fix(image): add response-header timeout and idempotent retry

Give the image client a configurable ResponseHeaderTimeout (default 4m,
capped at the overall timeout) so a stalled upstream fails fast, and
retry such timeouts when an idempotency key is set so a slow-but-safe
request can recover instead of terminating. Raise the Next.js API proxy
maxDuration to 1200s to accommodate long generations.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

server/internal/infrastructure/piagent/image_generation.go

@@ -25,8 +25,9 @@ import (
 )
 
 const (
-	defaultImageGenerationTimeout = 10 * time.Minute
-	maxImageGenerationTimeout     = 10 * time.Minute
+	defaultImageGenerationTimeout               = 10 * time.Minute
+	defaultImageGenerationResponseHeaderTimeout = 4 * time.Minute
+	maxImageGenerationTimeout                   = 10 * time.Minute
 )
 
 type ImageGenerator interface {
@@ -47,10 +48,11 @@ type ImageRequestOptions struct {
 }
 
 type ImageClientOptions struct {
-	BaseURL             string
-	APIKey              string
-	InputImageTransport string
-	TimeoutSeconds      int
+	BaseURL                      string
+	APIKey                       string
+	InputImageTransport          string
+	TimeoutSeconds               int
+	ResponseHeaderTimeoutSeconds int
 }
 
 type ImageClient struct {
@@ -71,6 +73,13 @@ func NewImageClient(opts ImageClientOptions) *ImageClient {
 	if timeout > maxImageGenerationTimeout {
 		timeout = maxImageGenerationTimeout
 	}
+	responseHeaderTimeout := time.Duration(opts.ResponseHeaderTimeoutSeconds) * time.Second
+	if responseHeaderTimeout <= 0 {
+		responseHeaderTimeout = defaultImageGenerationResponseHeaderTimeout
+	}
+	if responseHeaderTimeout > timeout {
+		responseHeaderTimeout = timeout
+	}
 
 	return &ImageClient{
 		baseURL:             strings.TrimRight(opts.BaseURL, "/"),
@@ -83,6 +92,7 @@ func NewImageClient(opts ImageClientOptions) *ImageClient {
 				DialContext:           (&net.Dialer{Timeout: 30 * time.Second, KeepAlive: 30 * time.Second}).DialContext,
 				DisableKeepAlives:     true,
 				ForceAttemptHTTP2:     false,
+				ResponseHeaderTimeout: responseHeaderTimeout,
 				TLSHandshakeTimeout:   10 * time.Second,
 				ExpectContinueTimeout: time.Second,
 			},
@@ -147,9 +157,13 @@ func (c *ImageClient) Generate(ctx context.Context, prompt string, opts ImageReq
 			)
 			return image, nil
 		}
+		if ctx.Err() != nil {
+			return GeneratedImage{}, ctx.Err()
+		}
 		lastErr = err
-		retryable := isRetryableImageGenerationError(err)
-		terminal := isImageGenerationTerminalError(err)
+		timeoutRetryable := idempotencyKey != "" && isImageGenerationTimeoutError(err)
+		retryable := isRetryableImageGenerationError(err) || timeoutRetryable
+		terminal := isImageGenerationTerminalError(err) && !timeoutRetryable
 		logx.Errorf(
 			"gpt image generation failed model=%s size=%s count=%d prompt_runes=%d attempt=%d elapsed=%s timeout=%s retryable=%t terminal=%t idempotency_key=%s err=%v",
 			model,
@@ -537,6 +551,13 @@ func isImageGenerationTerminalError(err error) bool {
 	if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
 		return true
 	}
+	return isImageGenerationTimeoutError(err)
+}
+
+func isImageGenerationTimeoutError(err error) bool {
+	if err == nil {
+		return false
+	}
 	var netErr net.Error
 	if errors.As(err, &netErr) && netErr.Timeout() {
 		return true