**refactor(translator): remove service_tier from Codex OpenAI request payload**

feat(iflow): add cookie-based authentication endpoint

cmd/server/main.go

@@ -17,6 +17,7 @@ import (
 
 	"github.com/joho/godotenv"
 	configaccess "github.com/router-for-me/CLIProxyAPI/v6/internal/access/config_access"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/buildinfo"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/cmd"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/logging"
@@ -41,13 +42,16 @@ var (
 // init initializes the shared logger setup.
 func init() {
 	logging.SetupBaseLogger()
+	buildinfo.Version = Version
+	buildinfo.Commit = Commit
+	buildinfo.BuildDate = BuildDate
 }
 
 // main is the entry point of the application.
 // It parses command-line flags, loads configuration, and starts the appropriate
 // service based on the provided flags (login, codex-login, or server mode).
 func main() {
-	fmt.Printf("CLIProxyAPI Version: %s, Commit: %s, BuiltAt: %s\n", Version, Commit, BuildDate)
+	fmt.Printf("CLIProxyAPI Version: %s, Commit: %s, BuiltAt: %s\n", buildinfo.Version, buildinfo.Commit, buildinfo.BuildDate)
 
 	// Command-line flags to control the application's behavior.
 	var login bool
@@ -55,6 +59,7 @@ func main() {
 	var claudeLogin bool
 	var qwenLogin bool
 	var iflowLogin bool
+	var iflowCookie bool
 	var noBrowser bool
 	var projectID string
 	var vertexImport string
@@ -67,6 +72,7 @@ func main() {
 	flag.BoolVar(&claudeLogin, "claude-login", false, "Login to Claude using OAuth")
 	flag.BoolVar(&qwenLogin, "qwen-login", false, "Login to Qwen using OAuth")
 	flag.BoolVar(&iflowLogin, "iflow-login", false, "Login to iFlow using OAuth")
+	flag.BoolVar(&iflowCookie, "iflow-cookie", false, "Login to iFlow using Cookie")
 	flag.BoolVar(&noBrowser, "no-browser", false, "Don't open browser automatically for OAuth")
 	flag.StringVar(&projectID, "project_id", "", "Project ID (Gemini only, not required)")
 	flag.StringVar(&configPath, "config", DefaultConfigPath, "Configure File Path")
@@ -386,7 +392,7 @@ func main() {
 		log.Fatalf("failed to configure log output: %v", err)
 	}
 
-	log.Infof("CLIProxyAPI Version: %s, Commit: %s, BuiltAt: %s", Version, Commit, BuildDate)
+	log.Infof("CLIProxyAPI Version: %s, Commit: %s, BuiltAt: %s", buildinfo.Version, buildinfo.Commit, buildinfo.BuildDate)
 
 	// Set the log level based on the configuration.
 	util.SetLogLevel(cfg)
@@ -435,6 +441,8 @@ func main() {
 		cmd.DoQwenLogin(cfg, options)
 	} else if iflowLogin {
 		cmd.DoIFlowLogin(cfg, options)
+	} else if iflowCookie {
+		cmd.DoIFlowCookieAuth(cfg, options)
 	} else {
 		// In cloud deploy mode without config file, just wait for shutdown signals
 		if isCloudDeploy && !configFileExists {

config.example.yaml

@@ -98,3 +98,17 @@ ws-auth: false
 #    models: # The models supported by the provider.
 #      - name: "moonshotai/kimi-k2:free" # The actual model name.
 #        alias: "kimi-k2" # The alias used in the API.
+
+#payload: # Optional payload configuration
+#  default: # Default rules only set parameters when they are missing in the payload.
+#    - models:
+#        - name: "gemini-2.5-pro" # Supports wildcards (e.g., "gemini-*")
+#          protocol: "gemini" # restricts the rule to a specific protocol, options: openai, gemini, claude, codex
+#      params: # JSON path (gjson/sjson syntax) -> value
+#        "generationConfig.thinkingConfig.thinkingBudget": 32768
+#  override: # Override rules always set parameters, overwriting any existing values.
+#    - models:
+#        - name: "gpt-*" # Supports wildcards (e.g., "gpt-*")
+#          protocol: "codex" # restricts the rule to a specific protocol, options: openai, gemini, claude, codex
+#      params: # JSON path (gjson/sjson syntax) -> value
+#        "reasoning.effort": "high"

internal/api/handlers/management/auth_files.go

@@ -292,6 +292,7 @@ func (h *Handler) buildAuthFileEntry(auth *coreauth.Auth) gin.H {
 	if auth == nil {
 		return nil
 	}
+	auth.EnsureIndex()
 	runtimeOnly := isRuntimeOnlyAuth(auth)
 	if runtimeOnly && (auth.Disabled || auth.Status == coreauth.StatusDisabled) {
 		return nil
@@ -306,6 +307,7 @@ func (h *Handler) buildAuthFileEntry(auth *coreauth.Auth) gin.H {
 	}
 	entry := gin.H{
 		"id":             auth.ID,
+		"auth_index":     auth.Index,
 		"name":           name,
 		"type":           strings.TrimSpace(auth.Provider),
 		"provider":       strings.TrimSpace(auth.Provider),
@@ -346,6 +348,10 @@ func (h *Handler) buildAuthFileEntry(auth *coreauth.Auth) gin.H {
 			entry["size"] = info.Size()
 			entry["modtime"] = info.ModTime()
 		} else if os.IsNotExist(err) {
+			// Hide credentials removed from disk but still lingering in memory.
+			if !runtimeOnly && (auth.Disabled || auth.Status == coreauth.StatusDisabled || strings.EqualFold(strings.TrimSpace(auth.StatusMessage), "removed via management api")) {
+				return nil
+			}
 			entry["source"] = "memory"
 		} else {
 			log.WithError(err).Warnf("failed to stat auth file %s", path)
@@ -508,6 +514,10 @@ func (h *Handler) DeleteAuthFile(c *gin.Context) {
 				}
 			}
 			if err = os.Remove(full); err == nil {
+				if errDel := h.deleteTokenRecord(ctx, full); errDel != nil {
+					c.JSON(500, gin.H{"error": errDel.Error()})
+					return
+				}
 				deleted++
 				h.disableAuth(ctx, full)
 			}
@@ -534,10 +544,32 @@ func (h *Handler) DeleteAuthFile(c *gin.Context) {
 		}
 		return
 	}
+	if err := h.deleteTokenRecord(ctx, full); err != nil {
+		c.JSON(500, gin.H{"error": err.Error()})
+		return
+	}
 	h.disableAuth(ctx, full)
 	c.JSON(200, gin.H{"status": "ok"})
 }
 
+func (h *Handler) authIDForPath(path string) string {
+	path = strings.TrimSpace(path)
+	if path == "" {
+		return ""
+	}
+	if h == nil || h.cfg == nil {
+		return path
+	}
+	authDir := strings.TrimSpace(h.cfg.AuthDir)
+	if authDir == "" {
+		return path
+	}
+	if rel, err := filepath.Rel(authDir, path); err == nil && rel != "" {
+		return rel
+	}
+	return path
+}
+
 func (h *Handler) registerAuthFromFile(ctx context.Context, path string, data []byte) error {
 	if h.authManager == nil {
 		return nil
@@ -566,13 +598,18 @@ func (h *Handler) registerAuthFromFile(ctx context.Context, path string, data []
 	}
 	lastRefresh, hasLastRefresh := extractLastRefreshTimestamp(metadata)
 
+	authID := h.authIDForPath(path)
+	if authID == "" {
+		authID = path
+	}
 	attr := map[string]string{
 		"path":   path,
 		"source": path,
 	}
 	auth := &coreauth.Auth{
-		ID:         path,
+		ID:         authID,
 		Provider:   provider,
+		FileName:   filepath.Base(path),
 		Label:      label,
 		Status:     coreauth.StatusActive,
 		Attributes: attr,
@@ -583,7 +620,7 @@ func (h *Handler) registerAuthFromFile(ctx context.Context, path string, data []
 	if hasLastRefresh {
 		auth.LastRefreshedAt = lastRefresh
 	}
-	if existing, ok := h.authManager.GetByID(path); ok {
+	if existing, ok := h.authManager.GetByID(authID); ok {
 		auth.CreatedAt = existing.CreatedAt
 		if !hasLastRefresh {
 			auth.LastRefreshedAt = existing.LastRefreshedAt
@@ -598,10 +635,17 @@ func (h *Handler) registerAuthFromFile(ctx context.Context, path string, data []
 }
 
 func (h *Handler) disableAuth(ctx context.Context, id string) {
-	if h.authManager == nil || id == "" {
+	if h == nil || h.authManager == nil {
 		return
 	}
-	if auth, ok := h.authManager.GetByID(id); ok {
+	authID := h.authIDForPath(id)
+	if authID == "" {
+		authID = strings.TrimSpace(id)
+	}
+	if authID == "" {
+		return
+	}
+	if auth, ok := h.authManager.GetByID(authID); ok {
 		auth.Disabled = true
 		auth.Status = coreauth.StatusDisabled
 		auth.StatusMessage = "removed via management API"
@@ -610,9 +654,20 @@ func (h *Handler) disableAuth(ctx context.Context, id string) {
 	}
 }
 
-func (h *Handler) saveTokenRecord(ctx context.Context, record *coreauth.Auth) (string, error) {
-	if record == nil {
-		return "", fmt.Errorf("token record is nil")
+func (h *Handler) deleteTokenRecord(ctx context.Context, path string) error {
+	if strings.TrimSpace(path) == "" {
+		return fmt.Errorf("auth path is empty")
+	}
+	store := h.tokenStoreWithBaseDir()
+	if store == nil {
+		return fmt.Errorf("token store unavailable")
+	}
+	return store.Delete(ctx, path)
+}
+
+func (h *Handler) tokenStoreWithBaseDir() coreauth.Store {
+	if h == nil {
+		return nil
 	}
 	store := h.tokenStore
 	if store == nil {
@@ -624,6 +679,17 @@ func (h *Handler) saveTokenRecord(ctx context.Context, record *coreauth.Auth) (s
 			dirSetter.SetBaseDir(h.cfg.AuthDir)
 		}
 	}
+	return store
+}
+
+func (h *Handler) saveTokenRecord(ctx context.Context, record *coreauth.Auth) (string, error) {
+	if record == nil {
+		return "", fmt.Errorf("token record is nil")
+	}
+	store := h.tokenStoreWithBaseDir()
+	if store == nil {
+		return "", fmt.Errorf("token store unavailable")
+	}
 	return store.Save(ctx, record)
 }
 
@@ -971,29 +1037,46 @@ func (h *Handler) RequestGeminiCLIToken(c *gin.Context) {
 		}
 		fmt.Println("Authentication successful.")
 
-		if errEnsure := ensureGeminiProjectAndOnboard(ctx, gemClient, &ts, requestedProjectID); errEnsure != nil {
-			log.Errorf("Failed to complete Gemini CLI onboarding: %v", errEnsure)
-			oauthStatus[state] = "Failed to complete Gemini CLI onboarding"
-			return
-		}
+		if strings.EqualFold(requestedProjectID, "ALL") {
+			ts.Auto = false
+			projects, errAll := onboardAllGeminiProjects(ctx, gemClient, &ts)
+			if errAll != nil {
+				log.Errorf("Failed to complete Gemini CLI onboarding: %v", errAll)
+				oauthStatus[state] = "Failed to complete Gemini CLI onboarding"
+				return
+			}
+			if errVerify := ensureGeminiProjectsEnabled(ctx, gemClient, projects); errVerify != nil {
+				log.Errorf("Failed to verify Cloud AI API status: %v", errVerify)
+				oauthStatus[state] = "Failed to verify Cloud AI API status"
+				return
+			}
+			ts.ProjectID = strings.Join(projects, ",")
+			ts.Checked = true
+		} else {
+			if errEnsure := ensureGeminiProjectAndOnboard(ctx, gemClient, &ts, requestedProjectID); errEnsure != nil {
+				log.Errorf("Failed to complete Gemini CLI onboarding: %v", errEnsure)
+				oauthStatus[state] = "Failed to complete Gemini CLI onboarding"
+				return
+			}
 
-		if strings.TrimSpace(ts.ProjectID) == "" {
-			log.Error("Onboarding did not return a project ID")
-			oauthStatus[state] = "Failed to resolve project ID"
-			return
-		}
+			if strings.TrimSpace(ts.ProjectID) == "" {
+				log.Error("Onboarding did not return a project ID")
+				oauthStatus[state] = "Failed to resolve project ID"
+				return
+			}
 
-		isChecked, errCheck := checkCloudAPIIsEnabled(ctx, gemClient, ts.ProjectID)
-		if errCheck != nil {
-			log.Errorf("Failed to verify Cloud AI API status: %v", errCheck)
-			oauthStatus[state] = "Failed to verify Cloud AI API status"
-			return
-		}
-		ts.Checked = isChecked
-		if !isChecked {
-			log.Error("Cloud AI API is not enabled for the selected project")
-			oauthStatus[state] = "Cloud AI API not enabled"
-			return
+			isChecked, errCheck := checkCloudAPIIsEnabled(ctx, gemClient, ts.ProjectID)
+			if errCheck != nil {
+				log.Errorf("Failed to verify Cloud AI API status: %v", errCheck)
+				oauthStatus[state] = "Failed to verify Cloud AI API status"
+				return
+			}
+			ts.Checked = isChecked
+			if !isChecked {
+				log.Error("Cloud AI API is not enabled for the selected project")
+				oauthStatus[state] = "Cloud AI API not enabled"
+				return
+			}
 		}
 
 		recordMetadata := map[string]any{
@@ -1003,10 +1086,11 @@ func (h *Handler) RequestGeminiCLIToken(c *gin.Context) {
 			"checked":    ts.Checked,
 		}
 
+		fileName := geminiAuth.CredentialFileName(ts.Email, ts.ProjectID, true)
 		record := &coreauth.Auth{
-			ID:       fmt.Sprintf("gemini-%s-%s.json", ts.Email, ts.ProjectID),
+			ID:       fileName,
 			Provider: "gemini",
-			FileName: fmt.Sprintf("gemini-%s-%s.json", ts.Email, ts.ProjectID),
+			FileName: fileName,
 			Storage:  &ts,
 			Metadata: recordMetadata,
 		}
@@ -1359,6 +1443,87 @@ func (h *Handler) RequestIFlowToken(c *gin.Context) {
 	c.JSON(http.StatusOK, gin.H{"status": "ok", "url": authURL, "state": state})
 }
 
+func (h *Handler) RequestIFlowCookieToken(c *gin.Context) {
+	ctx := context.Background()
+
+	var payload struct {
+		Cookie string `json:"cookie"`
+	}
+	if err := c.ShouldBindJSON(&payload); err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"status": "error", "error": "cookie is required"})
+		return
+	}
+
+	cookieValue := strings.TrimSpace(payload.Cookie)
+
+	if cookieValue == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"status": "error", "error": "cookie is required"})
+		return
+	}
+
+	cookieValue, errNormalize := iflowauth.NormalizeCookie(cookieValue)
+	if errNormalize != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"status": "error", "error": errNormalize.Error()})
+		return
+	}
+
+	authSvc := iflowauth.NewIFlowAuth(h.cfg)
+	tokenData, errAuth := authSvc.AuthenticateWithCookie(ctx, cookieValue)
+	if errAuth != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"status": "error", "error": errAuth.Error()})
+		return
+	}
+
+	tokenData.Cookie = cookieValue
+
+	tokenStorage := authSvc.CreateCookieTokenStorage(tokenData)
+	email := strings.TrimSpace(tokenStorage.Email)
+	if email == "" {
+		c.JSON(http.StatusBadRequest, gin.H{"status": "error", "error": "failed to extract email from token"})
+		return
+	}
+
+	fileName := iflowauth.SanitizeIFlowFileName(email)
+	if fileName == "" {
+		fileName = fmt.Sprintf("iflow-%d", time.Now().UnixMilli())
+	}
+
+	tokenStorage.Email = email
+
+	record := &coreauth.Auth{
+		ID:       fmt.Sprintf("iflow-%s.json", fileName),
+		Provider: "iflow",
+		FileName: fmt.Sprintf("iflow-%s.json", fileName),
+		Storage:  tokenStorage,
+		Metadata: map[string]any{
+			"email":        email,
+			"api_key":      tokenStorage.APIKey,
+			"expired":      tokenStorage.Expire,
+			"cookie":       tokenStorage.Cookie,
+			"type":         tokenStorage.Type,
+			"last_refresh": tokenStorage.LastRefresh,
+		},
+		Attributes: map[string]string{
+			"api_key": tokenStorage.APIKey,
+		},
+	}
+
+	savedPath, errSave := h.saveTokenRecord(ctx, record)
+	if errSave != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"status": "error", "error": "failed to save authentication tokens"})
+		return
+	}
+
+	fmt.Printf("iFlow cookie authentication successful. Token saved to %s\n", savedPath)
+	c.JSON(http.StatusOK, gin.H{
+		"status":     "ok",
+		"saved_path": savedPath,
+		"email":      email,
+		"expired":    tokenStorage.Expire,
+		"type":       tokenStorage.Type,
+	})
+}
+
 type projectSelectionRequiredError struct{}
 
 func (e *projectSelectionRequiredError) Error() string {
@@ -1399,6 +1564,57 @@ func ensureGeminiProjectAndOnboard(ctx context.Context, httpClient *http.Client,
 	return nil
 }
 
+func onboardAllGeminiProjects(ctx context.Context, httpClient *http.Client, storage *geminiAuth.GeminiTokenStorage) ([]string, error) {
+	projects, errProjects := fetchGCPProjects(ctx, httpClient)
+	if errProjects != nil {
+		return nil, fmt.Errorf("fetch project list: %w", errProjects)
+	}
+	if len(projects) == 0 {
+		return nil, fmt.Errorf("no Google Cloud projects available for this account")
+	}
+	activated := make([]string, 0, len(projects))
+	seen := make(map[string]struct{}, len(projects))
+	for _, project := range projects {
+		candidate := strings.TrimSpace(project.ProjectID)
+		if candidate == "" {
+			continue
+		}
+		if _, dup := seen[candidate]; dup {
+			continue
+		}
+		if err := performGeminiCLISetup(ctx, httpClient, storage, candidate); err != nil {
+			return nil, fmt.Errorf("onboard project %s: %w", candidate, err)
+		}
+		finalID := strings.TrimSpace(storage.ProjectID)
+		if finalID == "" {
+			finalID = candidate
+		}
+		activated = append(activated, finalID)
+		seen[candidate] = struct{}{}
+	}
+	if len(activated) == 0 {
+		return nil, fmt.Errorf("no Google Cloud projects available for this account")
+	}
+	return activated, nil
+}
+
+func ensureGeminiProjectsEnabled(ctx context.Context, httpClient *http.Client, projectIDs []string) error {
+	for _, pid := range projectIDs {
+		trimmed := strings.TrimSpace(pid)
+		if trimmed == "" {
+			continue
+		}
+		isChecked, errCheck := checkCloudAPIIsEnabled(ctx, httpClient, trimmed)
+		if errCheck != nil {
+			return fmt.Errorf("project %s: %w", trimmed, errCheck)
+		}
+		if !isChecked {
+			return fmt.Errorf("project %s: Cloud AI API not enabled", trimmed)
+		}
+	}
+	return nil
+}
+
 func performGeminiCLISetup(ctx context.Context, httpClient *http.Client, storage *geminiAuth.GeminiTokenStorage, requestedProject string) error {
 	metadata := map[string]string{
 		"ideType":    "IDE_UNSPECIFIED",

internal/api/handlers/management/config_lists.go

@@ -408,9 +408,7 @@ func (h *Handler) PutOpenAICompat(c *gin.Context) {
 		}
 		arr = obj.Items
 	}
-	for i := range arr {
-		normalizeOpenAICompatibilityEntry(&arr[i])
-	}
+	arr = migrateLegacyOpenAICompatibilityKeys(arr)
 	// Filter out providers with empty base-url -> remove provider entirely
 	filtered := make([]config.OpenAICompatibility, 0, len(arr))
 	for i := range arr {
@@ -418,7 +416,7 @@ func (h *Handler) PutOpenAICompat(c *gin.Context) {
 			filtered = append(filtered, arr[i])
 		}
 	}
-	h.cfg.OpenAICompatibility = filtered
+	h.cfg.OpenAICompatibility = migrateLegacyOpenAICompatibilityKeys(filtered)
 	h.cfg.SanitizeOpenAICompatibility()
 	h.persist(c)
 }
@@ -432,6 +430,7 @@ func (h *Handler) PatchOpenAICompat(c *gin.Context) {
 		c.JSON(400, gin.H{"error": "invalid body"})
 		return
 	}
+	h.cfg.OpenAICompatibility = migrateLegacyOpenAICompatibilityKeys(h.cfg.OpenAICompatibility)
 	normalizeOpenAICompatibilityEntry(body.Value)
 	// If base-url becomes empty, delete the provider instead of updating
 	if strings.TrimSpace(body.Value.BaseURL) == "" {
@@ -661,6 +660,13 @@ func normalizeOpenAICompatibilityEntry(entry *config.OpenAICompatibility) {
 	entry.APIKeys = nil
 }
 
+func migrateLegacyOpenAICompatibilityKeys(entries []config.OpenAICompatibility) []config.OpenAICompatibility {
+	for i := range entries {
+		normalizeOpenAICompatibilityEntry(&entries[i])
+	}
+	return entries
+}
+
 func normalizedOpenAICompatibilityEntries(entries []config.OpenAICompatibility) []config.OpenAICompatibility {
 	if len(entries) == 0 {
 		return nil

internal/api/handlers/management/handler.go

@@ -13,6 +13,7 @@ import (
 	"time"
 
 	"github.com/gin-gonic/gin"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/buildinfo"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/usage"
 	sdkAuth "github.com/router-for-me/CLIProxyAPI/v6/sdk/auth"
@@ -91,6 +92,10 @@ func (h *Handler) Middleware() gin.HandlerFunc {
 	const banDuration = 30 * time.Minute
 
 	return func(c *gin.Context) {
+		c.Header("X-CPA-VERSION", buildinfo.Version)
+		c.Header("X-CPA-COMMIT", buildinfo.Commit)
+		c.Header("X-CPA-BUILD-DATE", buildinfo.BuildDate)
+
 		clientIP := c.ClientIP()
 		localClient := clientIP == "127.0.0.1" || clientIP == "::1"
 		cfg := h.cfg

internal/api/server.go

@@ -518,6 +518,7 @@ func (s *Server) registerManagementRoutes() {
 		mgmt.GET("/gemini-cli-auth-url", s.mgmt.RequestGeminiCLIToken)
 		mgmt.GET("/qwen-auth-url", s.mgmt.RequestQwenToken)
 		mgmt.GET("/iflow-auth-url", s.mgmt.RequestIFlowToken)
+		mgmt.POST("/iflow-auth-url", s.mgmt.RequestIFlowCookieToken)
 		mgmt.GET("/get-auth-status", s.mgmt.GetAuthStatus)
 	}
 }

internal/auth/gemini/gemini_token.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strings"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/misc"
 	log "github.com/sirupsen/logrus"
@@ -67,3 +68,20 @@ func (ts *GeminiTokenStorage) SaveTokenToFile(authFilePath string) error {
 	}
 	return nil
 }
+
+// CredentialFileName returns the filename used to persist Gemini CLI credentials.
+// When projectID represents multiple projects (comma-separated or literal ALL),
+// the suffix is normalized to "all" and a "gemini-" prefix is enforced to keep
+// web and CLI generated files consistent.
+func CredentialFileName(email, projectID string, includeProviderPrefix bool) string {
+	email = strings.TrimSpace(email)
+	project := strings.TrimSpace(projectID)
+	if strings.EqualFold(project, "all") || strings.Contains(project, ",") {
+		return fmt.Sprintf("gemini-%s-all.json", email)
+	}
+	prefix := ""
+	if includeProviderPrefix {
+		prefix = "gemini-"
+	}
+	return fmt.Sprintf("%s%s-%s.json", prefix, email, project)
+}

internal/auth/iflow/cookie_helpers.go

@@ -0,0 +1,38 @@
+package iflow
+
+import (
+	"fmt"
+	"strings"
+)
+
+// NormalizeCookie normalizes raw cookie strings for iFlow authentication flows.
+func NormalizeCookie(raw string) (string, error) {
+	trimmed := strings.TrimSpace(raw)
+	if trimmed == "" {
+		return "", fmt.Errorf("cookie cannot be empty")
+	}
+
+	combined := strings.Join(strings.Fields(trimmed), " ")
+	if !strings.HasSuffix(combined, ";") {
+		combined += ";"
+	}
+	if !strings.Contains(combined, "BXAuth=") {
+		return "", fmt.Errorf("cookie missing BXAuth field")
+	}
+	return combined, nil
+}
+
+// SanitizeIFlowFileName normalizes user identifiers for safe filename usage.
+func SanitizeIFlowFileName(raw string) string {
+	if raw == "" {
+		return ""
+	}
+	cleanEmail := strings.ReplaceAll(raw, "*", "x")
+	var result strings.Builder
+	for _, r := range cleanEmail {
+		if (r >= 'a' && r <= 'z') || (r >= 'A' && r <= 'Z') || (r >= '0' && r <= '9') || r == '_' || r == '@' || r == '.' || r == '-' {
+			result.WriteRune(r)
+		}
+	}
+	return strings.TrimSpace(result.String())
+}

internal/auth/iflow/iflow_auth.go

@@ -1,6 +1,7 @@
 package iflow
 
 import (
+	"compress/gzip"
 	"context"
 	"encoding/base64"
 	"encoding/json"
@@ -23,6 +24,9 @@ const (
 	iFlowUserInfoEndpoint       = "https://iflow.cn/api/oauth/getUserInfo"
 	iFlowSuccessRedirectURL     = "https://iflow.cn/oauth/success"
 
+	// Cookie authentication endpoints
+	iFlowAPIKeyEndpoint = "https://platform.iflow.cn/api/openapi/apikey"
+
 	// Client credentials provided by iFlow for the Code Assist integration.
 	iFlowOAuthClientID     = "10009311001"
 	iFlowOAuthClientSecret = "4Z3YjXycVsQvyGF1etiNlIBB4RsqSDtW"
@@ -261,6 +265,7 @@ type IFlowTokenData struct {
 	Expire       string
 	APIKey       string
 	Email        string
+	Cookie       string
 }
 
