feat(util): implement dynamic thinking suffix normalization and refactor budget resolution logic

- Added support for parsing and normalizing dynamic thinking model suffixes.
- Centralized budget resolution across executors and payload helpers.
- Retired legacy Gemini-specific thinking handlers in favor of unified logic.
- Updated executors to use metadata-based thinking configuration.
- Added `ResolveOriginalModel` utility for resolving normalized upstream models using request metadata.
- Updated executors (Gemini, Codex, iFlow, OpenAI, Qwen) to incorporate upstream model resolution and substitute model values in payloads and request URLs.
- Ensured fallbacks handle cases with missing or malformed metadata to derive models robustly.
- Refactored upstream model resolution to dynamically incorporate metadata for selecting and normalizing models.
- Improved handling of thinking configurations and model overrides in executors.
- Removed hardcoded thinking model entries and migrated logic to metadata-based resolution.
- Updated payload mutations to always include the resolved model.

cmd/server/main.go

@@ -139,7 +139,8 @@ func main() {
 
 	wd, err := os.Getwd()
 	if err != nil {
-		log.Fatalf("failed to get working directory: %v", err)
+		log.Errorf("failed to get working directory: %v", err)
+		return
 	}
 
 	// Load environment variables from .env if present.
@@ -233,13 +234,15 @@ func main() {
 		})
 		cancel()
 		if err != nil {
-			log.Fatalf("failed to initialize postgres token store: %v", err)
+			log.Errorf("failed to initialize postgres token store: %v", err)
+			return
 		}
 		examplePath := filepath.Join(wd, "config.example.yaml")
 		ctx, cancel = context.WithTimeout(context.Background(), 30*time.Second)
 		if errBootstrap := pgStoreInst.Bootstrap(ctx, examplePath); errBootstrap != nil {
 			cancel()
-			log.Fatalf("failed to bootstrap postgres-backed config: %v", errBootstrap)
+			log.Errorf("failed to bootstrap postgres-backed config: %v", errBootstrap)
+			return
 		}
 		cancel()
 		configFilePath = pgStoreInst.ConfigPath()
@@ -262,7 +265,8 @@ func main() {
 		if strings.Contains(resolvedEndpoint, "://") {
 			parsed, errParse := url.Parse(resolvedEndpoint)
 			if errParse != nil {
-				log.Fatalf("failed to parse object store endpoint %q: %v", objectStoreEndpoint, errParse)
+				log.Errorf("failed to parse object store endpoint %q: %v", objectStoreEndpoint, errParse)
+				return
 			}
 			switch strings.ToLower(parsed.Scheme) {
 			case "http":
@@ -270,10 +274,12 @@ func main() {
 			case "https":
 				useSSL = true
 			default:
-				log.Fatalf("unsupported object store scheme %q (only http and https are allowed)", parsed.Scheme)
+				log.Errorf("unsupported object store scheme %q (only http and https are allowed)", parsed.Scheme)
+				return
 			}
 			if parsed.Host == "" {
-				log.Fatalf("object store endpoint %q is missing host information", objectStoreEndpoint)
+				log.Errorf("object store endpoint %q is missing host information", objectStoreEndpoint)
+				return
 			}
 			resolvedEndpoint = parsed.Host
 			if parsed.Path != "" && parsed.Path != "/" {
@@ -292,13 +298,15 @@ func main() {
 		}
 		objectStoreInst, err = store.NewObjectTokenStore(objCfg)
 		if err != nil {
-			log.Fatalf("failed to initialize object token store: %v", err)
+			log.Errorf("failed to initialize object token store: %v", err)
+			return
 		}
 		examplePath := filepath.Join(wd, "config.example.yaml")
 		ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
 		if errBootstrap := objectStoreInst.Bootstrap(ctx, examplePath); errBootstrap != nil {
 			cancel()
-			log.Fatalf("failed to bootstrap object-backed config: %v", errBootstrap)
+			log.Errorf("failed to bootstrap object-backed config: %v", errBootstrap)
+			return
 		}
 		cancel()
 		configFilePath = objectStoreInst.ConfigPath()
@@ -323,7 +331,8 @@ func main() {
 		gitStoreInst = store.NewGitTokenStore(gitStoreRemoteURL, gitStoreUser, gitStorePassword)
 		gitStoreInst.SetBaseDir(authDir)
 		if errRepo := gitStoreInst.EnsureRepository(); errRepo != nil {
-			log.Fatalf("failed to prepare git token store: %v", errRepo)
+			log.Errorf("failed to prepare git token store: %v", errRepo)
+			return
 		}
 		configFilePath = gitStoreInst.ConfigPath()
 		if configFilePath == "" {
@@ -332,17 +341,21 @@ func main() {
 		if _, statErr := os.Stat(configFilePath); errors.Is(statErr, fs.ErrNotExist) {
 			examplePath := filepath.Join(wd, "config.example.yaml")
 			if _, errExample := os.Stat(examplePath); errExample != nil {
-				log.Fatalf("failed to find template config file: %v", errExample)
+				log.Errorf("failed to find template config file: %v", errExample)
+				return
 			}
 			if errCopy := misc.CopyConfigTemplate(examplePath, configFilePath); errCopy != nil {
-				log.Fatalf("failed to bootstrap git-backed config: %v", errCopy)
+				log.Errorf("failed to bootstrap git-backed config: %v", errCopy)
+				return
 			}
 			if errCommit := gitStoreInst.PersistConfig(context.Background()); errCommit != nil {
-				log.Fatalf("failed to commit initial git-backed config: %v", errCommit)
+				log.Errorf("failed to commit initial git-backed config: %v", errCommit)
+				return
 			}
 			log.Infof("git-backed config initialized from template: %s", configFilePath)
 		} else if statErr != nil {
-			log.Fatalf("failed to inspect git-backed config: %v", statErr)
+			log.Errorf("failed to inspect git-backed config: %v", statErr)
+			return
 		}
 		cfg, err = config.LoadConfigOptional(configFilePath, isCloudDeploy)
 		if err == nil {
@@ -355,13 +368,15 @@ func main() {
 	} else {
 		wd, err = os.Getwd()
 		if err != nil {
-			log.Fatalf("failed to get working directory: %v", err)
+			log.Errorf("failed to get working directory: %v", err)
+			return
 		}
 		configFilePath = filepath.Join(wd, "config.yaml")
 		cfg, err = config.LoadConfigOptional(configFilePath, isCloudDeploy)
 	}
 	if err != nil {
-		log.Fatalf("failed to load config: %v", err)
+		log.Errorf("failed to load config: %v", err)
+		return
 	}
 	if cfg == nil {
 		cfg = &config.Config{}
@@ -391,7 +406,8 @@ func main() {
 	coreauth.SetQuotaCooldownDisabled(cfg.DisableCooling)
 
 	if err = logging.ConfigureLogOutput(cfg.LoggingToFile); err != nil {
-		log.Fatalf("failed to configure log output: %v", err)
+		log.Errorf("failed to configure log output: %v", err)
+		return
 	}
 
 	log.Infof("CLIProxyAPI Version: %s, Commit: %s, BuiltAt: %s", buildinfo.Version, buildinfo.Commit, buildinfo.BuildDate)
@@ -400,7 +416,8 @@ func main() {
 	util.SetLogLevel(cfg)
 
 	if resolvedAuthDir, errResolveAuthDir := util.ResolveAuthDir(cfg.AuthDir); errResolveAuthDir != nil {
-		log.Fatalf("failed to resolve auth directory: %v", errResolveAuthDir)
+		log.Errorf("failed to resolve auth directory: %v", errResolveAuthDir)
+		return
 	} else {
 		cfg.AuthDir = resolvedAuthDir
 	}

config.example.yaml

@@ -1,3 +1,7 @@
+# Server host/interface to bind to. Default is empty ("") to bind all interfaces (IPv4 + IPv6).
+# Use "127.0.0.1" or "localhost" to restrict access to local machine only.
+host: ""
+
 # Server port
 port: 8317
 

internal/api/handlers/management/auth_files.go

@@ -713,14 +713,16 @@ func (h *Handler) RequestAnthropicToken(c *gin.Context) {
 	// Generate PKCE codes
 	pkceCodes, err := claude.GeneratePKCECodes()
 	if err != nil {
-		log.Fatalf("Failed to generate PKCE codes: %v", err)
+		log.Errorf("Failed to generate PKCE codes: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate PKCE codes"})
 		return
 	}
 
