**feat(logging, middleware): add error-based logging support and error log management**

- Introduced `logOnErrorOnly` mode to enable logging only for error responses when request logging is disabled.
- Added endpoints to list and download error logs (`/request-error-logs`).
- Implemented error log file cleanup to retain only the newest 10 logs.
- Refactored `ResponseWriterWrapper` to support forced logging for error responses.
- Enhanced middleware to capture data for upstream error persistence.
- Improved log file naming and error log filename generation.

internal/api/handlers/management/logs.go

@@ -139,6 +139,126 @@ func (h *Handler) DeleteLogs(c *gin.Context) {
 	})
 }
 
+// GetRequestErrorLogs lists error request log files when RequestLog is disabled.
+// It returns an empty list when RequestLog is enabled.
+func (h *Handler) GetRequestErrorLogs(c *gin.Context) {
+	if h == nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "handler unavailable"})
+		return
+	}
+	if h.cfg == nil {
+		c.JSON(http.StatusServiceUnavailable, gin.H{"error": "configuration unavailable"})
+		return
+	}
+	if h.cfg.RequestLog {
+		c.JSON(http.StatusOK, gin.H{"files": []any{}})
+		return
+	}
+
+	dir := h.logDirectory()
+	if strings.TrimSpace(dir) == "" {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "log directory not configured"})
+		return
+	}
+
+	entries, err := os.ReadDir(dir)
+	if err != nil {
+		if os.IsNotExist(err) {
+			c.JSON(http.StatusOK, gin.H{"files": []any{}})
+			return
+		}
+		c.JSON(http.StatusInternalServerError, gin.H{"error": fmt.Sprintf("failed to list request error logs: %v", err)})
+		return
+	}
+
+	type errorLog struct {
+		Name     string `json:"name"`
+		Size     int64  `json:"size"`
+		Modified int64  `json:"modified"`
+	}
+
+	files := make([]errorLog, 0, len(entries))
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+		name := entry.Name()
+		if !strings.HasPrefix(name, "error-") || !strings.HasSuffix(name, ".log") {
+			continue
+		}
+		info, errInfo := entry.Info()
+		if errInfo != nil {
+			c.JSON(http.StatusInternalServerError, gin.H{"error": fmt.Sprintf("failed to read log info for %s: %v", name, errInfo)})
+			return
+		}
+		files = append(files, errorLog{
+			Name:     name,
+			Size:     info.Size(),
+			Modified: info.ModTime().Unix(),
+		})
+	}
+
+	sort.Slice(files, func(i, j int) bool { return files[i].Modified > files[j].Modified })
+
+	c.JSON(http.StatusOK, gin.H{"files": files})
+}
+
+// DownloadRequestErrorLog downloads a specific error request log file by name.
+func (h *Handler) DownloadRequestErrorLog(c *gin.Context) {
+	if h == nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "handler unavailable"})
+		return
+	}
+	if h.cfg == nil {
+		c.JSON(http.StatusServiceUnavailable, gin.H{"error": "configuration unavailable"})
+		return
+	}
+
+	dir := h.logDirectory()
+	if strings.TrimSpace(dir) == "" {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "log directory not configured"})
+		return
+	}
+
+	name := strings.TrimSpace(c.Param("name"))
+	if name == "" || strings.Contains(name, "/") || strings.Contains(name, "\\") {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid log file name"})
+		return
+	}
+	if !strings.HasPrefix(name, "error-") || !strings.HasSuffix(name, ".log") {
+		c.JSON(http.StatusNotFound, gin.H{"error": "log file not found"})
+		return
+	}
+
+	dirAbs, errAbs := filepath.Abs(dir)
+	if errAbs != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": fmt.Sprintf("failed to resolve log directory: %v", errAbs)})
+		return
+	}
+	fullPath := filepath.Clean(filepath.Join(dirAbs, name))
+	prefix := dirAbs + string(os.PathSeparator)
+	if !strings.HasPrefix(fullPath, prefix) {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid log file path"})
+		return
+	}
+
+	info, errStat := os.Stat(fullPath)
+	if errStat != nil {
+		if os.IsNotExist(errStat) {
+			c.JSON(http.StatusNotFound, gin.H{"error": "log file not found"})
+			return
+		}
+		c.JSON(http.StatusInternalServerError, gin.H{"error": fmt.Sprintf("failed to read log file: %v", errStat)})
+		return
+	}
+	if info.IsDir() {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid log file"})
+		return
+	}
+
+	c.FileAttachment(fullPath, name)
+}
+
 func (h *Handler) logDirectory() string {
 	if h == nil {
 		return ""
@@ -215,7 +335,9 @@ func (acc *logAccumulator) consumeFile(path string) error {
 		}
 		return err
 	}
-	defer file.Close()
+	defer func() {
+		_ = file.Close()
+	}()
 
 	scanner := bufio.NewScanner(file)
 	buf := make([]byte, 0, logScannerInitialBuffer)

internal/api/middleware/request_logging.go

@@ -15,8 +15,8 @@ import (
 
 // RequestLoggingMiddleware creates a Gin middleware that logs HTTP requests and responses.
 // It captures detailed information about the request and response, including headers and body,
-// and uses the provided RequestLogger to record this data. If logging is disabled in the
-// logger, the middleware has minimal overhead.
+// and uses the provided RequestLogger to record this data. When logging is disabled in the
+// logger, it still captures data so that upstream errors can be persisted.
 func RequestLoggingMiddleware(logger logging.RequestLogger) gin.HandlerFunc {
 	return func(c *gin.Context) {
 		if logger == nil {
@@ -30,12 +30,6 @@ func RequestLoggingMiddleware(logger logging.RequestLogger) gin.HandlerFunc {
 			return
 		}
 
-		// Early return if logging is disabled (zero overhead)
-		if !logger.IsEnabled() {
-			c.Next()
-			return
-		}
-
 		// Capture request information
 		requestInfo, err := captureRequestInfo(c)
 		if err != nil {
@@ -47,6 +41,9 @@ func RequestLoggingMiddleware(logger logging.RequestLogger) gin.HandlerFunc {
 