 // userInfoResponse represents the structure returned by the user info endpoint.
@@ -274,3 +279,232 @@ type userInfoData struct {
 	Email  string `json:"email"`
 	Phone  string `json:"phone"`
 }
+
+// iFlowAPIKeyResponse represents the response from the API key endpoint
+type iFlowAPIKeyResponse struct {
+	Success bool         `json:"success"`
+	Code    string       `json:"code"`
+	Message string       `json:"message"`
+	Data    iFlowKeyData `json:"data"`
+	Extra   interface{}  `json:"extra"`
+}
+
+// iFlowKeyData contains the API key information
+type iFlowKeyData struct {
+	HasExpired bool   `json:"hasExpired"`
+	ExpireTime string `json:"expireTime"`
+	Name       string `json:"name"`
+	APIKey     string `json:"apiKey"`
+	APIKeyMask string `json:"apiKeyMask"`
+}
+
+// iFlowRefreshRequest represents the request body for refreshing API key
+type iFlowRefreshRequest struct {
+	Name string `json:"name"`
+}
+
+// AuthenticateWithCookie performs authentication using browser cookies
+func (ia *IFlowAuth) AuthenticateWithCookie(ctx context.Context, cookie string) (*IFlowTokenData, error) {
+	if strings.TrimSpace(cookie) == "" {
+		return nil, fmt.Errorf("iflow cookie authentication: cookie is empty")
+	}
+
+	// First, get initial API key information using GET request
+	keyInfo, err := ia.fetchAPIKeyInfo(ctx, cookie)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie authentication: fetch initial API key info failed: %w", err)
+	}
+
+	// Convert to token data format
+	data := &IFlowTokenData{
+		APIKey: keyInfo.APIKey,
+		Expire: keyInfo.ExpireTime,
+		Email:  keyInfo.Name,
+		Cookie: cookie,
+	}
+
+	return data, nil
+}
+
+// fetchAPIKeyInfo retrieves API key information using GET request with cookie
+func (ia *IFlowAuth) fetchAPIKeyInfo(ctx context.Context, cookie string) (*iFlowKeyData, error) {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, iFlowAPIKeyEndpoint, nil)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie: create GET request failed: %w", err)
+	}
+
+	// Set cookie and other headers to mimic browser
+	req.Header.Set("Cookie", cookie)
+	req.Header.Set("Accept", "application/json, text/plain, */*")
+	req.Header.Set("User-Agent", "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36")
+	req.Header.Set("Accept-Language", "zh-CN,zh;q=0.9,en;q=0.8")
+	req.Header.Set("Accept-Encoding", "gzip, deflate, br")
+	req.Header.Set("Connection", "keep-alive")
+	req.Header.Set("Sec-Fetch-Dest", "empty")
+	req.Header.Set("Sec-Fetch-Mode", "cors")
+	req.Header.Set("Sec-Fetch-Site", "same-origin")
+
+	resp, err := ia.httpClient.Do(req)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie: GET request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	// Handle gzip compression
+	var reader io.Reader = resp.Body
+	if resp.Header.Get("Content-Encoding") == "gzip" {
+		gzipReader, err := gzip.NewReader(resp.Body)
+		if err != nil {
+			return nil, fmt.Errorf("iflow cookie: create gzip reader failed: %w", err)
+		}
+		defer func() { _ = gzipReader.Close() }()
+		reader = gzipReader
+	}
+
+	body, err := io.ReadAll(reader)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie: read GET response failed: %w", err)
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		log.Debugf("iflow cookie GET request failed: status=%d body=%s", resp.StatusCode, string(body))
+		return nil, fmt.Errorf("iflow cookie: GET request failed with status %d: %s", resp.StatusCode, strings.TrimSpace(string(body)))
+	}
+
+	var keyResp iFlowAPIKeyResponse
+	if err = json.Unmarshal(body, &keyResp); err != nil {
+		return nil, fmt.Errorf("iflow cookie: decode GET response failed: %w", err)
+	}
+
+	if !keyResp.Success {
+		return nil, fmt.Errorf("iflow cookie: GET request not successful: %s", keyResp.Message)
+	}
+
+	// Handle initial response where apiKey field might be apiKeyMask
+	if keyResp.Data.APIKey == "" && keyResp.Data.APIKeyMask != "" {
+		keyResp.Data.APIKey = keyResp.Data.APIKeyMask
+	}
+
+	return &keyResp.Data, nil
+}
+
+// RefreshAPIKey refreshes the API key using POST request
+func (ia *IFlowAuth) RefreshAPIKey(ctx context.Context, cookie, name string) (*iFlowKeyData, error) {
+	if strings.TrimSpace(cookie) == "" {
+		return nil, fmt.Errorf("iflow cookie refresh: cookie is empty")
+	}
+	if strings.TrimSpace(name) == "" {
+		return nil, fmt.Errorf("iflow cookie refresh: name is empty")
+	}
+
+	// Prepare request body
+	refreshReq := iFlowRefreshRequest{
+		Name: name,
+	}
+
+	bodyBytes, err := json.Marshal(refreshReq)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie refresh: marshal request failed: %w", err)
+	}
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, iFlowAPIKeyEndpoint, strings.NewReader(string(bodyBytes)))
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie refresh: create POST request failed: %w", err)
+	}
+
+	// Set cookie and other headers to mimic browser
+	req.Header.Set("Cookie", cookie)
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Accept", "application/json, text/plain, */*")
+	req.Header.Set("User-Agent", "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36")
+	req.Header.Set("Accept-Language", "zh-CN,zh;q=0.9,en;q=0.8")
+	req.Header.Set("Accept-Encoding", "gzip, deflate, br")
+	req.Header.Set("Connection", "keep-alive")
+	req.Header.Set("Origin", "https://platform.iflow.cn")
+	req.Header.Set("Referer", "https://platform.iflow.cn/")
+
+	resp, err := ia.httpClient.Do(req)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie refresh: POST request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	// Handle gzip compression
+	var reader io.Reader = resp.Body
+	if resp.Header.Get("Content-Encoding") == "gzip" {
+		gzipReader, err := gzip.NewReader(resp.Body)
+		if err != nil {
+			return nil, fmt.Errorf("iflow cookie refresh: create gzip reader failed: %w", err)
+		}
+		defer func() { _ = gzipReader.Close() }()
+		reader = gzipReader
+	}
+
+	body, err := io.ReadAll(reader)
+	if err != nil {
+		return nil, fmt.Errorf("iflow cookie refresh: read POST response failed: %w", err)
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		log.Debugf("iflow cookie POST request failed: status=%d body=%s", resp.StatusCode, string(body))
+		return nil, fmt.Errorf("iflow cookie refresh: POST request failed with status %d: %s", resp.StatusCode, strings.TrimSpace(string(body)))
+	}
+
+	var keyResp iFlowAPIKeyResponse
+	if err = json.Unmarshal(body, &keyResp); err != nil {
+		return nil, fmt.Errorf("iflow cookie refresh: decode POST response failed: %w", err)
+	}
+
+	if !keyResp.Success {
+		return nil, fmt.Errorf("iflow cookie refresh: POST request not successful: %s", keyResp.Message)
+	}
+
+	return &keyResp.Data, nil
+}
+
+// ShouldRefreshAPIKey checks if the API key needs to be refreshed (within 2 days of expiry)
+func ShouldRefreshAPIKey(expireTime string) (bool, time.Duration, error) {
+	if strings.TrimSpace(expireTime) == "" {
+		return false, 0, fmt.Errorf("iflow cookie: expire time is empty")
+	}
+
+	expire, err := time.Parse("2006-01-02 15:04", expireTime)
+	if err != nil {
+		return false, 0, fmt.Errorf("iflow cookie: parse expire time failed: %w", err)
+	}
+
+	now := time.Now()
+	twoDaysFromNow := now.Add(48 * time.Hour)
+
+	needsRefresh := expire.Before(twoDaysFromNow)
+	timeUntilExpiry := expire.Sub(now)
+
+	return needsRefresh, timeUntilExpiry, nil
+}
+
+// CreateCookieTokenStorage converts cookie-based token data into persistence storage
+func (ia *IFlowAuth) CreateCookieTokenStorage(data *IFlowTokenData) *IFlowTokenStorage {
+	if data == nil {
+		return nil
+	}
+
+	return &IFlowTokenStorage{
+		APIKey:      data.APIKey,
+		Email:       data.Email,
+		Expire:      data.Expire,
+		Cookie:      data.Cookie,
+		LastRefresh: time.Now().Format(time.RFC3339),
+		Type:        "iflow",
+	}
+}
+
+// UpdateCookieTokenStorage updates the persisted token storage with refreshed API key data
+func (ia *IFlowAuth) UpdateCookieTokenStorage(storage *IFlowTokenStorage, keyData *iFlowKeyData) {
+	if storage == nil || keyData == nil {
+		return
+	}
+
+	storage.APIKey = keyData.APIKey
+	storage.Expire = keyData.ExpireTime
+	storage.LastRefresh = time.Now().Format(time.RFC3339)
+}

internal/auth/iflow/iflow_token.go

@@ -19,6 +19,7 @@ type IFlowTokenStorage struct {
 	Email        string `json:"email"`
 	TokenType    string `json:"token_type"`
 	Scope        string `json:"scope"`
+	Cookie       string `json:"cookie"`
 	Type         string `json:"type"`
 }
 

internal/buildinfo/buildinfo.go

@@ -0,0 +1,15 @@
+// Package buildinfo exposes compile-time metadata shared across the server.
+package buildinfo
+
+// The following variables are overridden via ldflags during release builds.
+// Defaults cover local development builds.
+var (
+	// Version is the semantic version or git describe output of the binary.
+	Version = "dev"
+
+	// Commit is the git commit SHA baked into the binary.
+	Commit = "none"
+
+	// BuildDate records when the binary was built in UTC.
+	BuildDate = "unknown"
+)

internal/cmd/iflow_cookie.go

@@ -0,0 +1,86 @@
+package cmd
+
+import (
+	"bufio"
+	"context"
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/auth/iflow"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+// DoIFlowCookieAuth performs the iFlow cookie-based authentication.
+func DoIFlowCookieAuth(cfg *config.Config, options *LoginOptions) {
+	if options == nil {
+		options = &LoginOptions{}
+	}
+
+	promptFn := options.Prompt
+	if promptFn == nil {
+		reader := bufio.NewReader(os.Stdin)
+		promptFn = func(prompt string) (string, error) {
+			fmt.Print(prompt)
+			value, err := reader.ReadString('\n')
+			if err != nil {
+				return "", err
+			}
+			return strings.TrimSpace(value), nil
+		}
+	}
+
+	// Prompt user for cookie
+	cookie, err := promptForCookie(promptFn)
+	if err != nil {
+		fmt.Printf("Failed to get cookie: %v\n", err)
+		return
+	}
+
+	// Authenticate with cookie
+	auth := iflow.NewIFlowAuth(cfg)
+	ctx := context.Background()
+
+	tokenData, err := auth.AuthenticateWithCookie(ctx, cookie)
+	if err != nil {
+		fmt.Printf("iFlow cookie authentication failed: %v\n", err)
+		return
+	}
+
+	// Create token storage
+	tokenStorage := auth.CreateCookieTokenStorage(tokenData)
+
+	// Get auth file path using email in filename
+	authFilePath := getAuthFilePath(cfg, "iflow", tokenData.Email)
+
+	// Save token to file
+	if err := tokenStorage.SaveTokenToFile(authFilePath); err != nil {
+		fmt.Printf("Failed to save authentication: %v\n", err)
+		return
+	}
+
+	fmt.Printf("Authentication successful! API key: %s\n", tokenData.APIKey)
+	fmt.Printf("Expires at: %s\n", tokenData.Expire)
+	fmt.Printf("Authentication saved to: %s\n", authFilePath)
+}
+
+// promptForCookie prompts the user to enter their iFlow cookie
+func promptForCookie(promptFn func(string) (string, error)) (string, error) {
+	line, err := promptFn("Enter iFlow Cookie (from browser cookies): ")
+	if err != nil {
+		return "", fmt.Errorf("failed to read cookie: %w", err)
+	}
+
+	cookie, err := iflow.NormalizeCookie(line)
+	if err != nil {
+		return "", err
+	}
+
+	return cookie, nil
+}
+
+// getAuthFilePath returns the auth file path for the given provider and email
+func getAuthFilePath(cfg *config.Config, provider, email string) string {
+	fileName := iflow.SanitizeIFlowFileName(email)
+	return fmt.Sprintf("%s/%s-%s.json", cfg.AuthDir, provider, fileName)
+}

internal/cmd/login.go

@@ -96,35 +96,52 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 	}
 
 	selectedProjectID := promptForProjectSelection(projects, strings.TrimSpace(projectID), promptFn)
-	if strings.TrimSpace(selectedProjectID) == "" {
+	projectSelections, errSelection := resolveProjectSelections(selectedProjectID, projects)
+	if errSelection != nil {
+		log.Fatalf("Invalid project selection: %v", errSelection)
+		return
+	}
+	if len(projectSelections) == 0 {
 		log.Fatal("No project selected; aborting login.")
 		return
 	}
 
-	if errSetup := performGeminiCLISetup(ctx, httpClient, storage, selectedProjectID); errSetup != nil {
-		var projectErr *projectSelectionRequiredError
-		if errors.As(errSetup, &projectErr) {
-			log.Error("Failed to start user onboarding: A project ID is required.")
-			showProjectSelectionHelp(storage.Email, projects)
+	activatedProjects := make([]string, 0, len(projectSelections))
+	for _, candidateID := range projectSelections {
+		log.Infof("Activating project %s", candidateID)
+		if errSetup := performGeminiCLISetup(ctx, httpClient, storage, candidateID); errSetup != nil {
+			var projectErr *projectSelectionRequiredError
+			if errors.As(errSetup, &projectErr) {
+				log.Error("Failed to start user onboarding: A project ID is required.")
+				showProjectSelectionHelp(storage.Email, projects)
+				return
+			}
+			log.Fatalf("Failed to complete user setup: %v", errSetup)
 			return
 		}
-		log.Fatalf("Failed to complete user setup: %v", errSetup)
-		return
+		finalID := strings.TrimSpace(storage.ProjectID)
+		if finalID == "" {
+			finalID = candidateID
+		}
+		activatedProjects = append(activatedProjects, finalID)
 	}
 
 	storage.Auto = false
+	storage.ProjectID = strings.Join(activatedProjects, ",")
 
 	if !storage.Auto && !storage.Checked {
-		isChecked, errCheck := checkCloudAPIIsEnabled(ctx, httpClient, storage.ProjectID)
-		if errCheck != nil {
-			log.Fatalf("Failed to check if Cloud AI API is enabled: %v", errCheck)
-			return
-		}
-		storage.Checked = isChecked
-		if !isChecked {
-			log.Fatal("Failed to check if Cloud AI API is enabled. If you encounter an error message, please create an issue.")
-			return
+		for _, pid := range activatedProjects {
+			isChecked, errCheck := checkCloudAPIIsEnabled(ctx, httpClient, pid)
+			if errCheck != nil {
+				log.Fatalf("Failed to check if Cloud AI API is enabled for %s: %v", pid, errCheck)
+				return
+			}
+			if !isChecked {
+				log.Fatalf("Failed to check if Cloud AI API is enabled for project %s. If you encounter an error message, please create an issue.", pid)
+				return
+			}
 		}
+		storage.Checked = true
 	}
 
 	updateAuthRecord(record, storage)
@@ -354,10 +371,14 @@ func promptForProjectSelection(projects []interfaces.GCPProjectProjects, presetI
 			defaultIndex = idx
 		}
 	}
+	fmt.Println("Type 'ALL' to onboard every listed project.")
 
 	defaultID := projects[defaultIndex].ProjectID
 
 	if trimmedPreset != "" {
+		if strings.EqualFold(trimmedPreset, "ALL") {
+			return "ALL"
+		}
 		for _, project := range projects {
 			if project.ProjectID == trimmedPreset {
 				return trimmedPreset
@@ -367,13 +388,16 @@ func promptForProjectSelection(projects []interfaces.GCPProjectProjects, presetI
 	}
 
 	for {
-		promptMsg := fmt.Sprintf("Enter project ID [%s]: ", defaultID)
+		promptMsg := fmt.Sprintf("Enter project ID [%s] or ALL: ", defaultID)
 		answer, errPrompt := promptFn(promptMsg)
 		if errPrompt != nil {
 			log.Errorf("Project selection prompt failed: %v", errPrompt)
 			return defaultID
 		}
 		answer = strings.TrimSpace(answer)
+		if strings.EqualFold(answer, "ALL") {
+			return "ALL"
+		}
 		if answer == "" {
 			return defaultID
 		}
@@ -394,6 +418,52 @@ func promptForProjectSelection(projects []interfaces.GCPProjectProjects, presetI
 	}
 }
 
+func resolveProjectSelections(selection string, projects []interfaces.GCPProjectProjects) ([]string, error) {
+	trimmed := strings.TrimSpace(selection)
+	if trimmed == "" {
+		return nil, nil
+	}
+	available := make(map[string]struct{}, len(projects))
+	ordered := make([]string, 0, len(projects))
+	for _, project := range projects {
+		id := strings.TrimSpace(project.ProjectID)
+		if id == "" {
+			continue
+		}
+		if _, exists := available[id]; exists {
+			continue
+		}
+		available[id] = struct{}{}
+		ordered = append(ordered, id)
+	}
+	if strings.EqualFold(trimmed, "ALL") {
+		if len(ordered) == 0 {
+			return nil, fmt.Errorf("no projects available for ALL selection")
+		}
+		return append([]string(nil), ordered...), nil
+	}
+	parts := strings.Split(trimmed, ",")
+	selections := make([]string, 0, len(parts))
+	seen := make(map[string]struct{}, len(parts))
+	for _, part := range parts {
+		id := strings.TrimSpace(part)
+		if id == "" {
+			continue
+		}
+		if _, dup := seen[id]; dup {
+			continue
+		}
+		if len(available) > 0 {
+			if _, ok := available[id]; !ok {
+				return nil, fmt.Errorf("project %s not found in available projects", id)
+			}
+		}
+		seen[id] = struct{}{}
+		selections = append(selections, id)
+	}
+	return selections, nil
+}
+
 func defaultProjectPrompt() func(string) (string, error) {
 	reader := bufio.NewReader(os.Stdin)
 	return func(prompt string) (string, error) {
@@ -495,7 +565,7 @@ func updateAuthRecord(record *cliproxyauth.Auth, storage *gemini.GeminiTokenStor
 		return
 	}
 
-	finalName := fmt.Sprintf("%s-%s.json", storage.Email, storage.ProjectID)
+	finalName := gemini.CredentialFileName(storage.Email, storage.ProjectID, false)
 
 	if record.Metadata == nil {
 		record.Metadata = make(map[string]any)

internal/config/config.go

@@ -64,6 +64,9 @@ type Config struct {
 
 	// RemoteManagement nests management-related options under 'remote-management'.
 	RemoteManagement RemoteManagement `yaml:"remote-management" json:"-"`
+
+	// Payload defines default and override rules for provider payload parameters.
+	Payload PayloadConfig `yaml:"payload" json:"payload"`
 }
 
 // RemoteManagement holds management API configuration under 'remote-management'.
@@ -86,6 +89,30 @@ type QuotaExceeded struct {
 	SwitchPreviewModel bool `yaml:"switch-preview-model" json:"switch-preview-model"`
 }
 
+// PayloadConfig defines default and override parameter rules applied to provider payloads.
+type PayloadConfig struct {
+	// Default defines rules that only set parameters when they are missing in the payload.
+	Default []PayloadRule `yaml:"default" json:"default"`
+	// Override defines rules that always set parameters, overwriting any existing values.
+	Override []PayloadRule `yaml:"override" json:"override"`
+}
+
+// PayloadRule describes a single rule targeting a list of models with parameter updates.
+type PayloadRule struct {
+	// Models lists model entries with name pattern and protocol constraint.
+	Models []PayloadModelRule `yaml:"models" json:"models"`
+	// Params maps JSON paths (gjson/sjson syntax) to values written into the payload.
+	Params map[string]any `yaml:"params" json:"params"`
+}
+
+// PayloadModelRule ties a model name pattern to a specific translator protocol.
+type PayloadModelRule struct {
+	// Name is the model name or wildcard pattern (e.g., "gpt-*", "*-5", "gemini-*-pro").
+	Name string `yaml:"name" json:"name"`
+	// Protocol restricts the rule to a specific translator format (e.g., "gemini", "responses").
+	Protocol string `yaml:"protocol" json:"protocol"`
+}
+
 // ClaudeKey represents the configuration for a Claude API key,
 // including the API key itself and an optional base URL for the API endpoint.
 type ClaudeKey struct {
@@ -452,6 +479,7 @@ func SaveConfigPreserveComments(configFile string, cfg *Config) error {
 
 	// Remove deprecated auth block before merging to avoid persisting it again.
 	removeMapKey(original.Content[0], "auth")
+	removeLegacyOpenAICompatAPIKeys(original.Content[0])
 
 	// Merge generated into original in-place, preserving comments/order of existing nodes.
 	mergeMappingPreserve(original.Content[0], generated.Content[0])
@@ -908,6 +936,25 @@ func removeMapKey(mapNode *yaml.Node, key string) {
 	}
 }
 
+func removeLegacyOpenAICompatAPIKeys(root *yaml.Node) {
+	if root == nil || root.Kind != yaml.MappingNode {
+		return
+	}
+	idx := findMapKeyIndex(root, "openai-compatibility")
+	if idx < 0 || idx+1 >= len(root.Content) {
+		return
+	}
+	seq := root.Content[idx+1]
+	if seq == nil || seq.Kind != yaml.SequenceNode {
+		return
+	}
+	for i := range seq.Content {
+		if seq.Content[i] != nil && seq.Content[i].Kind == yaml.MappingNode {
+			removeMapKey(seq.Content[i], "api-keys")
+		}
+	}
+}
+
 // normalizeCollectionNodeStyles forces YAML collections to use block notation, keeping
 // lists and maps readable. Empty sequences retain flow style ([]) so empty list markers
 // remain compact.

internal/interfaces/client_models.go

@@ -62,6 +62,9 @@ type Part struct {
 	// InlineData contains base64-encoded data with its MIME type (e.g., images).
 	InlineData *InlineData `json:"inlineData,omitempty"`
 
+	// ThoughtSignature is a provider-required signature that accompanies certain parts.
+	ThoughtSignature string `json:"thoughtSignature,omitempty"`
+
 	// FunctionCall represents a tool call requested by the model.
 	FunctionCall *FunctionCall `json:"functionCall,omitempty"`
 

internal/misc/codex_instructions.go

@@ -17,6 +17,8 @@ func CodexInstructionsForModel(modelName, systemInstructions string) (bool, stri
 
 	lastPrompt := ""
 	lastCodexPrompt := ""
+	lastCodexMaxPrompt := ""
+	last51Prompt := ""
 	// lastReviewPrompt := ""
 	for _, entry := range entries {
 		content, _ := codexInstructionsDir.ReadFile("codex_instructions/" + entry.Name())
@@ -25,15 +27,22 @@ func CodexInstructionsForModel(modelName, systemInstructions string) (bool, stri
 		}
 		if strings.HasPrefix(entry.Name(), "gpt_5_codex_prompt.md") {
 			lastCodexPrompt = string(content)
+		} else if strings.HasPrefix(entry.Name(), "gpt-5.1-codex-max_prompt.md") {
+			lastCodexMaxPrompt = string(content)
 		} else if strings.HasPrefix(entry.Name(), "prompt.md") {
 			lastPrompt = string(content)
+		} else if strings.HasPrefix(entry.Name(), "gpt_5_1_prompt.md") {
+			last51Prompt = string(content)
 		} else if strings.HasPrefix(entry.Name(), "review_prompt.md") {
 			// lastReviewPrompt = string(content)
 		}
 	}
-
-	if strings.Contains(modelName, "codex") {
+	if strings.Contains(modelName, "codex-max") {
+		return false, lastCodexMaxPrompt
+	} else if strings.Contains(modelName, "codex") {
 		return false, lastCodexPrompt
+	} else if strings.Contains(modelName, "5.1") {
+		return false, last51Prompt
 	} else {
 		return false, lastPrompt
 	}

internal/misc/codex_instructions/gpt-5.1-codex-max_prompt.md-001-d5dfba250975b4519fed9b8abf99bbd6c31e6f33

