Merge pull request #3009 from liulinhuai/fix/images-upstream-error-passthrough

fix(images): surface real upstream error instead of generic 502

backend/internal/service/openai_images.go

@@ -645,7 +645,7 @@ func (s *OpenAIGatewayService) forwardOpenAIImagesAPIKey(
 				RetryableOnSameAccount: account.IsPoolMode() && account.IsPoolModeRetryableStatus(resp.StatusCode),
 			}
 		}
-		return s.handleErrorResponse(upstreamCtx, resp, c, account, forwardBody)
+		return s.handleOpenAIImagesErrorResponse(upstreamCtx, resp, c, account, upstreamModel)
 	}
 	defer func() { _ = resp.Body.Close() }()
 

backend/internal/service/openai_images_responses.go

@@ -579,6 +579,176 @@ func openAIImagesUpstreamErrorFromGJSON(errorObj gjson.Result, upstreamRequestID
 	}
 }
 
+// openAIImagesErrorTypeForStatus returns an OpenAI-style error type when the
+// upstream body does not provide one of its own.
+func openAIImagesErrorTypeForStatus(status int) string {
+	switch {
+	case status == http.StatusBadRequest:
+		return "invalid_request_error"
+	case status == http.StatusUnauthorized:
+		return "authentication_error"
+	case status == http.StatusForbidden:
+		return "permission_error"
+	case status == http.StatusNotFound:
+		return "not_found_error"
+	case status == http.StatusTooManyRequests:
+		return "rate_limit_error"
+	case status >= 500:
+		return "api_error"
+	default:
+		return "upstream_error"
+	}
+}
+
+// openAIImagesUpstreamErrorFromHTTP builds an OpenAIImagesUpstreamError from a
+// non-2xx upstream HTTP response, preserving the real status code, type, code,
+// message and param so the client sees the actual upstream error instead of a
+// generic 502.
+func openAIImagesUpstreamErrorFromHTTP(statusCode int, header http.Header, body []byte) *OpenAIImagesUpstreamError {
+	errType := strings.TrimSpace(gjson.GetBytes(body, "error.type").String())
+	code := strings.TrimSpace(extractUpstreamErrorCode(body))
+	param := strings.TrimSpace(gjson.GetBytes(body, "error.param").String())
+	message := sanitizeUpstreamErrorMessage(strings.TrimSpace(extractUpstreamErrorMessage(body)))
+	if message == "" {
+		message = fmt.Sprintf("Upstream request failed (status %d)", statusCode)
+	}
+	if errType == "" {
+		errType = openAIImagesErrorTypeForStatus(statusCode)
+	}
+	requestID := ""
+	if header != nil {
+		requestID = strings.TrimSpace(header.Get("x-request-id"))
+	}
+	return &OpenAIImagesUpstreamError{
+		StatusCode:        statusCode,
+		ErrorType:         errType,
+		Code:              code,
+		Message:           message,
+		Param:             param,
+		UpstreamRequestID: requestID,
+	}
+}
+
+// handleOpenAIImagesErrorResponse is the non-failover error handler for the
+// images endpoints (/v1/images/generations and /v1/images/edits). Unlike the
+// generic handleErrorResponse — which collapses every non-failover upstream
+// error into a generic 502 "Upstream request failed" — it surfaces the real
+// upstream status code and error message/type/code/param to the client. This
+// mirrors how the Chat Completions and Messages compat paths use
+// handleCompatErrorResponse.
+//
+// It returns an *OpenAIImagesUpstreamError (already written to the client) so
+// the images handler treats it as a terminal user-facing error rather than
+// re-writing a fallback response.
+func (s *OpenAIGatewayService) handleOpenAIImagesErrorResponse(
+	ctx context.Context,
+	resp *http.Response,
+	c *gin.Context,
+	account *Account,
+	requestedModel ...string,
+) (*OpenAIForwardResult, error) {
+	body := s.readUpstreamErrorBody(resp)
+
+	upstreamMsg := sanitizeUpstreamErrorMessage(strings.TrimSpace(extractUpstreamErrorMessage(body)))
+	upstreamDetail := ""
+	if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
+		maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
+		if maxBytes <= 0 {
