Method QueryText

model/openrouter.go:131–239

(question string, writer io.Writer, history []*RawMessage, prompt string, knowledgeMessages []*RawMessage, toolSession *ToolSession, lang string)

}

func (p *OpenRouterModelProvider) QueryText(question string, writer io.Writer, history []*RawMessage, prompt string, knowledgeMessages []*RawMessage, toolSession *ToolSession, lang string) (*ModelResult, error) {
	client := p.getProxyClientFromToken()

	ctx := context.Background()
	flusher, ok := writer.(http.Flusher)
	if !ok {
		return nil, fmt.Errorf(i18n.Translate(lang, "model:writer does not implement http.Flusher"))
	}

	model := p.subType
	if model == "" {
		model = openrouter.Gpt35Turbo
	}

	tokenCount, err := GetTokenSize(model, question)
	if err != nil {
		return nil, err
	}

	contextLength := getContextLength(p.subType)

	if strings.HasPrefix(question, "$OpenAgentDryRun$") {
		modelResult, err := getDefaultModelResult(model, question, "")
		if err != nil {
			return nil, fmt.Errorf(i18n.Translate(lang, "model:cannot calculate tokens"))
		}
		if contextLength > modelResult.TotalTokenCount {
			return modelResult, nil
		} else {
			return nil, fmt.Errorf(i18n.Translate(lang, "model:exceed max tokens"))
		}
	}

	maxTokens := contextLength - tokenCount
	if maxTokens < 0 {
		return nil, fmt.Errorf(i18n.Translate(lang, "model:The token count: [%d] exceeds the model: [%s]'s maximum token count: [%d]"), tokenCount, model, contextLength)
	}

	temperature := p.temperature
	topP := p.topP

	respStream, err := client.CreateChatCompletionStream(
		ctx,
		&openrouter.ChatCompletionRequest{
			Model: p.subType,
			Messages: []openrouter.ChatCompletionMessage{
				{
					Role:    openrouter.ChatMessageRoleSystem,
					Content: "You are a helpful assistant.",
				},
				{
					Role:    openrouter.ChatMessageRoleUser,
					Content: question,
				},
			},
			Stream:      false,
			Temperature: temperature,
			TopP:        topP,