 		// Create response writer wrapper
 		wrapper := NewResponseWriterWrapper(c.Writer, logger, requestInfo)
+		if !logger.IsEnabled() {
+			wrapper.logOnErrorOnly = true
+		}
 		c.Writer = wrapper
 
 		// Process the request

internal/api/middleware/response_writer.go

@@ -5,6 +5,7 @@ package middleware
 
 import (
 	"bytes"
+	"net/http"
 	"strings"
 
 	"github.com/gin-gonic/gin"
@@ -24,15 +25,16 @@ type RequestInfo struct {
 // It is designed to handle both standard and streaming responses, ensuring that logging operations do not block the client response.
 type ResponseWriterWrapper struct {
 	gin.ResponseWriter
-	body         *bytes.Buffer              // body is a buffer to store the response body for non-streaming responses.
-	isStreaming  bool                       // isStreaming indicates whether the response is a streaming type (e.g., text/event-stream).
-	streamWriter logging.StreamingLogWriter // streamWriter is a writer for handling streaming log entries.
-	chunkChannel chan []byte                // chunkChannel is a channel for asynchronously passing response chunks to the logger.
-	streamDone   chan struct{}              // streamDone signals when the streaming goroutine completes.
-	logger       logging.RequestLogger      // logger is the instance of the request logger service.
-	requestInfo  *RequestInfo               // requestInfo holds the details of the original request.
-	statusCode   int                        // statusCode stores the HTTP status code of the response.
-	headers      map[string][]string        // headers stores the response headers.
+	body           *bytes.Buffer              // body is a buffer to store the response body for non-streaming responses.
+	isStreaming    bool                       // isStreaming indicates whether the response is a streaming type (e.g., text/event-stream).
+	streamWriter   logging.StreamingLogWriter // streamWriter is a writer for handling streaming log entries.
+	chunkChannel   chan []byte                // chunkChannel is a channel for asynchronously passing response chunks to the logger.
+	streamDone     chan struct{}              // streamDone signals when the streaming goroutine completes.
+	logger         logging.RequestLogger      // logger is the instance of the request logger service.
+	requestInfo    *RequestInfo               // requestInfo holds the details of the original request.
+	statusCode     int                        // statusCode stores the HTTP status code of the response.
+	headers        map[string][]string        // headers stores the response headers.
+	logOnErrorOnly bool                       // logOnErrorOnly enables logging only when an error response is detected.
 }
 
 // NewResponseWriterWrapper creates and initializes a new ResponseWriterWrapper.
@@ -192,12 +194,34 @@ func (w *ResponseWriterWrapper) processStreamingChunks(done chan struct{}) {
 // For non-streaming responses, it logs the complete request and response details,
 // including any API-specific request/response data stored in the Gin context.
 func (w *ResponseWriterWrapper) Finalize(c *gin.Context) error {
-	if !w.logger.IsEnabled() {
+	if w.logger == nil {
+		return nil
+	}
+
+	finalStatusCode := w.statusCode
+	if finalStatusCode == 0 {
+		if statusWriter, ok := w.ResponseWriter.(interface{ Status() int }); ok {
+			finalStatusCode = statusWriter.Status()
+		} else {
+			finalStatusCode = 200
+		}
+	}
+
+	var slicesAPIResponseError []*interfaces.ErrorMessage
+	apiResponseError, isExist := c.Get("API_RESPONSE_ERROR")
+	if isExist {
+		if apiErrors, ok := apiResponseError.([]*interfaces.ErrorMessage); ok {
+			slicesAPIResponseError = apiErrors
+		}
+	}
+
+	hasAPIError := len(slicesAPIResponseError) > 0 || finalStatusCode >= http.StatusBadRequest
+	forceLog := w.logOnErrorOnly && hasAPIError && !w.logger.IsEnabled()
+	if !w.logger.IsEnabled() && !forceLog {
 		return nil
 	}
 
 	if w.isStreaming {
-		// Close streaming channel and writer
 		if w.chunkChannel != nil {
 			close(w.chunkChannel)
 			w.chunkChannel = nil
@@ -209,80 +233,98 @@ func (w *ResponseWriterWrapper) Finalize(c *gin.Context) error {
 		}
 
 		if w.streamWriter != nil {
-			err := w.streamWriter.Close()
+			if err := w.streamWriter.Close(); err != nil {
+				w.streamWriter = nil
+				return err
+			}
 			w.streamWriter = nil
-			return err
 		}
-	} else {
-		// Capture final status code and headers if not already captured
-		finalStatusCode := w.statusCode
-		if finalStatusCode == 0 {
-			// Get status from underlying ResponseWriter if available
-			if statusWriter, ok := w.ResponseWriter.(interface{ Status() int }); ok {
-				finalStatusCode = statusWriter.Status()
-			} else {
-				finalStatusCode = 200 // Default
-			}
+		if forceLog {
+			return w.logRequest(finalStatusCode, w.cloneHeaders(), w.body.Bytes(), w.extractAPIRequest(c), w.extractAPIResponse(c), slicesAPIResponseError, forceLog)
 		}
+		return nil
+	}
 
-		// Ensure we have the latest headers before finalizing
-		w.ensureHeadersCaptured()
+	return w.logRequest(finalStatusCode, w.cloneHeaders(), w.body.Bytes(), w.extractAPIRequest(c), w.extractAPIResponse(c), slicesAPIResponseError, forceLog)
+}
 
-		// Use the captured headers as the final headers
-		finalHeaders := make(map[string][]string)
-		for key, values := range w.headers {
-			// Make a copy of the values slice to avoid reference issues
-			headerValues := make([]string, len(values))
-			copy(headerValues, values)
-			finalHeaders[key] = headerValues
-		}
+func (w *ResponseWriterWrapper) cloneHeaders() map[string][]string {
+	w.ensureHeadersCaptured()
 
-		var apiRequestBody []byte
-		apiRequest, isExist := c.Get("API_REQUEST")
-		if isExist {
-			var ok bool
-			apiRequestBody, ok = apiRequest.([]byte)
-			if !ok {
-				apiRequestBody = nil
-			}
-		}
+	finalHeaders := make(map[string][]string, len(w.headers))
+	for key, values := range w.headers {
+		headerValues := make([]string, len(values))
+		copy(headerValues, values)
+		finalHeaders[key] = headerValues
+	}
 