 	// Generate random state parameter
 	state, err := misc.GenerateRandomState()
 	if err != nil {
-		log.Fatalf("Failed to generate state parameter: %v", err)
+		log.Errorf("Failed to generate state parameter: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate state parameter"})
 		return
 	}
 
@@ -730,7 +732,8 @@ func (h *Handler) RequestAnthropicToken(c *gin.Context) {
 	// Generate authorization URL (then override redirect_uri to reuse server port)
 	authURL, state, err := anthropicAuth.GenerateAuthURL(state, pkceCodes)
 	if err != nil {
-		log.Fatalf("Failed to generate authorization URL: %v", err)
+		log.Errorf("Failed to generate authorization URL: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate authorization url"})
 		return
 	}
 
@@ -872,7 +875,7 @@ func (h *Handler) RequestAnthropicToken(c *gin.Context) {
 		}
 		savedPath, errSave := h.saveTokenRecord(ctx, record)
 		if errSave != nil {
-			log.Fatalf("Failed to save authentication tokens: %v", errSave)
+			log.Errorf("Failed to save authentication tokens: %v", errSave)
 			oauthStatus[state] = "Failed to save authentication tokens"
 			return
 		}
@@ -1045,7 +1048,7 @@ func (h *Handler) RequestGeminiCLIToken(c *gin.Context) {
 		gemAuth := geminiAuth.NewGeminiAuth()
 		gemClient, errGetClient := gemAuth.GetAuthenticatedClient(ctx, &ts, h.cfg, true)
 		if errGetClient != nil {
-			log.Fatalf("failed to get authenticated client: %v", errGetClient)
+			log.Errorf("failed to get authenticated client: %v", errGetClient)
 			oauthStatus[state] = "Failed to get authenticated client"
 			return
 		}
@@ -1110,7 +1113,7 @@ func (h *Handler) RequestGeminiCLIToken(c *gin.Context) {
 		}
 		savedPath, errSave := h.saveTokenRecord(ctx, record)
 		if errSave != nil {
-			log.Fatalf("Failed to save token to file: %v", errSave)
+			log.Errorf("Failed to save token to file: %v", errSave)
 			oauthStatus[state] = "Failed to save token to file"
 			return
 		}
@@ -1131,14 +1134,16 @@ func (h *Handler) RequestCodexToken(c *gin.Context) {
 	// Generate PKCE codes
 	pkceCodes, err := codex.GeneratePKCECodes()
 	if err != nil {
-		log.Fatalf("Failed to generate PKCE codes: %v", err)
+		log.Errorf("Failed to generate PKCE codes: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate PKCE codes"})
 		return
 	}
 
 	// Generate random state parameter
 	state, err := misc.GenerateRandomState()
 	if err != nil {
-		log.Fatalf("Failed to generate state parameter: %v", err)
+		log.Errorf("Failed to generate state parameter: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate state parameter"})
 		return
 	}
 
@@ -1148,7 +1153,8 @@ func (h *Handler) RequestCodexToken(c *gin.Context) {
 	// Generate authorization URL
 	authURL, err := openaiAuth.GenerateAuthURL(state, pkceCodes)
 	if err != nil {
-		log.Fatalf("Failed to generate authorization URL: %v", err)
+		log.Errorf("Failed to generate authorization URL: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate authorization url"})
 		return
 	}
 
@@ -1283,7 +1289,7 @@ func (h *Handler) RequestCodexToken(c *gin.Context) {
 		savedPath, errSave := h.saveTokenRecord(ctx, record)
 		if errSave != nil {
 			oauthStatus[state] = "Failed to save authentication tokens"
-			log.Fatalf("Failed to save authentication tokens: %v", errSave)
+			log.Errorf("Failed to save authentication tokens: %v", errSave)
 			return
 		}
 		fmt.Printf("Authentication successful! Token saved to %s\n", savedPath)
@@ -1318,7 +1324,8 @@ func (h *Handler) RequestAntigravityToken(c *gin.Context) {
 
 	state, errState := misc.GenerateRandomState()
 	if errState != nil {
-		log.Fatalf("Failed to generate state parameter: %v", errState)
+		log.Errorf("Failed to generate state parameter: %v", errState)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate state parameter"})
 		return
 	}
 
@@ -1514,7 +1521,7 @@ func (h *Handler) RequestAntigravityToken(c *gin.Context) {
 		}
 		savedPath, errSave := h.saveTokenRecord(ctx, record)
 		if errSave != nil {
-			log.Fatalf("Failed to save token to file: %v", errSave)
+			log.Errorf("Failed to save token to file: %v", errSave)
 			oauthStatus[state] = "Failed to save token to file"
 			return
 		}
@@ -1543,7 +1550,8 @@ func (h *Handler) RequestQwenToken(c *gin.Context) {
 	// Generate authorization URL
 	deviceFlow, err := qwenAuth.InitiateDeviceFlow(ctx)
 	if err != nil {
-		log.Fatalf("Failed to generate authorization URL: %v", err)
+		log.Errorf("Failed to generate authorization URL: %v", err)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to generate authorization url"})
 		return
 	}
 	authURL := deviceFlow.VerificationURIComplete
@@ -1570,7 +1578,7 @@ func (h *Handler) RequestQwenToken(c *gin.Context) {
 		}
 		savedPath, errSave := h.saveTokenRecord(ctx, record)
 		if errSave != nil {
-			log.Fatalf("Failed to save authentication tokens: %v", errSave)
+			log.Errorf("Failed to save authentication tokens: %v", errSave)
 			oauthStatus[state] = "Failed to save authentication tokens"
 			return
 		}
@@ -1674,7 +1682,7 @@ func (h *Handler) RequestIFlowToken(c *gin.Context) {
 		savedPath, errSave := h.saveTokenRecord(ctx, record)
 		if errSave != nil {
 			oauthStatus[state] = "Failed to save authentication tokens"
-			log.Fatalf("Failed to save authentication tokens: %v", errSave)
+			log.Errorf("Failed to save authentication tokens: %v", errSave)
 			return
 		}
 
@@ -2103,6 +2111,7 @@ func checkCloudAPIIsEnabled(ctx context.Context, httpClient *http.Client, projec
 				continue
 			}
 		}
+		_ = resp.Body.Close()
 		return false, fmt.Errorf("project activation required: %s", errMessage)
 	}
 	return true, nil

internal/api/modules/amp/fallback_handlers.go

@@ -133,8 +133,8 @@ func (fh *FallbackHandler) WrapHandler(handler gin.HandlerFunc) gin.HandlerFunc
 			return
 		}
 
-		// Normalize model (handles Gemini thinking suffixes)
-		normalizedModel, _ := util.NormalizeGeminiThinkingModel(modelName)
+		// Normalize model (handles dynamic thinking suffixes)
+		normalizedModel, _ := util.NormalizeThinkingModel(modelName)
 
 		// Track resolved model for logging (may change if mapping is applied)
 		resolvedModel := normalizedModel

internal/api/modules/amp/routes.go

@@ -156,6 +156,7 @@ func (m *AmpModule) registerManagementRoutes(engine *gin.Engine, baseHandler *ha
 	rootMiddleware := []gin.HandlerFunc{m.managementAvailabilityMiddleware(), noCORSMiddleware(), m.localhostOnlyMiddleware()}
 	engine.GET("/threads/*path", append(rootMiddleware, proxyHandler)...)
 	engine.GET("/threads.rss", append(rootMiddleware, proxyHandler)...)
+	engine.GET("/news.rss", append(rootMiddleware, proxyHandler)...)
 
