chat history with no messages; connector version mismatch where the service returns null instead of throwing; rate-limited or empty completion response.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/e8b8326e9c78cd87. Report an issue: GitHub.