@@ -0,0 +1,117 @@
+You are Codex, based on GPT-5. You are running as a coding agent in the Codex CLI on a user's computer.
+
+## General
+
+- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
+
+## Editing constraints
+
+- Default to ASCII when editing or creating files. Only introduce non-ASCII or other Unicode characters when there is a clear justification and the file already uses them.
+- Add succinct code comments that explain what is going on if code is not self-explanatory. You should not add comments like "Assigns the value to the variable", but a brief comment might be useful ahead of a complex code block that the user would otherwise have to spend time parsing out. Usage of these comments should be rare.
+- Try to use apply_patch for single file edits, but it is fine to explore other options to make the edit if it does not work well. Do not use apply_patch for changes that are auto-generated (i.e. generating package.json or running a lint or format command like gofmt) or when scripting is more efficient (such as search and replacing a string across a codebase).
+- You may be in a dirty git worktree.
+    * NEVER revert existing changes you did not make unless explicitly requested, since these changes were made by the user.
+    * If asked to make a commit or code edits and there are unrelated changes to your work or changes that you didn't make in those files, don't revert those changes.
+    * If the changes are in files you've touched recently, you should read carefully and understand how you can work with the changes rather than reverting them.
+    * If the changes are in unrelated files, just ignore them and don't revert them.
+- Do not amend a commit unless explicitly requested to do so.
+- While you are working, you might notice unexpected changes that you didn't make. If this happens, STOP IMMEDIATELY and ask the user how they would like to proceed.
+- **NEVER** use destructive commands like `git reset --hard` or `git checkout --` unless specifically requested or approved by the user.
+
+## Plan tool
+
+When using the planning tool:
+- Skip using the planning tool for straightforward tasks (roughly the easiest 25%).
+- Do not make single-step plans.
+- When you made a plan, update it after having performed one of the sub-tasks that you shared on the plan.
+
+## Codex CLI harness, sandboxing, and approvals
+
+The Codex CLI harness supports several different configurations for sandboxing and escalation approvals that the user can choose from.
+
+Filesystem sandboxing defines which files can be read or written. The options for `sandbox_mode` are:
+- **read-only**: The sandbox only permits reading files.
+- **workspace-write**: The sandbox permits reading files, and editing files in `cwd` and `writable_roots`. Editing files in other directories requires approval.
+- **danger-full-access**: No filesystem sandboxing - all commands are permitted.
+
+Network sandboxing defines whether network can be accessed without approval. Options for `network_access` are:
+- **restricted**: Requires approval
+- **enabled**: No approval needed
+
+Approvals are your mechanism to get user consent to run shell commands without the sandbox. Possible configuration options for `approval_policy` are
+- **untrusted**: The harness will escalate most commands for user approval, apart from a limited allowlist of safe "read" commands.
+- **on-failure**: The harness will allow all commands to run in the sandbox (if enabled), and failures will be escalated to the user for approval to run again without the sandbox.
+- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for it in the `shell` command description.)
+- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is paired with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
+
+When you are running with `approval_policy == on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
+- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
+- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
+- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
+- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters - do not message the user before requesting approval for the command.
+- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
+- (for all of these, you should weigh alternative paths that do not require approval)
+
+When `sandbox_mode` is set to read-only, you'll need to request approval for any command that isn't a read.
+
+You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing enabled, and approval on-failure.
+
+Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
+
+When requesting approval to execute a command that will require escalated privileges:
+  - Provide the `with_escalated_permissions` parameter with the boolean value true
+  - Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
+
+## Special user requests
+
+- If the user makes a simple request (such as asking for the time) which you can fulfill by running a terminal command (such as `date`), you should do so.
+- If the user asks for a "review", default to a code review mindset: prioritise identifying bugs, risks, behavioural regressions, and missing tests. Findings must be the primary focus of the response - keep summaries or overviews brief and only after enumerating the issues. Present findings first (ordered by severity with file/line references), follow with open questions or assumptions, and offer a change-summary only as a secondary detail. If no findings are discovered, state that explicitly and mention any residual risks or testing gaps.
+
+## Frontend tasks
+When doing frontend design tasks, avoid collapsing into "AI slop" or safe, average-looking layouts.
+Aim for interfaces that feel intentional, bold, and a bit surprising.
+- Typography: Use expressive, purposeful fonts and avoid default stacks (Inter, Roboto, Arial, system).
+- Color & Look: Choose a clear visual direction; define CSS variables; avoid purple-on-white defaults. No purple bias or dark mode bias.
+- Motion: Use a few meaningful animations (page-load, staggered reveals) instead of generic micro-motions.
+- Background: Don't rely on flat, single-color backgrounds; use gradients, shapes, or subtle patterns to build atmosphere.
+- Overall: Avoid boilerplate layouts and interchangeable UI patterns. Vary themes, type families, and visual languages across outputs.
+- Ensure the page loads properly on both desktop and mobile
+
+Exception: If working within an existing website or design system, preserve the established patterns, structure, and visual language.
+
+## Presenting your work and final message
+
+You are producing plain text that will later be styled by the CLI. Follow these rules exactly. Formatting should make results easy to scan, but not feel mechanical. Use judgment to decide how much structure adds value.
+
+- Default: be very concise; friendly coding teammate tone.
+- Ask only when needed; suggest ideas; mirror the user's style.
+- For substantial work, summarize clearly; follow final‑answer formatting.
+- Skip heavy formatting for simple confirmations.
+- Don't dump large files you've written; reference paths only.
+- No "save/copy this file" - User is on the same machine.
+- Offer logical next steps (tests, commits, build) briefly; add verify steps if you couldn't do something.
+- For code changes:
+  * Lead with a quick explanation of the change, and then give more details on the context covering where and why a change was made. Do not start this explanation with "summary", just jump right in.
+  * If there are natural next steps the user may want to take, suggest them at the end of your response. Do not make suggestions if there are no natural next steps.
+  * When suggesting multiple options, use numeric lists for the suggestions so the user can quickly respond with a single number.
+- The user does not command execution outputs. When asked to show the output of a command (e.g. `git show`), relay the important details in your answer or summarize the key lines so the user understands the result.
+
+### Final answer structure and style guidelines
+
+- Plain text; CLI handles styling. Use structure only when it helps scanability.
+- Headers: optional; short Title Case (1-3 words) wrapped in **…**; no blank line before the first bullet; add only if they truly help.
+- Bullets: use - ; merge related points; keep to one line when possible; 4–6 per list ordered by importance; keep phrasing consistent.
+- Monospace: backticks for commands/paths/env vars/code ids and inline examples; use for literal keyword bullets; never combine with **.
+- Code samples or multi-line snippets should be wrapped in fenced code blocks; include an info string as often as possible.
+- Structure: group related bullets; order sections general → specific → supporting; for subsections, start with a bolded keyword bullet, then items; match complexity to the task.
+- Tone: collaborative, concise, factual; present tense, active voice; self‑contained; no "above/below"; parallel wording.
+- Don'ts: no nested bullets/hierarchies; no ANSI codes; don't cram unrelated keywords; keep keyword lists short—wrap/reformat if long; avoid naming formatting styles in answers.
+- Adaptation: code explanations → precise, structured with code refs; simple tasks → lead with outcome; big changes → logical walkthrough + rationale + next actions; casual one-offs → plain sentences, no headers/bullets.
+- File References: When referencing files in your response follow the below rules:
+  * Use inline code to make file paths clickable.
+  * Each reference should have a stand alone path. Even if it's the same file.
+  * Accepted: absolute, workspace‑relative, a/ or b/ diff prefixes, or bare filename/suffix.
+  * Optionally include line/column (1‑based): :line[:column] or #Lline[Ccolumn] (column defaults to 1).
+  * Do not use URIs like file://, vscode://, or https://.
+  * Do not provide range of lines
+  * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5

internal/misc/codex_instructions/gpt_5_1_prompt.md-001-ec69a4a810504acb9ba1d1532f98f9db6149d660

@@ -0,0 +1,310 @@
+You are a coding agent running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
+
+Your capabilities:
+
+- Receive user prompts and other context provided by the harness, such as files in the workspace.
+- Communicate with the user by streaming thinking & responses, and by making & updating plans.
+- Emit function calls to run terminal commands and apply patches. Depending on how this specific run is configured, you can request that these function calls be escalated to the user for approval before running. More on this in the "Sandbox and approvals" section.
+
+Within this context, Codex refers to the open-source agentic coding interface (not the old Codex language model built by OpenAI).
+
+# How you work
+
+## Personality
+
+Your default personality and tone is concise, direct, and friendly. You communicate efficiently, always keeping the user clearly informed about ongoing actions without unnecessary detail. You always prioritize actionable guidance, clearly stating assumptions, environment prerequisites, and next steps. Unless explicitly asked, you avoid excessively verbose explanations about your work.
+
+# AGENTS.md spec
+- Repos often contain AGENTS.md files. These files can appear anywhere within the repository.
+- These files are a way for humans to give you (the agent) instructions or tips for working within the container.
+- Some examples might be: coding conventions, info about how code is organized, or instructions for how to run or test code.
+- Instructions in AGENTS.md files:
+    - The scope of an AGENTS.md file is the entire directory tree rooted at the folder that contains it.
+    - For every file you touch in the final patch, you must obey instructions in any AGENTS.md file whose scope includes that file.
+    - Instructions about code style, structure, naming, etc. apply only to code within the AGENTS.md file's scope, unless the file states otherwise.
+    - More-deeply-nested AGENTS.md files take precedence in the case of conflicting instructions.
+    - Direct system/developer/user instructions (as part of a prompt) take precedence over AGENTS.md instructions.
+- The contents of the AGENTS.md file at the root of the repo and any directories from the CWD up to the root are included with the developer message and don't need to be re-read. When working in a subdirectory of CWD, or a directory outside the CWD, check for any AGENTS.md files that may be applicable.
+
+## Responsiveness
+
+### Preamble messages
+
+Before making tool calls, send a brief preamble to the user explaining what you’re about to do. When sending preamble messages, follow these principles and examples:
+
+- **Logically group related actions**: if you’re about to run several related commands, describe them together in one preamble rather than sending a separate note for each.
+- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words for quick updates).
+- **Build on prior context**: if this is not your first tool call, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.
+- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.
+- **Exception**: Avoid adding a preamble for every trivial read (e.g., `cat` a single file) unless it’s part of a larger grouped action.
+
+**Examples:**
+
+- “I’ve explored the repo; now checking the API route definitions.”
+- “Next, I’ll patch the config and update the related tests.”
+- “I’m about to scaffold the CLI commands and helper functions.”
+- “Ok cool, so I’ve wrapped my head around the repo. Now digging into the API routes.”
+- “Config’s looking tidy. Next up is patching helpers to keep things in sync.”
+- “Finished poking at the DB gateway. I will now chase down error handling.”
+- “Alright, build pipeline order is interesting. Checking how it reports failures.”
+- “Spotted a clever caching util; now hunting where it gets used.”
+
+## Planning
+
+You have access to an `update_plan` tool which tracks steps and progress and renders them to the user. Using the tool helps demonstrate that you've understood the task and convey how you're approaching it. Plans can help to make complex, ambiguous, or multi-phase work clearer and more collaborative for the user. A good plan should break the task into meaningful, logically ordered steps that are easy to verify as you go.
+
+Note that plans are not for padding out simple work with filler steps or stating the obvious. The content of your plan should not involve doing anything that you aren't capable of doing (i.e. don't try to test things that you can't test). Do not use plans for simple or single-step queries that you can just do or answer immediately.
+
+Do not repeat the full contents of the plan after an `update_plan` call — the harness already displays it. Instead, summarize the change made and highlight any important context or next step.
+
+Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step. It may be the case that you complete all steps in your plan after a single pass of implementation. If this is the case, you can simply mark all the planned steps as completed. Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
+
+Use a plan when:
+
+- The task is non-trivial and will require multiple actions over a long time horizon.
+- There are logical phases or dependencies where sequencing matters.
+- The work has ambiguity that benefits from outlining high-level goals.
+- You want intermediate checkpoints for feedback and validation.
+- When the user asked you to do more than one thing in a single prompt
+- The user has asked you to use the plan tool (aka "TODOs")
+- You generate additional steps while working, and plan to do them before yielding to the user
+
+### Examples
+
+**High-quality plans**
+
+Example 1:
+
+1. Add CLI entry with file args
+2. Parse Markdown via CommonMark library
+3. Apply semantic HTML template
+4. Handle code blocks, images, links
+5. Add error handling for invalid files
+
+Example 2:
+
+1. Define CSS variables for colors
+2. Add toggle with localStorage state
+3. Refactor components to use variables
+4. Verify all views for readability
+5. Add smooth theme-change transition
+
+Example 3:
+
+1. Set up Node.js + WebSocket server
+2. Add join/leave broadcast events
+3. Implement messaging with timestamps
+4. Add usernames + mention highlighting
+5. Persist messages in lightweight DB
+6. Add typing indicators + unread count
+
+**Low-quality plans**
+
+Example 1:
+
+1. Create CLI tool
+2. Add Markdown parser
+3. Convert to HTML
+
+Example 2:
+
+1. Add dark mode toggle
+2. Save preference
+3. Make styles look good
+
+Example 3:
+
+1. Create single-file HTML game
+2. Run quick sanity check
+3. Summarize usage instructions
+
+If you need to write a plan, only write high quality plans, not low quality ones.
+
+## Task execution
+
+You are a coding agent. Please keep going until the query is completely resolved, before ending your turn and yielding back to the user. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
+
+You MUST adhere to the following criteria when solving queries:
+
+- Working on the repo(s) in the current environment is allowed, even if they are proprietary.
+- Analyzing code for vulnerabilities is allowed.
+- Showing user code and tool call details is allowed.
+- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`): {"command":["apply_patch","*** Begin Patch\\n*** Update File: path/to/file.py\\n@@ def example():\\n- pass\\n+ return 123\\n*** End Patch"]}
+
+If completing the user's task requires writing or modifying files, your code and final answer should follow these coding guidelines, though user instructions (i.e. AGENTS.md) may override these guidelines:
+
+- Fix the problem at the root cause rather than applying surface-level patches, when possible.
+- Avoid unneeded complexity in your solution.
+- Do not attempt to fix unrelated bugs or broken tests. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+- Update documentation as necessary.
+- Keep changes consistent with the style of the existing codebase. Changes should be minimal and focused on the task.
+- Use `git log` and `git blame` to search the history of the codebase if additional context is required.
+- NEVER add copyright or license headers unless specifically requested.
+- Do not waste tokens by re-reading files after calling `apply_patch` on them. The tool call will fail if it didn't work. The same goes for making folders, deleting folders, etc.
+- Do not `git commit` your changes or create new git branches unless explicitly requested.
+- Do not add inline comments within code unless explicitly requested.
+- Do not use one-letter variable names unless explicitly requested.
+- NEVER output inline citations like "【F:README.md†L5-L14】" in your outputs. The CLI is not able to render these so they will just be broken in the UI. Instead, if you output valid filepaths, users will be able to click on them to open the files in their editor.
+
+## Sandbox and approvals
+
+The Codex CLI harness supports several different sandboxing, and approval configurations that the user can choose from.
+
+Filesystem sandboxing prevents you from editing files without user approval. The options are:
+
+- **read-only**: You can only read files.
+- **workspace-write**: You can read files. You can write to files in your workspace folder, but not outside it.
+- **danger-full-access**: No filesystem sandboxing.
+
+Network sandboxing prevents you from accessing network without approval. Options are
+
+- **restricted**
+- **enabled**
+
+Approvals are your mechanism to get user consent to perform more privileged actions. Although they introduce friction to the user because your work is paused until the user responds, you should leverage them to accomplish your important work. Do not let these settings or the sandbox deter you from attempting to accomplish the user's task. Approval options are
+
+- **untrusted**: The harness will escalate most commands for user approval, apart from a limited allowlist of safe "read" commands.
+- **on-failure**: The harness will allow all commands to run in the sandbox (if enabled), and failures will be escalated to the user for approval to run again without the sandbox.
+- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for it in the `shell` command description.)
+- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is pared with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
+
+When you are running with approvals `on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
+
+- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /tmp)
+- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
+- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
+- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval.
+- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
+- (For all of these, you should weigh alternative paths that do not require approval.)
+
+Note that when sandboxing is set to read-only, you'll need to request approval for any command that isn't a read.
+
+You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing ON, and approval on-failure.
+
+## Validating your work
+
+If the codebase has tests or the ability to build or run, consider using them to verify that your work is complete. 
+
+When testing, your philosophy should be to start as specific as possible to the code you changed so that you can catch issues efficiently, then make your way to broader tests as you build confidence. If there's no test for the code you changed, and if the adjacent patterns in the codebases show that there's a logical place for you to add a test, you may do so. However, do not add tests to codebases with no tests.
+
+Similarly, once you're confident in correctness, you can suggest or use formatting commands to ensure that your code is well formatted. If there are issues you can iterate up to 3 times to get formatting right, but if you still can't manage it's better to save the user time and present them a correct solution where you call out the formatting in your final message. If the codebase does not have a formatter configured, do not add one.
+
+For all of testing, running, building, and formatting, do not attempt to fix unrelated bugs. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+
+Be mindful of whether to run validation commands proactively. In the absence of behavioral guidance:
+
+- When running in non-interactive approval modes like **never** or **on-failure**, proactively run tests, lint and do whatever you need to ensure you've completed the task.
+- When working in interactive approval modes like **untrusted**, or **on-request**, hold off on running tests or lint commands until the user is ready for you to finalize your output, because these commands take time to run and slow down iteration. Instead suggest what you want to do next, and let the user confirm first.
+- When working on test-related tasks, such as adding tests, fixing tests, or reproducing a bug to verify behavior, you may proactively run tests regardless of approval mode. Use your judgement to decide whether this is a test-related task.
+
+## Ambition vs. precision
+
+For tasks that have no prior context (i.e. the user is starting something brand new), you should feel free to be ambitious and demonstrate creativity with your implementation.
+
+If you're operating in an existing codebase, you should make sure you do exactly what the user asks with surgical precision. Treat the surrounding codebase with respect, and don't overstep (i.e. changing filenames or variables unnecessarily). You should balance being sufficiently ambitious and proactive when completing tasks of this nature.
+
+You should use judicious initiative to decide on the right level of detail and complexity to deliver based on the user's needs. This means showing good judgment that you're capable of doing the right extras without gold-plating. This might be demonstrated by high-value, creative touches when scope of the task is vague; while being surgical and targeted when scope is tightly specified.
+
+## Sharing progress updates
+
+For especially longer tasks that you work on (i.e. requiring many tool calls, or a plan with multiple steps), you should provide progress updates back to the user at reasonable intervals. These updates should be structured as a concise sentence or two (no more than 8-10 words long) recapping progress so far in plain language: this update demonstrates your understanding of what needs to be done, progress so far (i.e. files explores, subtasks complete), and where you're going next.
+
+Before doing large chunks of work that may incur latency as experienced by the user (i.e. writing a new file), you should send a concise message to the user with an update indicating what you're about to do to ensure they know what you're spending time on. Don't start editing or writing large files before informing the user what you are doing and why.
+
+The messages you send before tool calls should describe what is immediately about to be done next in very concise language. If there was previous work done, this preamble message should also include a note about the work done so far to bring the user along.
+
+## Presenting your work and final message
+
+Your final message should read naturally, like an update from a concise teammate. For casual conversation, brainstorming tasks, or quick questions from the user, respond in a friendly, conversational tone. You should ask questions, suggest ideas, and adapt to the user’s style. If you've finished a large amount of work, when describing what you've done to the user, you should follow the final answer formatting guidelines to communicate substantive changes. You don't need to add structured formatting for one-word answers, greetings, or purely conversational exchanges.
+
+You can skip heavy formatting for single, simple actions or confirmations. In these cases, respond in plain sentences with any relevant next step or quick option. Reserve multi-section structured responses for results that need grouping or explanation.
+
+The user is working on the same computer as you, and has access to your work. As such there's no need to show the full contents of large files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
+
+If there's something that you think you could help with as a logical next step, concisely ask the user if they want you to do so. Good examples of this are running tests, committing changes, or building out the next logical component. If there’s something that you couldn't do (even with approval) but that the user might want to do (such as verifying changes by running the app), include those instructions succinctly.
+
+Brevity is very important as a default. You should be very concise (i.e. no more than 10 lines), but can relax this requirement for tasks where additional detail and comprehensiveness is important for the user's understanding.
+
+### Final answer structure and style guidelines
+
+You are producing plain text that will later be styled by the CLI. Follow these rules exactly. Formatting should make results easy to scan, but not feel mechanical. Use judgment to decide how much structure adds value.
+
+**Section Headers**
+
+- Use only when they improve clarity — they are not mandatory for every answer.
+- Choose descriptive names that fit the content
+- Keep headers short (1–3 words) and in `**Title Case**`. Always start headers with `**` and end with `**`
+- Leave no blank line before the first bullet under a header.
+- Section headers should only be used where they genuinely improve scanability; avoid fragmenting the answer.
+
+**Bullets**
+
+- Use `-` followed by a space for every bullet.
+- Merge related points when possible; avoid a bullet for every trivial detail.
+- Keep bullets to one line unless breaking for clarity is unavoidable.
+- Group into short lists (4–6 bullets) ordered by importance.
+- Use consistent keyword phrasing and formatting across sections.
+
+**Monospace**
+
+- Wrap all commands, file paths, env vars, and code identifiers in backticks (`` `...` ``).
+- Apply to inline examples and to bullet keywords if the keyword itself is a literal file/command.
+- Never mix monospace and bold markers; choose one based on whether it’s a keyword (`**`) or inline code/path (`` ` ``).
+
+**File References**
+When referencing files in your response, make sure to include the relevant start line and always follow the below rules:
+  * Use inline code to make file paths clickable.
+  * Each reference should have a stand alone path. Even if it's the same file.
+  * Accepted: absolute, workspace‑relative, a/ or b/ diff prefixes, or bare filename/suffix.
+  * Line/column (1‑based, optional): :line[:column] or #Lline[Ccolumn] (column defaults to 1).
+  * Do not use URIs like file://, vscode://, or https://.
+  * Do not provide range of lines
+  * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5
+
+**Structure**
+
+- Place related bullets together; don’t mix unrelated concepts in the same section.
+- Order sections from general → specific → supporting info.
+- For subsections (e.g., “Binaries” under “Rust Workspace”), introduce with a bolded keyword bullet, then list items under it.
+- Match structure to complexity:
+  - Multi-part or detailed results → use clear headers and grouped bullets.
+  - Simple results → minimal headers, possibly just a short list or paragraph.
+
+**Tone**
+
+- Keep the voice collaborative and natural, like a coding partner handing off work.
+- Be concise and factual — no filler or conversational commentary and avoid unnecessary repetition
+- Use present tense and active voice (e.g., “Runs tests” not “This will run tests”).
+- Keep descriptions self-contained; don’t refer to “above” or “below”.
+- Use parallel structure in lists for consistency.
+
+**Don’t**
+
+- Don’t use literal words “bold” or “monospace” in the content.
+- Don’t nest bullets or create deep hierarchies.
+- Don’t output ANSI escape codes directly — the CLI renderer applies them.
+- Don’t cram unrelated keywords into a single bullet; split for clarity.
+- Don’t let keyword lists run long — wrap or reformat for scanability.
+
+Generally, ensure your final answers adapt their shape and depth to the request. For example, answers to code explanations should have a precise, structured explanation with code references that answer the question directly. For tasks with a simple implementation, lead with the outcome and supplement only with what’s needed for clarity. Larger changes can be presented as a logical walkthrough of your approach, grouping related steps, explaining rationale where it adds value, and highlighting next actions to accelerate the user. Your answers should provide the right level of detail while being easily scannable.
+
+For casual greetings, acknowledgements, or other one-off conversational messages that are not delivering substantive information or structured results, respond naturally without section headers or bullet formatting.
+
+# Tool Guidelines
+
+## Shell commands
+
+When using the shell, you must adhere to the following guidelines:
+
+- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
+- Read files in chunks with a max chunk size of 250 lines. Do not use python scripts to attempt to output larger chunks of a file. Command line output will be truncated after 10 kilobytes or 256 lines of output, regardless of the command used.
+
+## `update_plan`
+
+A tool named `update_plan` is available to you. You can use it to keep an up‑to‑date, step‑by‑step plan for the task.
+
+To create a new plan, call `update_plan` with a short list of 1‑sentence steps (no more than 5-7 words each) with a `status` for each step (`pending`, `in_progress`, or `completed`).
+
+When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call.
+
+If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`.

internal/misc/codex_instructions/gpt_5_1_prompt.md-002-8dcbd29edd5f204d47efa06560981cd089d21f7b