 	// Root-level auth routes for CLI login flow
 	// Amp uses multiple auth routes: /auth/cli-login, /auth/callback, /auth/sign-in, /auth/logout

internal/api/server.go

@@ -300,7 +300,7 @@ func NewServer(cfg *config.Config, authManager *auth.Manager, accessManager *sdk
 
 	// Create HTTP server
 	s.server = &http.Server{
-		Addr:    fmt.Sprintf(":%d", cfg.Port),
+		Addr:    fmt.Sprintf("%s:%d", cfg.Host, cfg.Port),
 		Handler: engine,
 	}
 

internal/auth/gemini/gemini_auth.go

@@ -76,7 +76,8 @@ func (g *GeminiAuth) GetAuthenticatedClient(ctx context.Context, ts *GeminiToken
 			auth := &proxy.Auth{User: username, Password: password}
 			dialer, errSOCKS5 := proxy.SOCKS5("tcp", proxyURL.Host, auth, proxy.Direct)
 			if errSOCKS5 != nil {
-				log.Fatalf("create SOCKS5 dialer failed: %v", errSOCKS5)
+				log.Errorf("create SOCKS5 dialer failed: %v", errSOCKS5)
+				return nil, fmt.Errorf("create SOCKS5 dialer failed: %w", errSOCKS5)
 			}
 			transport = &http.Transport{
 				DialContext: func(ctx context.Context, network, addr string) (net.Conn, error) {
@@ -238,7 +239,11 @@ func (g *GeminiAuth) getTokenFromWeb(ctx context.Context, config *oauth2.Config,
 	// Start the server in a goroutine.
 	go func() {
 		if err := server.ListenAndServe(); !errors.Is(err, http.ErrServerClosed) {
-			log.Fatalf("ListenAndServe(): %v", err)
+			log.Errorf("ListenAndServe(): %v", err)
+			select {
+			case errChan <- err:
+			default:
+			}
 		}
 	}()
 

internal/cmd/login.go

@@ -65,20 +65,20 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 	authenticator := sdkAuth.NewGeminiAuthenticator()
 	record, errLogin := authenticator.Login(ctx, cfg, loginOpts)
 	if errLogin != nil {
-		log.Fatalf("Gemini authentication failed: %v", errLogin)
+		log.Errorf("Gemini authentication failed: %v", errLogin)
 		return
 	}
 
 	storage, okStorage := record.Storage.(*gemini.GeminiTokenStorage)
 	if !okStorage || storage == nil {
-		log.Fatal("Gemini authentication failed: unsupported token storage")
+		log.Error("Gemini authentication failed: unsupported token storage")
 		return
 	}
 
 	geminiAuth := gemini.NewGeminiAuth()
 	httpClient, errClient := geminiAuth.GetAuthenticatedClient(ctx, storage, cfg, options.NoBrowser)
 	if errClient != nil {
-		log.Fatalf("Gemini authentication failed: %v", errClient)
+		log.Errorf("Gemini authentication failed: %v", errClient)
 		return
 	}
 
@@ -86,7 +86,7 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 
 	projects, errProjects := fetchGCPProjects(ctx, httpClient)
 	if errProjects != nil {
-		log.Fatalf("Failed to get project list: %v", errProjects)
+		log.Errorf("Failed to get project list: %v", errProjects)
 		return
 	}
 
@@ -98,11 +98,11 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 	selectedProjectID := promptForProjectSelection(projects, strings.TrimSpace(projectID), promptFn)
 	projectSelections, errSelection := resolveProjectSelections(selectedProjectID, projects)
 	if errSelection != nil {
-		log.Fatalf("Invalid project selection: %v", errSelection)
+		log.Errorf("Invalid project selection: %v", errSelection)
 		return
 	}
 	if len(projectSelections) == 0 {
-		log.Fatal("No project selected; aborting login.")
+		log.Error("No project selected; aborting login.")
 		return
 	}
 
@@ -116,7 +116,7 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 				showProjectSelectionHelp(storage.Email, projects)
 				return
 			}
-			log.Fatalf("Failed to complete user setup: %v", errSetup)
+			log.Errorf("Failed to complete user setup: %v", errSetup)
 			return
 		}
 		finalID := strings.TrimSpace(storage.ProjectID)
@@ -133,11 +133,11 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 		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)
+				log.Errorf("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)
+				log.Errorf("Failed to check if Cloud AI API is enabled for project %s. If you encounter an error message, please create an issue.", pid)
 				return
 			}
 		}
@@ -153,7 +153,7 @@ func DoLogin(cfg *config.Config, projectID string, options *LoginOptions) {
 
 	savedPath, errSave := store.Save(ctx, record)
 	if errSave != nil {
-		log.Fatalf("Failed to save token to file: %v", errSave)
+		log.Errorf("Failed to save token to file: %v", errSave)
 		return
 	}
 
@@ -555,6 +555,7 @@ func checkCloudAPIIsEnabled(ctx context.Context, httpClient *http.Client, projec
 				continue
 			}
 		}
+		_ = resp.Body.Close()
 		return false, fmt.Errorf("project activation required: %s", errMessage)
 	}
 	return true, nil

internal/cmd/run.go

@@ -45,12 +45,13 @@ func StartService(cfg *config.Config, configPath string, localPassword string) {
 
 	service, err := builder.Build()
 	if err != nil {
-		log.Fatalf("failed to build proxy service: %v", err)
+		log.Errorf("failed to build proxy service: %v", err)
+		return
 	}
 
 	err = service.Run(runCtx)
 	if err != nil && !errors.Is(err, context.Canceled) {
-		log.Fatalf("proxy service exited with error: %v", err)
+		log.Errorf("proxy service exited with error: %v", err)
 	}
 }
 