+			maxBytes = 2048
+		}
+		upstreamDetail = truncateString(string(body), maxBytes)
+	}
+	setOpsUpstreamError(c, resp.StatusCode, upstreamMsg, upstreamDetail)
+
+	if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
+		logger.LegacyPrintf("service.openai_gateway",
+			"OpenAI images upstream error %d (account=%d platform=%s type=%s): %s",
+			resp.StatusCode,
+			account.ID,
+			account.Platform,
+			account.Type,
+			truncateForLog(body, s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes),
+		)
+	}
+
+	// Honor admin-configured error passthrough rules first.
+	if status, errType, errMsg, matched := applyErrorPassthroughRule(
+		c,
+		account.Platform,
+		resp.StatusCode,
+		body,
+		http.StatusBadGateway,
+		"upstream_error",
+		"Upstream request failed",
+	); matched {
+		upErr := &OpenAIImagesUpstreamError{
+			StatusCode:        status,
+			ErrorType:         errType,
+			Message:           errMsg,
+			UpstreamRequestID: strings.TrimSpace(resp.Header.Get("x-request-id")),
+		}
+		writeOpenAIImagesUpstreamErrorResponse(c, upErr)
+		return nil, upErr
+	}
+
+	// If the account is not configured to handle this status code, fall back to
+	// a generic gateway error without exposing upstream internals (mirrors
+	// handleCompatErrorResponse).
+	if !account.ShouldHandleErrorCode(resp.StatusCode) {
+		appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
+			Platform:           account.Platform,
+			AccountID:          account.ID,
+			AccountName:        account.Name,
+			UpstreamStatusCode: resp.StatusCode,
+			UpstreamRequestID:  resp.Header.Get("x-request-id"),
+			Kind:               "http_error",
+			Message:            upstreamMsg,
+			Detail:             upstreamDetail,
+		})
+		upErr := &OpenAIImagesUpstreamError{
+			StatusCode:        http.StatusInternalServerError,
+			ErrorType:         "upstream_error",
+			Message:           "Upstream gateway error",
+			UpstreamRequestID: strings.TrimSpace(resp.Header.Get("x-request-id")),
+		}
+		writeOpenAIImagesUpstreamErrorResponse(c, upErr)
+		return nil, upErr
+	}
+
+	// Track rate limits / decide whether to disable the account (secondary failover).
+	var modelForCooldown string
+	if len(requestedModel) > 0 {
+		modelForCooldown = strings.TrimSpace(requestedModel[0])
+	}
+	shouldDisable := s.handleOpenAIAccountUpstreamError(ctx, account, resp.StatusCode, resp.Header, body, modelForCooldown)
+	kind := "http_error"
+	if shouldDisable {
+		kind = "failover"
+	}
+	appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
+		Platform:           account.Platform,
+		AccountID:          account.ID,
+		AccountName:        account.Name,
+		UpstreamStatusCode: resp.StatusCode,
+		UpstreamRequestID:  resp.Header.Get("x-request-id"),
+		Kind:               kind,
+		Message:            upstreamMsg,
+		Detail:             upstreamDetail,
+	})
+	if shouldDisable {
+		return nil, &UpstreamFailoverError{
+			StatusCode:             resp.StatusCode,
+			ResponseBody:           body,
+			RetryableOnSameAccount: account.IsPoolMode() && account.IsPoolModeRetryableStatus(resp.StatusCode),
+		}
+	}
+
+	// Surface the real upstream error to the client.
+	upErr := openAIImagesUpstreamErrorFromHTTP(resp.StatusCode, resp.Header, body)
+	writeOpenAIImagesUpstreamErrorResponse(c, upErr)
+	return nil, upErr
+}
+
 func buildOpenAIImagesAPIResponse(
 	results []openAIResponsesImageResult,
 	createdAt int64,
@@ -1195,7 +1365,7 @@ func (s *OpenAIGatewayService) forwardOpenAIImagesOAuth(
 				RetryableOnSameAccount: account.IsPoolMode() && account.IsPoolModeRetryableStatus(resp.StatusCode),
 			}
 		}
-		return s.handleErrorResponse(upstreamCtx, resp, c, account, responsesBody)
+		return s.handleOpenAIImagesErrorResponse(upstreamCtx, resp, c, account, requestModel)
 	}
 	defer func() { _ = resp.Body.Close() }()
 