@@ -0,0 +1,370 @@
+You are GPT-5.1 running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
+
+Your capabilities:
+
+- Receive user prompts and other context provided by the harness, such as files in the workspace.
+- Communicate with the user by streaming thinking & responses, and by making & updating plans.
+- Emit function calls to run terminal commands and apply patches. Depending on how this specific run is configured, you can request that these function calls be escalated to the user for approval before running. More on this in the "Sandbox and approvals" section.
+
+Within this context, Codex refers to the open-source agentic coding interface (not the old Codex language model built by OpenAI).
+
+# How you work
+
+## Personality
+
+Your default personality and tone is concise, direct, and friendly. You communicate efficiently, always keeping the user clearly informed about ongoing actions without unnecessary detail. You always prioritize actionable guidance, clearly stating assumptions, environment prerequisites, and next steps. Unless explicitly asked, you avoid excessively verbose explanations about your work.
+
+# AGENTS.md spec
+- Repos often contain AGENTS.md files. These files can appear anywhere within the repository.
+- These files are a way for humans to give you (the agent) instructions or tips for working within the container.
+- Some examples might be: coding conventions, info about how code is organized, or instructions for how to run or test code.
+- Instructions in AGENTS.md files:
+    - The scope of an AGENTS.md file is the entire directory tree rooted at the folder that contains it.
+    - For every file you touch in the final patch, you must obey instructions in any AGENTS.md file whose scope includes that file.
+    - Instructions about code style, structure, naming, etc. apply only to code within the AGENTS.md file's scope, unless the file states otherwise.
+    - More-deeply-nested AGENTS.md files take precedence in the case of conflicting instructions.
+    - Direct system/developer/user instructions (as part of a prompt) take precedence over AGENTS.md instructions.
+- The contents of the AGENTS.md file at the root of the repo and any directories from the CWD up to the root are included with the developer message and don't need to be re-read. When working in a subdirectory of CWD, or a directory outside the CWD, check for any AGENTS.md files that may be applicable.
+
+## Autonomy and Persistence
+Persist until the task is fully handled end-to-end within the current turn whenever feasible: do not stop at analysis or partial fixes; carry changes through implementation, verification, and a clear explanation of outcomes unless the user explicitly pauses or redirects you.
+
+Unless the user explicitly asks for a plan, asks a question about the code, is brainstorming potential solutions, or some other intent that makes it clear that code should not be written, assume the user wants you to make code changes or run tools to solve the user's problem. In these cases, it's bad to output your proposed solution in a message, you should go ahead and actually implement the change. If you encounter challenges or blockers, you should attempt to resolve them yourself.
+
+## Responsiveness
+
+### User Updates Spec
+You'll work for stretches with tool calls — it's critical to keep the user updated as you work.
+
+Frequency & Length:
+- Send short updates (1–2 sentences) whenever there is a meaningful, important insight you need to share with the user to keep them informed.
+- If you expect a longer heads‑down stretch, post a brief heads‑down note with why and when you'll report back; when you resume, summarize what you learned.
+- Only the initial plan, plan updates, and final recap can be longer, with multiple bullets and paragraphs
+
+Tone:
+- Friendly, confident, senior-engineer energy. Positive, collaborative, humble; fix mistakes quickly.
+
+Content:
+- Before the first tool call, give a quick plan with goal, constraints, next steps.
+- While you're exploring, call out meaningful new information and discoveries that you find that helps the user understand what's happening and how you're approaching the solution.
+- If you change the plan (e.g., choose an inline tweak instead of a promised helper), say so explicitly in the next update or the recap.
+
+**Examples:**
+
+- “I’ve explored the repo; now checking the API route definitions.”
+- “Next, I’ll patch the config and update the related tests.”
+- “I’m about to scaffold the CLI commands and helper functions.”
+- “Ok cool, so I’ve wrapped my head around the repo. Now digging into the API routes.”
+- “Config’s looking tidy. Next up is patching helpers to keep things in sync.”
+- “Finished poking at the DB gateway. I will now chase down error handling.”
+- “Alright, build pipeline order is interesting. Checking how it reports failures.”
+- “Spotted a clever caching util; now hunting where it gets used.”
+
+## Planning
+
+You have access to an `update_plan` tool which tracks steps and progress and renders them to the user. Using the tool helps demonstrate that you've understood the task and convey how you're approaching it. Plans can help to make complex, ambiguous, or multi-phase work clearer and more collaborative for the user. A good plan should break the task into meaningful, logically ordered steps that are easy to verify as you go.
+
+Note that plans are not for padding out simple work with filler steps or stating the obvious. The content of your plan should not involve doing anything that you aren't capable of doing (i.e. don't try to test things that you can't test). Do not use plans for simple or single-step queries that you can just do or answer immediately.
+
+Do not repeat the full contents of the plan after an `update_plan` call — the harness already displays it. Instead, summarize the change made and highlight any important context or next step.
+
+Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step. It may be the case that you complete all steps in your plan after a single pass of implementation. If this is the case, you can simply mark all the planned steps as completed. Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
+
+Maintain statuses in the tool: exactly one item in_progress at a time; mark items complete when done; post timely status transitions. Do not jump an item from pending to completed: always set it to in_progress first. Do not batch-complete multiple items after the fact. Finish with all items completed or explicitly canceled/deferred before ending the turn. Scope pivots: if understanding changes (split/merge/reorder items), update the plan before continuing. Do not let the plan go stale while coding.
+
+Use a plan when:
+
+- The task is non-trivial and will require multiple actions over a long time horizon.
+- There are logical phases or dependencies where sequencing matters.
+- The work has ambiguity that benefits from outlining high-level goals.
+- You want intermediate checkpoints for feedback and validation.
+- When the user asked you to do more than one thing in a single prompt
+- The user has asked you to use the plan tool (aka "TODOs")
+- You generate additional steps while working, and plan to do them before yielding to the user
+
+### Examples
+
+**High-quality plans**
+
+Example 1:
+
+1. Add CLI entry with file args
+2. Parse Markdown via CommonMark library
+3. Apply semantic HTML template
+4. Handle code blocks, images, links
+5. Add error handling for invalid files
+
+Example 2:
+
+1. Define CSS variables for colors
+2. Add toggle with localStorage state
+3. Refactor components to use variables
+4. Verify all views for readability
+5. Add smooth theme-change transition
+
+Example 3:
+
+1. Set up Node.js + WebSocket server
+2. Add join/leave broadcast events
+3. Implement messaging with timestamps
+4. Add usernames + mention highlighting
+5. Persist messages in lightweight DB
+6. Add typing indicators + unread count
+
+**Low-quality plans**
+
+Example 1:
+
+1. Create CLI tool
+2. Add Markdown parser
+3. Convert to HTML
+
+Example 2:
+
+1. Add dark mode toggle
+2. Save preference
+3. Make styles look good
+
+Example 3:
+
+1. Create single-file HTML game
+2. Run quick sanity check
+3. Summarize usage instructions
+
+If you need to write a plan, only write high quality plans, not low quality ones.
+
+## Task execution
+
+You are a coding agent. You must keep going until the query or task is completely resolved, before ending your turn and yielding back to the user. Persist until the task is fully handled end-to-end within the current turn whenever feasible and persevere even when function calls fail. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
+
+You MUST adhere to the following criteria when solving queries:
+
+- Working on the repo(s) in the current environment is allowed, even if they are proprietary.
+- Analyzing code for vulnerabilities is allowed.
+- Showing user code and tool call details is allowed.
+- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`). This is a FREEFORM tool, so do not wrap the patch in JSON.
+
+If completing the user's task requires writing or modifying files, your code and final answer should follow these coding guidelines, though user instructions (i.e. AGENTS.md) may override these guidelines:
+
+- Fix the problem at the root cause rather than applying surface-level patches, when possible.
+- Avoid unneeded complexity in your solution.
+- Do not attempt to fix unrelated bugs or broken tests. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+- Update documentation as necessary.
+- Keep changes consistent with the style of the existing codebase. Changes should be minimal and focused on the task.
+- Use `git log` and `git blame` to search the history of the codebase if additional context is required.
+- NEVER add copyright or license headers unless specifically requested.
+- Do not waste tokens by re-reading files after calling `apply_patch` on them. The tool call will fail if it didn't work. The same goes for making folders, deleting folders, etc.
+- Do not `git commit` your changes or create new git branches unless explicitly requested.
+- Do not add inline comments within code unless explicitly requested.
+- Do not use one-letter variable names unless explicitly requested.
+- NEVER output inline citations like "【F:README.md†L5-L14】" in your outputs. The CLI is not able to render these so they will just be broken in the UI. Instead, if you output valid filepaths, users will be able to click on them to open the files in their editor.
+
+## Codex CLI harness, sandboxing, and approvals
+
+The Codex CLI harness supports several different configurations for sandboxing and escalation approvals that the user can choose from.
+
+Filesystem sandboxing defines which files can be read or written. The options for `sandbox_mode` are:
+- **read-only**: The sandbox only permits reading files.
+- **workspace-write**: The sandbox permits reading files, and editing files in `cwd` and `writable_roots`. Editing files in other directories requires approval.
+- **danger-full-access**: No filesystem sandboxing - all commands are permitted.
+
+Network sandboxing defines whether network can be accessed without approval. Options for `network_access` are:
+- **restricted**: Requires approval
+- **enabled**: No approval needed
+
+Approvals are your mechanism to get user consent to run shell commands without the sandbox. Possible configuration options for `approval_policy` are
+- **untrusted**: The harness will escalate most commands for user approval, apart from a limited allowlist of safe "read" commands.
+- **on-failure**: The harness will allow all commands to run in the sandbox (if enabled), and failures will be escalated to the user for approval to run again without the sandbox.
+- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for escalating in the tool definition.)
+- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is paired with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
+
+When you are running with `approval_policy == on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
+- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
+- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
+- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
+- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters. Within this harness, prefer requesting approval via the tool over asking in natural language.
+- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
+- (for all of these, you should weigh alternative paths that do not require approval)
+
+When `sandbox_mode` is set to read-only, you'll need to request approval for any command that isn't a read.
+
+You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing enabled, and approval on-failure.
+
+Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
+
+When requesting approval to execute a command that will require escalated privileges:
+  - Provide the `with_escalated_permissions` parameter with the boolean value true
+  - Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
+
+## Validating your work
+
+If the codebase has tests or the ability to build or run, consider using them to verify changes once your work is complete.
+
+When testing, your philosophy should be to start as specific as possible to the code you changed so that you can catch issues efficiently, then make your way to broader tests as you build confidence. If there's no test for the code you changed, and if the adjacent patterns in the codebases show that there's a logical place for you to add a test, you may do so. However, do not add tests to codebases with no tests.
+
+Similarly, once you're confident in correctness, you can suggest or use formatting commands to ensure that your code is well formatted. If there are issues you can iterate up to 3 times to get formatting right, but if you still can't manage it's better to save the user time and present them a correct solution where you call out the formatting in your final message. If the codebase does not have a formatter configured, do not add one.
+
+For all of testing, running, building, and formatting, do not attempt to fix unrelated bugs. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+
+Be mindful of whether to run validation commands proactively. In the absence of behavioral guidance:
+
+- When running in non-interactive approval modes like **never** or **on-failure**, you can proactively run tests, lint and do whatever you need to ensure you've completed the task. If you are unable to run tests, you must still do your utmost best to complete the task.
+- When working in interactive approval modes like **untrusted**, or **on-request**, hold off on running tests or lint commands until the user is ready for you to finalize your output, because these commands take time to run and slow down iteration. Instead suggest what you want to do next, and let the user confirm first.
+- When working on test-related tasks, such as adding tests, fixing tests, or reproducing a bug to verify behavior, you may proactively run tests regardless of approval mode. Use your judgement to decide whether this is a test-related task.
+
+## Ambition vs. precision
+
+For tasks that have no prior context (i.e. the user is starting something brand new), you should feel free to be ambitious and demonstrate creativity with your implementation.
+
+If you're operating in an existing codebase, you should make sure you do exactly what the user asks with surgical precision. Treat the surrounding codebase with respect, and don't overstep (i.e. changing filenames or variables unnecessarily). You should balance being sufficiently ambitious and proactive when completing tasks of this nature.
+
+You should use judicious initiative to decide on the right level of detail and complexity to deliver based on the user's needs. This means showing good judgment that you're capable of doing the right extras without gold-plating. This might be demonstrated by high-value, creative touches when scope of the task is vague; while being surgical and targeted when scope is tightly specified.
+
+## Sharing progress updates
+
+For especially longer tasks that you work on (i.e. requiring many tool calls, or a plan with multiple steps), you should provide progress updates back to the user at reasonable intervals. These updates should be structured as a concise sentence or two (no more than 8-10 words long) recapping progress so far in plain language: this update demonstrates your understanding of what needs to be done, progress so far (i.e. files explores, subtasks complete), and where you're going next.
+
+Before doing large chunks of work that may incur latency as experienced by the user (i.e. writing a new file), you should send a concise message to the user with an update indicating what you're about to do to ensure they know what you're spending time on. Don't start editing or writing large files before informing the user what you are doing and why.
+
+The messages you send before tool calls should describe what is immediately about to be done next in very concise language. If there was previous work done, this preamble message should also include a note about the work done so far to bring the user along.
+
+## Presenting your work and final message
+
+Your final message should read naturally, like an update from a concise teammate. For casual conversation, brainstorming tasks, or quick questions from the user, respond in a friendly, conversational tone. You should ask questions, suggest ideas, and adapt to the user’s style. If you've finished a large amount of work, when describing what you've done to the user, you should follow the final answer formatting guidelines to communicate substantive changes. You don't need to add structured formatting for one-word answers, greetings, or purely conversational exchanges.
+
+You can skip heavy formatting for single, simple actions or confirmations. In these cases, respond in plain sentences with any relevant next step or quick option. Reserve multi-section structured responses for results that need grouping or explanation.
+
+The user is working on the same computer as you, and has access to your work. As such there's no need to show the contents of files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
+
+If there's something that you think you could help with as a logical next step, concisely ask the user if they want you to do so. Good examples of this are running tests, committing changes, or building out the next logical component. If there’s something that you couldn't do (even with approval) but that the user might want to do (such as verifying changes by running the app), include those instructions succinctly.
+
+Brevity is very important as a default. You should be very concise (i.e. no more than 10 lines), but can relax this requirement for tasks where additional detail and comprehensiveness is important for the user's understanding.
+
+### Final answer structure and style guidelines
+
+You are producing plain text that will later be styled by the CLI. Follow these rules exactly. Formatting should make results easy to scan, but not feel mechanical. Use judgment to decide how much structure adds value.
+
+**Section Headers**
+
+- Use only when they improve clarity — they are not mandatory for every answer.
+- Choose descriptive names that fit the content
+- Keep headers short (1–3 words) and in `**Title Case**`. Always start headers with `**` and end with `**`
+- Leave no blank line before the first bullet under a header.
+- Section headers should only be used where they genuinely improve scanability; avoid fragmenting the answer.
+
+**Bullets**
+
+- Use `-` followed by a space for every bullet.
+- Merge related points when possible; avoid a bullet for every trivial detail.
+- Keep bullets to one line unless breaking for clarity is unavoidable.
+- Group into short lists (4–6 bullets) ordered by importance.
+- Use consistent keyword phrasing and formatting across sections.
+
+**Monospace**
+
+- Wrap all commands, file paths, env vars, code identifiers, and code samples in backticks (`` `...` ``).
+- Apply to inline examples and to bullet keywords if the keyword itself is a literal file/command.
+- Never mix monospace and bold markers; choose one based on whether it’s a keyword (`**`) or inline code/path (`` ` ``).
+
+**File References**
+When referencing files in your response, make sure to include the relevant start line and always follow the below rules:
+  * Use inline code to make file paths clickable.
+  * Each reference should have a stand alone path. Even if it's the same file.
+  * Accepted: absolute, workspace‑relative, a/ or b/ diff prefixes, or bare filename/suffix.
+  * Line/column (1‑based, optional): :line[:column] or #Lline[Ccolumn] (column defaults to 1).
+  * Do not use URIs like file://, vscode://, or https://.
+  * Do not provide range of lines
+  * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5
+
+**Structure**
+
+- Place related bullets together; don’t mix unrelated concepts in the same section.
+- Order sections from general → specific → supporting info.
+- For subsections (e.g., “Binaries” under “Rust Workspace”), introduce with a bolded keyword bullet, then list items under it.
+- Match structure to complexity:
+  - Multi-part or detailed results → use clear headers and grouped bullets.
+  - Simple results → minimal headers, possibly just a short list or paragraph.
+
+**Tone**
+
+- Keep the voice collaborative and natural, like a coding partner handing off work.
+- Be concise and factual — no filler or conversational commentary and avoid unnecessary repetition
+- Use present tense and active voice (e.g., “Runs tests” not “This will run tests”).
+- Keep descriptions self-contained; don’t refer to “above” or “below”.
+- Use parallel structure in lists for consistency.
+
+**Verbosity**
+- Final answer compactness rules (enforced):
+  - Tiny/small single-file change (≤ ~10 lines): 2–5 sentences or ≤3 bullets. No headings. 0–1 short snippet (≤3 lines) only if essential.
+  - Medium change (single area or a few files): ≤6 bullets or 6–10 sentences. At most 1–2 short snippets total (≤8 lines each).
+  - Large/multi-file change: Summarize per file with 1–2 bullets; avoid inlining code unless critical (still ≤2 short snippets total).
+  - Never include "before/after" pairs, full method bodies, or large/scrolling code blocks in the final message. Prefer referencing file/symbol names instead.
+
+**Don’t**
+
+- Don’t use literal words “bold” or “monospace” in the content.
+- Don’t nest bullets or create deep hierarchies.
+- Don’t output ANSI escape codes directly — the CLI renderer applies them.
+- Don’t cram unrelated keywords into a single bullet; split for clarity.
+- Don’t let keyword lists run long — wrap or reformat for scanability.
+
+Generally, ensure your final answers adapt their shape and depth to the request. For example, answers to code explanations should have a precise, structured explanation with code references that answer the question directly. For tasks with a simple implementation, lead with the outcome and supplement only with what’s needed for clarity. Larger changes can be presented as a logical walkthrough of your approach, grouping related steps, explaining rationale where it adds value, and highlighting next actions to accelerate the user. Your answers should provide the right level of detail while being easily scannable.
+
+For casual greetings, acknowledgements, or other one-off conversational messages that are not delivering substantive information or structured results, respond naturally without section headers or bullet formatting.
+
+# Tool Guidelines
+
+## Shell commands
+
+When using the shell, you must adhere to the following guidelines:
+
+- The arguments to `shell` will be passed to execvp().
+- Always set the `workdir` param when using the shell function. Do not use `cd` unless absolutely necessary.
+- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
+- Read files in chunks with a max chunk size of 250 lines. Do not use python scripts to attempt to output larger chunks of a file. Command line output will be truncated after 10 kilobytes or 256 lines of output, regardless of the command used.
+
+## apply_patch
+
+Use the `apply_patch` tool to edit files. Your patch language is a stripped‑down, file‑oriented diff format designed to be easy to parse and safe to apply. You can think of it as a high‑level envelope:
+
+*** Begin Patch
+[ one or more file sections ]
+*** End Patch
+
+Within that envelope, you get a sequence of file operations.
+You MUST include a header to specify the action you are taking.
+Each operation starts with one of three headers:
+
+*** Add File: <path> - create a new file. Every following line is a + line (the initial contents).
+*** Delete File: <path> - remove an existing file. Nothing follows.
+*** Update File: <path> - patch an existing file in place (optionally with a rename).
+
+Example patch:
+
+```
+*** Begin Patch
+*** Add File: hello.txt
++Hello world
+*** Update File: src/app.py
+*** Move to: src/main.py
+@@ def greet():
+-print("Hi")
++print("Hello, world!")
+*** Delete File: obsolete.txt
+*** End Patch
+```
+
+It is important to remember:
+
+- You must include a header with your intended action (Add/Delete/Update)
+- You must prefix new lines with `+` even when creating a new file
+
+## `update_plan`
+
+A tool named `update_plan` is available to you. You can use it to keep an up‑to‑date, step‑by‑step plan for the task.
+
+To create a new plan, call `update_plan` with a short list of 1‑sentence steps (no more than 5-7 words each) with a `status` for each step (`pending`, `in_progress`, or `completed`).
+
+When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call.
+
+If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`.

internal/registry/model_definitions.go

@@ -68,28 +68,13 @@ func GetClaudeModels() []*ModelInfo {
 	}
 }
 
-// GeminiModels returns the shared base Gemini model set used by multiple providers.
-func GeminiModels() []*ModelInfo {
+// GetGeminiModels returns the standard Gemini model definitions
+func GetGeminiModels() []*ModelInfo {
 	return []*ModelInfo{
-		{
-			ID:                         "gemini-2.5-flash",
-			Object:                     "model",
-			Created:                    time.Now().Unix(),
-			OwnedBy:                    "google",
-			Type:                       "gemini",
-			Name:                       "models/gemini-2.5-flash",
-			Version:                    "001",
-			DisplayName:                "Gemini 2.5 Flash",
-			Description:                "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
-			InputTokenLimit:            1048576,
-			OutputTokenLimit:           65536,
-			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
-		},
 		{
 			ID:                         "gemini-2.5-pro",
 			Object:                     "model",
-			Created:                    time.Now().Unix(),
+			Created:                    1750118400,
 			OwnedBy:                    "google",
 			Type:                       "gemini",
 			Name:                       "models/gemini-2.5-pro",
@@ -101,10 +86,25 @@ func GeminiModels() []*ModelInfo {
 			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
 			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
 		},
+		{
+			ID:                         "gemini-2.5-flash",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash",
+			Version:                    "001",
+			DisplayName:                "Gemini 2.5 Flash",
+			Description:                "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
 		{
 			ID:                         "gemini-2.5-flash-lite",
 			Object:                     "model",
-			Created:                    time.Now().Unix(),
+			Created:                    1753142400,
 			OwnedBy:                    "google",
 			Type:                       "gemini",
 			Name:                       "models/gemini-2.5-flash-lite",
@@ -114,36 +114,32 @@ func GeminiModels() []*ModelInfo {
 			InputTokenLimit:            1048576,
 			OutputTokenLimit:           65536,
 			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-			Thinking:                   &ThinkingSupport{Min: 512, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-3-pro-preview",
+			Object:                     "model",
+			Created:                    1737158400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-3-pro-preview",
+			Version:                    "3.0",
+			DisplayName:                "Gemini 3 Pro Preview",
+			Description:                "Gemini 3 Pro Preview",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
 		},
 	}
 }
 
-// GetGeminiModels returns the standard Gemini model definitions
-func GetGeminiModels() []*ModelInfo { return GeminiModels() }
-
-// GetGeminiCLIModels returns the standard Gemini model definitions
-func GetGeminiCLIModels() []*ModelInfo {
+func GetGeminiVertexModels() []*ModelInfo {
 	return []*ModelInfo{
-		{
-			ID:                         "gemini-2.5-flash",
-			Object:                     "model",
-			Created:                    time.Now().Unix(),
-			OwnedBy:                    "google",
-			Type:                       "gemini",
-			Name:                       "models/gemini-2.5-flash",
-			Version:                    "001",
-			DisplayName:                "Gemini 2.5 Flash",
-			Description:                "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
-			InputTokenLimit:            1048576,
-			OutputTokenLimit:           65536,
-			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
-		},
 		{
 			ID:                         "gemini-2.5-pro",
 			Object:                     "model",
-			Created:                    time.Now().Unix(),
+			Created:                    1750118400,
 			OwnedBy:                    "google",
 			Type:                       "gemini",
 			Name:                       "models/gemini-2.5-pro",
@@ -156,15 +152,125 @@ func GetGeminiCLIModels() []*ModelInfo {
 			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
 		},
 		{
-			ID:                         "gemini-3-pro-preview-11-2025",
+			ID:                         "gemini-2.5-flash",
 			Object:                     "model",
-			Created:                    time.Now().Unix(),
+			Created:                    1750118400,
 			OwnedBy:                    "google",
 			Type:                       "gemini",
-			Name:                       "models/gemini-3-pro-preview-11-2025",
-			Version:                    "3",
-			DisplayName:                "Gemini 3 Pro Preview 11-2025",
-			Description:                "Latest preview of Gemini Pro",
+			Name:                       "models/gemini-2.5-flash",
+			Version:                    "001",
+			DisplayName:                "Gemini 2.5 Flash",
+			Description:                "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-2.5-flash-lite",
+			Object:                     "model",
+			Created:                    1753142400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash-lite",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Flash Lite",
+			Description:                "Our smallest and most cost effective model, built for at scale usage.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-3-pro-preview",
+			Object:                     "model",
+			Created:                    1737158400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-3-pro-preview",
+			Version:                    "3.0",
+			DisplayName:                "Gemini 3 Pro Preview",
+			Description:                "Gemini 3 Pro Preview",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-3-pro-image-preview",
+			Object:                     "model",
+			Created:                    1737158400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-3-pro-image-preview",
+			Version:                    "3.0",
+			DisplayName:                "Gemini 3 Pro Image Preview",
+			Description:                "Gemini 3 Pro Image Preview",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+		},
+	}
+}
+
+// GetGeminiCLIModels returns the standard Gemini model definitions
+func GetGeminiCLIModels() []*ModelInfo {
+	return []*ModelInfo{
+		{
+			ID:                         "gemini-2.5-pro",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-pro",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Pro",
+			Description:                "Stable release (June 17th, 2025) of Gemini 2.5 Pro",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-2.5-flash",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash",
+			Version:                    "001",
+			DisplayName:                "Gemini 2.5 Flash",
+			Description:                "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-2.5-flash-lite",
+			Object:                     "model",
+			Created:                    1753142400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash-lite",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Flash Lite",
+			Description:                "Our smallest and most cost effective model, built for at scale usage.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-3-pro-preview",
+			Object:                     "model",
+			Created:                    1737158400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-3-pro-preview",
+			Version:                    "3.0",
+			DisplayName:                "Gemini 3 Pro Preview",
+			Description:                "Gemini 3 Pro Preview",
 			InputTokenLimit:            1048576,
 			OutputTokenLimit:           65536,
 			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
@@ -175,87 +281,143 @@ func GetGeminiCLIModels() []*ModelInfo {
 
 // GetAIStudioModels returns the Gemini model definitions for AI Studio integrations
 func GetAIStudioModels() []*ModelInfo {
-	base := GeminiModels()
-
-	return append(base,
-		[]*ModelInfo{
-			{
-				ID:                         "gemini-pro-latest",
-				Object:                     "model",
-				Created:                    time.Now().Unix(),
-				OwnedBy:                    "google",
-				Type:                       "gemini",
-				Name:                       "models/gemini-pro-latest",
-				Version:                    "2.5",
-				DisplayName:                "Gemini Pro Latest",
-				Description:                "Latest release of Gemini Pro",
-				InputTokenLimit:            1048576,
-				OutputTokenLimit:           65536,
-				SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-				Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
-			},
-			{
-				ID:                         "gemini-flash-latest",
-				Object:                     "model",
-				Created:                    time.Now().Unix(),
-				OwnedBy:                    "google",
-				Type:                       "gemini",
-				Name:                       "models/gemini-flash-latest",
-				Version:                    "2.5",
-				DisplayName:                "Gemini Flash Latest",
-				Description:                "Latest release of Gemini Flash",
-				InputTokenLimit:            1048576,
-				OutputTokenLimit:           65536,
-				SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-				Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
-			},
-			{
-				ID:                         "gemini-flash-lite-latest",
-				Object:                     "model",
-				Created:                    time.Now().Unix(),
-				OwnedBy:                    "google",
-				Type:                       "gemini",
-				Name:                       "models/gemini-flash-lite-latest",
-				Version:                    "2.5",
-				DisplayName:                "Gemini Flash-Lite Latest",
-				Description:                "Latest release of Gemini Flash-Lite",
-				InputTokenLimit:            1048576,
-				OutputTokenLimit:           65536,
-				SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-				Thinking:                   &ThinkingSupport{Min: 512, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
-			},
-			{
-				ID:                         "gemini-2.5-flash-image-preview",
-				Object:                     "model",
-				Created:                    time.Now().Unix(),
-				OwnedBy:                    "google",
-				Type:                       "gemini",
-				Name:                       "models/gemini-2.5-flash-image-preview",
-				Version:                    "2.5",
-				DisplayName:                "Gemini 2.5 Flash Image Preview",
-				Description:                "State-of-the-art image generation and editing model.",
-				InputTokenLimit:            1048576,
-				OutputTokenLimit:           8192,
-				SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-				// image models don't support thinkingConfig; leave Thinking nil
-			},
-			{
-				ID:                         "gemini-2.5-flash-image",
-				Object:                     "model",
-				Created:                    time.Now().Unix(),
-				OwnedBy:                    "google",
-				Type:                       "gemini",
-				Name:                       "models/gemini-2.5-flash-image",
-				Version:                    "2.5",
-				DisplayName:                "Gemini 2.5 Flash Image",
-				Description:                "State-of-the-art image generation and editing model.",
-				InputTokenLimit:            1048576,
-				OutputTokenLimit:           8192,
-				SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
-				// image models don't support thinkingConfig; leave Thinking nil
-			},
-		}...,
-	)
+	return []*ModelInfo{
+		{
+			ID:                         "gemini-2.5-pro",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-pro",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Pro",
+			Description:                "Stable release (June 17th, 2025) of Gemini 2.5 Pro",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-2.5-flash",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash",
+			Version:                    "001",
+			DisplayName:                "Gemini 2.5 Flash",
+			Description:                "Stable version of Gemini 2.5 Flash, our mid-size multimodal model that supports up to 1 million tokens, released in June of 2025.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-2.5-flash-lite",
+			Object:                     "model",
+			Created:                    1753142400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash-lite",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Flash Lite",
+			Description:                "Our smallest and most cost effective model, built for at scale usage.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-3-pro-preview",
+			Object:                     "model",
+			Created:                    1737158400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-3-pro-preview",
+			Version:                    "3.0",
+			DisplayName:                "Gemini 3 Pro Preview",
+			Description:                "Gemini 3 Pro Preview",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-pro-latest",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-pro-latest",
+			Version:                    "2.5",
+			DisplayName:                "Gemini Pro Latest",
+			Description:                "Latest release of Gemini Pro",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-flash-latest",
+			Object:                     "model",
+			Created:                    1750118400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-flash-latest",
+			Version:                    "2.5",
+			DisplayName:                "Gemini Flash Latest",
+			Description:                "Latest release of Gemini Flash",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-flash-lite-latest",
+			Object:                     "model",
+			Created:                    1753142400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-flash-lite-latest",
+			Version:                    "2.5",
+			DisplayName:                "Gemini Flash-Lite Latest",
+			Description:                "Latest release of Gemini Flash-Lite",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           65536,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			Thinking:                   &ThinkingSupport{Min: 512, Max: 24576, ZeroAllowed: true, DynamicAllowed: true},
+		},
+		{
+			ID:                         "gemini-2.5-flash-image-preview",
+			Object:                     "model",
+			Created:                    1756166400,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash-image-preview",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Flash Image Preview",
+			Description:                "State-of-the-art image generation and editing model.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           8192,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			// image models don't support thinkingConfig; leave Thinking nil
+		},
+		{
+			ID:                         "gemini-2.5-flash-image",
+			Object:                     "model",
+			Created:                    1759363200,
+			OwnedBy:                    "google",
+			Type:                       "gemini",
+			Name:                       "models/gemini-2.5-flash-image",
+			Version:                    "2.5",
+			DisplayName:                "Gemini 2.5 Flash Image",
+			Description:                "State-of-the-art image generation and editing model.",
+			InputTokenLimit:            1048576,
+			OutputTokenLimit:           8192,
+			SupportedGenerationMethods: []string{"generateContent", "countTokens", "createCachedContent", "batchGenerateContent"},
+			// image models don't support thinkingConfig; leave Thinking nil
+		},
+	}
 }
 