internal/cmd/vertex_import.go

@@ -29,30 +29,30 @@ func DoVertexImport(cfg *config.Config, keyPath string) {
 	}
 	rawPath := strings.TrimSpace(keyPath)
 	if rawPath == "" {
-		log.Fatalf("vertex-import: missing service account key path")
+		log.Errorf("vertex-import: missing service account key path")
 		return
 	}
 	data, errRead := os.ReadFile(rawPath)
 	if errRead != nil {
-		log.Fatalf("vertex-import: read file failed: %v", errRead)
+		log.Errorf("vertex-import: read file failed: %v", errRead)
 		return
 	}
 	var sa map[string]any
 	if errUnmarshal := json.Unmarshal(data, &sa); errUnmarshal != nil {
-		log.Fatalf("vertex-import: invalid service account json: %v", errUnmarshal)
+		log.Errorf("vertex-import: invalid service account json: %v", errUnmarshal)
 		return
 	}
 	// Validate and normalize private_key before saving
 	normalizedSA, errFix := vertex.NormalizeServiceAccountMap(sa)
 	if errFix != nil {
-		log.Fatalf("vertex-import: %v", errFix)
+		log.Errorf("vertex-import: %v", errFix)
 		return
 	}
 	sa = normalizedSA
 	email, _ := sa["client_email"].(string)
 	projectID, _ := sa["project_id"].(string)
 	if strings.TrimSpace(projectID) == "" {
-		log.Fatalf("vertex-import: project_id missing in service account json")
+		log.Errorf("vertex-import: project_id missing in service account json")
 		return
 	}
 	if strings.TrimSpace(email) == "" {
@@ -92,7 +92,7 @@ func DoVertexImport(cfg *config.Config, keyPath string) {
 	}
 	path, errSave := store.Save(context.Background(), record)
 	if errSave != nil {
-		log.Fatalf("vertex-import: save credential failed: %v", errSave)
+		log.Errorf("vertex-import: save credential failed: %v", errSave)
 		return
 	}
 	fmt.Printf("Vertex credentials imported: %s\n", path)

internal/config/config.go

@@ -20,6 +20,9 @@ import (
 // Config represents the application's configuration, loaded from a YAML file.
 type Config struct {
 	config.SDKConfig `yaml:",inline"`
+	// Host is the network host/interface on which the API server will bind.
+	// Default is empty ("") to bind all interfaces (IPv4 + IPv6). Use "127.0.0.1" or "localhost" for local-only access.
+	Host string `yaml:"host" json:"-"`
 	// Port is the network port on which the API server will listen.
 	Port int `yaml:"port" json:"-"`
 
@@ -320,6 +323,7 @@ func LoadConfigOptional(configFile string, optional bool) (*Config, error) {
 	// Unmarshal the YAML data into the Config struct.
 	var cfg Config
 	// Set defaults before unmarshal so that absent keys keep defaults.
+	cfg.Host = "" // Default empty: binds to all interfaces (IPv4 + IPv6)
 	cfg.LoggingToFile = false
 	cfg.UsageStatisticsEnabled = false
 	cfg.DisableCooling = false

internal/registry/model_definitions.go

@@ -26,60 +26,6 @@ func GetClaudeModels() []*ModelInfo {
 			DisplayName:         "Claude 4.5 Sonnet",
 			ContextLength:       200000,
 			MaxCompletionTokens: 64000,
-		},
-		{
-			ID:                  "claude-sonnet-4-5-thinking",
-			Object:              "model",
-			Created:             1759104000, // 2025-09-29
-			OwnedBy:             "anthropic",
-			Type:                "claude",
-			DisplayName:         "Claude 4.5 Sonnet Thinking",
-			ContextLength:       200000,
-			MaxCompletionTokens: 64000,
-			Thinking:            &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
-		},
-		{
-			ID:                  "claude-opus-4-5-thinking",
-			Object:              "model",
-			Created:             1761955200, // 2025-11-01
-			OwnedBy:             "anthropic",
-			Type:                "claude",
-			DisplayName:         "Claude 4.5 Opus Thinking",
-			ContextLength:       200000,
-			MaxCompletionTokens: 64000,
-			Thinking:            &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
-		},
-		{
-			ID:                  "claude-opus-4-5-thinking-low",
-			Object:              "model",
-			Created:             1761955200, // 2025-11-01
-			OwnedBy:             "anthropic",
-			Type:                "claude",
-			DisplayName:         "Claude 4.5 Opus Thinking Low",
-			ContextLength:       200000,
-			MaxCompletionTokens: 64000,
-			Thinking:            &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
-		},
-		{
-			ID:                  "claude-opus-4-5-thinking-medium",
-			Object:              "model",
-			Created:             1761955200, // 2025-11-01
-			OwnedBy:             "anthropic",
-			Type:                "claude",
-			DisplayName:         "Claude 4.5 Opus Thinking Medium",
-			ContextLength:       200000,
-			MaxCompletionTokens: 64000,
-			Thinking:            &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
-		},
-		{
-			ID:                  "claude-opus-4-5-thinking-high",
-			Object:              "model",
-			Created:             1761955200, // 2025-11-01
-			OwnedBy:             "anthropic",
-			Type:                "claude",
-			DisplayName:         "Claude 4.5 Opus Thinking High",
-			ContextLength:       200000,
-			MaxCompletionTokens: 64000,
 			Thinking:            &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
 		},
 		{
@@ -92,6 +38,7 @@ func GetClaudeModels() []*ModelInfo {
 			Description:         "Premium model combining maximum intelligence with practical performance",
 			ContextLength:       200000,
 			MaxCompletionTokens: 64000,
+			Thinking:            &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
 		},
 		{
 			ID:                  "claude-opus-4-1-20250805",
@@ -530,58 +477,6 @@ func GetOpenAIModels() []*ModelInfo {
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
 		},
-		{
-			ID:                  "gpt-5-minimal",
-			Object:              "model",
-			Created:             1754524800,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-08-07",
-			DisplayName:         "GPT 5 Minimal",
-			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-low",
-			Object:              "model",
-			Created:             1754524800,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-08-07",
-			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-medium",
-			Object:              "model",
-			Created:             1754524800,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-08-07",
-			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-high",
-			Object:              "model",
-			Created:             1754524800,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-08-07",
-			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-codex",
 			Object:              "model",
@@ -595,45 +490,6 @@ func GetOpenAIModels() []*ModelInfo {
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
 		},
-		{
-			ID:                  "gpt-5-codex-low",
-			Object:              "model",
-			Created:             1757894400,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-09-15",
-			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-codex-medium",
-			Object:              "model",
-			Created:             1757894400,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-09-15",
-			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-codex-high",
-			Object:              "model",
-			Created:             1757894400,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-09-15",
-			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-codex-mini",
 			Object:              "model",
@@ -647,32 +503,6 @@ func GetOpenAIModels() []*ModelInfo {
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
 		},
-		{
-			ID:                  "gpt-5-codex-mini-medium",
-			Object:              "model",
-			Created:             1762473600,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-11-07",
-			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-codex-mini-high",
-			Object:              "model",
-			Created:             1762473600,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5-2025-11-07",
-			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",
 			Object:              "model",
@@ -686,58 +516,6 @@ func GetOpenAIModels() []*ModelInfo {
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
 		},
-		{
-			ID:                  "gpt-5.1-none",
-			Object:              "model",
-			Created:             1762905600,
-			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:             1762905600,
-			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:             1762905600,
-			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:             1762905600,
-			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",
@@ -745,47 +523,8 @@ func GetOpenAIModels() []*ModelInfo {
 			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:             1762905600,
-			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:             1762905600,
-			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:             1762905600,
-			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.",
+			DisplayName:         "GPT 5.1 Codex",
+			Description:         "Stable version of GPT 5.1 Codex, The best model for coding and agentic tasks across domains.",
 			ContextLength:       400000,
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
@@ -797,39 +536,12 @@ func GetOpenAIModels() []*ModelInfo {
 			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.",
+			DisplayName:         "GPT 5.1 Codex Mini",
+			Description:         "Stable version of GPT 5.1 Codex Mini: cheaper, faster, but less capable version of GPT 5.1 Codex.",
 			ContextLength:       400000,
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
 		},
-		{
-			ID:                  "gpt-5.1-codex-mini-medium",
-			Object:              "model",
-			Created:             1762905600,
-			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:             1762905600,
-			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",
@@ -837,60 +549,8 @@ func GetOpenAIModels() []*ModelInfo {
 			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:             1763424000,
-			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:             1763424000,
-			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:             1763424000,
-			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:             1763424000,
-			OwnedBy:             "openai",
-			Type:                "openai",
-			Version:             "gpt-5.1-max",
-			DisplayName:         "GPT 5 Codex Max XHigh",
-			Description:         "Stable version of GPT 5 Codex Max XHigh",
+			DisplayName:         "GPT 5.1 Codex Max",
+			Description:         "Stable version of GPT 5.1 Codex Max",
 			ContextLength:       400000,
 			MaxCompletionTokens: 128000,
 			SupportedParameters: []string{"tools"},
@@ -991,16 +651,19 @@ func GetIFlowModels() []*ModelInfo {
 type AntigravityModelConfig struct {
 	Thinking            *ThinkingSupport
 	MaxCompletionTokens int
+	Name                string
 }
 