-		var apiResponseBody []byte
-		apiResponse, isExist := c.Get("API_RESPONSE")
-		if isExist {
-			var ok bool
-			apiResponseBody, ok = apiResponse.([]byte)
-			if !ok {
-				apiResponseBody = nil
-			}
-		}
+	return finalHeaders
+}
 
-		var slicesAPIResponseError []*interfaces.ErrorMessage
-		apiResponseError, isExist := c.Get("API_RESPONSE_ERROR")
-		if isExist {
-			var ok bool
-			slicesAPIResponseError, ok = apiResponseError.([]*interfaces.ErrorMessage)
-			if !ok {
-				slicesAPIResponseError = nil
-			}
-		}
+func (w *ResponseWriterWrapper) extractAPIRequest(c *gin.Context) []byte {
+	apiRequest, isExist := c.Get("API_REQUEST")
+	if !isExist {
+		return nil
+	}
+	data, ok := apiRequest.([]byte)
+	if !ok || len(data) == 0 {
+		return nil
+	}
+	return data
+}
 
-		// Log complete non-streaming response
-		return w.logger.LogRequest(
+func (w *ResponseWriterWrapper) extractAPIResponse(c *gin.Context) []byte {
+	apiResponse, isExist := c.Get("API_RESPONSE")
+	if !isExist {
+		return nil
+	}
+	data, ok := apiResponse.([]byte)
+	if !ok || len(data) == 0 {
+		return nil
+	}
+	return data
+}
+
+func (w *ResponseWriterWrapper) logRequest(statusCode int, headers map[string][]string, body []byte, apiRequestBody, apiResponseBody []byte, apiResponseErrors []*interfaces.ErrorMessage, forceLog bool) error {
+	if w.requestInfo == nil {
+		return nil
+	}
+
+	var requestBody []byte
+	if len(w.requestInfo.Body) > 0 {
+		requestBody = w.requestInfo.Body
+	}
+
+	if loggerWithOptions, ok := w.logger.(interface {
+		LogRequestWithOptions(string, string, map[string][]string, []byte, int, map[string][]string, []byte, []byte, []byte, []*interfaces.ErrorMessage, bool) error
+	}); ok {
+		return loggerWithOptions.LogRequestWithOptions(
 			w.requestInfo.URL,
 			w.requestInfo.Method,
 			w.requestInfo.Headers,
-			w.requestInfo.Body,
-			finalStatusCode,
-			finalHeaders,
-			w.body.Bytes(),
+			requestBody,
+			statusCode,
+			headers,
+			body,
 			apiRequestBody,
 			apiResponseBody,
-			slicesAPIResponseError,
+			apiResponseErrors,
+			forceLog,
 		)
 	}
 
-	return nil
+	return w.logger.LogRequest(
+		w.requestInfo.URL,
+		w.requestInfo.Method,
+		w.requestInfo.Headers,
+		requestBody,
+		statusCode,
+		headers,
+		body,
+		apiRequestBody,
+		apiResponseBody,
+		apiResponseErrors,
+	)
 }
 
 // Status returns the HTTP response status code captured by the wrapper.

internal/api/server.go

@@ -509,6 +509,8 @@ func (s *Server) registerManagementRoutes() {
 
 		mgmt.GET("/logs", s.mgmt.GetLogs)
 		mgmt.DELETE("/logs", s.mgmt.DeleteLogs)
+		mgmt.GET("/request-error-logs", s.mgmt.GetRequestErrorLogs)
+		mgmt.GET("/request-error-logs/:name", s.mgmt.DownloadRequestErrorLog)
 		mgmt.GET("/request-log", s.mgmt.GetRequestLog)
 		mgmt.PUT("/request-log", s.mgmt.PutRequestLog)
 		mgmt.PATCH("/request-log", s.mgmt.PutRequestLog)

internal/logging/request_logger.go

@@ -12,6 +12,7 @@ import (
 	"os"
 	"path/filepath"
 	"regexp"
+	"sort"
 	"strings"
 	"time"
 
@@ -156,17 +157,30 @@ func (l *FileRequestLogger) SetEnabled(enabled bool) {
 // Returns:
 //   - error: An error if logging fails, nil otherwise
 func (l *FileRequestLogger) LogRequest(url, method string, requestHeaders map[string][]string, body []byte, statusCode int, responseHeaders map[string][]string, response, apiRequest, apiResponse []byte, apiResponseErrors []*interfaces.ErrorMessage) error {
-	if !l.enabled {
+	return l.logRequest(url, method, requestHeaders, body, statusCode, responseHeaders, response, apiRequest, apiResponse, apiResponseErrors, false)
+}
+
+// LogRequestWithOptions logs a request with optional forced logging behavior.
+// The force flag allows writing error logs even when regular request logging is disabled.
+func (l *FileRequestLogger) LogRequestWithOptions(url, method string, requestHeaders map[string][]string, body []byte, statusCode int, responseHeaders map[string][]string, response, apiRequest, apiResponse []byte, apiResponseErrors []*interfaces.ErrorMessage, force bool) error {
+	return l.logRequest(url, method, requestHeaders, body, statusCode, responseHeaders, response, apiRequest, apiResponse, apiResponseErrors, force)
+}
+
+func (l *FileRequestLogger) logRequest(url, method string, requestHeaders map[string][]string, body []byte, statusCode int, responseHeaders map[string][]string, response, apiRequest, apiResponse []byte, apiResponseErrors []*interfaces.ErrorMessage, force bool) error {
+	if !l.enabled && !force {
 		return nil
 	}
 
 	// Ensure logs directory exists
-	if err := l.ensureLogsDir(); err != nil {
-		return fmt.Errorf("failed to create logs directory: %w", err)
+	if errEnsure := l.ensureLogsDir(); errEnsure != nil {
+		return fmt.Errorf("failed to create logs directory: %w", errEnsure)
 	}
 
 	// Generate filename
 	filename := l.generateFilename(url)
+	if force && !l.enabled {
+		filename = l.generateErrorFilename(url)
+	}
 	filePath := filepath.Join(l.logsDir, filename)
 