backend/internal/service/openai_images_test.go

@@ -656,6 +656,64 @@ func TestOpenAIGatewayServiceForwardImages_OAuthPassesNAndReturnsAllImages(t *te
 	require.Equal(t, "draw a cat 3", gjson.Get(rec.Body.String(), "data.2.revised_prompt").String())
 }
 
+func TestOpenAIGatewayServiceForwardImages_OAuthUpstreamHTTPErrorSurfacesRealError(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	body := []byte(`{"model":"gpt-image-2","prompt":"draw a cat","response_format":"b64_json"}`)
+
+	// The non-failover upstream error path is shared by /generations and /edits;
+	// use /generations here so the request parses without an uploaded image.
+	req := httptest.NewRequest(http.MethodPost, "/v1/images/generations", bytes.NewReader(body))
+	req.Header.Set("Content-Type", "application/json")
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = req
+	c.Set("api_key", &APIKey{ID: 42})
+
+	svc := &OpenAIGatewayService{}
+	parsed, err := svc.ParseOpenAIImagesRequest(c, body)
+	require.NoError(t, err)
+
+	svc.httpUpstream = &httpUpstreamRecorder{
+		resp: &http.Response{
+			StatusCode: http.StatusBadRequest,
+			Header: http.Header{
+				"Content-Type": []string{"application/json"},
+				"X-Request-Id": []string{"req_img_badreq"},
+			},
+			Body: io.NopCloser(strings.NewReader(
+				`{"error":{"message":"Invalid value for 'size': expected one of 1024x1024, 1536x1024.","type":"invalid_request_error","param":"size","code":"unknown_parameter"}}`,
+			)),
+		},
+	}
+
+	account := &Account{
+		ID:       1,
+		Name:     "openai-oauth",
+		Platform: PlatformOpenAI,
+		Type:     AccountTypeOAuth,
+		Credentials: map[string]any{
+			"access_token": "token-123",
+		},
+	}
+
+	result, err := svc.ForwardImages(context.Background(), c, account, body, parsed, "")
+	require.Nil(t, result)
+
+	var upstreamErr *OpenAIImagesUpstreamError
+	require.ErrorAs(t, err, &upstreamErr)
+	require.Equal(t, http.StatusBadRequest, upstreamErr.StatusCode)
+	require.Equal(t, "invalid_request_error", upstreamErr.ErrorType)
+	require.Equal(t, "unknown_parameter", upstreamErr.Code)
+
+	// The client must receive the actual upstream status code and message instead
+	// of a generic 502 "Upstream request failed".
+	require.Equal(t, http.StatusBadRequest, rec.Code)
+	require.Equal(t, "invalid_request_error", gjson.Get(rec.Body.String(), "error.type").String())
+	require.Equal(t, "unknown_parameter", gjson.Get(rec.Body.String(), "error.code").String())
+	require.Equal(t, "size", gjson.Get(rec.Body.String(), "error.param").String())
+	require.Contains(t, gjson.Get(rec.Body.String(), "error.message").String(), "Invalid value for 'size'")
+}
+
 func TestOpenAIGatewayServiceForwardImages_OAuthNonStreamModerationBlockedReturnsClientError(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 	body := []byte(`{"model":"gpt-image-2","prompt":"draw blocked image","response_format":"b64_json"}`)