 // GetAntigravityModelConfig returns static configuration for antigravity models.
 // Keys use the ALIASED model names (after modelName2Alias conversion) for direct lookup.
 func GetAntigravityModelConfig() map[string]*AntigravityModelConfig {
 	return map[string]*AntigravityModelConfig{
-		"gemini-2.5-flash":                  {Thinking: &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true}},
-		"gemini-2.5-flash-lite":             {Thinking: &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true}},
-		"gemini-3-pro-preview":              {Thinking: &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true}},
-		"gemini-claude-sonnet-4-5-thinking": {Thinking: &ThinkingSupport{Min: 1024, Max: 200000, ZeroAllowed: false, DynamicAllowed: true}, MaxCompletionTokens: 64000},
-		"gemini-claude-opus-4-5-thinking":   {Thinking: &ThinkingSupport{Min: 1024, Max: 200000, ZeroAllowed: false, DynamicAllowed: true}, MaxCompletionTokens: 64000},
+		"gemini-2.5-flash":                        {Thinking: &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true}, Name: "models/gemini-2.5-flash"},
+		"gemini-2.5-flash-lite":                   {Thinking: &ThinkingSupport{Min: 0, Max: 24576, ZeroAllowed: true, DynamicAllowed: true}, Name: "models/gemini-2.5-flash-lite"},
+		"gemini-2.5-computer-use-preview-10-2025": {Name: "models/gemini-2.5-computer-use-preview-10-2025"},
+		"gemini-3-pro-preview":                    {Thinking: &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true}, Name: "models/gemini-3-pro-preview"},
+		"gemini-3-pro-image-preview":              {Thinking: &ThinkingSupport{Min: 128, Max: 32768, ZeroAllowed: false, DynamicAllowed: true}, Name: "models/gemini-3-pro-image-preview"},
+		"gemini-claude-sonnet-4-5-thinking":       {Thinking: &ThinkingSupport{Min: 1024, Max: 200000, ZeroAllowed: false, DynamicAllowed: true}, MaxCompletionTokens: 64000},
+		"gemini-claude-opus-4-5-thinking":         {Thinking: &ThinkingSupport{Min: 1024, Max: 200000, ZeroAllowed: false, DynamicAllowed: true}, MaxCompletionTokens: 64000},
 	}
 }

internal/runtime/executor/aistudio_executor.go

@@ -309,7 +309,9 @@ func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts c
 	to := sdktranslator.FromString("gemini")
 	payload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
 	payload = applyThinkingMetadata(payload, req.Metadata, req.Model)
+	payload = util.ApplyDefaultThinkingIfNeeded(req.Model, payload)
 	payload = util.ConvertThinkingLevelToBudget(payload)
+	payload = util.NormalizeGeminiThinkingBudget(req.Model, payload)
 	payload = util.StripThinkingConfigIfUnsupported(req.Model, payload)
 	payload = fixGeminiImageAspectRatio(req.Model, payload)
 	payload = applyPayloadConfig(e.cfg, req.Model, payload)

internal/runtime/executor/antigravity_executor.go

@@ -27,18 +27,18 @@ import (
 )
 
 const (
-	antigravityBaseURLDaily        = "https://daily-cloudcode-pa.sandbox.googleapis.com"
-	antigravityBaseURLAutopush     = "https://autopush-cloudcode-pa.sandbox.googleapis.com"
-	antigravityBaseURLProd         = "https://cloudcode-pa.googleapis.com"
-	antigravityStreamPath          = "/v1internal:streamGenerateContent"
-	antigravityGeneratePath        = "/v1internal:generateContent"
-	antigravityModelsPath          = "/v1internal:fetchAvailableModels"
-	antigravityClientID            = "1071006060591-tmhssin2h21lcre235vtolojh4g403ep.apps.googleusercontent.com"
-	antigravityClientSecret        = "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"
-	defaultAntigravityAgent        = "antigravity/1.11.5 windows/amd64"
-	antigravityAuthType            = "antigravity"
-	refreshSkew                    = 3000 * time.Second
-	streamScannerBuffer        int = 20_971_520
+	antigravityBaseURLDaily = "https://daily-cloudcode-pa.sandbox.googleapis.com"
+	// antigravityBaseURLAutopush     = "https://autopush-cloudcode-pa.sandbox.googleapis.com"
+	antigravityBaseURLProd      = "https://cloudcode-pa.googleapis.com"
+	antigravityStreamPath       = "/v1internal:streamGenerateContent"
+	antigravityGeneratePath     = "/v1internal:generateContent"
+	antigravityModelsPath       = "/v1internal:fetchAvailableModels"
+	antigravityClientID         = "1071006060591-tmhssin2h21lcre235vtolojh4g403ep.apps.googleusercontent.com"
+	antigravityClientSecret     = "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"
+	defaultAntigravityAgent     = "antigravity/1.11.5 windows/amd64"
+	antigravityAuthType         = "antigravity"
+	refreshSkew                 = 3000 * time.Second
+	streamScannerBuffer     int = 20_971_520
 )
 
 var randSource = rand.New(rand.NewSource(time.Now().UnixNano()))
@@ -77,6 +77,7 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 
 	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
 	translated = normalizeAntigravityThinking(req.Model, translated)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
@@ -171,6 +172,7 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
 	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
 	translated = normalizeAntigravityThinking(req.Model, translated)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
@@ -373,9 +375,14 @@ func FetchAntigravityModels(ctx context.Context, auth *cliproxyauth.Auth, cfg *c
 		for originalName := range result.Map() {
 			aliasName := modelName2Alias(originalName)
 			if aliasName != "" {
+				cfg := modelConfig[aliasName]
+				modelName := aliasName
+				if cfg != nil && cfg.Name != "" {
+					modelName = cfg.Name
+				}
 				modelInfo := &registry.ModelInfo{
 					ID:          aliasName,
-					Name:        aliasName,
+					Name:        modelName,
 					Description: aliasName,
 					DisplayName: aliasName,
 					Version:     aliasName,
@@ -385,7 +392,7 @@ func FetchAntigravityModels(ctx context.Context, auth *cliproxyauth.Auth, cfg *c
 					Type:        antigravityAuthType,
 				}
 				// Look up Thinking support from static config using alias name
-				if cfg, ok := modelConfig[aliasName]; ok {
+				if cfg != nil {
 					if cfg.Thinking != nil {
 						modelInfo.Thinking = cfg.Thinking
 					}
@@ -536,6 +543,9 @@ func (e *AntigravityExecutor) buildRequest(ctx context.Context, auth *cliproxyau
 		strJSON = util.DeleteKey(strJSON, "minLength")
 		strJSON = util.DeleteKey(strJSON, "maxLength")
 		strJSON = util.DeleteKey(strJSON, "exclusiveMinimum")
+		strJSON = util.DeleteKey(strJSON, "exclusiveMaximum")
+		strJSON = util.DeleteKey(strJSON, "$ref")
+		strJSON = util.DeleteKey(strJSON, "$defs")
 