 	// Decompress response if needed
@@ -184,6 +198,12 @@ func (l *FileRequestLogger) LogRequest(url, method string, requestHeaders map[st
 		return fmt.Errorf("failed to write log file: %w", err)
 	}
 
+	if force && !l.enabled {
+		if errCleanup := l.cleanupOldErrorLogs(); errCleanup != nil {
+			log.WithError(errCleanup).Warn("failed to clean up old error logs")
+		}
+	}
+
 	return nil
 }
 
@@ -239,6 +259,11 @@ func (l *FileRequestLogger) LogStreamingRequest(url, method string, headers map[
 	return writer, nil
 }
 
+// generateErrorFilename creates a filename with an error prefix to differentiate forced error logs.
+func (l *FileRequestLogger) generateErrorFilename(url string) string {
+	return fmt.Sprintf("error-%s", l.generateFilename(url))
+}
+
 // ensureLogsDir creates the logs directory if it doesn't exist.
 //
 // Returns:
@@ -312,6 +337,52 @@ func (l *FileRequestLogger) sanitizeForFilename(path string) string {
 	return sanitized
 }
 
+// cleanupOldErrorLogs keeps only the newest 10 forced error log files.
+func (l *FileRequestLogger) cleanupOldErrorLogs() error {
+	entries, errRead := os.ReadDir(l.logsDir)
+	if errRead != nil {
+		return errRead
+	}
+
+	type logFile struct {
+		name    string
+		modTime time.Time
+	}
+
+	var files []logFile
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+		name := entry.Name()
+		if !strings.HasPrefix(name, "error-") || !strings.HasSuffix(name, ".log") {
+			continue
+		}
+		info, errInfo := entry.Info()
+		if errInfo != nil {
+			log.WithError(errInfo).Warn("failed to read error log info")
+			continue
+		}
+		files = append(files, logFile{name: name, modTime: info.ModTime()})
+	}
+
+	if len(files) <= 10 {
+		return nil
+	}
+
+	sort.Slice(files, func(i, j int) bool {
+		return files[i].modTime.After(files[j].modTime)
+	})
+
+	for _, file := range files[10:] {
+		if errRemove := os.Remove(filepath.Join(l.logsDir, file.name)); errRemove != nil {
+			log.WithError(errRemove).Warnf("failed to remove old error log: %s", file.name)
+		}
+	}
+
+	return nil
+}
+
 // formatLogContent creates the complete log content for non-streaming requests.
 //
 // Parameters:

internal/runtime/executor/aistudio_executor.go

@@ -319,6 +319,9 @@ func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts c
 	payload = util.StripThinkingConfigIfUnsupported(req.Model, payload)
 	payload = fixGeminiImageAspectRatio(req.Model, payload)
 	payload = applyPayloadConfig(e.cfg, req.Model, payload)
+	payload, _ = sjson.DeleteBytes(payload, "generationConfig.maxOutputTokens")
+	payload, _ = sjson.DeleteBytes(payload, "generationConfig.responseMimeType")
+	payload, _ = sjson.DeleteBytes(payload, "generationConfig.responseJsonSchema")
 	metadataAction := "generateContent"
 	if req.Metadata != nil {
 		if action, _ := req.Metadata["action"].(string); action == "countTokens" {

internal/runtime/executor/antigravity_executor.go

@@ -26,16 +26,18 @@ import (
 )
 
 const (
-	antigravityBaseURL          = "https://daily-cloudcode-pa.sandbox.googleapis.com"
-	antigravityStreamPath       = "/v1internal:streamGenerateContent"
-	antigravityGeneratePath     = "/v1internal:generateContent"
-	antigravityModelsPath       = "/v1internal:fetchAvailableModels"
-	antigravityClientID         = "1071006060591-tmhssin2h21lcre235vtolojh4g403ep.apps.googleusercontent.com"
-	antigravityClientSecret     = "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"
-	defaultAntigravityAgent     = "antigravity/1.11.3 windows/amd64"
-	antigravityAuthType         = "antigravity"
-	refreshSkew                 = 5 * time.Minute
-	streamScannerBuffer     int = 20_971_520
+	antigravityBaseURLDaily        = "https://daily-cloudcode-pa.sandbox.googleapis.com"
+	antigravityBaseURLAutopush     = "https://autopush-cloudcode-pa.sandbox.googleapis.com"
+	antigravityBaseURLProd         = "https://cloudcode-pa.googleapis.com"
+	antigravityStreamPath          = "/v1internal:streamGenerateContent"
+	antigravityGeneratePath        = "/v1internal:generateContent"
+	antigravityModelsPath          = "/v1internal:fetchAvailableModels"
+	antigravityClientID            = "1071006060591-tmhssin2h21lcre235vtolojh4g403ep.apps.googleusercontent.com"
+	antigravityClientSecret        = "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"
+	defaultAntigravityAgent        = "antigravity/1.11.5 windows/amd64"
+	antigravityAuthType            = "antigravity"
+	refreshSkew                    = 3000 * time.Second
+	streamScannerBuffer        int = 20_971_520
 )
 
 var randSource = rand.New(rand.NewSource(time.Now().UnixNano()))
@@ -73,43 +75,76 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au
 	to := sdktranslator.FromString("antigravity")
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 
-	httpReq, errReq := e.buildRequest(ctx, auth, token, req.Model, translated, false, opts.Alt)
-	if errReq != nil {
-		return resp, errReq
-	}
-
+	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
-	httpResp, errDo := httpClient.Do(httpReq)
-	if errDo != nil {
-		recordAPIResponseError(ctx, e.cfg, errDo)
-		return resp, errDo
-	}
-	defer func() {
+
+	var lastStatus int
+	var lastBody []byte
+	var lastErr error
+
+	for idx, baseURL := range baseURLs {
+		httpReq, errReq := e.buildRequest(ctx, auth, token, req.Model, translated, false, opts.Alt, baseURL)
+		if errReq != nil {
+			err = errReq
+			return resp, err
+		}
+
+		httpResp, errDo := httpClient.Do(httpReq)
+		if errDo != nil {
+			recordAPIResponseError(ctx, e.cfg, errDo)
+			lastStatus = 0
+			lastBody = nil
+			lastErr = errDo
+			if idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: request error on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+				continue
+			}
+			err = errDo
+			return resp, err
+		}
+
+		recordAPIResponseMetadata(ctx, e.cfg, httpResp.StatusCode, httpResp.Header.Clone())
+		bodyBytes, errRead := io.ReadAll(httpResp.Body)
 		if errClose := httpResp.Body.Close(); errClose != nil {
 			log.Errorf("antigravity executor: close response body error: %v", errClose)
 		}
-	}()
+		if errRead != nil {
+			recordAPIResponseError(ctx, e.cfg, errRead)
+			err = errRead
+			return resp, err
+		}
+		appendAPIResponseChunk(ctx, e.cfg, bodyBytes)
 
