wavetermdev/waveterm · error

cannot continue with a different model, model:%q, cont-model

Error message

cannot continue with a different model, model:%q, cont-model:%q

What it means

RunOpenAIChatStep validates a provided continuation (cont *uctypes.WaveContinueResponse) at openai-backend.go:500: if uctypes.AreModelsCompatible(chat.APIType, chatOpts.Config.Model, cont.Model) is false, it returns "cannot continue with a different model". A continuation response must have been produced by the same (compatible) model as the request being made.

Source

Thrown at pkg/aiusechat/openai/openai-backend.go:500

	}
	if !uctypes.AreModelsCompatible(chat.APIType, chat.Model, chatOpts.Config.Model) {
		return nil, nil, nil, fmt.Errorf("model mismatch: chat has %s, chatOpts has %s", chat.Model, chatOpts.Config.Model)
	}
	if chat.APIVersion != chatOpts.Config.APIVersion {
		return nil, nil, nil, fmt.Errorf("API version mismatch: chat has %s, chatOpts has %s", chat.APIVersion, chatOpts.Config.APIVersion)
	}

	// Context with timeout if provided.
	if chatOpts.Config.TimeoutMs > 0 {
		var cancel context.CancelFunc
		ctx, cancel = context.WithTimeout(ctx, time.Duration(chatOpts.Config.TimeoutMs)*time.Millisecond)
		defer cancel()
	}

	// Validate continuation if provided
	if cont != nil {
		if !uctypes.AreModelsCompatible(chat.APIType, chatOpts.Config.Model, cont.Model) {
			return nil, nil, nil, fmt.Errorf("cannot continue with a different model, model:%q, cont-model:%q", chatOpts.Config.Model, cont.Model)
		}
	}

	// Convert GenAIMessages to input objects (OpenAIMessage or OpenAIFunctionCallInput)
	var inputs []any
	for _, genMsg := range chat.NativeMessages {
		// Cast to OpenAIChatMessage
		chatMsg, ok := genMsg.(*OpenAIChatMessage)
		if !ok {
			return nil, nil, nil, fmt.Errorf("expected OpenAIChatMessage, got %T", genMsg)
		}

		// Convert to appropriate input type based on what's populated
		if chatMsg.Message != nil {
			// Clean message to remove preview URLs
			cleanedMsg := chatMsg.Message.cleanAndCopy()
			inputs = append(inputs, *cleanedMsg)
		} else if chatMsg.FunctionCall != nil {

View on GitHub (pinned to a4447c1563)

Solutions

  1. Drop the cont argument (pass nil) and start a fresh step with the new model
  2. Align chatOpts.Config.Model with cont.Model (if that's the model you intend to continue with)
  3. Re-generate the continuation under the current model instead of reusing the old one
  4. Check AreModelsCompatible: if the models should be compatible, fix the model strings (aliases/version suffixes) rather than the logic

Example fix

// before
cont := loadSavedContinuation() // cont.Model = "gpt-4o"
chatOpts.Config.Model = "gpt-4.1"
RunOpenAIChatStep(ctx, sse, chatOpts, cont) // mismatch
// after
if cont != nil && cont.Model != chatOpts.Config.Model {
    cont = nil // discard stale continuation
}
RunOpenAIChatStep(ctx, sse, chatOpts, cont)
Defensive patterns

Strategy: validation

Validate before calling

if cont != nil {
    chat := chatstore.DefaultChatStore.Get(chatOpts.ChatId)
    if chat == nil || !uctypes.AreModelsCompatible(chat.APIType, chatOpts.Config.Model, cont.Model) {
        cont = nil // discard incompatible continuation
    }
}

Type guard

func continuationValid(chatOpts uctypes.WaveChatOpts, cont *uctypes.WaveContinueResponse) bool {
    if cont == nil {
        return true
    }
    chat := chatstore.DefaultChatStore.Get(chatOpts.ChatId)
    return chat != nil && uctypes.AreModelsCompatible(chat.APIType, chatOpts.Config.Model, cont.Model)
}

Try / catch

if _, _, _, err := RunOpenAIChatStep(ctx, sse, chatOpts, cont); err != nil {
    if strings.Contains(err.Error(), "cannot continue with a different model") {
        // retry once with cont = nil
        _, _, _, err = RunOpenAIChatStep(ctx, sse, chatOpts, nil)
        return err
    }
    return err
}

Prevention

When it happens

Trigger: Passing a cont (e.g. a follow-up/next-page response handle) whose cont.Model was produced under a different model than chatOpts.Config.Model — typically after the user switched models between a stopped turn and its continuation.

Common situations: Model changed in settings between the original response and clicking 'continue'; replaying an old saved continuation record against a chat now configured with a newer model; concurrent chats sharing a continuation object by mistake.

Related errors

AI-assisted analysis of wavetermdev/waveterm@a4447c1563 (2026-09-01). Data as JSON: /api/errors/cf1068e47a013400. Report an issue: GitHub.