 		paths = make([]string, 0)
 		util.Walk(gjson.Parse(strJSON), "", "anyOf", &paths)
@@ -651,7 +661,7 @@ func buildBaseURL(auth *cliproxyauth.Auth) string {
 	if baseURLs := antigravityBaseURLFallbackOrder(auth); len(baseURLs) > 0 {
 		return baseURLs[0]
 	}
-	return antigravityBaseURLAutopush
+	return antigravityBaseURLDaily
 }
 
 func resolveHost(base string) string {
@@ -687,7 +697,7 @@ func antigravityBaseURLFallbackOrder(auth *cliproxyauth.Auth) []string {
 	}
 	return []string{
 		antigravityBaseURLDaily,
-		antigravityBaseURLAutopush,
+		// antigravityBaseURLAutopush,
 		antigravityBaseURLProd,
 	}
 }
@@ -831,9 +841,12 @@ func normalizeAntigravityThinking(model string, payload []byte) []byte {
 		effectiveMax, setDefaultMax := antigravityEffectiveMaxTokens(model, payload)
 		if effectiveMax > 0 && normalized >= effectiveMax {
 			normalized = effectiveMax - 1
-			if normalized < 1 {
-				normalized = 1
-			}
+		}
+		minBudget := antigravityMinThinkingBudget(model)
+		if minBudget > 0 && normalized >= 0 && normalized < minBudget {
+			// Budget is below minimum, remove thinking config entirely
+			payload, _ = sjson.DeleteBytes(payload, "request.generationConfig.thinkingConfig")
+			return payload
 		}
 		if setDefaultMax {
 			if res, errSet := sjson.SetBytes(payload, "request.generationConfig.maxOutputTokens", effectiveMax); errSet == nil {
@@ -861,3 +874,12 @@ func antigravityEffectiveMaxTokens(model string, payload []byte) (max int, fromM
 	}
 	return 0, false
 }
+
+// antigravityMinThinkingBudget returns the minimum thinking budget for a model.
+// Falls back to -1 if no model info is found.
+func antigravityMinThinkingBudget(model string) int {
+	if modelInfo := registry.GetGlobalRegistry().GetModelInfo(model); modelInfo != nil && modelInfo.Thinking != nil {
+		return modelInfo.Thinking.Min
+	}
+	return -1
+}

internal/runtime/executor/claude_executor.go

@@ -54,15 +54,22 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	// Use streaming translation to preserve function calling, except for claude.
 	stream := from != to
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
-	modelForUpstream := req.Model
-	if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
-		body, _ = sjson.SetBytes(body, "model", modelOverride)
-		modelForUpstream = modelOverride
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
 	}
-	// Inject thinking config based on model suffix for thinking variants
-	body = e.injectThinkingConfig(req.Model, body)
+	if modelOverride := e.resolveUpstreamModel(upstreamModel, auth); modelOverride != "" {
+		upstreamModel = modelOverride
+	} else if !strings.EqualFold(upstreamModel, req.Model) {
+		if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
+			upstreamModel = modelOverride
+		}
+	}
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
+	// Inject thinking config based on model metadata for thinking variants
+	body = e.injectThinkingConfig(req.Model, req.Metadata, body)
 
-	if !strings.HasPrefix(modelForUpstream, "claude-3-5-haiku") {
+	if !strings.HasPrefix(upstreamModel, "claude-3-5-haiku") {
 		body = checkSystemInstructions(body)
 	}
 	body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -161,11 +168,20 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("claude")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-	if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
-		body, _ = sjson.SetBytes(body, "model", modelOverride)
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
 	}
-	// Inject thinking config based on model suffix for thinking variants
-	body = e.injectThinkingConfig(req.Model, body)
+	if modelOverride := e.resolveUpstreamModel(upstreamModel, auth); modelOverride != "" {
+		upstreamModel = modelOverride
+	} else if !strings.EqualFold(upstreamModel, req.Model) {
+		if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
+			upstreamModel = modelOverride
+		}
+	}
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
+	// Inject thinking config based on model metadata for thinking variants
+	body = e.injectThinkingConfig(req.Model, req.Metadata, body)
 	body = checkSystemInstructions(body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 
@@ -295,13 +311,20 @@ func (e *ClaudeExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Aut
 	// Use streaming translation to preserve function calling, except for claude.
 	stream := from != to
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
-	modelForUpstream := req.Model
-	if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
-		body, _ = sjson.SetBytes(body, "model", modelOverride)
-		modelForUpstream = modelOverride
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
 	}
+	if modelOverride := e.resolveUpstreamModel(upstreamModel, auth); modelOverride != "" {
+		upstreamModel = modelOverride
+	} else if !strings.EqualFold(upstreamModel, req.Model) {
+		if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
+			upstreamModel = modelOverride
+		}
+	}
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
-	if !strings.HasPrefix(modelForUpstream, "claude-3-5-haiku") {
+	if !strings.HasPrefix(upstreamModel, "claude-3-5-haiku") {
 		body = checkSystemInstructions(body)
 	}
 
@@ -427,25 +450,15 @@ func extractAndRemoveBetas(body []byte) ([]string, []byte) {
 	return betas, body
 }
 
-// injectThinkingConfig adds thinking configuration based on model name suffix
-func (e *ClaudeExecutor) injectThinkingConfig(modelName string, body []byte) []byte {
+// injectThinkingConfig adds thinking configuration based on metadata or legacy suffixes.
+func (e *ClaudeExecutor) injectThinkingConfig(modelName string, metadata map[string]any, body []byte) []byte {
 	// Only inject if thinking config is not already present
 	if gjson.GetBytes(body, "thinking").Exists() {
 		return body
 	}
 
-	var budgetTokens int
-	switch {
-	case strings.HasSuffix(modelName, "-thinking-low"):
-		budgetTokens = 1024
-	case strings.HasSuffix(modelName, "-thinking-medium"):
-		budgetTokens = 8192
-	case strings.HasSuffix(modelName, "-thinking-high"):
-		budgetTokens = 24576
-	case strings.HasSuffix(modelName, "-thinking"):
-		// Default thinking without suffix uses medium budget
-		budgetTokens = 8192
-	default:
+	budgetTokens, ok := resolveClaudeThinkingBudget(modelName, metadata)
+	if !ok || budgetTokens <= 0 {
 		return body
 	}
 
@@ -454,6 +467,44 @@ func (e *ClaudeExecutor) injectThinkingConfig(modelName string, body []byte) []b
 	return body
 }
 