-	recordAPIResponseMetadata(ctx, e.cfg, httpResp.StatusCode, httpResp.Header.Clone())
-	bodyBytes, errRead := io.ReadAll(httpResp.Body)
-	if errRead != nil {
-		recordAPIResponseError(ctx, e.cfg, errRead)
-		return resp, errRead
-	}
-	appendAPIResponseChunk(ctx, e.cfg, bodyBytes)
+		if httpResp.StatusCode < http.StatusOK || httpResp.StatusCode >= http.StatusMultipleChoices {
+			log.Debugf("antigravity executor: upstream error status: %d, body: %s", httpResp.StatusCode, summarizeErrorBody(httpResp.Header.Get("Content-Type"), bodyBytes))
+			lastStatus = httpResp.StatusCode
+			lastBody = append([]byte(nil), bodyBytes...)
+			lastErr = nil
+			if httpResp.StatusCode == http.StatusTooManyRequests && idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: rate limited on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+				continue
+			}
+			err = statusErr{code: httpResp.StatusCode, msg: string(bodyBytes)}
+			return resp, err
+		}
 
-	if httpResp.StatusCode < http.StatusOK || httpResp.StatusCode >= http.StatusMultipleChoices {
-		log.Debugf("antigravity executor: upstream error status: %d, body: %s", httpResp.StatusCode, summarizeErrorBody(httpResp.Header.Get("Content-Type"), bodyBytes))
-		err = statusErr{code: httpResp.StatusCode, msg: string(bodyBytes)}
-		return resp, err
+		reporter.publish(ctx, parseAntigravityUsage(bodyBytes))
+		var param any
+		converted := sdktranslator.TranslateNonStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), translated, bodyBytes, &param)
+		resp = cliproxyexecutor.Response{Payload: []byte(converted)}
+		reporter.ensurePublished(ctx)
+		return resp, nil
 	}
 
-	reporter.publish(ctx, parseAntigravityUsage(bodyBytes))
-	var param any
-	converted := sdktranslator.TranslateNonStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), translated, bodyBytes, &param)
-	resp = cliproxyexecutor.Response{Payload: []byte(converted)}
-	reporter.ensurePublished(ctx)
-	return resp, nil
+	switch {
+	case lastStatus != 0:
+		err = statusErr{code: lastStatus, msg: string(lastBody)}
+	case lastErr != nil:
+		err = lastErr
+	default:
+		err = statusErr{code: http.StatusServiceUnavailable, msg: "antigravity executor: no base url available"}
+	}
+	return resp, err
 }
 
 // ExecuteStream handles streaming requests via the antigravity upstream.
@@ -131,75 +166,121 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 	to := sdktranslator.FromString("antigravity")
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
-	httpReq, errReq := e.buildRequest(ctx, auth, token, req.Model, translated, true, opts.Alt)
-	if errReq != nil {
-		return nil, errReq
-	}
-
+	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
-	httpResp, errDo := httpClient.Do(httpReq)
-	if errDo != nil {
-		recordAPIResponseError(ctx, e.cfg, errDo)
-		return nil, errDo
-	}
-	recordAPIResponseMetadata(ctx, e.cfg, httpResp.StatusCode, httpResp.Header.Clone())
-	if httpResp.StatusCode < http.StatusOK || httpResp.StatusCode >= http.StatusMultipleChoices {
-		bodyBytes, _ := io.ReadAll(httpResp.Body)
-		appendAPIResponseChunk(ctx, e.cfg, bodyBytes)
-		if errClose := httpResp.Body.Close(); errClose != nil {
-			log.Errorf("antigravity executor: close response body error: %v", errClose)
-		}
-		err = statusErr{code: httpResp.StatusCode, msg: string(bodyBytes)}
-		return nil, err
-	}
 
-	out := make(chan cliproxyexecutor.StreamChunk)
-	stream = out
-	go func() {
-		defer close(out)
-		defer func() {
+	var lastStatus int
+	var lastBody []byte
+	var lastErr error
+
+	for idx, baseURL := range baseURLs {
+		httpReq, errReq := e.buildRequest(ctx, auth, token, req.Model, translated, true, opts.Alt, baseURL)
+		if errReq != nil {
+			err = errReq
+			return nil, err
+		}
+
+		httpResp, errDo := httpClient.Do(httpReq)
+		if errDo != nil {
+			recordAPIResponseError(ctx, e.cfg, errDo)
+			lastStatus = 0
+			lastBody = nil
+			lastErr = errDo
+			if idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: request error on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+				continue
+			}
+			err = errDo
+			return nil, err
+		}
+		recordAPIResponseMetadata(ctx, e.cfg, httpResp.StatusCode, httpResp.Header.Clone())
+		if httpResp.StatusCode < http.StatusOK || httpResp.StatusCode >= http.StatusMultipleChoices {
+			bodyBytes, errRead := io.ReadAll(httpResp.Body)
 			if errClose := httpResp.Body.Close(); errClose != nil {
 				log.Errorf("antigravity executor: close response body error: %v", errClose)
 			}
-		}()
-		scanner := bufio.NewScanner(httpResp.Body)
-		scanner.Buffer(nil, streamScannerBuffer)
-		var param any
-		for scanner.Scan() {
-			line := scanner.Bytes()
-			appendAPIResponseChunk(ctx, e.cfg, line)
-
-			// Filter usage metadata for all models
-			// Only retain usage statistics in the terminal chunk
-			line = FilterSSEUsageMetadata(line)
-
-			payload := jsonPayload(line)
-			if payload == nil {
+			if errRead != nil {
+				recordAPIResponseError(ctx, e.cfg, errRead)
+				lastStatus = 0
+				lastBody = nil
+				lastErr = errRead
+				if idx+1 < len(baseURLs) {
+					log.Debugf("antigravity executor: read error on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+					continue
+				}
+				err = errRead
+				return nil, err
+			}
+			appendAPIResponseChunk(ctx, e.cfg, bodyBytes)
+			lastStatus = httpResp.StatusCode
+			lastBody = append([]byte(nil), bodyBytes...)
+			lastErr = nil
+			if httpResp.StatusCode == http.StatusTooManyRequests && idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: rate limited on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
 				continue
 			}
+			err = statusErr{code: httpResp.StatusCode, msg: string(bodyBytes)}
+			return nil, err
+		}
 
