feat: support response api

main.go

@@ -30,6 +30,23 @@ type chatCompletionReq struct {
 	Stream          bool          `json:"stream"`
 }
 
+type ResponseReasoning struct {
+	Effort  string `json:"effort"`
+	Summary string `json:"summary,omitempty"` // auto, concise, detailed
+}
+type chatResponseReq struct {
+	Model       string            `json:"model"`
+	Input       []ChatMessage     `json:"input"`
+	Temperature *float64          `json:"temperature,omitempty"`
+	Reasoning   ResponseReasoning `json:"reasoning,omitempty"`
+	Stream      bool              `json:"stream"`
+}
+
+type responseStreamEvent struct {
+	Type  string `json:"type"`
+	Delta string `json:"delta"`
+}
+
 type streamEvent struct {
 	Choices []struct {
 		Delta struct {
@@ -43,7 +60,7 @@ type streamEvent struct {
 	} `json:"error,omitempty"`
 }
 
-func streamChatText(
+func streamChatCompletions(
 	ctx context.Context,
 	client *http.Client,
 	baseURL string,
@@ -176,11 +193,137 @@ func streamChatText(
 	return pr, nil
 }
 
+func streamChatResponses(
+	ctx context.Context,
+	client *http.Client,
+	baseURL string,
+	apiKey string,
+	model string,
+	reasoningEffort string,
+	temperature *float64,
+	msgs []ChatMessage,
+) (io.ReadCloser, error) {
+	if msgs == nil || len(msgs) == 0 {
+		return nil, errors.New("missing messages")
+	}
+	endpoint := strings.TrimRight(baseURL, "/") + "/responses"
+
+	body := chatResponseReq{
+		Model:       model,
+		Input:       msgs,
+		Temperature: temperature,
+		Reasoning:   ResponseReasoning{Effort: reasoningEffort},
+		Stream:      true,
+	}
+
+	payload, err := json.Marshal(body)
+	if err != nil {
+		return nil, err
+	}
+
+	pr, pw := io.Pipe()
+
+	go func() {
+		defer func(pw *io.PipeWriter) {
+			err := pw.Close()
+			if err != nil {
+				log.Println(err)
+			}
+		}(pw)
+
+		req, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, bytes.NewReader(payload))
+		if err != nil {
+			_ = pw.CloseWithError(err)
+			return
+		}
+
+		req.Header.Set("Content-Type", "application/json")
+		req.Header.Set("Accept", "text/event-stream")
+		req.Header.Set("Authorization", "Bearer "+apiKey)
+
+		resp, err := client.Do(req)
+		if err != nil {
+			_ = pw.CloseWithError(err)
+			return
+		}
+		defer func(Body io.ReadCloser) {
+			err := Body.Close()
+			if err != nil {
+				log.Println(err)
+			}
+		}(resp.Body)
+
+		if resp.StatusCode < 200 || resp.StatusCode > 299 {
+			b, _ := io.ReadAll(resp.Body)
+			_ = pw.CloseWithError(fmt.Errorf("HTTP %d: %s", resp.StatusCode, strings.TrimSpace(string(b))))
+			return
+		}
+
+		sc := bufio.NewScanner(resp.Body)
+		sc.Buffer(make([]byte, 0, 64*1024), 2*1024*1024)
+
+		var dataLines []string
+		flushEvent := func() bool {
+			if len(dataLines) == 0 {
+				_ = pw.Close()
+				return true
+			}
+			data := strings.Join(dataLines, "\n")
+			dataLines = dataLines[:0]
+
+			var evt responseStreamEvent
+			if err := json.Unmarshal([]byte(data), &evt); err != nil {
+				_ = pw.CloseWithError(fmt.Errorf("failed to unmarshal event: %w, data=%q", err, data))
+				return false
+			}
+			if evt.Type == "response.output_text.delta" {
+				if _, err := io.WriteString(pw, evt.Delta); err != nil {
+					_ = pw.CloseWithError(err)
+					return false
+				}
+			}
+			return true
+		}
+
+		for sc.Scan() {
+			line := sc.Text()
+
+			if line == "" {
+				if ok := flushEvent(); !ok {
+					return
+				}
+				continue
+			}
+
+			if strings.HasPrefix(line, "data:") {
+				v := strings.TrimSpace(strings.TrimPrefix(line, "data:"))
+				dataLines = append(dataLines, v)
+			}
+			if strings.HasPrefix(line, "event: response.completed") {
+				break
+			}
+		}
+
+		if len(dataLines) > 0 {
+			_ = flushEvent()
+			return
+		}
+
+		if err := sc.Err(); err != nil {
+			_ = pw.CloseWithError(err)
+			return
+		}
+	}()
+
+	return pr, nil
+}
+
 // Translate src to lang
 func Translate(
 	ctx context.Context,
 	client *http.Client,
 	baseURL string,
+	responseApi bool,
 	apiKey string,
 	model string,
 	reasoningEffort string,
@@ -201,12 +344,17 @@ func Translate(
 			Content: src,
 		},
 	}
-	return streamChatText(ctx, client, baseURL, apiKey, model, reasoningEffort, temperature, msgs)
+	if responseApi {
+		return streamChatResponses(ctx, client, baseURL, apiKey, model, reasoningEffort, temperature, msgs)
+	}
+
+	return streamChatCompletions(ctx, client, baseURL, apiKey, model, reasoningEffort, temperature, msgs)
 }
 
 var cli struct {
 	Config          string   `short:"c" default:"~/.config/translate.toml" help:"Path to config file"`
 	BaseURL         string   `short:"b" help:"LLM API base URL" default:"https://api.openai.com/v1"`
+	ResponseApi     bool     `default:"false" help:"Use /v1/responses or /v1/chat/completions"`
 	ApiKey          string   `short:"k" help:"LLM API Key"`
 	Model           string   `short:"m" help:"LLM model" default:"gpt-5-nano"`
 	ReasoningEffort string   `help:"LLM reasoning effort (note that some LLMs may not support certain settings)" default:"minimal"`
@@ -239,7 +387,7 @@ func main() {
 		}
 	}
 
-	stream, err := Translate(context.Background(), http.DefaultClient, cli.BaseURL, cli.ApiKey, cli.Model,
+	stream, err := Translate(context.Background(), http.DefaultClient, cli.BaseURL, cli.ResponseApi, cli.ApiKey, cli.Model,
 		cli.ReasoningEffort, cli.Temperature, src, lang)
 
 	if err != nil {