+func resolveClaudeThinkingBudget(modelName string, metadata map[string]any) (int, bool) {
+	budget, include, effort, matched := util.ThinkingFromMetadata(metadata)
+	if matched {
+		if include != nil && !*include {
+			return 0, false
+		}
+		if budget != nil {
+			normalized := util.NormalizeThinkingBudget(modelName, *budget)
+			if normalized > 0 {
+				return normalized, true
+			}
+			return 0, false
+		}
+		if effort != nil {
+			if derived, ok := util.ThinkingEffortToBudget(modelName, *effort); ok && derived > 0 {
+				return derived, true
+			}
+		}
+	}
+	return claudeBudgetFromSuffix(modelName)
+}
+
+func claudeBudgetFromSuffix(modelName string) (int, bool) {
+	lower := strings.ToLower(strings.TrimSpace(modelName))
+	switch {
+	case strings.HasSuffix(lower, "-thinking-low"):
+		return 1024, true
+	case strings.HasSuffix(lower, "-thinking-medium"):
+		return 8192, true
+	case strings.HasSuffix(lower, "-thinking-high"):
+		return 24576, true
+	case strings.HasSuffix(lower, "-thinking"):
+		return 8192, true
+	default:
+		return 0, false
+	}
+}
+
 // ensureMaxTokensForThinking ensures max_tokens > thinking.budget_tokens when thinking is enabled.
 // Anthropic API requires this constraint; violating it returns a 400 error.
 // This function should be called after all thinking configuration is finalized.
@@ -491,35 +542,45 @@ func ensureMaxTokensForThinking(modelName string, body []byte) []byte {
 }
 
 func (e *ClaudeExecutor) resolveUpstreamModel(alias string, auth *cliproxyauth.Auth) string {
-	if alias == "" {
+	trimmed := strings.TrimSpace(alias)
+	if trimmed == "" {
 		return ""
 	}
-	// Hardcoded mappings for thinking models to actual Claude model names
-	switch alias {
-	case "claude-opus-4-5-thinking", "claude-opus-4-5-thinking-low", "claude-opus-4-5-thinking-medium", "claude-opus-4-5-thinking-high":
-		return "claude-opus-4-5-20251101"
-	case "claude-sonnet-4-5-thinking":
-		return "claude-sonnet-4-5-20250929"
-	}
+
 	entry := e.resolveClaudeConfig(auth)
 	if entry == nil {
 		return ""
 	}
+
+	normalizedModel, metadata := util.NormalizeThinkingModel(trimmed)
+
+	// Candidate names to match against configured aliases/names.
+	candidates := []string{strings.TrimSpace(normalizedModel)}
+	if !strings.EqualFold(normalizedModel, trimmed) {
+		candidates = append(candidates, trimmed)
+	}
+	if original := util.ResolveOriginalModel(normalizedModel, metadata); original != "" && !strings.EqualFold(original, normalizedModel) {
+		candidates = append(candidates, original)
+	}
+
 	for i := range entry.Models {
 		model := entry.Models[i]
 		name := strings.TrimSpace(model.Name)
 		modelAlias := strings.TrimSpace(model.Alias)
-		if modelAlias != "" {
-			if strings.EqualFold(modelAlias, alias) {
+
+		for _, candidate := range candidates {
+			if candidate == "" {
+				continue
+			}
+			if modelAlias != "" && strings.EqualFold(modelAlias, candidate) {
 				if name != "" {
 					return name
 				}
-				return alias
+				return candidate
+			}
+			if name != "" && strings.EqualFold(name, candidate) {
+				return name
 			}
-			continue
-		}
-		if name != "" && strings.EqualFold(name, alias) {
-			return name
 		}
 	}
 	return ""

internal/runtime/executor/codex_executor.go

@@ -49,14 +49,14 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("codex")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-
-	body = e.setReasoningEffortByAlias(req.Model, body)
-
+	body = applyReasoningEffortMetadata(body, req.Metadata, req.Model)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
-
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	body, _ = sjson.SetBytes(body, "stream", true)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 
@@ -142,13 +142,16 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("codex")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
-	body = e.setReasoningEffortByAlias(req.Model, body)
+	body = applyReasoningEffortMetadata(body, req.Metadata, req.Model)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/responses"
 	httpReq, err := e.cacheHelper(ctx, from, url, req, body)
@@ -235,14 +238,16 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 }
 
 func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("codex")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 
 	modelForCounting := req.Model
 
-	body = e.setReasoningEffortByAlias(req.Model, body)
-
+	body = applyReasoningEffortMetadata(body, req.Metadata, req.Model)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 	body, _ = sjson.SetBytes(body, "stream", false)
 
@@ -261,83 +266,6 @@ 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

@@ -64,6 +64,8 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	to := sdktranslator.FromString("gemini-cli")
 	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
 	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
@@ -199,6 +201,8 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	to := sdktranslator.FromString("gemini-cli")
 	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
 	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
 	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
 	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)

internal/runtime/executor/gemini_executor.go

@@ -75,14 +75,19 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	// Official Gemini API via API key or OAuth bearer
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	body = applyThinkingMetadata(body, req.Metadata, req.Model)
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -91,7 +96,7 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 		}
 	}
 	baseURL := resolveGeminiBaseURL(auth)
-	url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, req.Model, action)
+	url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, upstreamModel, action)
 	if opts.Alt != "" && action != "countTokens" {
 		url = url + fmt.Sprintf("?$alt=%s", opts.Alt)
 	}
@@ -165,16 +170,21 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	body = applyThinkingMetadata(body, req.Metadata, req.Model)
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	baseURL := resolveGeminiBaseURL(auth)
-	url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, req.Model, "streamGenerateContent")
+	url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, upstreamModel, "streamGenerateContent")
 	if opts.Alt == "" {
 		url = url + "?alt=sse"
 	} else {

internal/runtime/executor/gemini_vertex_executor.go

@@ -105,10 +105,12 @@ func (e *GeminiVertexExecutor) CountTokens(ctx context.Context, auth *cliproxyau
 
 // countTokensWithServiceAccount handles token counting using service account credentials.
 func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, projectID, location string, saJSON []byte) (cliproxyexecutor.Response, error) {
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
 		if budgetOverride != nil {
 			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
 			budgetOverride = &norm
@@ -117,13 +119,14 @@ func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context
 	}
 	translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq)
 	translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq)
+	translatedReq, _ = sjson.SetBytes(translatedReq, "model", upstreamModel)
 	respCtx := context.WithValue(ctx, "alt", opts.Alt)
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig")
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "safetySettings")
 
 	baseURL := vertexBaseURL(location)
-	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, "countTokens")
+	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, upstreamModel, "countTokens")
 
 	httpReq, errNewReq := http.NewRequestWithContext(respCtx, http.MethodPost, url, bytes.NewReader(translatedReq))
 	if errNewReq != nil {
@@ -191,10 +194,12 @@ func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context
 
 // countTokensWithAPIKey handles token counting using API key credentials.
 func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, apiKey, baseURL string) (cliproxyexecutor.Response, error) {
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
 		if budgetOverride != nil {
 			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
 			budgetOverride = &norm
@@ -203,6 +208,7 @@ func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *
 	}
 	translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq)
 	translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq)
+	translatedReq, _ = sjson.SetBytes(translatedReq, "model", upstreamModel)
 	respCtx := context.WithValue(ctx, "alt", opts.Alt)
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig")
@@ -286,19 +292,24 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
 		if budgetOverride != nil {
 			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -307,7 +318,7 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 		}
 	}
 	baseURL := vertexBaseURL(location)
-	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, action)
+	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, upstreamModel, action)
 	if opts.Alt != "" && action != "countTokens" {
 		url = url + fmt.Sprintf("?$alt=%s", opts.Alt)
 	}
@@ -381,19 +392,24 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
 		if budgetOverride != nil {
 			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -406,7 +422,7 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 	if baseURL == "" {
 		baseURL = "https://generativelanguage.googleapis.com"
 	}
-	url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, req.Model, action)
+	url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, upstreamModel, action)
 	if opts.Alt != "" && action != "countTokens" {
 		url = url + fmt.Sprintf("?$alt=%s", opts.Alt)
 	}
@@ -477,22 +493,27 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-	if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
 		if budgetOverride != nil {
 			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	baseURL := vertexBaseURL(location)
-	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, "streamGenerateContent")
+	url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, upstreamModel, "streamGenerateContent")
 	if opts.Alt == "" {
 		url = url + "?alt=sse"
 	} else {
@@ -589,25 +610,30 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
 