-			if detail, ok := parseAntigravityStreamUsage(payload); ok {
-				reporter.publish(ctx, detail)
-			}
+		out := make(chan cliproxyexecutor.StreamChunk)
+		stream = out
+		go func(resp *http.Response) {
+			defer close(out)
+			defer func() {
+				if errClose := resp.Body.Close(); errClose != nil {
+					log.Errorf("antigravity executor: close response body error: %v", errClose)
+				}
+			}()
+			scanner := bufio.NewScanner(resp.Body)
+			scanner.Buffer(nil, streamScannerBuffer)
+			var param any
+			for scanner.Scan() {
+				line := scanner.Bytes()
+				appendAPIResponseChunk(ctx, e.cfg, line)
 
-			chunks := sdktranslator.TranslateStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), translated, bytes.Clone(payload), &param)
-			for i := range chunks {
-				out <- cliproxyexecutor.StreamChunk{Payload: []byte(chunks[i])}
+				// Filter usage metadata for all models
+				// Only retain usage statistics in the terminal chunk
+				line = FilterSSEUsageMetadata(line)
+
+				payload := jsonPayload(line)
+				if payload == nil {
+					continue
+				}
+
+				if detail, ok := parseAntigravityStreamUsage(payload); ok {
+					reporter.publish(ctx, detail)
+				}
+
+				chunks := sdktranslator.TranslateStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), translated, bytes.Clone(payload), &param)
+				for i := range chunks {
+					out <- cliproxyexecutor.StreamChunk{Payload: []byte(chunks[i])}
+				}
 			}
-		}
-		tail := sdktranslator.TranslateStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), translated, []byte("[DONE]"), &param)
-		for i := range tail {
-			out <- cliproxyexecutor.StreamChunk{Payload: []byte(tail[i])}
-		}
-		if errScan := scanner.Err(); errScan != nil {
-			recordAPIResponseError(ctx, e.cfg, errScan)
-			reporter.publishFailure(ctx)
-			out <- cliproxyexecutor.StreamChunk{Err: errScan}
-		} else {
-			reporter.ensurePublished(ctx)
-		}
-	}()
-	return stream, nil
+			tail := sdktranslator.TranslateStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), translated, []byte("[DONE]"), &param)
+			for i := range tail {
+				out <- cliproxyexecutor.StreamChunk{Payload: []byte(tail[i])}
+			}
+			if errScan := scanner.Err(); errScan != nil {
+				recordAPIResponseError(ctx, e.cfg, errScan)
+				reporter.publishFailure(ctx)
+				out <- cliproxyexecutor.StreamChunk{Err: errScan}
+			} else {
+				reporter.ensurePublished(ctx)
+			}
+		}(httpResp)
+		return stream, nil
+	}
+
+	switch {
+	case lastStatus != 0:
+		err = statusErr{code: lastStatus, msg: string(lastBody)}
+	case lastErr != nil:
+		err = lastErr
+	default:
+		err = statusErr{code: http.StatusServiceUnavailable, msg: "antigravity executor: no base url available"}
+	}
+	return nil, err
 }
 
 // Refresh refreshes the OAuth token using the refresh token.
@@ -230,57 +311,72 @@ func FetchAntigravityModels(ctx context.Context, auth *cliproxyauth.Auth, cfg *c
 		auth = updatedAuth
 	}
 
-	modelsURL := buildBaseURL(auth) + antigravityModelsPath
-	httpReq, errReq := http.NewRequestWithContext(ctx, http.MethodPost, modelsURL, bytes.NewReader([]byte(`{}`)))
-	if errReq != nil {
-		return nil
-	}
-	httpReq.Header.Set("Content-Type", "application/json")
-	httpReq.Header.Set("Authorization", "Bearer "+token)
-	httpReq.Header.Set("User-Agent", resolveUserAgent(auth))
-	if host := resolveHost(auth); host != "" {
-		httpReq.Host = host
-	}
-
+	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, cfg, auth, 0)
-	httpResp, errDo := httpClient.Do(httpReq)
-	if errDo != nil {
-		return nil
-	}
-	defer func() {
+
+	for idx, baseURL := range baseURLs {
+		modelsURL := baseURL + antigravityModelsPath
+		httpReq, errReq := http.NewRequestWithContext(ctx, http.MethodPost, modelsURL, bytes.NewReader([]byte(`{}`)))
+		if errReq != nil {
+			return nil
+		}
+		httpReq.Header.Set("Content-Type", "application/json")
+		httpReq.Header.Set("Authorization", "Bearer "+token)
+		httpReq.Header.Set("User-Agent", resolveUserAgent(auth))
+		if host := resolveHost(baseURL); host != "" {
+			httpReq.Host = host
+		}
+
+		httpResp, errDo := httpClient.Do(httpReq)
+		if errDo != nil {
+			if idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: models request error on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+				continue
+			}
+			return nil
+		}
+
+		bodyBytes, errRead := io.ReadAll(httpResp.Body)
 		if errClose := httpResp.Body.Close(); errClose != nil {
 			log.Errorf("antigravity executor: close response body error: %v", errClose)
 		}
-	}()
-
-	bodyBytes, errRead := io.ReadAll(httpResp.Body)
-	if errRead != nil {
-		return nil
-	}
-	if httpResp.StatusCode < http.StatusOK || httpResp.StatusCode >= http.StatusMultipleChoices {
-		return nil
-	}
-
-	result := gjson.GetBytes(bodyBytes, "models")
-	if !result.Exists() {
-		return nil
-	}
-
-	now := time.Now().Unix()
-	models := make([]*registry.ModelInfo, 0, len(result.Map()))
-	for id := range result.Map() {
-		id = modelName2Alias(id)
-		if id != "" {
-			models = append(models, &registry.ModelInfo{
-				ID:      id,
-				Object:  "model",
-				Created: now,
-				OwnedBy: antigravityAuthType,
-				Type:    antigravityAuthType,
-			})
+		if errRead != nil {
+			if idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: models read error on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+				continue
+			}
+			return nil
 		}
+		if httpResp.StatusCode < http.StatusOK || httpResp.StatusCode >= http.StatusMultipleChoices {
+			if httpResp.StatusCode == http.StatusTooManyRequests && idx+1 < len(baseURLs) {
+				log.Debugf("antigravity executor: models request rate limited on base url %s, retrying with fallback base url: %s", baseURL, baseURLs[idx+1])
+				continue
+			}
+			return nil
+		}
+
+		result := gjson.GetBytes(bodyBytes, "models")
+		if !result.Exists() {
+			return nil
+		}
+
+		now := time.Now().Unix()
+		models := make([]*registry.ModelInfo, 0, len(result.Map()))
+		for id := range result.Map() {
+			id = modelName2Alias(id)
+			if id != "" {
+				models = append(models, &registry.ModelInfo{
+					ID:      id,
+					Object:  "model",
+					Created: now,
+					OwnedBy: antigravityAuthType,
+					Type:    antigravityAuthType,
+				})
+			}
+		}
+		return models
 	}
-	return models
+	return nil
 }
 