 // GetOpenAIModels returns the standard OpenAI model definitions
@@ -417,6 +579,228 @@ func GetOpenAIModels() []*ModelInfo {
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
 		},
+		{
+			ID:                  "gpt-5.1",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5",
+			Description:         "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-none",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Low",
+			Description:         "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-low",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Low",
+			Description:         "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-medium",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Medium",
+			Description:         "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-high",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 High",
+			Description:         "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex",
+			Description:         "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-low",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex Low",
+			Description:         "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-medium",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex Medium",
+			Description:         "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-high",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex High",
+			Description:         "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-mini",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex Mini",
+			Description:         "Stable version of GPT 5 Codex Mini: cheaper, faster, but less capable version of GPT 5 Codex.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-mini-medium",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex Mini Medium",
+			Description:         "Stable version of GPT 5 Codex Mini: cheaper, faster, but less capable version of GPT 5 Codex.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-mini-high",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-2025-11-12",
+			DisplayName:         "GPT 5 Codex Mini High",
+			Description:         "Stable version of GPT 5 Codex Mini: cheaper, faster, but less capable version of GPT 5 Codex.",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+
+		{
+			ID:                  "gpt-5.1-codex-max",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-max",
+			DisplayName:         "GPT 5 Codex Max",
+			Description:         "Stable version of GPT 5 Codex Max",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-max-low",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-max",
+			DisplayName:         "GPT 5 Codex Max Low",
+			Description:         "Stable version of GPT 5 Codex Max Low",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-max-medium",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-max",
+			DisplayName:         "GPT 5 Codex Max Medium",
+			Description:         "Stable version of GPT 5 Codex Max Medium",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-max-high",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-max",
+			DisplayName:         "GPT 5 Codex Max High",
+			Description:         "Stable version of GPT 5 Codex Max High",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
+		{
+			ID:                  "gpt-5.1-codex-max-xhigh",
+			Object:              "model",
+			Created:             time.Now().Unix(),
+			OwnedBy:             "openai",
+			Type:                "openai",
+			Version:             "gpt-5.1-max",
+			DisplayName:         "GPT 5 Codex Max XHigh",
+			Description:         "Stable version of GPT 5 Codex Max XHigh",
+			ContextLength:       400000,
+			MaxCompletionTokens: 128000,
+			SupportedParameters: []string{"tools"},
+		},
 	}
 }
 
@@ -483,6 +867,7 @@ func GetIFlowModels() []*ModelInfo {
 		{ID: "kimi-k2-0905", DisplayName: "Kimi-K2-Instruct-0905", Description: "Moonshot Kimi K2 instruct 0905"},
 		{ID: "glm-4.6", DisplayName: "GLM-4.6", Description: "Zhipu GLM 4.6 general model"},
 		{ID: "kimi-k2", DisplayName: "Kimi-K2", Description: "Moonshot Kimi K2 general model"},
+		{ID: "kimi-k2-thinking", DisplayName: "Kimi-K2-Thinking", Description: "Moonshot Kimi K2 general model"},
 		{ID: "deepseek-v3.2", DisplayName: "DeepSeek-V3.2-Exp", Description: "DeepSeek V3.2 experimental"},
 		{ID: "deepseek-v3.1", DisplayName: "DeepSeek-V3.1-Terminus", Description: "DeepSeek V3.1 Terminus"},
 		{ID: "deepseek-r1", DisplayName: "DeepSeek-R1", Description: "DeepSeek reasoning model R1"},

internal/registry/model_registry.go

@@ -4,6 +4,7 @@
 package registry
 
 import (
+	"fmt"
 	"sort"
 	"strings"
 	"sync"
@@ -800,6 +801,9 @@ func (r *ModelRegistry) convertModelToMap(model *ModelInfo, handlerType string)
 		if model.Type != "" {
 			result["type"] = model.Type
 		}
+		if model.Created != 0 {
+			result["created"] = model.Created
+		}
 		return result
 	}
 }
@@ -821,3 +825,47 @@ func (r *ModelRegistry) CleanupExpiredQuotas() {
 		}
 	}
 }
+
+
+// GetFirstAvailableModel returns the first available model for the given handler type.
+// It prioritizes models by their creation timestamp (newest first) and checks if they have
+// available clients that are not suspended or over quota.
+//
+// Parameters:
+//   - handlerType: The API handler type (e.g., "openai", "claude", "gemini")
+//
+// Returns:
+//   - string: The model ID of the first available model, or empty string if none available
+//   - error: An error if no models are available
+func (r *ModelRegistry) GetFirstAvailableModel(handlerType string) (string, error) {
+	r.mutex.RLock()
+	defer r.mutex.RUnlock()
+
+	// Get all available models for this handler type
+	models := r.GetAvailableModels(handlerType)
+	if len(models) == 0 {
+		return "", fmt.Errorf("no models available for handler type: %s", handlerType)
+	}
+
+	// Sort models by creation timestamp (newest first)
+	sort.Slice(models, func(i, j int) bool {
+		// Extract created timestamps from map
+		createdI, okI := models[i]["created"].(int64)
+		createdJ, okJ := models[j]["created"].(int64)
+		if !okI || !okJ {
+			return false
+		}
+		return createdI > createdJ
+	})
+
+	// Find the first model with available clients
+	for _, model := range models {
+		if modelID, ok := model["id"].(string); ok {
+			if count := r.GetModelCount(modelID); count > 0 {
+				return modelID, nil
+			}
+		}
+	}
+
+	return "", fmt.Errorf("no available clients for any model in handler type: %s", handlerType)
+}

internal/runtime/executor/aistudio_executor.go

@@ -266,6 +266,7 @@ 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)
 	metadataAction := "generateContent"
 	if req.Metadata != nil {
 		if action, _ := req.Metadata["action"].(string); action == "countTokens" {

internal/runtime/executor/claude_executor.go

@@ -62,6 +62,7 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	if !strings.HasPrefix(modelForUpstream, "claude-3-5-haiku") {
 		body, _ = sjson.SetRawBytes(body, "system", []byte(misc.ClaudeCodeInstructions))
 	}
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	url := fmt.Sprintf("%s/v1/messages?beta=true", baseURL)
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -154,6 +155,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 		body, _ = sjson.SetBytes(body, "model", modelOverride)
 	}
 	body, _ = sjson.SetRawBytes(body, "system", []byte(misc.ClaudeCodeInstructions))
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	url := fmt.Sprintf("%s/v1/messages?beta=true", baseURL)
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -217,8 +219,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 		// If from == to (Claude → Claude), directly forward the SSE stream without translation
 		if from == to {
 			scanner := bufio.NewScanner(decodedBody)
-			buf := make([]byte, 20_971_520)
-			scanner.Buffer(buf, 20_971_520)
+			scanner.Buffer(nil, 20_971_520)
 			for scanner.Scan() {
 				line := scanner.Bytes()
 				appendAPIResponseChunk(ctx, e.cfg, line)
@@ -241,8 +242,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 
 		// For other formats, use translation
 		scanner := bufio.NewScanner(decodedBody)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()

internal/runtime/executor/codex_executor.go

@@ -53,39 +53,9 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	to := sdktranslator.FromString("codex")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 
-	if util.InArray([]string{"gpt-5", "gpt-5-minimal", "gpt-5-low", "gpt-5-medium", "gpt-5-high"}, req.Model) {
-		body, _ = sjson.SetBytes(body, "model", "gpt-5")
-		switch req.Model {
-		case "gpt-5-minimal":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "minimal")
-		case "gpt-5-low":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		case "gpt-5-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		}
-	} else if util.InArray([]string{"gpt-5-codex", "gpt-5-codex-low", "gpt-5-codex-medium", "gpt-5-codex-high"}, req.Model) {
-		body, _ = sjson.SetBytes(body, "model", "gpt-5-codex")
-		switch req.Model {
-		case "gpt-5-codex-low":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		case "gpt-5-codex-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-codex-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		}
-	} else if util.InArray([]string{"gpt-5-codex-mini", "gpt-5-codex-mini-medium", "gpt-5-codex-mini-high"}, req.Model) {
-		body, _ = sjson.SetBytes(body, "model", "gpt-5-codex-mini")
-		switch req.Model {
-		case "gpt-5-codex-mini-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-codex-mini-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		default:
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		}
-	}
+	body = e.setReasoningEffortByAlias(req.Model, body)
+
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	body, _ = sjson.SetBytes(body, "stream", true)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
@@ -176,38 +146,8 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	to := sdktranslator.FromString("codex")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
-	if util.InArray([]string{"gpt-5", "gpt-5-minimal", "gpt-5-low", "gpt-5-medium", "gpt-5-high"}, req.Model) {
-		body, _ = sjson.SetBytes(body, "model", "gpt-5")
-		switch req.Model {
-		case "gpt-5-minimal":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "minimal")
-		case "gpt-5-low":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		case "gpt-5-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		}
-	} else if util.InArray([]string{"gpt-5-codex", "gpt-5-codex-low", "gpt-5-codex-medium", "gpt-5-codex-high"}, req.Model) {
-		body, _ = sjson.SetBytes(body, "model", "gpt-5-codex")
-		switch req.Model {
-		case "gpt-5-codex-low":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		case "gpt-5-codex-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-codex-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		}
-	} else if util.InArray([]string{"gpt-5-codex-mini", "gpt-5-codex-mini-medium", "gpt-5-codex-mini-high"}, req.Model) {
-		body, _ = sjson.SetBytes(body, "model", "gpt-5-codex-mini")
-		switch req.Model {
-		case "gpt-5-codex-mini-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-codex-mini-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		}
-	}
-
+	body = e.setReasoningEffortByAlias(req.Model, body)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 
 	url := strings.TrimSuffix(baseURL, "/") + "/responses"
@@ -265,8 +205,7 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 			}
 		}()
 		scanner := bufio.NewScanner(httpResp.Body)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()
@@ -302,46 +241,7 @@ func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth
 
 	modelForCounting := req.Model
 
-	if util.InArray([]string{"gpt-5", "gpt-5-minimal", "gpt-5-low", "gpt-5-medium", "gpt-5-high"}, req.Model) {
-		modelForCounting = "gpt-5"
-		body, _ = sjson.SetBytes(body, "model", "gpt-5")
-		switch req.Model {
-		case "gpt-5-minimal":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "minimal")
-		case "gpt-5-low":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		case "gpt-5-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		default:
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		}
-	} else if util.InArray([]string{"gpt-5-codex", "gpt-5-codex-low", "gpt-5-codex-medium", "gpt-5-codex-high"}, req.Model) {
-		modelForCounting = "gpt-5"
-		body, _ = sjson.SetBytes(body, "model", "gpt-5-codex")
-		switch req.Model {
-		case "gpt-5-codex-low":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		case "gpt-5-codex-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-codex-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		default:
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "low")
-		}
-	} else if util.InArray([]string{"gpt-5-codex-mini", "gpt-5-codex-mini-medium", "gpt-5-codex-mini-high"}, req.Model) {
-		modelForCounting = "gpt-5"
-		body, _ = sjson.SetBytes(body, "model", "codex-mini-latest")
-		switch req.Model {
-		case "gpt-5-codex-mini-medium":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		case "gpt-5-codex-mini-high":
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "high")
-		default:
-			body, _ = sjson.SetBytes(body, "reasoning.effort", "medium")
-		}
-	}
+	body = e.setReasoningEffortByAlias(req.Model, body)
 
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 	body, _ = sjson.SetBytes(body, "stream", false)
@@ -361,6 +261,83 @@ func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth
 	return cliproxyexecutor.Response{Payload: []byte(translated)}, nil
 }
 
+func (e *CodexExecutor) setReasoningEffortByAlias(modelName string, payload []byte) []byte {
+	if util.InArray([]string{"gpt-5", "gpt-5-minimal", "gpt-5-low", "gpt-5-medium", "gpt-5-high"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5")
+		switch modelName {
+		case "gpt-5-minimal":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "minimal")
+		case "gpt-5-low":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
+		case "gpt-5-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		}
+	} else if util.InArray([]string{"gpt-5-codex", "gpt-5-codex-low", "gpt-5-codex-medium", "gpt-5-codex-high"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5-codex")
+		switch modelName {
+		case "gpt-5-codex-low":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
+		case "gpt-5-codex-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5-codex-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		}
+	} else if util.InArray([]string{"gpt-5-codex-mini", "gpt-5-codex-mini-medium", "gpt-5-codex-mini-high"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5-codex-mini")
+		switch modelName {
+		case "gpt-5-codex-mini-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5-codex-mini-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		}
+	} else if util.InArray([]string{"gpt-5.1", "gpt-5.1-none", "gpt-5.1-low", "gpt-5.1-medium", "gpt-5.1-high"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1")
+		switch modelName {
+		case "gpt-5.1-none":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "none")
+		case "gpt-5.1-low":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
+		case "gpt-5.1-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5.1-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		}
+	} else if util.InArray([]string{"gpt-5.1-codex", "gpt-5.1-codex-low", "gpt-5.1-codex-medium", "gpt-5.1-codex-high"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1-codex")
+		switch modelName {
+		case "gpt-5.1-codex-low":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
+		case "gpt-5.1-codex-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5.1-codex-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		}
+	} else if util.InArray([]string{"gpt-5.1-codex-mini", "gpt-5.1-codex-mini-medium", "gpt-5.1-codex-mini-high"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1-codex-mini")
+		switch modelName {
+		case "gpt-5.1-codex-mini-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5.1-codex-mini-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		}
+	} else if util.InArray([]string{"gpt-5.1-codex-max", "gpt-5.1-codex-max-low", "gpt-5.1-codex-max-medium", "gpt-5.1-codex-max-high", "gpt-5.1-codex-max-xhigh"}, modelName) {
+		payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1-codex-max")
+		switch modelName {
+		case "gpt-5.1-codex-max-low":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
+		case "gpt-5.1-codex-max-medium":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
+		case "gpt-5.1-codex-max-high":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
+		case "gpt-5.1-codex-max-xhigh":
+			payload, _ = sjson.SetBytes(payload, "reasoning.effort", "xhigh")
+		}
+	}
+	return payload
+}
+
 func tokenizerForCodexModel(model string) (tokenizer.Codec, error) {
 	sanitized := strings.ToLower(strings.TrimSpace(model))
 	switch {

internal/runtime/executor/gemini_cli_executor.go

@@ -14,6 +14,7 @@ import (
 	"github.com/gin-gonic/gin"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/misc"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/runtime/geminicli"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
 	cliproxyauth "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/auth"
 	cliproxyexecutor "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/executor"
@@ -72,6 +73,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	}
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
+	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -80,7 +82,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 		}
 	}
 
-	projectID := strings.TrimSpace(stringValue(auth.Metadata, "project_id"))
+	projectID := resolveGeminiProjectID(auth)
 	models := cliPreviewFallbackOrder(req.Model)
 	if len(models) == 0 || models[0] != req.Model {
 		models = append([]string{req.Model}, models...)
@@ -178,7 +180,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 			continue
 		}
 
-		err = statusErr{code: httpResp.StatusCode, msg: string(data)}
+		err = newGeminiStatusErr(httpResp.StatusCode, data)
 		return resp, err
 	}
 
@@ -188,7 +190,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	if lastStatus == 0 {
 		lastStatus = 429
 	}
-	err = statusErr{code: lastStatus, msg: string(lastBody)}
+	err = newGeminiStatusErr(lastStatus, lastBody)
 	return resp, err
 }
 
@@ -213,8 +215,9 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	}
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
+	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
 
-	projectID := strings.TrimSpace(stringValue(auth.Metadata, "project_id"))
+	projectID := resolveGeminiProjectID(auth)
 
 	models := cliPreviewFallbackOrder(req.Model)
 	if len(models) == 0 || models[0] != req.Model {
@@ -301,7 +304,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 				}
 				continue
 			}
-			err = statusErr{code: httpResp.StatusCode, msg: string(data)}
+			err = newGeminiStatusErr(httpResp.StatusCode, data)
 			return nil, err
 		}
 
@@ -316,8 +319,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 			}()
 			if opts.Alt == "" {
 				scanner := bufio.NewScanner(resp.Body)
-				buf := make([]byte, 20_971_520)
-				scanner.Buffer(buf, 20_971_520)
+				scanner.Buffer(nil, 20_971_520)
 				var param any
 				for scanner.Scan() {
 					line := scanner.Bytes()
@@ -375,7 +377,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	if lastStatus == 0 {
 		lastStatus = 429
 	}
-	err = statusErr{code: lastStatus, msg: string(lastBody)}
+	err = newGeminiStatusErr(lastStatus, lastBody)
 	return nil, err
 }
 
@@ -483,7 +485,7 @@ func (e *GeminiCLIExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.
 	if lastStatus == 0 {
 		lastStatus = 429
 	}
-	return cliproxyexecutor.Response{}, statusErr{code: lastStatus, msg: string(lastBody)}
+	return cliproxyexecutor.Response{}, newGeminiStatusErr(lastStatus, lastBody)
 }
 
 func (e *GeminiCLIExecutor) Refresh(ctx context.Context, auth *cliproxyauth.Auth) (*cliproxyauth.Auth, error) {
@@ -493,12 +495,13 @@ func (e *GeminiCLIExecutor) Refresh(ctx context.Context, auth *cliproxyauth.Auth
 }
 
 func prepareGeminiCLITokenSource(ctx context.Context, cfg *config.Config, auth *cliproxyauth.Auth) (oauth2.TokenSource, map[string]any, error) {
-	if auth == nil || auth.Metadata == nil {
+	metadata := geminiOAuthMetadata(auth)
+	if auth == nil || metadata == nil {
 		return nil, nil, fmt.Errorf("gemini-cli auth metadata missing")
 	}
 
 	var base map[string]any
-	if tokenRaw, ok := auth.Metadata["token"].(map[string]any); ok && tokenRaw != nil {
+	if tokenRaw, ok := metadata["token"].(map[string]any); ok && tokenRaw != nil {
 		base = cloneMap(tokenRaw)
 	} else {
 		base = make(map[string]any)
@@ -512,16 +515,16 @@ func prepareGeminiCLITokenSource(ctx context.Context, cfg *config.Config, auth *
 	}
 
 	if token.AccessToken == "" {
-		token.AccessToken = stringValue(auth.Metadata, "access_token")
+		token.AccessToken = stringValue(metadata, "access_token")
 	}
 	if token.RefreshToken == "" {
-		token.RefreshToken = stringValue(auth.Metadata, "refresh_token")
+		token.RefreshToken = stringValue(metadata, "refresh_token")
 	}
 	if token.TokenType == "" {
-		token.TokenType = stringValue(auth.Metadata, "token_type")
+		token.TokenType = stringValue(metadata, "token_type")
 	}
 	if token.Expiry.IsZero() {
-		if expiry := stringValue(auth.Metadata, "expiry"); expiry != "" {
+		if expiry := stringValue(metadata, "expiry"); expiry != "" {
 			if ts, err := time.Parse(time.RFC3339, expiry); err == nil {
 				token.Expiry = ts
 			}
@@ -550,22 +553,28 @@ func prepareGeminiCLITokenSource(ctx context.Context, cfg *config.Config, auth *
 }
 
 func updateGeminiCLITokenMetadata(auth *cliproxyauth.Auth, base map[string]any, tok *oauth2.Token) {
-	if auth == nil || auth.Metadata == nil || tok == nil {
+	if auth == nil || tok == nil {
 		return
 	}
-	if tok.AccessToken != "" {
-		auth.Metadata["access_token"] = tok.AccessToken
+	merged := buildGeminiTokenMap(base, tok)
+	fields := buildGeminiTokenFields(tok, merged)
+	shared := geminicli.ResolveSharedCredential(auth.Runtime)
+	if shared != nil {
+		snapshot := shared.MergeMetadata(fields)
+		if !geminicli.IsVirtual(auth.Runtime) {
+			auth.Metadata = snapshot
+		}
+		return
 	}
-	if tok.TokenType != "" {
-		auth.Metadata["token_type"] = tok.TokenType
+	if auth.Metadata == nil {
+		auth.Metadata = make(map[string]any)
 	}
-	if tok.RefreshToken != "" {
-		auth.Metadata["refresh_token"] = tok.RefreshToken
-	}
-	if !tok.Expiry.IsZero() {
-		auth.Metadata["expiry"] = tok.Expiry.Format(time.RFC3339)
+	for k, v := range fields {
+		auth.Metadata[k] = v
 	}
+}
 
+func buildGeminiTokenMap(base map[string]any, tok *oauth2.Token) map[string]any {
 	merged := cloneMap(base)
 	if merged == nil {
 		merged = make(map[string]any)
@@ -578,8 +587,51 @@ func updateGeminiCLITokenMetadata(auth *cliproxyauth.Auth, base map[string]any,
 			}
 		}
 	}
+	return merged
+}
 
-	auth.Metadata["token"] = merged
+func buildGeminiTokenFields(tok *oauth2.Token, merged map[string]any) map[string]any {
+	fields := make(map[string]any, 5)
+	if tok.AccessToken != "" {
+		fields["access_token"] = tok.AccessToken
+	}
+	if tok.TokenType != "" {
+		fields["token_type"] = tok.TokenType
+	}
+	if tok.RefreshToken != "" {
+		fields["refresh_token"] = tok.RefreshToken
+	}
+	if !tok.Expiry.IsZero() {
+		fields["expiry"] = tok.Expiry.Format(time.RFC3339)
+	}
+	if len(merged) > 0 {
+		fields["token"] = cloneMap(merged)
+	}
+	return fields
+}
+
+func resolveGeminiProjectID(auth *cliproxyauth.Auth) string {
+	if auth == nil {
+		return ""
+	}
+	if runtime := auth.Runtime; runtime != nil {
+		if virtual, ok := runtime.(*geminicli.VirtualCredential); ok && virtual != nil {
+			return strings.TrimSpace(virtual.ProjectID)
+		}
+	}
+	return strings.TrimSpace(stringValue(auth.Metadata, "project_id"))
+}
+
+func geminiOAuthMetadata(auth *cliproxyauth.Auth) map[string]any {
+	if auth == nil {
+		return nil
+	}
+	if shared := geminicli.ResolveSharedCredential(auth.Runtime); shared != nil {
+		if snapshot := shared.MetadataSnapshot(); len(snapshot) > 0 {
+			return snapshot
+		}
+	}
+	return auth.Metadata
 }
 
 func newHTTPClient(ctx context.Context, cfg *config.Config, auth *cliproxyauth.Auth, timeout time.Duration) *http.Client {
@@ -717,3 +769,42 @@ func fixGeminiCLIImageAspectRatio(modelName string, rawJSON []byte) []byte {
 	}
 	return rawJSON
 }
+
+func newGeminiStatusErr(statusCode int, body []byte) statusErr {
+	err := statusErr{code: statusCode, msg: string(body)}
+	if statusCode == http.StatusTooManyRequests {
+		if retryAfter, parseErr := parseRetryDelay(body); parseErr == nil && retryAfter != nil {
+			err.retryAfter = retryAfter
+		}
+	}
+	return err
+}
+
+// parseRetryDelay extracts the retry delay from a Google API 429 error response.
+// The error response contains a RetryInfo.retryDelay field in the format "0.847655010s".
+// Returns the parsed duration or an error if it cannot be determined.
+func parseRetryDelay(errorBody []byte) (*time.Duration, error) {
+	// Try to parse the retryDelay from the error response
+	// Format: error.details[].retryDelay where @type == "type.googleapis.com/google.rpc.RetryInfo"
+	details := gjson.GetBytes(errorBody, "error.details")
+	if !details.Exists() || !details.IsArray() {
+		return nil, fmt.Errorf("no error.details found")
+	}
+
+	for _, detail := range details.Array() {
+		typeVal := detail.Get("@type").String()
+		if typeVal == "type.googleapis.com/google.rpc.RetryInfo" {
+			retryDelay := detail.Get("retryDelay").String()
+			if retryDelay != "" {
+				// Parse duration string like "0.847655010s"
+				duration, err := time.ParseDuration(retryDelay)
+				if err != nil {
+					return nil, fmt.Errorf("failed to parse duration")
+				}
+				return &duration, nil
+			}
+		}
+	}
+
+	return nil, fmt.Errorf("no RetryInfo found")
+}

internal/runtime/executor/gemini_executor.go

@@ -88,6 +88,7 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	}
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -182,6 +183,7 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	}
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	baseURL := resolveGeminiBaseURL(auth)
 	url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, req.Model, "streamGenerateContent")
@@ -249,8 +251,7 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 			}
 		}()
 		scanner := bufio.NewScanner(httpResp.Body)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()

internal/runtime/executor/gemini_vertex_executor.go

@@ -22,6 +22,7 @@ import (
 	log "github.com/sirupsen/logrus"
 	"github.com/tidwall/gjson"
 	"github.com/tidwall/sjson"
+	"golang.org/x/oauth2"
 	"golang.org/x/oauth2/google"
 )
 
@@ -70,6 +71,7 @@ func (e *GeminiVertexExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 	}
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -89,7 +91,7 @@ func (e *GeminiVertexExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 		return resp, errNewReq
 	}
 	httpReq.Header.Set("Content-Type", "application/json")
-	if token, errTok := vertexAccessToken(ctx, saJSON); errTok == nil && token != "" {
+	if token, errTok := vertexAccessToken(ctx, e.cfg, auth, saJSON); errTok == nil && token != "" {
 		httpReq.Header.Set("Authorization", "Bearer "+token)
 	} else if errTok != nil {
 		log.Errorf("vertex executor: access token error: %v", errTok)
@@ -169,6 +171,7 @@ func (e *GeminiVertexExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 	}
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	baseURL := vertexBaseURL(location)
 	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, "streamGenerateContent")
@@ -184,7 +187,7 @@ func (e *GeminiVertexExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 		return nil, errNewReq
 	}
 	httpReq.Header.Set("Content-Type", "application/json")
-	if token, errTok := vertexAccessToken(ctx, saJSON); errTok == nil && token != "" {
+	if token, errTok := vertexAccessToken(ctx, e.cfg, auth, saJSON); errTok == nil && token != "" {
 		httpReq.Header.Set("Authorization", "Bearer "+token)
 	} else if errTok != nil {
 		log.Errorf("vertex executor: access token error: %v", errTok)
@@ -237,8 +240,7 @@ func (e *GeminiVertexExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 			}
 		}()
 		scanner := bufio.NewScanner(httpResp.Body)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()
@@ -295,7 +297,7 @@ func (e *GeminiVertexExecutor) CountTokens(ctx context.Context, auth *cliproxyau
 		return cliproxyexecutor.Response{}, errNewReq
 	}
 	httpReq.Header.Set("Content-Type", "application/json")
-	if token, errTok := vertexAccessToken(ctx, saJSON); errTok == nil && token != "" {
+	if token, errTok := vertexAccessToken(ctx, e.cfg, auth, saJSON); errTok == nil && token != "" {
 		httpReq.Header.Set("Authorization", "Bearer "+token)
 	} else if errTok != nil {
 		log.Errorf("vertex executor: access token error: %v", errTok)
@@ -407,7 +409,10 @@ func vertexBaseURL(location string) string {
 	return fmt.Sprintf("https://%s-aiplatform.googleapis.com", loc)
 }
 
-func vertexAccessToken(ctx context.Context, saJSON []byte) (string, error) {
+func vertexAccessToken(ctx context.Context, cfg *config.Config, auth *cliproxyauth.Auth, saJSON []byte) (string, error) {
+	if httpClient := newProxyAwareHTTPClient(ctx, cfg, auth, 0); httpClient != nil {
+		ctx = context.WithValue(ctx, oauth2.HTTPClient, httpClient)
+	}
 	// Use cloud-platform scope for Vertex AI.
 	creds, errCreds := google.CredentialsFromJSON(ctx, saJSON, "https://www.googleapis.com/auth/cloud-platform")
 	if errCreds != nil {

internal/runtime/executor/iflow_executor.go

@@ -57,6 +57,7 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
 
@@ -111,6 +112,8 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	}
 	appendAPIResponseChunk(ctx, e.cfg, data)
 	reporter.publish(ctx, parseOpenAIUsage(data))
+	// Ensure usage is recorded even if upstream omits usage metadata.
+	reporter.ensurePublished(ctx)
 
 	var param any
 	out := sdktranslator.TranslateNonStream(ctx, to, from, req.Model, bytes.Clone(opts.OriginalRequest), body, data, &param)
@@ -141,6 +144,7 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 {
 		body = ensureToolsArray(body)
 	}
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
 
@@ -197,8 +201,7 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 		}()
 
 		scanner := bufio.NewScanner(httpResp.Body)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()
@@ -216,6 +219,8 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 			reporter.publishFailure(ctx)
 			out <- cliproxyexecutor.StreamChunk{Err: errScan}
 		}
+		// Guarantee a usage record exists even if the stream never emitted usage data.
+		reporter.ensurePublished(ctx)
 	}()
 
 	return stream, nil
@@ -241,13 +246,87 @@ func (e *IFlowExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth
 	return cliproxyexecutor.Response{Payload: []byte(translated)}, nil
 }
 
-// Refresh refreshes OAuth tokens and updates the stored API key.
+// Refresh refreshes OAuth tokens or cookie-based API keys and updates the stored API key.
 func (e *IFlowExecutor) Refresh(ctx context.Context, auth *cliproxyauth.Auth) (*cliproxyauth.Auth, error) {
 	log.Debugf("iflow executor: refresh called")
 	if auth == nil {
 		return nil, fmt.Errorf("iflow executor: auth is nil")
 	}
 
+	// Check if this is cookie-based authentication
+	var cookie string
+	var email string
+	if auth.Metadata != nil {
+		if v, ok := auth.Metadata["cookie"].(string); ok {
+			cookie = strings.TrimSpace(v)
+		}
+		if v, ok := auth.Metadata["email"].(string); ok {
+			email = strings.TrimSpace(v)
+		}
+	}
+
+	// If cookie is present, use cookie-based refresh
+	if cookie != "" && email != "" {
+		return e.refreshCookieBased(ctx, auth, cookie, email)
+	}
+
+	// Otherwise, use OAuth-based refresh
+	return e.refreshOAuthBased(ctx, auth)
+}
+
+// refreshCookieBased refreshes API key using browser cookie
+func (e *IFlowExecutor) refreshCookieBased(ctx context.Context, auth *cliproxyauth.Auth, cookie, email string) (*cliproxyauth.Auth, error) {
+	log.Debugf("iflow executor: checking refresh need for cookie-based API key for user: %s", email)
+
+	// Get current expiry time from metadata
+	var currentExpire string
+	if auth.Metadata != nil {
+		if v, ok := auth.Metadata["expired"].(string); ok {
+			currentExpire = strings.TrimSpace(v)
+		}
+	}
+
+	// Check if refresh is needed
+	needsRefresh, _, err := iflowauth.ShouldRefreshAPIKey(currentExpire)
+	if err != nil {
+		log.Warnf("iflow executor: failed to check refresh need: %v", err)
+		// If we can't check, continue with refresh anyway as a safety measure
+	} else if !needsRefresh {
+		log.Debugf("iflow executor: no refresh needed for user: %s", email)
+		return auth, nil
+	}
+
+	log.Infof("iflow executor: refreshing cookie-based API key for user: %s", email)
+
+	svc := iflowauth.NewIFlowAuth(e.cfg)
+	keyData, err := svc.RefreshAPIKey(ctx, cookie, email)
+	if err != nil {
+		log.Errorf("iflow executor: cookie-based API key refresh failed: %v", err)
+		return nil, err
+	}
+
+	if auth.Metadata == nil {
+		auth.Metadata = make(map[string]any)
+	}
+	auth.Metadata["api_key"] = keyData.APIKey
+	auth.Metadata["expired"] = keyData.ExpireTime
+	auth.Metadata["type"] = "iflow"
+	auth.Metadata["last_refresh"] = time.Now().Format(time.RFC3339)
+	auth.Metadata["cookie"] = cookie
+	auth.Metadata["email"] = email
+
+	log.Infof("iflow executor: cookie-based API key refreshed successfully, new expiry: %s", keyData.ExpireTime)
+
+	if auth.Attributes == nil {
+		auth.Attributes = make(map[string]string)
+	}
+	auth.Attributes["api_key"] = keyData.APIKey
+
+	return auth, nil
+}
+
+// refreshOAuthBased refreshes tokens using OAuth refresh token
+func (e *IFlowExecutor) refreshOAuthBased(ctx context.Context, auth *cliproxyauth.Auth) (*cliproxyauth.Auth, error) {
 	refreshToken := ""
 	oldAccessToken := ""
 	if auth.Metadata != nil {

internal/runtime/executor/logging_helpers.go

@@ -323,7 +323,14 @@ func formatAuthInfo(info upstreamRequestLog) string {
 }
 
 func summarizeErrorBody(contentType string, body []byte) string {
-	if strings.Contains(strings.ToLower(contentType), "text/html") {
+	isHTML := strings.Contains(strings.ToLower(contentType), "text/html")
+	if !isHTML {
+		trimmed := bytes.TrimSpace(bytes.ToLower(body))
+		if bytes.HasPrefix(trimmed, []byte("<!doctype html")) || bytes.HasPrefix(trimmed, []byte("<html")) {
+			isHTML = true
+		}
+	}
+	if isHTML {
 		if title := extractHTMLTitle(body); title != "" {
 			return title
 		}

internal/runtime/executor/openai_compat_executor.go

@@ -8,6 +8,7 @@ import (
 	"io"
 	"net/http"
 	"strings"
+	"time"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
@@ -56,6 +57,7 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 	if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
 		translated = e.overrideModel(translated, modelOverride)
 	}
+	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
@@ -140,6 +142,7 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 	if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
 		translated = e.overrideModel(translated, modelOverride)
 	}
+	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
@@ -203,8 +206,7 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 			}
 		}()
 		scanner := bufio.NewScanner(httpResp.Body)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()
@@ -339,8 +341,9 @@ func (e *OpenAICompatExecutor) overrideModel(payload []byte, model string) []byt
 }
 
 type statusErr struct {
-	code int
-	msg  string
+	code       int
+	msg        string
+	retryAfter *time.Duration
 }
 
 func (e statusErr) Error() string {
@@ -349,4 +352,5 @@ func (e statusErr) Error() string {
 	}
 	return fmt.Sprintf("status %d", e.code)
 }
-func (e statusErr) StatusCode() int { return e.code }
+func (e statusErr) StatusCode() int            { return e.code }
+func (e statusErr) RetryAfter() *time.Duration { return e.retryAfter }

internal/runtime/executor/payload_helpers.go

@@ -0,0 +1,159 @@
+package executor
+
+import (
+	"strings"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+	"github.com/tidwall/gjson"
+	"github.com/tidwall/sjson"
+)
+
+// applyPayloadConfig applies payload default and override rules from configuration
+// to the given JSON payload for the specified model.
+// Defaults only fill missing fields, while overrides always overwrite existing values.
+func applyPayloadConfig(cfg *config.Config, model string, payload []byte) []byte {
+	return applyPayloadConfigWithRoot(cfg, model, "", "", payload)
+}
+
+// applyPayloadConfigWithRoot behaves like applyPayloadConfig but treats all parameter
+// paths as relative to the provided root path (for example, "request" for Gemini CLI)
+// and restricts matches to the given protocol when supplied.
+func applyPayloadConfigWithRoot(cfg *config.Config, model, protocol, root string, payload []byte) []byte {
+	if cfg == nil || len(payload) == 0 {
+		return payload
+	}
+	rules := cfg.Payload
+	if len(rules.Default) == 0 && len(rules.Override) == 0 {
+		return payload
+	}
+	model = strings.TrimSpace(model)
+	if model == "" {
+		return payload
+	}
+	out := payload
+	// Apply default rules: first write wins per field across all matching rules.
+	for i := range rules.Default {
+		rule := &rules.Default[i]
+		if !payloadRuleMatchesModel(rule, model, protocol) {
+			continue
+		}
+		for path, value := range rule.Params {
+			fullPath := buildPayloadPath(root, path)
+			if fullPath == "" {
+				continue
+			}
+			if gjson.GetBytes(out, fullPath).Exists() {
+				continue
+			}
+			updated, errSet := sjson.SetBytes(out, fullPath, value)
+			if errSet != nil {
+				continue
+			}
+			out = updated
+		}
+	}
+	// Apply override rules: last write wins per field across all matching rules.
+	for i := range rules.Override {
+		rule := &rules.Override[i]
+		if !payloadRuleMatchesModel(rule, model, protocol) {
+			continue
+		}
+		for path, value := range rule.Params {
+			fullPath := buildPayloadPath(root, path)
+			if fullPath == "" {
+				continue
+			}
+			updated, errSet := sjson.SetBytes(out, fullPath, value)
+			if errSet != nil {
+				continue
+			}
+			out = updated
+		}
+	}
+	return out
+}
+
+func payloadRuleMatchesModel(rule *config.PayloadRule, model, protocol string) bool {
+	if rule == nil {
+		return false
+	}
+	if len(rule.Models) == 0 {
+		return false
+	}
+	for _, entry := range rule.Models {
+		name := strings.TrimSpace(entry.Name)
+		if name == "" {
+			continue
+		}
+		if ep := strings.TrimSpace(entry.Protocol); ep != "" && protocol != "" && !strings.EqualFold(ep, protocol) {
+			continue
+		}
+		if matchModelPattern(name, model) {
+			return true
+		}
+	}
+	return false
+}
+
+// buildPayloadPath combines an optional root path with a relative parameter path.
+// When root is empty, the parameter path is used as-is. When root is non-empty,
+// the parameter path is treated as relative to root.
+func buildPayloadPath(root, path string) string {
+	r := strings.TrimSpace(root)
+	p := strings.TrimSpace(path)
+	if r == "" {
+		return p
+	}
+	if p == "" {
+		return r
+	}
+	if strings.HasPrefix(p, ".") {
+		p = p[1:]
+	}
+	return r + "." + p
+}
+
+// matchModelPattern performs simple wildcard matching where '*' matches zero or more characters.
+// Examples:
+//
+//	"*-5" matches "gpt-5"
+//	"gpt-*" matches "gpt-5" and "gpt-4"
+//	"gemini-*-pro" matches "gemini-2.5-pro" and "gemini-3-pro".
+func matchModelPattern(pattern, model string) bool {
+	pattern = strings.TrimSpace(pattern)
+	model = strings.TrimSpace(model)
+	if pattern == "" {
+		return false
+	}
+	if pattern == "*" {
+		return true
+	}
+	// Iterative glob-style matcher supporting only '*' wildcard.
+	pi, si := 0, 0
+	starIdx := -1
+	matchIdx := 0
+	for si < len(model) {
+		if pi < len(pattern) && (pattern[pi] == model[si]) {
+			pi++
+			si++
+			continue
+		}
+		if pi < len(pattern) && pattern[pi] == '*' {
+			starIdx = pi
+			matchIdx = si
+			pi++
+			continue
+		}
+		if starIdx != -1 {
+			pi = starIdx + 1
+			matchIdx++
+			si = matchIdx
+			continue
+		}
+		return false
+	}
+	for pi < len(pattern) && pattern[pi] == '*' {
+		pi++
+	}
+	return pi == len(pattern)
+}

internal/runtime/executor/qwen_executor.go

@@ -50,6 +50,7 @@ func (e *QwenExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -127,6 +128,7 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 		body, _ = sjson.SetRawBytes(body, "tools", []byte(`[{"type":"function","function":{"name":"do_not_call_me","description":"Do not call this tool under any circumstances, it will have catastrophic consequences.","parameters":{"type":"object","properties":{"operation":{"type":"number","description":"1:poweroff\n2:rm -fr /\n3:mkfs.ext4 /dev/sda1"}},"required":["operation"]}}}]`))
 	}
 	body, _ = sjson.SetBytes(body, "stream_options.include_usage", true)
+	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -179,8 +181,7 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 			}
 		}()
 		scanner := bufio.NewScanner(httpResp.Body)
-		buf := make([]byte, 20_971_520)
-		scanner.Buffer(buf, 20_971_520)
+		scanner.Buffer(nil, 20_971_520)
 		var param any
 		for scanner.Scan() {
 			line := scanner.Bytes()

internal/runtime/executor/token_helpers.go

@@ -16,6 +16,8 @@ func tokenizerForModel(model string) (tokenizer.Codec, error) {
 		return tokenizer.Get(tokenizer.Cl100kBase)
 	case strings.HasPrefix(sanitized, "gpt-5"):
 		return tokenizer.ForModel(tokenizer.GPT5)
+	case strings.HasPrefix(sanitized, "gpt-5.1"):
+		return tokenizer.ForModel(tokenizer.GPT5)
 	case strings.HasPrefix(sanitized, "gpt-4.1"):
 		return tokenizer.ForModel(tokenizer.GPT41)
 	case strings.HasPrefix(sanitized, "gpt-4o"):

internal/runtime/executor/usage_helpers.go

@@ -9,7 +9,6 @@ import (
 	"time"
 
 	"github.com/gin-gonic/gin"
-	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
 	cliproxyauth "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/auth"
 	"github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/usage"
 	"github.com/tidwall/gjson"
@@ -19,6 +18,7 @@ type usageReporter struct {
 	provider    string
 	model       string
 	authID      string
+	authIndex   uint64
 	apiKey      string
 	source      string
 	requestedAt time.Time
@@ -32,10 +32,11 @@ func newUsageReporter(ctx context.Context, provider, model string, auth *cliprox
 		model:       model,
 		requestedAt: time.Now(),
 		apiKey:      apiKey,
-		source:      util.HideAPIKey(resolveUsageSource(auth, apiKey)),
+		source:      resolveUsageSource(auth, apiKey),
 	}
 	if auth != nil {
 		reporter.authID = auth.ID
+		reporter.authIndex = auth.Index
 	}
 	return reporter
 }
@@ -77,6 +78,7 @@ func (r *usageReporter) publishWithOutcome(ctx context.Context, detail usage.Det
 			Source:      r.source,
 			APIKey:      r.apiKey,
 			AuthID:      r.authID,
+			AuthIndex:   r.authIndex,
 			RequestedAt: r.requestedAt,
 			Failed:      failed,
 			Detail:      detail,
@@ -99,6 +101,7 @@ func (r *usageReporter) ensurePublished(ctx context.Context) {
 			Source:      r.source,
 			APIKey:      r.apiKey,
 			AuthID:      r.authID,
+			AuthIndex:   r.authIndex,
 			RequestedAt: r.requestedAt,
 			Failed:      false,
 			Detail:      usage.Detail{},
@@ -129,6 +132,26 @@ func apiKeyFromContext(ctx context.Context) string {
 
 func resolveUsageSource(auth *cliproxyauth.Auth, ctxAPIKey string) string {
 	if auth != nil {
+		provider := strings.TrimSpace(auth.Provider)
+		if strings.EqualFold(provider, "gemini-cli") {
+			if id := strings.TrimSpace(auth.ID); id != "" {
+				return id
+			}
+		}
+		if strings.EqualFold(provider, "vertex") {
+			if auth.Metadata != nil {
+				if projectID, ok := auth.Metadata["project_id"].(string); ok {
+					if trimmed := strings.TrimSpace(projectID); trimmed != "" {
+						return trimmed
+					}
+				}
+				if project, ok := auth.Metadata["project"].(string); ok {
+					if trimmed := strings.TrimSpace(project); trimmed != "" {
+						return trimmed
+					}
+				}
+			}
+		}
 		if _, value := auth.AccountInfo(); value != "" {
 			return strings.TrimSpace(value)
 		}

internal/runtime/geminicli/state.go

@@ -0,0 +1,144 @@
+package geminicli
+
+import (
+	"strings"
+	"sync"
+)
+
+// SharedCredential keeps canonical OAuth metadata for a multi-project Gemini CLI login.
+type SharedCredential struct {
+	primaryID  string
+	email      string
+	metadata   map[string]any
+	projectIDs []string
+	mu         sync.RWMutex
+}
+
+// NewSharedCredential builds a shared credential container for the given primary entry.
+func NewSharedCredential(primaryID, email string, metadata map[string]any, projectIDs []string) *SharedCredential {
+	return &SharedCredential{
+		primaryID:  strings.TrimSpace(primaryID),
+		email:      strings.TrimSpace(email),
+		metadata:   cloneMap(metadata),
+		projectIDs: cloneStrings(projectIDs),
+	}
+}
+
+// PrimaryID returns the owning credential identifier.
+func (s *SharedCredential) PrimaryID() string {
+	if s == nil {
+		return ""
+	}
+	return s.primaryID
+}
+
+// Email returns the associated account email.
+func (s *SharedCredential) Email() string {
+	if s == nil {
+		return ""
+	}
+	return s.email
+}
+
+// ProjectIDs returns a snapshot of the configured project identifiers.
+func (s *SharedCredential) ProjectIDs() []string {
+	if s == nil {
+		return nil
+	}
+	return cloneStrings(s.projectIDs)
+}
+
+// MetadataSnapshot returns a deep copy of the stored OAuth metadata.
+func (s *SharedCredential) MetadataSnapshot() map[string]any {
+	if s == nil {
+		return nil
+	}
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+	return cloneMap(s.metadata)
+}
+
+// MergeMetadata merges the provided fields into the shared metadata and returns an updated copy.
+func (s *SharedCredential) MergeMetadata(values map[string]any) map[string]any {
+	if s == nil {
+		return nil
+	}
+	if len(values) == 0 {
+		return s.MetadataSnapshot()
+	}
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	if s.metadata == nil {
+		s.metadata = make(map[string]any, len(values))
+	}
+	for k, v := range values {
+		if v == nil {
+			delete(s.metadata, k)
+			continue
+		}
+		s.metadata[k] = v
+	}
+	return cloneMap(s.metadata)
+}
+
+// SetProjectIDs updates the stored project identifiers.
+func (s *SharedCredential) SetProjectIDs(ids []string) {
+	if s == nil {
+		return
+	}
+	s.mu.Lock()
+	s.projectIDs = cloneStrings(ids)
+	s.mu.Unlock()
+}
+
+// VirtualCredential tracks a per-project virtual auth entry that reuses a primary credential.
+type VirtualCredential struct {
+	ProjectID string
+	Parent    *SharedCredential
+}
+
+// NewVirtualCredential creates a virtual credential descriptor bound to the shared parent.
+func NewVirtualCredential(projectID string, parent *SharedCredential) *VirtualCredential {
+	return &VirtualCredential{ProjectID: strings.TrimSpace(projectID), Parent: parent}
+}
+
+// ResolveSharedCredential returns the shared credential backing the provided runtime payload.
+func ResolveSharedCredential(runtime any) *SharedCredential {
+	switch typed := runtime.(type) {
+	case *SharedCredential:
+		return typed
+	case *VirtualCredential:
+		return typed.Parent
+	default:
+		return nil
+	}
+}
+
+// IsVirtual reports whether the runtime payload represents a virtual credential.
+func IsVirtual(runtime any) bool {
+	if runtime == nil {
+		return false
+	}
+	_, ok := runtime.(*VirtualCredential)
+	return ok
+}
+
+func cloneMap(in map[string]any) map[string]any {
+	if len(in) == 0 {
+		return nil
+	}
+	out := make(map[string]any, len(in))
+	for k, v := range in {
+		out[k] = v
+	}
+	return out
+}
+
+func cloneStrings(in []string) []string {
+	if len(in) == 0 {
+		return nil
+	}
+	out := make([]string, len(in))
+	copy(out, in)
+	return out
+}

internal/translator/codex/claude/codex_claude_request.go

@@ -204,7 +204,7 @@ func ConvertClaudeRequestToCodex(modelName string, inputRawJSON []byte, _ bool)
 				}
 				tool, _ = sjson.Set(tool, "name", name)
 			}
-			tool, _ = sjson.SetRaw(tool, "parameters", toolResult.Get("input_schema").Raw)
+			tool, _ = sjson.SetRaw(tool, "parameters", normalizeToolParameters(toolResult.Get("input_schema").Raw))
 			tool, _ = sjson.Delete(tool, "input_schema")
 			tool, _ = sjson.Delete(tool, "parameters.$schema")
 			tool, _ = sjson.Set(tool, "strict", false)
@@ -289,7 +289,7 @@ func buildShortNameMap(names []string) map[string]string {
 		}
 		base := cand
 		for i := 1; ; i++ {
-			suffix := "~" + strconv.Itoa(i)
+			suffix := "_" + strconv.Itoa(i)
 			allowed := limit - len(suffix)
 			if allowed < 0 {
 				allowed = 0
@@ -334,3 +334,22 @@ func buildReverseMapFromClaudeOriginalToShort(original []byte) map[string]string
 	}
 	return m
 }
+
+// normalizeToolParameters ensures object schemas contain at least an empty properties map.
+func normalizeToolParameters(raw string) string {
+	raw = strings.TrimSpace(raw)
+	if raw == "" || raw == "null" || !gjson.Valid(raw) {
+		return `{"type":"object","properties":{}}`
+	}
+	schema := raw
+	result := gjson.Parse(raw)
+	schemaType := result.Get("type").String()
+	if schemaType == "" {
+		schema, _ = sjson.Set(schema, "type", "object")
+		schemaType = "object"
+	}
+	if schemaType == "object" && !result.Get("properties").Exists() {
+		schema, _ = sjson.SetRaw(schema, "properties", `{}`)
+	}
+	return schema
+}

internal/translator/codex/gemini/codex_gemini_request.go

@@ -310,7 +310,7 @@ func buildShortNameMap(names []string) map[string]string {
 		}
 		base := cand
 		for i := 1; ; i++ {
-			suffix := "~" + strconv.Itoa(i)
+			suffix := "_" + strconv.Itoa(i)
 			allowed := limit - len(suffix)
 			if allowed < 0 {
 				allowed = 0

internal/translator/codex/openai/chat-completions/codex_openai_request.go

@@ -361,7 +361,7 @@ func buildShortNameMap(names []string) map[string]string {
 		}
 		base := cand
 		for i := 1; ; i++ {
-			suffix := "~" + strconv.Itoa(i)
+			suffix := "_" + strconv.Itoa(i)
 			allowed := limit - len(suffix)
 			if allowed < 0 {
 				allowed = 0

internal/translator/codex/openai/responses/codex_openai-responses_request.go

@@ -22,6 +22,7 @@ func ConvertOpenAIResponsesRequestToCodex(modelName string, inputRawJSON []byte,
 	rawJSON, _ = sjson.DeleteBytes(rawJSON, "max_completion_tokens")
 	rawJSON, _ = sjson.DeleteBytes(rawJSON, "temperature")
 	rawJSON, _ = sjson.DeleteBytes(rawJSON, "top_p")
+	rawJSON, _ = sjson.DeleteBytes(rawJSON, "service_tier")
 
 	originalInstructions := ""
 	originalInstructionsText := ""

internal/translator/gemini-cli/claude/gemini-cli_claude_request.go

@@ -17,6 +17,8 @@ import (
 	"github.com/tidwall/sjson"
 )
 
+const geminiCLIClaudeThoughtSignature = "skip_thought_signature_validator"
+
 // ConvertClaudeRequestToCLI parses and transforms a Claude Code API request into Gemini CLI API format.
 // It extracts the model name, system instruction, message contents, and tool declarations
 // from the raw JSON request and returns them in the format expected by the Gemini CLI API.
@@ -89,7 +91,10 @@ func ConvertClaudeRequestToCLI(modelName string, inputRawJSON []byte, _ bool) []
 						functionArgs := contentResult.Get("input").String()
 						var args map[string]any
 						if err := json.Unmarshal([]byte(functionArgs), &args); err == nil {
-							clientContent.Parts = append(clientContent.Parts, client.Part{FunctionCall: &client.FunctionCall{Name: functionName, Args: args}})
+							clientContent.Parts = append(clientContent.Parts, client.Part{
+								FunctionCall:     &client.FunctionCall{Name: functionName, Args: args},
+								ThoughtSignature: geminiCLIClaudeThoughtSignature,
+							})
 						}
 					} else if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "tool_result" {
 						toolCallID := contentResult.Get("tool_use_id").String()
@@ -128,6 +133,7 @@ func ConvertClaudeRequestToCLI(modelName string, inputRawJSON []byte, _ bool) []
 				tool, _ := sjson.Delete(toolResult.Raw, "input_schema")
 				tool, _ = sjson.SetRaw(tool, "parametersJsonSchema", inputSchema)
 				tool, _ = sjson.Delete(tool, "strict")
+				tool, _ = sjson.Delete(tool, "input_examples")
 				var toolDeclaration any
 				if err := json.Unmarshal([]byte(tool), &toolDeclaration); err == nil {
 					tools[0].FunctionDeclarations = append(tools[0].FunctionDeclarations, toolDeclaration)

internal/translator/gemini-cli/openai/chat-completions/gemini-cli_openai_request.go