 func (e *AntigravityExecutor) ensureAccessToken(ctx context.Context, auth *cliproxyauth.Auth) (string, *cliproxyauth.Auth, error) {
@@ -366,12 +462,15 @@ func (e *AntigravityExecutor) refreshToken(ctx context.Context, auth *cliproxyau
 	return auth, nil
 }
 
-func (e *AntigravityExecutor) buildRequest(ctx context.Context, auth *cliproxyauth.Auth, token, modelName string, payload []byte, stream bool, alt string) (*http.Request, error) {
+func (e *AntigravityExecutor) buildRequest(ctx context.Context, auth *cliproxyauth.Auth, token, modelName string, payload []byte, stream bool, alt, baseURL string) (*http.Request, error) {
 	if token == "" {
 		return nil, statusErr{code: http.StatusUnauthorized, msg: "missing access token"}
 	}
 
-	base := buildBaseURL(auth)
+	base := strings.TrimSuffix(baseURL, "/")
+	if base == "" {
+		base = buildBaseURL(auth)
+	}
 	path := antigravityGeneratePath
 	if stream {
 		path = antigravityStreamPath
@@ -405,7 +504,7 @@ func (e *AntigravityExecutor) buildRequest(ctx context.Context, auth *cliproxyau
 	} else {
 		httpReq.Header.Set("Accept", "application/json")
 	}
-	if host := resolveHost(auth); host != "" {
+	if host := resolveHost(base); host != "" {
 		httpReq.Host = host
 	}
 
@@ -489,26 +588,13 @@ func int64Value(value any) (int64, bool) {
 }
 
 func buildBaseURL(auth *cliproxyauth.Auth) string {
-	if auth != nil {
-		if auth.Attributes != nil {
-			if v := strings.TrimSpace(auth.Attributes["base_url"]); v != "" {
-				return strings.TrimSuffix(v, "/")
-			}
-		}
-		if auth.Metadata != nil {
-			if v, ok := auth.Metadata["base_url"].(string); ok {
-				v = strings.TrimSpace(v)
-				if v != "" {
-					return strings.TrimSuffix(v, "/")
-				}
-			}
-		}
+	if baseURLs := antigravityBaseURLFallbackOrder(auth); len(baseURLs) > 0 {
+		return baseURLs[0]
 	}
-	return antigravityBaseURL
+	return antigravityBaseURLAutopush
 }
 
-func resolveHost(auth *cliproxyauth.Auth) string {
-	base := buildBaseURL(auth)
+func resolveHost(base string) string {
 	parsed, errParse := url.Parse(base)
 	if errParse != nil {
 		return ""
@@ -535,6 +621,37 @@ func resolveUserAgent(auth *cliproxyauth.Auth) string {
 	return defaultAntigravityAgent
 }
 
+func antigravityBaseURLFallbackOrder(auth *cliproxyauth.Auth) []string {
+	if base := resolveCustomAntigravityBaseURL(auth); base != "" {
+		return []string{base}
+	}
+	return []string{
+		antigravityBaseURLDaily,
+		antigravityBaseURLAutopush,
+		// antigravityBaseURLProd,
+	}
+}
+
+func resolveCustomAntigravityBaseURL(auth *cliproxyauth.Auth) string {
+	if auth == nil {
+		return ""
+	}
+	if auth.Attributes != nil {
+		if v := strings.TrimSpace(auth.Attributes["base_url"]); v != "" {
+			return strings.TrimSuffix(v, "/")
+		}
+	}
+	if auth.Metadata != nil {
+		if v, ok := auth.Metadata["base_url"].(string); ok {
+			v = strings.TrimSpace(v)
+			if v != "" {
+				return strings.TrimSuffix(v, "/")
+			}
+		}
+	}
+	return ""
+}
+
 func geminiToAntigravity(modelName string, payload []byte) []byte {
 	template, _ := sjson.Set(string(payload), "model", modelName)
 	template, _ = sjson.Set(template, "userAgent", "antigravity")
@@ -544,18 +661,13 @@ func geminiToAntigravity(modelName string, payload []byte) []byte {
 
 	template, _ = sjson.Delete(template, "request.safetySettings")
 	template, _ = sjson.Set(template, "request.toolConfig.functionCallingConfig.mode", "VALIDATED")
-
-	gjson.Get(template, "request.contents").ForEach(func(key, content gjson.Result) bool {
-		if content.Get("role").String() == "model" {
-			content.Get("parts").ForEach(func(partKey, part gjson.Result) bool {
-				if part.Get("functionCall").Exists() {
-					template, _ = sjson.Set(template, fmt.Sprintf("request.contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
-				}
-				return true
-			})
+	template, _ = sjson.Delete(template, "request.generationConfig.maxOutputTokens")
+	if !strings.HasPrefix(modelName, "gemini-3-") {
+		if thinkingLevel := gjson.Get(template, "request.generationConfig.thinkingConfig.thinkingLevel"); thinkingLevel.Exists() {
+			template, _ = sjson.Delete(template, "request.generationConfig.thinkingConfig.thinkingLevel")
+			template, _ = sjson.Set(template, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
 		}
-		return true
-	})
+	}
 
 	if strings.HasPrefix(modelName, "claude-sonnet-") {
 		gjson.Get(template, "request.tools").ForEach(func(key, tool gjson.Result) bool {
@@ -604,7 +716,7 @@ func modelName2Alias(modelName string) string {
 		return "gemini-claude-sonnet-4-5"
 	case "claude-sonnet-4-5-thinking":
 		return "gemini-claude-sonnet-4-5-thinking"
-	case "chat_20706", "chat_23310", "gemini-2.5-flash-thinking", "gemini-3-pro-low":
+	case "chat_20706", "chat_23310", "gemini-2.5-flash-thinking", "gemini-3-pro-low", "gemini-2.5-pro":
 		return ""
 	default:
 		return modelName

internal/translator/antigravity/gemini/antigravity_gemini_request.go

@@ -98,6 +98,20 @@ func ConvertGeminiRequestToAntigravity(_ string, inputRawJSON []byte, _ bool) []
 		}
 	}
 
+	gjson.GetBytes(rawJSON, "request.contents").ForEach(func(key, content gjson.Result) bool {
+		if content.Get("role").String() == "model" {
+			content.Get("parts").ForEach(func(partKey, part gjson.Result) bool {
+				if part.Get("functionCall").Exists() {
+					rawJSON, _ = sjson.SetBytes(rawJSON, fmt.Sprintf("request.contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				} else if part.Get("thoughtSignature").Exists() {
+					rawJSON, _ = sjson.SetBytes(rawJSON, fmt.Sprintf("request.contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				}
+				return true
+			})
+		}
+		return true
+	})
+
 	return common.AttachDefaultSafetySettings(rawJSON, "request.safetySettings")
 }
 

internal/translator/gemini-cli/gemini/gemini-cli_gemini_request.go

@@ -98,6 +98,20 @@ func ConvertGeminiRequestToGeminiCLI(_ string, inputRawJSON []byte, _ bool) []by
 		}
 	}
 
+	gjson.GetBytes(rawJSON, "request.contents").ForEach(func(key, content gjson.Result) bool {
+		if content.Get("role").String() == "model" {
+			content.Get("parts").ForEach(func(partKey, part gjson.Result) bool {
+				if part.Get("functionCall").Exists() {
+					rawJSON, _ = sjson.SetBytes(rawJSON, fmt.Sprintf("request.contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				} else if part.Get("thoughtSignature").Exists() {
+					rawJSON, _ = sjson.SetBytes(rawJSON, fmt.Sprintf("request.contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				}
+				return true
+			})
+		}
+		return true
+	})
+
 	return common.AttachDefaultSafetySettings(rawJSON, "request.safetySettings")
 }
 

internal/translator/gemini/gemini-cli/gemini_gemini-cli_request.go

@@ -46,5 +46,19 @@ func ConvertGeminiCLIRequestToGemini(_ string, inputRawJSON []byte, _ bool) []by
 		}
 	}
 
+	gjson.GetBytes(rawJSON, "contents").ForEach(func(key, content gjson.Result) bool {
+		if content.Get("role").String() == "model" {
+			content.Get("parts").ForEach(func(partKey, part gjson.Result) bool {
+				if part.Get("functionCall").Exists() {
+					rawJSON, _ = sjson.SetBytes(rawJSON, fmt.Sprintf("contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				} else if part.Get("thoughtSignature").Exists() {
+					rawJSON, _ = sjson.SetBytes(rawJSON, fmt.Sprintf("contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				}
+				return true
+			})
+		}
+		return true
+	})
+
 	return common.AttachDefaultSafetySettings(rawJSON, "safetySettings")
 }

internal/translator/gemini/gemini/gemini_gemini_request.go

@@ -72,6 +72,20 @@ func ConvertGeminiRequestToGemini(_ string, inputRawJSON []byte, _ bool) []byte
 		return true
 	})
 
+	gjson.GetBytes(out, "contents").ForEach(func(key, content gjson.Result) bool {
+		if content.Get("role").String() == "model" {
+			content.Get("parts").ForEach(func(partKey, part gjson.Result) bool {
+				if part.Get("functionCall").Exists() {
+					out, _ = sjson.SetBytes(out, fmt.Sprintf("contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				} else if part.Get("thoughtSignature").Exists() {
+					out, _ = sjson.SetBytes(out, fmt.Sprintf("contents.%d.parts.%d.thoughtSignature", key.Int(), partKey.Int()), "skip_thought_signature_validator")
+				}
+				return true
+			})
+		}
+		return true
+	})
+
 	out = common.AttachDefaultSafetySettings(out, "safetySettings")
 
 	return out

internal/translator/gemini/openai/responses/gemini_openai-responses_response.go

@@ -433,12 +433,18 @@ func ConvertGeminiResponseToOpenAIResponses(_ context.Context, modelName string,
 			// output tokens
 			if v := um.Get("candidatesTokenCount"); v.Exists() {
 				completed, _ = sjson.Set(completed, "response.usage.output_tokens", v.Int())
+			} else {
+				completed, _ = sjson.Set(completed, "response.usage.output_tokens", 0)
 			}
 			if v := um.Get("thoughtsTokenCount"); v.Exists() {
 				completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", v.Int())
+			} else {
+				completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", 0)
 			}
 			if v := um.Get("totalTokenCount"); v.Exists() {
 				completed, _ = sjson.Set(completed, "response.usage.total_tokens", v.Int())
+			} else {
+				completed, _ = sjson.Set(completed, "response.usage.total_tokens", 0)
 			}
 		}