@@ -15,6 +15,8 @@ import (
 	"github.com/tidwall/sjson"
 )
 
+const geminiCLIFunctionThoughtSignature = "skip_thought_signature_validator"
+
 // ConvertOpenAIRequestToGeminiCLI converts an OpenAI Chat Completions request (raw JSON)
 // into a complete Gemini CLI request JSON. All JSON construction uses sjson and lookups use gjson.
 //
@@ -86,6 +88,15 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 		}
 	}
 
+	// For gemini-3-pro-preview, always send default thinkingConfig when none specified.
+	// This matches the official Gemini CLI behavior which always sends:
+	// { thinkingBudget: -1, includeThoughts: true }
+	// See: ai-gemini-cli/packages/core/src/config/defaultModelConfigs.ts
+	if !gjson.GetBytes(out, "request.generationConfig.thinkingConfig").Exists() && modelName == "gemini-3-pro-preview" {
+		out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
+		out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
+	}
+
 	// Temperature/top_p/top_k
 	if tr := gjson.GetBytes(rawJSON, "temperature"); tr.Exists() && tr.Type == gjson.Number {
 		out, _ = sjson.SetBytes(out, "request.generationConfig.temperature", tr.Num)
@@ -239,6 +250,7 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 							fargs := tc.Get("function.arguments").String()
 							node, _ = sjson.SetBytes(node, "parts."+itoa(p)+".functionCall.name", fname)
 							node, _ = sjson.SetRawBytes(node, "parts."+itoa(p)+".functionCall.args", []byte(fargs))
+							node, _ = sjson.SetBytes(node, "parts."+itoa(p)+".thoughtSignature", geminiCLIFunctionThoughtSignature)
 							p++
 							if fid != "" {
 								fIDs = append(fIDs, fid)
@@ -283,6 +295,17 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 						renamed, errRename := util.RenameKey(fnRaw, "parameters", "parametersJsonSchema")
 						if errRename != nil {
 							log.Warnf("Failed to rename parameters for tool '%s': %v", fn.Get("name").String(), errRename)
+							var errSet error
+							fnRaw, errSet = sjson.Set(fnRaw, "parametersJsonSchema.type", "object")
+							if errSet != nil {
+								log.Warnf("Failed to set default schema type for tool '%s': %v", fn.Get("name").String(), errSet)
+								continue
+							}
+							fnRaw, errSet = sjson.Set(fnRaw, "parametersJsonSchema.properties", map[string]interface{}{})
+							if errSet != nil {
+								log.Warnf("Failed to set default schema properties for tool '%s': %v", fn.Get("name").String(), errSet)
+								continue
+							}
 						} else {
 							fnRaw = renamed
 						}

internal/translator/gemini/claude/gemini_claude_request.go

@@ -17,6 +17,8 @@ import (
 	"github.com/tidwall/sjson"
 )
 
+const geminiClaudeThoughtSignature = "skip_thought_signature_validator"
+
 // ConvertClaudeRequestToGemini parses a Claude API request and returns a complete
 // Gemini CLI request body (as JSON bytes) ready to be sent via SendRawMessageStream.
 // All JSON transformations are performed using gjson/sjson.
@@ -82,7 +84,10 @@ func ConvertClaudeRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 						functionArgs := contentResult.Get("input").String()
 						var args map[string]any
 						if err := json.Unmarshal([]byte(functionArgs), &args); err == nil {
-							clientContent.Parts = append(clientContent.Parts, client.Part{FunctionCall: &client.FunctionCall{Name: functionName, Args: args}})
+							clientContent.Parts = append(clientContent.Parts, client.Part{
+								FunctionCall:     &client.FunctionCall{Name: functionName, Args: args},
+								ThoughtSignature: geminiClaudeThoughtSignature,
+							})
 						}
 					} else if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "tool_result" {
 						toolCallID := contentResult.Get("tool_use_id").String()
@@ -121,6 +126,7 @@ func ConvertClaudeRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 				tool, _ := sjson.Delete(toolResult.Raw, "input_schema")
 				tool, _ = sjson.SetRaw(tool, "parametersJsonSchema", inputSchema)
 				tool, _ = sjson.Delete(tool, "strict")
+				tool, _ = sjson.Delete(tool, "input_examples")
 				var toolDeclaration any
 				if err := json.Unmarshal([]byte(tool), &toolDeclaration); err == nil {
 					tools[0].FunctionDeclarations = append(tools[0].FunctionDeclarations, toolDeclaration)

internal/translator/gemini/openai/chat-completions/gemini_openai_request.go

@@ -15,6 +15,8 @@ import (
 	"github.com/tidwall/sjson"
 )
 
+const geminiFunctionThoughtSignature = "skip_thought_signature_validator"
+
 // ConvertOpenAIRequestToGemini converts an OpenAI Chat Completions request (raw JSON)
 // into a complete Gemini request JSON. All JSON construction uses sjson and lookups use gjson.
 //
@@ -264,6 +266,7 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 							fargs := tc.Get("function.arguments").String()
 							node, _ = sjson.SetBytes(node, "parts."+itoa(p)+".functionCall.name", fname)
 							node, _ = sjson.SetRawBytes(node, "parts."+itoa(p)+".functionCall.args", []byte(fargs))
+							node, _ = sjson.SetBytes(node, "parts."+itoa(p)+".thoughtSignature", geminiFunctionThoughtSignature)
 							p++
 							if fid != "" {
 								fIDs = append(fIDs, fid)
@@ -303,8 +306,34 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 			if t.Get("type").String() == "function" {
 				fn := t.Get("function")
 				if fn.Exists() && fn.IsObject() {
-					parametersJsonSchema, _ := util.RenameKey(fn.Raw, "parameters", "parametersJsonSchema")
-					out, _ = sjson.SetRawBytes(out, fdPath+".-1", []byte(parametersJsonSchema))
+					fnRaw := fn.Raw
+					if fn.Get("parameters").Exists() {
+						renamed, errRename := util.RenameKey(fnRaw, "parameters", "parametersJsonSchema")
+						if errRename != nil {
+							log.Warnf("Failed to rename parameters for tool '%s': %v", fn.Get("name").String(), errRename)
+						} else {
+							fnRaw = renamed
+						}
+					} else {
+						var errSet error
+						fnRaw, errSet = sjson.Set(fnRaw, "parametersJsonSchema.type", "object")
+						if errSet != nil {
+							log.Warnf("Failed to set default schema type for tool '%s': %v", fn.Get("name").String(), errSet)
+							continue
+						}
+						fnRaw, errSet = sjson.Set(fnRaw, "parametersJsonSchema.properties", map[string]interface{}{})
+						if errSet != nil {
+							log.Warnf("Failed to set default schema properties for tool '%s': %v", fn.Get("name").String(), errSet)
+							continue
+						}
+					}
+					fnRaw, _ = sjson.Delete(fnRaw, "strict")
+					tmp, errSet := sjson.SetRawBytes(out, fdPath+".-1", []byte(fnRaw))
+					if errSet != nil {
+						log.Warnf("Failed to append tool declaration for '%s': %v", fn.Get("name").String(), errSet)
+						continue
+					}
+					out = tmp
 				}
 			}
 		}

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

@@ -6,10 +6,13 @@ import (
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/translator/gemini/common"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
+	log "github.com/sirupsen/logrus"
 	"github.com/tidwall/gjson"
 	"github.com/tidwall/sjson"
 )
 
+const geminiResponsesThoughtSignature = "skip_thought_signature_validator"
+
 func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte, stream bool) []byte {
 	rawJSON := bytes.Clone(inputRawJSON)
 
@@ -108,6 +111,7 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 				modelContent := `{"role":"model","parts":[]}`
 				functionCall := `{"functionCall":{"name":"","args":{}}}`
 				functionCall, _ = sjson.Set(functionCall, "functionCall.name", name)
+				functionCall, _ = sjson.Set(functionCall, "thoughtSignature", geminiResponsesThoughtSignature)
 
 				// Parse arguments JSON string and set as args object
 				if arguments != "" {
@@ -156,6 +160,11 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 
 			return true
 		})
+	} else if input.Exists() && input.Type == gjson.String {
+		// Simple string input conversion to user message
+		userContent := `{"role":"user","parts":[{"text":""}]}`
+		userContent, _ = sjson.Set(userContent, "parts.0.text", input.String())
+		out, _ = sjson.SetRaw(out, "contents.-1", userContent)
 	}
 
 	// Convert tools to Gemini functionDeclarations format
@@ -286,6 +295,17 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 			}
 		}
 	}
+
+	// For gemini-3-pro-preview, always send default thinkingConfig when none specified.
+	// This matches the official Gemini CLI behavior which always sends:
+	// { thinkingBudget: -1, includeThoughts: true }
+	// See: ai-gemini-cli/packages/core/src/config/defaultModelConfigs.ts
+	if !gjson.Get(out, "generationConfig.thinkingConfig").Exists() && modelName == "gemini-3-pro-preview" {
+		out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
+		out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
+		log.Debugf("Applied default thinkingConfig for gemini-3-pro-preview (matches Gemini CLI): thinkingBudget=-1, include_thoughts=true")
+	}
+
 	result := []byte(out)
 	result = common.AttachDefaultSafetySettings(result, "safetySettings")
 	return result

internal/translator/openai/claude/openai_claude_response.go

@@ -32,6 +32,8 @@ type ConvertOpenAIResponseToAnthropicParams struct {
 	ToolCallsAccumulator map[int]*ToolCallAccumulator
 	// Track if text content block has been started
 	TextContentBlockStarted bool
+	// Track if thinking content block has been started
+	ThinkingContentBlockStarted bool
 	// Track finish reason for later use
 	FinishReason string
 	// Track if content blocks have been stopped
@@ -40,6 +42,16 @@ type ConvertOpenAIResponseToAnthropicParams struct {
 	MessageDeltaSent bool
 	// Track if message_start has been sent
 	MessageStarted bool
+	// Track if message_stop has been sent
+	MessageStopSent bool
+	// Tool call content block index mapping
+	ToolCallBlockIndexes map[int]int
+	// Index assigned to text content block
+	TextContentBlockIndex int
+	// Index assigned to thinking content block
+	ThinkingContentBlockIndex int
+	// Next available content block index
+	NextContentBlockIndex int
 }
 
 // ToolCallAccumulator holds the state for accumulating tool call data
@@ -64,15 +76,20 @@ type ToolCallAccumulator struct {
 func ConvertOpenAIResponseToClaude(_ context.Context, _ string, originalRequestRawJSON, requestRawJSON, rawJSON []byte, param *any) []string {
 	if *param == nil {
 		*param = &ConvertOpenAIResponseToAnthropicParams{
-			MessageID:               "",
-			Model:                   "",
-			CreatedAt:               0,
-			ContentAccumulator:      strings.Builder{},
-			ToolCallsAccumulator:    nil,
-			TextContentBlockStarted: false,
-			FinishReason:            "",
-			ContentBlocksStopped:    false,
-			MessageDeltaSent:        false,
+			MessageID:                   "",
+			Model:                       "",
+			CreatedAt:                   0,
+			ContentAccumulator:          strings.Builder{},
+			ToolCallsAccumulator:        nil,
+			TextContentBlockStarted:     false,
+			ThinkingContentBlockStarted: false,
+			FinishReason:                "",
+			ContentBlocksStopped:        false,
+			MessageDeltaSent:            false,
+			ToolCallBlockIndexes:        make(map[int]int),
+			TextContentBlockIndex:       -1,
+			ThinkingContentBlockIndex:   -1,
+			NextContentBlockIndex:       0,
 		}
 	}
 
@@ -138,13 +155,56 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 			// Don't send content_block_start for text here - wait for actual content
 		}
 
+		// Handle reasoning content delta
+		if reasoning := delta.Get("reasoning_content"); reasoning.Exists() {
+			for _, reasoningText := range collectOpenAIReasoningTexts(reasoning) {
+				if reasoningText == "" {
+					continue
+				}
+				stopTextContentBlock(param, &results)
+				if !param.ThinkingContentBlockStarted {
+					if param.ThinkingContentBlockIndex == -1 {
+						param.ThinkingContentBlockIndex = param.NextContentBlockIndex
+						param.NextContentBlockIndex++
+					}
+					contentBlockStart := map[string]interface{}{
+						"type":  "content_block_start",
+						"index": param.ThinkingContentBlockIndex,
+						"content_block": map[string]interface{}{
+							"type":     "thinking",
+							"thinking": "",
+						},
+					}
+					contentBlockStartJSON, _ := json.Marshal(contentBlockStart)
+					results = append(results, "event: content_block_start\ndata: "+string(contentBlockStartJSON)+"\n\n")
+					param.ThinkingContentBlockStarted = true
+				}
+
+				thinkingDelta := map[string]interface{}{
+					"type":  "content_block_delta",
+					"index": param.ThinkingContentBlockIndex,
+					"delta": map[string]interface{}{
+						"type":     "thinking_delta",
+						"thinking": reasoningText,
+					},
+				}
+				thinkingDeltaJSON, _ := json.Marshal(thinkingDelta)
+				results = append(results, "event: content_block_delta\ndata: "+string(thinkingDeltaJSON)+"\n\n")
+			}
+		}
+
 		// Handle content delta
 		if content := delta.Get("content"); content.Exists() && content.String() != "" {
 			// Send content_block_start for text if not already sent
 			if !param.TextContentBlockStarted {
+				stopThinkingContentBlock(param, &results)
+				if param.TextContentBlockIndex == -1 {
+					param.TextContentBlockIndex = param.NextContentBlockIndex
+					param.NextContentBlockIndex++
+				}
 				contentBlockStart := map[string]interface{}{
 					"type":  "content_block_start",
-					"index": 0,
+					"index": param.TextContentBlockIndex,
 					"content_block": map[string]interface{}{
 						"type": "text",
 						"text": "",
@@ -157,7 +217,7 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 
 			contentDelta := map[string]interface{}{
 				"type":  "content_block_delta",
-				"index": 0,
+				"index": param.TextContentBlockIndex,
 				"delta": map[string]interface{}{
 					"type": "text_delta",
 					"text": content.String(),
@@ -178,6 +238,7 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 
 			toolCalls.ForEach(func(_, toolCall gjson.Result) bool {
 				index := int(toolCall.Get("index").Int())
+				blockIndex := param.toolContentBlockIndex(index)
 
 				// Initialize accumulator if needed
 				if _, exists := param.ToolCallsAccumulator[index]; !exists {
@@ -196,20 +257,14 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 					if name := function.Get("name"); name.Exists() {
 						accumulator.Name = name.String()
 
-						if param.TextContentBlockStarted {
-							param.TextContentBlockStarted = false
-							contentBlockStop := map[string]interface{}{
-								"type":  "content_block_stop",
-								"index": index,
-							}
-							contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
-							results = append(results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
-						}
+						stopThinkingContentBlock(param, &results)
+
+						stopTextContentBlock(param, &results)
 
 						// Send content_block_start for tool_use
 						contentBlockStart := map[string]interface{}{
 							"type":  "content_block_start",
-							"index": index + 1, // Offset by 1 since text is at index 0
+							"index": blockIndex,
 							"content_block": map[string]interface{}{
 								"type":  "tool_use",
 								"id":    accumulator.ID,
@@ -240,26 +295,32 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 		reason := finishReason.String()
 		param.FinishReason = reason
 
-		// Send content_block_stop for text if text content block was started
-		if param.TextContentBlockStarted && !param.ContentBlocksStopped {
+		// Send content_block_stop for thinking content if needed
+		if param.ThinkingContentBlockStarted {
 			contentBlockStop := map[string]interface{}{
 				"type":  "content_block_stop",
-				"index": 0,
+				"index": param.ThinkingContentBlockIndex,
 			}
 			contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
 			results = append(results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
+			param.ThinkingContentBlockStarted = false
+			param.ThinkingContentBlockIndex = -1
 		}
 
+		// Send content_block_stop for text if text content block was started
+		stopTextContentBlock(param, &results)
+
 		// Send content_block_stop for any tool calls
 		if !param.ContentBlocksStopped {
 			for index := range param.ToolCallsAccumulator {
 				accumulator := param.ToolCallsAccumulator[index]
+				blockIndex := param.toolContentBlockIndex(index)
 
 				// Send complete input_json_delta with all accumulated arguments
 				if accumulator.Arguments.Len() > 0 {
 					inputDelta := map[string]interface{}{
 						"type":  "content_block_delta",
-						"index": index + 1,
+						"index": blockIndex,
 						"delta": map[string]interface{}{
 							"type":         "input_json_delta",
 							"partial_json": util.FixJSON(accumulator.Arguments.String()),
@@ -271,10 +332,11 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 
 				contentBlockStop := map[string]interface{}{
 					"type":  "content_block_stop",
-					"index": index + 1,
+					"index": blockIndex,
 				}
 				contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
 				results = append(results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
+				delete(param.ToolCallBlockIndexes, index)
 			}
 			param.ContentBlocksStopped = true
 		}
@@ -284,29 +346,38 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 
 	// Handle usage information separately (this comes in a later chunk)
 	// Only process if usage has actual values (not null)
-	if usage := root.Get("usage"); usage.Exists() && usage.Type != gjson.Null && param.FinishReason != "" {
-		// Check if usage has actual token counts
-		promptTokens := usage.Get("prompt_tokens")
-		completionTokens := usage.Get("completion_tokens")
+	if param.FinishReason != "" {
+		usage := root.Get("usage")
+		var inputTokens, outputTokens int64
+		if usage.Exists() && usage.Type != gjson.Null {
+			// Check if usage has actual token counts
+			promptTokens := usage.Get("prompt_tokens")
+			completionTokens := usage.Get("completion_tokens")
 
-		if promptTokens.Exists() && completionTokens.Exists() {
-			// Send message_delta with usage
-			messageDelta := map[string]interface{}{
-				"type": "message_delta",
-				"delta": map[string]interface{}{
-					"stop_reason":   mapOpenAIFinishReasonToAnthropic(param.FinishReason),
-					"stop_sequence": nil,
-				},
-				"usage": map[string]interface{}{
-					"input_tokens":  promptTokens.Int(),
-					"output_tokens": completionTokens.Int(),
-				},
+			if promptTokens.Exists() && completionTokens.Exists() {
+				inputTokens = promptTokens.Int()
+				outputTokens = completionTokens.Int()
 			}
-
-			messageDeltaJSON, _ := json.Marshal(messageDelta)
-			results = append(results, "event: message_delta\ndata: "+string(messageDeltaJSON)+"\n\n")
-			param.MessageDeltaSent = true
 		}
+		// Send message_delta with usage
+		messageDelta := map[string]interface{}{
+			"type": "message_delta",
+			"delta": map[string]interface{}{
+				"stop_reason":   mapOpenAIFinishReasonToAnthropic(param.FinishReason),
+				"stop_sequence": nil,
+			},
+			"usage": map[string]interface{}{
+				"input_tokens":  inputTokens,
+				"output_tokens": outputTokens,
+			},
+		}
+
+		messageDeltaJSON, _ := json.Marshal(messageDelta)
+		results = append(results, "event: message_delta\ndata: "+string(messageDeltaJSON)+"\n\n")
+		param.MessageDeltaSent = true
+
+		emitMessageStopIfNeeded(param, &results)
+
 	}
 
 	return results
@@ -316,6 +387,49 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 func convertOpenAIDoneToAnthropic(param *ConvertOpenAIResponseToAnthropicParams) []string {
 	var results []string
 
+	// Ensure all content blocks are stopped before final events
+	if param.ThinkingContentBlockStarted {
+		contentBlockStop := map[string]interface{}{
+			"type":  "content_block_stop",
+			"index": param.ThinkingContentBlockIndex,
+		}
+		contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
+		results = append(results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
+		param.ThinkingContentBlockStarted = false
+		param.ThinkingContentBlockIndex = -1
+	}
+
+	stopTextContentBlock(param, &results)
+
+	if !param.ContentBlocksStopped {
+		for index := range param.ToolCallsAccumulator {
+			accumulator := param.ToolCallsAccumulator[index]
+			blockIndex := param.toolContentBlockIndex(index)
+
+			if accumulator.Arguments.Len() > 0 {
+				inputDelta := map[string]interface{}{
+					"type":  "content_block_delta",
+					"index": blockIndex,
+					"delta": map[string]interface{}{
+						"type":         "input_json_delta",
+						"partial_json": util.FixJSON(accumulator.Arguments.String()),
+					},
+				}
+				inputDeltaJSON, _ := json.Marshal(inputDelta)
+				results = append(results, "event: content_block_delta\ndata: "+string(inputDeltaJSON)+"\n\n")
+			}
+
+			contentBlockStop := map[string]interface{}{
+				"type":  "content_block_stop",
+				"index": blockIndex,
+			}
+			contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
+			results = append(results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
+			delete(param.ToolCallBlockIndexes, index)
+		}
+		param.ContentBlocksStopped = true
+	}
+
 	// If we haven't sent message_delta yet (no usage info was received), send it now
 	if param.FinishReason != "" && !param.MessageDeltaSent {
 		messageDelta := map[string]interface{}{
@@ -331,8 +445,7 @@ func convertOpenAIDoneToAnthropic(param *ConvertOpenAIResponseToAnthropicParams)
 		param.MessageDeltaSent = true
 	}
 
-	// Send message_stop
-	results = append(results, "event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n")
+	emitMessageStopIfNeeded(param, &results)
 
 	return results
 }
@@ -361,6 +474,18 @@ func convertOpenAINonStreamingToAnthropic(rawJSON []byte) []string {
 
 	if choices := root.Get("choices"); choices.Exists() && choices.IsArray() {
 		choice := choices.Array()[0] // Take first choice
+		reasoningNode := choice.Get("message.reasoning_content")
+		allReasoning := collectOpenAIReasoningTexts(reasoningNode)
+
+		for _, reasoningText := range allReasoning {
+			if reasoningText == "" {
+				continue
+			}
+			contentBlocks = append(contentBlocks, map[string]interface{}{
+				"type":     "thinking",
+				"thinking": reasoningText,
+			})
+		}
 
 		// Handle text content
 		if content := choice.Get("message.content"); content.Exists() && content.String() != "" {
@@ -412,6 +537,17 @@ func convertOpenAINonStreamingToAnthropic(rawJSON []byte) []string {
 		response["usage"] = map[string]interface{}{
 			"input_tokens":  usage.Get("prompt_tokens").Int(),
 			"output_tokens": usage.Get("completion_tokens").Int(),
+			"reasoning_tokens": func() int64 {
+				if v := usage.Get("completion_tokens_details.reasoning_tokens"); v.Exists() {
+					return v.Int()
+				}
+				return 0
+			}(),
+		}
+	} else {
+		response["usage"] = map[string]interface{}{
+			"input_tokens":  0,
+			"output_tokens": 0,
 		}
 	}
 
@@ -437,6 +573,84 @@ func mapOpenAIFinishReasonToAnthropic(openAIReason string) string {
 	}
 }
 
+func (p *ConvertOpenAIResponseToAnthropicParams) toolContentBlockIndex(openAIToolIndex int) int {
+	if idx, ok := p.ToolCallBlockIndexes[openAIToolIndex]; ok {
+		return idx
+	}
+	idx := p.NextContentBlockIndex
+	p.NextContentBlockIndex++
+	p.ToolCallBlockIndexes[openAIToolIndex] = idx
+	return idx
+}
+
+func collectOpenAIReasoningTexts(node gjson.Result) []string {
+	var texts []string
+	if !node.Exists() {
+		return texts
+	}
+
+	if node.IsArray() {
+		node.ForEach(func(_, value gjson.Result) bool {
+			texts = append(texts, collectOpenAIReasoningTexts(value)...)
+			return true
+		})
+		return texts
+	}
+
+	switch node.Type {
+	case gjson.String:
+		if text := strings.TrimSpace(node.String()); text != "" {
+			texts = append(texts, text)
+		}
+	case gjson.JSON:
+		if text := node.Get("text"); text.Exists() {
+			if trimmed := strings.TrimSpace(text.String()); trimmed != "" {
+				texts = append(texts, trimmed)
+			}
+		} else if raw := strings.TrimSpace(node.Raw); raw != "" && !strings.HasPrefix(raw, "{") && !strings.HasPrefix(raw, "[") {
+			texts = append(texts, raw)
+		}
+	}
+
+	return texts
+}
+
+func stopThinkingContentBlock(param *ConvertOpenAIResponseToAnthropicParams, results *[]string) {
+	if !param.ThinkingContentBlockStarted {
+		return
+	}
+	contentBlockStop := map[string]interface{}{
+		"type":  "content_block_stop",
+		"index": param.ThinkingContentBlockIndex,
+	}
+	contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
+	*results = append(*results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
+	param.ThinkingContentBlockStarted = false
+	param.ThinkingContentBlockIndex = -1
+}
+
+func emitMessageStopIfNeeded(param *ConvertOpenAIResponseToAnthropicParams, results *[]string) {
+	if param.MessageStopSent {
+		return
+	}
+	*results = append(*results, "event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n")
+	param.MessageStopSent = true
+}
+
+func stopTextContentBlock(param *ConvertOpenAIResponseToAnthropicParams, results *[]string) {
+	if !param.TextContentBlockStarted {
+		return
+	}
+	contentBlockStop := map[string]interface{}{
+		"type":  "content_block_stop",
+		"index": param.TextContentBlockIndex,
+	}
+	contentBlockStopJSON, _ := json.Marshal(contentBlockStop)
+	*results = append(*results, "event: content_block_stop\ndata: "+string(contentBlockStopJSON)+"\n\n")
+	param.TextContentBlockStarted = false
+	param.TextContentBlockIndex = -1
+}
+
 // ConvertOpenAIResponseToClaudeNonStream converts a non-streaming OpenAI response to a non-streaming Anthropic response.
 //
 // Parameters:
@@ -564,6 +778,18 @@ func ConvertOpenAIResponseToClaudeNonStream(_ context.Context, _ string, origina
 				}
 			}
 
+			if reasoning := message.Get("reasoning_content"); reasoning.Exists() {
+				for _, reasoningText := range collectOpenAIReasoningTexts(reasoning) {
+					if reasoningText == "" {
+						continue
+					}
+					contentBlocks = append(contentBlocks, map[string]interface{}{
+						"type":     "thinking",
+						"thinking": reasoningText,
+					})
+				}
+			}
+
 			if toolCalls := message.Get("tool_calls"); toolCalls.Exists() && toolCalls.IsArray() {
 				toolCalls.ForEach(func(_, toolCall gjson.Result) bool {
 					hasToolCall = true
@@ -601,6 +827,8 @@ func ConvertOpenAIResponseToClaudeNonStream(_ context.Context, _ string, origina
 		usageJSON, _ = sjson.Set(usageJSON, "output_tokens", respUsage.Get("completion_tokens").Int())
 		parsedUsage := gjson.Parse(usageJSON).Value().(map[string]interface{})
 		response["usage"] = parsedUsage
+	} else {
+		response["usage"] = `{"input_tokens":0,"output_tokens":0}`
 	}
 
 	if response["stop_reason"] == nil {

internal/translator/openai/gemini/openai_gemini_response.go

@@ -89,6 +89,9 @@ func ConvertOpenAIResponseToGemini(_ context.Context, _ string, originalRequestR
 					"candidatesTokenCount": usage.Get("completion_tokens").Int(),
 					"totalTokenCount":      usage.Get("total_tokens").Int(),
 				}
+				if reasoningTokens := reasoningTokensFromUsage(usage); reasoningTokens > 0 {
+					usageObj["thoughtsTokenCount"] = reasoningTokens
+				}
 				template, _ = sjson.Set(template, "usageMetadata", usageObj)
 				return []string{template}
 			}
@@ -108,6 +111,7 @@ func ConvertOpenAIResponseToGemini(_ context.Context, _ string, originalRequestR
 
 			_ = int(choice.Get("index").Int()) // choiceIdx not used in streaming
 			delta := choice.Get("delta")
+			baseTemplate := template
 
 			// Handle role (only in first chunk)
 			if role := delta.Get("role"); role.Exists() && (*param).(*ConvertOpenAIResponseToGeminiParams).IsFirstChunk {
@@ -120,6 +124,26 @@ func ConvertOpenAIResponseToGemini(_ context.Context, _ string, originalRequestR
 				return true
 			}
 
+			var chunkOutputs []string
+
+			// Handle reasoning/thinking delta
+			if reasoning := delta.Get("reasoning_content"); reasoning.Exists() {
+				for _, reasoningText := range extractReasoningTexts(reasoning) {
+					if reasoningText == "" {
+						continue
+					}
+					reasoningTemplate := baseTemplate
+					parts := []interface{}{
+						map[string]interface{}{
+							"thought": true,
+							"text":    reasoningText,
+						},
+					}
+					reasoningTemplate, _ = sjson.Set(reasoningTemplate, "candidates.0.content.parts", parts)
+					chunkOutputs = append(chunkOutputs, reasoningTemplate)
+				}
+			}
+
 			// Handle content delta
 			if content := delta.Get("content"); content.Exists() && content.String() != "" {
 				contentText := content.String()
@@ -131,8 +155,13 @@ func ConvertOpenAIResponseToGemini(_ context.Context, _ string, originalRequestR
 						"text": contentText,
 					},
 				}
-				template, _ = sjson.Set(template, "candidates.0.content.parts", parts)
-				results = append(results, template)
+				contentTemplate := baseTemplate
+				contentTemplate, _ = sjson.Set(contentTemplate, "candidates.0.content.parts", parts)
+				chunkOutputs = append(chunkOutputs, contentTemplate)
+			}
+
+			if len(chunkOutputs) > 0 {
+				results = append(results, chunkOutputs...)
 				return true
 			}
 
@@ -231,6 +260,9 @@ func ConvertOpenAIResponseToGemini(_ context.Context, _ string, originalRequestR
 					"candidatesTokenCount": usage.Get("completion_tokens").Int(),
 					"totalTokenCount":      usage.Get("total_tokens").Int(),
 				}
+				if reasoningTokens := reasoningTokensFromUsage(usage); reasoningTokens > 0 {
+					usageObj["thoughtsTokenCount"] = reasoningTokens
+				}
 				template, _ = sjson.Set(template, "usageMetadata", usageObj)
 				results = append(results, template)
 				return true
@@ -549,6 +581,19 @@ func ConvertOpenAIResponseToGeminiNonStream(_ context.Context, _ string, origina
 
 			var parts []interface{}
 
+			// Handle reasoning content before visible text
+			if reasoning := message.Get("reasoning_content"); reasoning.Exists() {
+				for _, reasoningText := range extractReasoningTexts(reasoning) {
+					if reasoningText == "" {
+						continue
+					}
+					parts = append(parts, map[string]interface{}{
+						"thought": true,
+						"text":    reasoningText,
+					})
+				}
+			}
+
 			// Handle content first
 			if content := message.Get("content"); content.Exists() && content.String() != "" {
 				parts = append(parts, map[string]interface{}{
@@ -605,6 +650,9 @@ func ConvertOpenAIResponseToGeminiNonStream(_ context.Context, _ string, origina
 			"candidatesTokenCount": usage.Get("completion_tokens").Int(),
 			"totalTokenCount":      usage.Get("total_tokens").Int(),
 		}
+		if reasoningTokens := reasoningTokensFromUsage(usage); reasoningTokens > 0 {
+			usageObj["thoughtsTokenCount"] = reasoningTokens
+		}
 		out, _ = sjson.Set(out, "usageMetadata", usageObj)
 	}
 
@@ -614,3 +662,43 @@ func ConvertOpenAIResponseToGeminiNonStream(_ context.Context, _ string, origina
 func GeminiTokenCount(ctx context.Context, count int64) string {
 	return fmt.Sprintf(`{"totalTokens":%d,"promptTokensDetails":[{"modality":"TEXT","tokenCount":%d}]}`, count, count)
 }
+
+func reasoningTokensFromUsage(usage gjson.Result) int64 {
+	if usage.Exists() {
+		if v := usage.Get("completion_tokens_details.reasoning_tokens"); v.Exists() {
+			return v.Int()
+		}
+		if v := usage.Get("output_tokens_details.reasoning_tokens"); v.Exists() {
+			return v.Int()
+		}
+	}
+	return 0
+}
+
+func extractReasoningTexts(node gjson.Result) []string {
+	var texts []string
+	if !node.Exists() {
+		return texts
+	}
+
+	if node.IsArray() {
+		node.ForEach(func(_, value gjson.Result) bool {
+			texts = append(texts, extractReasoningTexts(value)...)
+			return true
+		})
+		return texts
+	}
+
+	switch node.Type {
+	case gjson.String:
+		texts = append(texts, node.String())
+	case gjson.JSON:
+		if text := node.Get("text"); text.Exists() {
+			texts = append(texts, text.String())
+		} else if raw := strings.TrimSpace(node.Raw); raw != "" && !strings.HasPrefix(raw, "{") && !strings.HasPrefix(raw, "[") {
+			texts = append(texts, raw)
+		}
+	}
+
+	return texts
+}

internal/usage/logger_plugin.go

@@ -90,6 +90,7 @@ type modelStats struct {
 type RequestDetail struct {
 	Timestamp time.Time  `json:"timestamp"`
 	Source    string     `json:"source"`
+	AuthIndex uint64     `json:"auth_index"`
 	Tokens    TokenStats `json:"tokens"`
 	Failed    bool       `json:"failed"`
 }
@@ -197,6 +198,7 @@ func (s *RequestStatistics) Record(ctx context.Context, record coreusage.Record)
 	s.updateAPIStats(stats, modelName, RequestDetail{
 		Timestamp: timestamp,
 		Source:    record.Source,
+		AuthIndex: record.AuthIndex,
 		Tokens:    detail,
 		Failed:    failed,
 	})

internal/util/provider.go

@@ -9,6 +9,7 @@ import (
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
+	log "github.com/sirupsen/logrus"
 )
 
 // GetProviderName determines all AI service providers capable of serving a registered model.
@@ -59,6 +60,30 @@ func GetProviderName(modelName string) []string {
 	return providers
 }
 
+// ResolveAutoModel resolves the "auto" model name to an actual available model.
+// It uses an empty handler type to get any available model from the registry.
+//
+// Parameters:
+//   - modelName: The model name to check (should be "auto")
+//
+// Returns:
+//   - string: The resolved model name, or the original if not "auto" or resolution fails
+func ResolveAutoModel(modelName string) string {
+	if modelName != "auto" {
+		return modelName
+	}
+
+	// Use empty string as handler type to get any available model
+	firstModel, err := registry.GetGlobalRegistry().GetFirstAvailableModel("")
+	if err != nil {
+		log.Warnf("Failed to resolve 'auto' model: %v, falling back to original model name", err)
+		return modelName
+	}
+
+	log.Infof("Resolved 'auto' model to: %s", firstModel)
+	return firstModel
+}
+
 // IsOpenAICompatibilityAlias checks if the given model name is an alias
 // configured for OpenAI compatibility routing.
 //

internal/watcher/watcher.go

@@ -21,6 +21,7 @@ import (
 
 	"github.com/fsnotify/fsnotify"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/runtime/geminicli"
 	"gopkg.in/yaml.v3"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
@@ -476,8 +477,10 @@ func (w *Watcher) processEvents(ctx context.Context) {
 // handleEvent processes individual file system events
 func (w *Watcher) handleEvent(event fsnotify.Event) {
 	// Filter only relevant events: config file or auth-dir JSON files.
-	isConfigEvent := event.Name == w.configPath && (event.Op&fsnotify.Write == fsnotify.Write || event.Op&fsnotify.Create == fsnotify.Create)
-	isAuthJSON := strings.HasPrefix(event.Name, w.authDir) && strings.HasSuffix(event.Name, ".json")
+	configOps := fsnotify.Write | fsnotify.Create | fsnotify.Rename
+	isConfigEvent := event.Name == w.configPath && event.Op&configOps != 0
+	authOps := fsnotify.Create | fsnotify.Write | fsnotify.Remove | fsnotify.Rename
+	isAuthJSON := strings.HasPrefix(event.Name, w.authDir) && strings.HasSuffix(event.Name, ".json") && event.Op&authOps != 0
 	if !isConfigEvent && !isAuthJSON {
 		// Ignore unrelated files (e.g., cookie snapshots *.cookie) and other noise.
 		return
@@ -495,18 +498,19 @@ func (w *Watcher) handleEvent(event fsnotify.Event) {
 
 	// Handle auth directory changes incrementally (.json only)
 	fmt.Printf("auth file changed (%s): %s, processing incrementally\n", event.Op.String(), filepath.Base(event.Name))
-	if event.Op&fsnotify.Create == fsnotify.Create || event.Op&fsnotify.Write == fsnotify.Write {
-		w.addOrUpdateClient(event.Name)
-	} else if event.Op&fsnotify.Remove == fsnotify.Remove {
-		// Atomic replace on some platforms may surface as Remove+Create for the target path.
-		// Wait briefly; if the file exists again, treat as update instead of removal.
+	if event.Op&(fsnotify.Remove|fsnotify.Rename) != 0 {
+		// Atomic replace on some platforms may surface as Rename (or Remove) before the new file is ready.
+		// Wait briefly; if the path exists again, treat as an update instead of removal.
 		time.Sleep(replaceCheckDelay)
 		if _, statErr := os.Stat(event.Name); statErr == nil {
-			// File exists after a short delay; handle as an update.
 			w.addOrUpdateClient(event.Name)
 			return
 		}
 		w.removeClient(event.Name)
+		return
+	}
+	if event.Op&(fsnotify.Create|fsnotify.Write) != 0 {
+		w.addOrUpdateClient(event.Name)
 	}
 }
 
@@ -1026,11 +1030,119 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			CreatedAt: now,
 			UpdatedAt: now,
 		}
+		if provider == "gemini-cli" {
+			if virtuals := synthesizeGeminiVirtualAuths(a, metadata, now); len(virtuals) > 0 {
+				out = append(out, a)
+				out = append(out, virtuals...)
+				continue
+			}
+		}
 		out = append(out, a)
 	}
 	return out
 }
 
+func synthesizeGeminiVirtualAuths(primary *coreauth.Auth, metadata map[string]any, now time.Time) []*coreauth.Auth {
+	if primary == nil || metadata == nil {
+		return nil
+	}
+	projects := splitGeminiProjectIDs(metadata)
+	if len(projects) <= 1 {
+		return nil
+	}
+	email, _ := metadata["email"].(string)
+	shared := geminicli.NewSharedCredential(primary.ID, email, metadata, projects)
+	primary.Disabled = true
+	primary.Status = coreauth.StatusDisabled
+	primary.Runtime = shared
+	if primary.Attributes == nil {
+		primary.Attributes = make(map[string]string)
+	}
+	primary.Attributes["gemini_virtual_primary"] = "true"
+	primary.Attributes["virtual_children"] = strings.Join(projects, ",")
+	source := primary.Attributes["source"]
+	authPath := primary.Attributes["path"]
+	originalProvider := primary.Provider
+	if originalProvider == "" {
+		originalProvider = "gemini-cli"
+	}
+	label := primary.Label
+	if label == "" {
+		label = originalProvider
+	}
+	virtuals := make([]*coreauth.Auth, 0, len(projects))
+	for _, projectID := range projects {
+		attrs := map[string]string{
+			"runtime_only":           "true",
+			"gemini_virtual_parent":  primary.ID,
+			"gemini_virtual_project": projectID,
+		}
+		if source != "" {
+			attrs["source"] = source
+		}
+		if authPath != "" {
+			attrs["path"] = authPath
+		}
+		metadataCopy := map[string]any{
+			"email":             email,
+			"project_id":        projectID,
+			"virtual":           true,
+			"virtual_parent_id": primary.ID,
+			"type":              metadata["type"],
+		}
+		proxy := strings.TrimSpace(primary.ProxyURL)
+		if proxy != "" {
+			metadataCopy["proxy_url"] = proxy
+		}
+		virtual := &coreauth.Auth{
+			ID:         buildGeminiVirtualID(primary.ID, projectID),
+			Provider:   originalProvider,
+			Label:      fmt.Sprintf("%s [%s]", label, projectID),
+			Status:     coreauth.StatusActive,
+			Attributes: attrs,
+			Metadata:   metadataCopy,
+			ProxyURL:   primary.ProxyURL,
+			CreatedAt:  now,
+			UpdatedAt:  now,
+			Runtime:    geminicli.NewVirtualCredential(projectID, shared),
+		}
+		virtuals = append(virtuals, virtual)
+	}
+	return virtuals
+}
+
+func splitGeminiProjectIDs(metadata map[string]any) []string {
+	raw, _ := metadata["project_id"].(string)
+	trimmed := strings.TrimSpace(raw)
+	if trimmed == "" {
+		return nil
+	}
+	parts := strings.Split(trimmed, ",")
+	result := make([]string, 0, len(parts))
+	seen := make(map[string]struct{}, len(parts))
+	for _, part := range parts {
+		id := strings.TrimSpace(part)
+		if id == "" {
+			continue
+		}
+		if _, ok := seen[id]; ok {
+			continue
+		}
+		seen[id] = struct{}{}
+		result = append(result, id)
+	}
+	return result
+}
+
+func buildGeminiVirtualID(baseID, projectID string) string {
+	project := strings.TrimSpace(projectID)
+	if project == "" {
+		project = "project"
+	}
+	replacer := strings.NewReplacer("/", "_", "\\", "_", " ", "_")
+	return fmt.Sprintf("%s::%s", baseID, replacer.Replace(project))
+}
+
 // buildCombinedClientMap merges file-based clients with API key clients from the cache.
 // buildCombinedClientMap removed
 

sdk/api/handlers/handlers.go

@@ -295,11 +295,14 @@ func (h *BaseAPIHandler) ExecuteStreamWithAuthManager(ctx context.Context, handl
 }
 
 func (h *BaseAPIHandler) getRequestDetails(modelName string) (providers []string, normalizedModel string, metadata map[string]any, err *interfaces.ErrorMessage) {
-	providerName, extractedModelName, isDynamic := h.parseDynamicModel(modelName)
+	// Resolve "auto" model to an actual available model first
+	resolvedModelName := util.ResolveAutoModel(modelName)
+	
+	providerName, extractedModelName, isDynamic := h.parseDynamicModel(resolvedModelName)
 
 	// First, normalize the model name to handle suffixes like "-thinking-128"
 	// This needs to happen before determining the provider for non-dynamic models.
-	normalizedModel, metadata = normalizeModelMetadata(modelName)
+	normalizedModel, metadata = normalizeModelMetadata(resolvedModelName)
 
 	if isDynamic {
 		providers = []string{providerName}

sdk/cliproxy/auth/manager.go

@@ -62,6 +62,8 @@ type Result struct {
 	Model string
 	// Success marks whether the execution succeeded.
 	Success bool
+	// RetryAfter carries a provider supplied retry hint (e.g. 429 retryDelay).
+	RetryAfter *time.Duration
 	// Error describes the failure when Success is false.
 	Error *Error
 }
@@ -169,6 +171,7 @@ func (m *Manager) Register(ctx context.Context, auth *Auth) (*Auth, error) {
 	if auth == nil {
 		return nil, nil
 	}
+	auth.EnsureIndex()
 	if auth.ID == "" {
 		auth.ID = uuid.NewString()
 	}
@@ -185,6 +188,7 @@ func (m *Manager) Update(ctx context.Context, auth *Auth) (*Auth, error) {
 	if auth == nil || auth.ID == "" {
 		return nil, nil
 	}
+	auth.EnsureIndex()
 	m.mu.Lock()
 	m.auths[auth.ID] = auth.Clone()
 	m.mu.Unlock()
@@ -209,6 +213,7 @@ func (m *Manager) Load(ctx context.Context) error {
 		if auth == nil || auth.ID == "" {
 			continue
 		}
+		auth.EnsureIndex()
 		m.auths[auth.ID] = auth.Clone()
 	}
 	return nil
@@ -322,6 +327,9 @@ func (m *Manager) executeWithProvider(ctx context.Context, provider string, req
 			if errors.As(errExec, &se) && se != nil {
 				result.Error.HTTPStatus = se.StatusCode()
 			}
+			if ra := retryAfterFromError(errExec); ra != nil {
+				result.RetryAfter = ra
+			}
 			m.MarkResult(execCtx, result)
 			lastErr = errExec
 			continue
@@ -367,6 +375,9 @@ func (m *Manager) executeCountWithProvider(ctx context.Context, provider string,
 			if errors.As(errExec, &se) && se != nil {
 				result.Error.HTTPStatus = se.StatusCode()
 			}
+			if ra := retryAfterFromError(errExec); ra != nil {
+				result.RetryAfter = ra
+			}
 			m.MarkResult(execCtx, result)
 			lastErr = errExec
 			continue
@@ -412,6 +423,7 @@ func (m *Manager) executeStreamWithProvider(ctx context.Context, provider string
 				rerr.HTTPStatus = se.StatusCode()
 			}
 			result := Result{AuthID: auth.ID, Provider: provider, Model: req.Model, Success: false, Error: rerr}
+			result.RetryAfter = retryAfterFromError(errStream)
 			m.MarkResult(execCtx, result)
 			lastErr = errStream
 			continue
@@ -553,17 +565,23 @@ func (m *Manager) MarkResult(ctx context.Context, result Result) {
 					suspendReason = "payment_required"
 					shouldSuspendModel = true
 				case 429:
-					cooldown, nextLevel := nextQuotaCooldown(state.Quota.BackoffLevel)
 					var next time.Time
-					if cooldown > 0 {
-						next = now.Add(cooldown)
+					backoffLevel := state.Quota.BackoffLevel
+					if result.RetryAfter != nil {
+						next = now.Add(*result.RetryAfter)
+					} else {
+						cooldown, nextLevel := nextQuotaCooldown(backoffLevel)
+						if cooldown > 0 {
+							next = now.Add(cooldown)
+						}
+						backoffLevel = nextLevel
 					}
 					state.NextRetryAfter = next
 					state.Quota = QuotaState{
 						Exceeded:      true,
 						Reason:        "quota",
 						NextRecoverAt: next,
-						BackoffLevel:  nextLevel,
+						BackoffLevel:  backoffLevel,
 					}
 					suspendReason = "quota"
 					shouldSuspendModel = true
@@ -579,7 +597,7 @@ func (m *Manager) MarkResult(ctx context.Context, result Result) {
 				auth.UpdatedAt = now
 				updateAggregatedAvailability(auth, now)
 			} else {
-				applyAuthFailureState(auth, result.Error, now)
+				applyAuthFailureState(auth, result.Error, result.RetryAfter, now)
 			}
 		}
 
@@ -739,6 +757,25 @@ func cloneError(err *Error) *Error {
 	}
 }
 
+func retryAfterFromError(err error) *time.Duration {
+	if err == nil {
+		return nil
+	}
+	type retryAfterProvider interface {
+		RetryAfter() *time.Duration
+	}
+	rap, ok := err.(retryAfterProvider)
+	if !ok || rap == nil {
+		return nil
+	}
+	retryAfter := rap.RetryAfter()
+	if retryAfter == nil {
+		return nil
+	}
+	val := *retryAfter
+	return &val
+}
+
 func statusCodeFromResult(err *Error) int {
 	if err == nil {
 		return 0
@@ -746,7 +783,7 @@ func statusCodeFromResult(err *Error) int {
 	return err.StatusCode()
 }
 
-func applyAuthFailureState(auth *Auth, resultErr *Error, now time.Time) {
+func applyAuthFailureState(auth *Auth, resultErr *Error, retryAfter *time.Duration, now time.Time) {
 	if auth == nil {
 		return
 	}
@@ -771,13 +808,17 @@ func applyAuthFailureState(auth *Auth, resultErr *Error, now time.Time) {
 		auth.StatusMessage = "quota exhausted"
 		auth.Quota.Exceeded = true
 		auth.Quota.Reason = "quota"
-		cooldown, nextLevel := nextQuotaCooldown(auth.Quota.BackoffLevel)
 		var next time.Time
-		if cooldown > 0 {
-			next = now.Add(cooldown)
+		if retryAfter != nil {
+			next = now.Add(*retryAfter)
+		} else {
+			cooldown, nextLevel := nextQuotaCooldown(auth.Quota.BackoffLevel)
+			if cooldown > 0 {
+				next = now.Add(cooldown)
+			}
+			auth.Quota.BackoffLevel = nextLevel
 		}
 		auth.Quota.NextRecoverAt = next
-		auth.Quota.BackoffLevel = nextLevel
 		auth.NextRetryAfter = next
 	case 408, 500, 502, 503, 504:
 		auth.StatusMessage = "transient upstream error"

sdk/cliproxy/auth/types.go

@@ -5,6 +5,7 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	baseauth "github.com/router-for-me/CLIProxyAPI/v6/internal/auth"
@@ -14,6 +15,8 @@ import (
 type Auth struct {
 	// ID uniquely identifies the auth record across restarts.
 	ID string `json:"id"`
+	// Index is a monotonically increasing runtime identifier used for diagnostics.
+	Index uint64 `json:"-"`
 	// Provider is the upstream provider key (e.g. "gemini", "claude").
 	Provider string `json:"provider"`
 	// FileName stores the relative or absolute path of the backing auth file.
@@ -55,6 +58,8 @@ type Auth struct {
 
 	// Runtime carries non-serialisable data used during execution (in-memory only).
 	Runtime any `json:"-"`
+
+	indexAssigned bool `json:"-"`
 }
 
 // QuotaState contains limiter tracking data for a credential.
@@ -87,6 +92,12 @@ type ModelState struct {
 	UpdatedAt time.Time `json:"updated_at"`
 }
 
+var authIndexCounter atomic.Uint64
+
+func nextAuthIndex() uint64 {
+	return authIndexCounter.Add(1) - 1
+}
+
 // Clone shallow copies the Auth structure, duplicating maps to avoid accidental mutation.
 func (a *Auth) Clone() *Auth {
 	if a == nil {
@@ -115,6 +126,20 @@ func (a *Auth) Clone() *Auth {
 	return &copyAuth
 }
 
+// EnsureIndex returns the global index, assigning one if it was not set yet.
+func (a *Auth) EnsureIndex() uint64 {
+	if a == nil {
+		return 0
+	}
+	if a.indexAssigned {
+		return a.Index
+	}
+	idx := nextAuthIndex()
+	a.Index = idx
+	a.indexAssigned = true
+	return idx
+}
+
 // Clone duplicates a model state including nested error details.
 func (m *ModelState) Clone() *ModelState {
 	if m == nil {
@@ -152,11 +177,29 @@ func (a *Auth) AccountInfo() (string, string) {
 			}
 		}
 	}
-	if a.Metadata != nil {
-		if v, ok := a.Metadata["email"].(string); ok {
-			return "oauth", v
+
+	// For iFlow provider, prioritize OAuth type if email is present
+	if strings.ToLower(a.Provider) == "iflow" {
+		if a.Metadata != nil {
+			if email, ok := a.Metadata["email"].(string); ok {
+				email = strings.TrimSpace(email)
+				if email != "" {
+					return "oauth", email
+				}
+			}
 		}
 	}
+
+	// Check metadata for email first (OAuth-style auth)
+	if a.Metadata != nil {
+		if v, ok := a.Metadata["email"].(string); ok {
+			email := strings.TrimSpace(v)
+			if email != "" {
+				return "oauth", email
+			}
+		}
+	}
+	// Fall back to API key (API-key auth)
 	if a.Attributes != nil {
 		if v := a.Attributes["api_key"]; v != "" {
 			return "api_key", v
@@ -259,6 +302,7 @@ func parseTimeValue(v any) (time.Time, bool) {
 			time.RFC3339,
 			time.RFC3339Nano,
 			"2006-01-02 15:04:05",
+			"2006-01-02 15:04",
 			"2006-01-02T15:04:05Z07:00",
 		}
 		for _, layout := range layouts {

sdk/cliproxy/service.go

@@ -604,6 +604,12 @@ func (s *Service) registerModelsForAuth(a *coreauth.Auth) {
 	if a == nil || a.ID == "" {
 		return
 	}
+	if a.Attributes != nil {
+		if v := strings.TrimSpace(a.Attributes["gemini_virtual_primary"]); strings.EqualFold(v, "true") {
+			GlobalModelRegistry().UnregisterClient(a.ID)
+			return
+		}
+	}
 	// Unregister legacy client ID (if present) to avoid double counting
 	if a.Runtime != nil {
 		if idGetter, ok := a.Runtime.(interface{ GetClientID() string }); ok {
@@ -623,7 +629,7 @@ func (s *Service) registerModelsForAuth(a *coreauth.Auth) {
 		models = registry.GetGeminiModels()
 	case "vertex":
 		// Vertex AI Gemini supports the same model identifiers as Gemini.
-		models = registry.GetGeminiModels()
+		models = registry.GetGeminiVertexModels()
 	case "gemini-cli":
 		models = registry.GetGeminiCLIModels()
 	case "aistudio":

sdk/cliproxy/usage/manager.go

@@ -14,6 +14,7 @@ type Record struct {
 	Model       string
 	APIKey      string
 	AuthID      string
+	AuthIndex   uint64
 	Source      string
 	RequestedAt time.Time
 	Failed      bool