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-	if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
 		if budgetOverride != nil {
 			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
+	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
+	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
 	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
 	body = fixGeminiImageAspectRatio(req.Model, body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 
 	// For API key auth, use simpler URL format without project/location
 	if baseURL == "" {
 		baseURL = "https://generativelanguage.googleapis.com"
 	}
-	url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, req.Model, "streamGenerateContent")
+	url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, upstreamModel, "streamGenerateContent")
 	if opts.Alt == "" {
 		url = url + "?alt=sse"
 	} else {

internal/runtime/executor/iflow_executor.go

@@ -57,6 +57,10 @@ 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 = applyReasoningEffortMetadataChatCompletions(body, req.Metadata, req.Model)
+	if upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata); upstreamModel != "" {
+		body, _ = sjson.SetBytes(body, "model", upstreamModel)
+	}
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
@@ -139,6 +143,10 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	to := sdktranslator.FromString("openai")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
+	body = applyReasoningEffortMetadataChatCompletions(body, req.Metadata, req.Model)
+	if upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata); upstreamModel != "" {
+		body, _ = sjson.SetBytes(body, "model", upstreamModel)
+	}
 	// Ensure tools array exists to avoid provider quirks similar to Qwen's behaviour.
 	toolsResult := gjson.GetBytes(body, "tools")
 	if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 {

internal/runtime/executor/logging_helpers.go

@@ -157,7 +157,7 @@ func appendAPIResponseChunk(ctx context.Context, cfg *config.Config, chunk []byt
 	if ginCtx == nil {
 		return
 	}
-	_, attempt := ensureAttempt(ginCtx)
+	attempts, attempt := ensureAttempt(ginCtx)
 	ensureResponseIntro(attempt)
 
 	if !attempt.headersWritten {
@@ -175,6 +175,8 @@ func appendAPIResponseChunk(ctx context.Context, cfg *config.Config, chunk []byt
 	}
 	attempt.response.WriteString(string(data))
 	attempt.bodyHasContent = true
+
+	updateAggregatedResponse(ginCtx, attempts)
 }
 
 func ginContextFrom(ctx context.Context) *gin.Context {

internal/runtime/executor/openai_compat_executor.go

@@ -58,6 +58,10 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 		translated = e.overrideModel(translated, modelOverride)
 	}
 	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated)
+	translated = applyReasoningEffortMetadataChatCompletions(translated, req.Metadata, req.Model)
+	if upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata); upstreamModel != "" {
+		translated, _ = sjson.SetBytes(translated, "model", upstreamModel)
+	}
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
@@ -143,6 +147,10 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 		translated = e.overrideModel(translated, modelOverride)
 	}
 	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated)
+	translated = applyReasoningEffortMetadataChatCompletions(translated, req.Metadata, req.Model)
+	if upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata); upstreamModel != "" {
+		translated, _ = sjson.SetBytes(translated, "model", upstreamModel)
+	}
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))

internal/runtime/executor/payload_helpers.go

@@ -12,8 +12,8 @@ import (
 // applyThinkingMetadata applies thinking config from model suffix metadata (e.g., -reasoning, -thinking-N)
 // for standard Gemini format payloads. It normalizes the budget when the model supports thinking.
 func applyThinkingMetadata(payload []byte, metadata map[string]any, model string) []byte {
-	budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(metadata)
-	if !ok {
+	budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(model, metadata)
+	if !ok || (budgetOverride == nil && includeOverride == nil) {
 		return payload
 	}
 	if !util.ModelSupportsThinking(model) {
@@ -29,17 +29,60 @@ func applyThinkingMetadata(payload []byte, metadata map[string]any, model string
 // applyThinkingMetadataCLI applies thinking config from model suffix metadata (e.g., -reasoning, -thinking-N)
 // for Gemini CLI format payloads (nested under "request"). It normalizes the budget when the model supports thinking.
 func applyThinkingMetadataCLI(payload []byte, metadata map[string]any, model string) []byte {
-	budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(metadata)
-	if !ok {
+	budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(model, metadata)
+	if !ok || (budgetOverride == nil && includeOverride == nil) {
 		return payload
 	}
-	if budgetOverride != nil && util.ModelSupportsThinking(model) {
+	if !util.ModelSupportsThinking(model) {
+		return payload
+	}
+	if budgetOverride != nil {
 		norm := util.NormalizeThinkingBudget(model, *budgetOverride)
 		budgetOverride = &norm
 	}
 	return util.ApplyGeminiCLIThinkingConfig(payload, budgetOverride, includeOverride)
 }
 
+// applyReasoningEffortMetadata applies reasoning effort overrides (reasoning.effort) when present in metadata.
+// It avoids overwriting an existing reasoning.effort field and only applies to models that support thinking.
+func applyReasoningEffortMetadata(payload []byte, metadata map[string]any, model string) []byte {
+	if len(metadata) == 0 {
+		return payload
+	}
+	if !util.ModelSupportsThinking(model) {
+		return payload
+	}
+	if gjson.GetBytes(payload, "reasoning.effort").Exists() {
+		return payload
+	}
+	if effort, ok := util.ReasoningEffortFromMetadata(metadata); ok && effort != "" {
+		if updated, err := sjson.SetBytes(payload, "reasoning.effort", effort); err == nil {
+			return updated
+		}
+	}
+	return payload
+}
+
+// applyReasoningEffortMetadataChatCompletions applies reasoning_effort (OpenAI chat completions field)
+// when present in metadata. It avoids overwriting an existing reasoning_effort field.
+func applyReasoningEffortMetadataChatCompletions(payload []byte, metadata map[string]any, model string) []byte {
+	if len(metadata) == 0 {
+		return payload
+	}
+	if !util.ModelSupportsThinking(model) {
+		return payload
+	}
+	if gjson.GetBytes(payload, "reasoning_effort").Exists() {
+		return payload
+	}
+	if effort, ok := util.ReasoningEffortFromMetadata(metadata); ok && effort != "" {
+		if updated, err := sjson.SetBytes(payload, "reasoning_effort", effort); err == nil {
+			return updated
+		}
+	}
+	return payload
+}
+
 // 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.

internal/runtime/executor/qwen_executor.go

@@ -12,6 +12,7 @@ import (
 
 	qwenauth "github.com/router-for-me/CLIProxyAPI/v6/internal/auth/qwen"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+	"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"
 	sdktranslator "github.com/router-for-me/CLIProxyAPI/v6/sdk/translator"
@@ -50,6 +51,10 @@ 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 = applyReasoningEffortMetadataChatCompletions(body, req.Metadata, req.Model)
+	if upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata); upstreamModel != "" {
+		body, _ = sjson.SetBytes(body, "model", upstreamModel)
+	}
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
@@ -121,6 +126,10 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 	to := sdktranslator.FromString("openai")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 
+	body = applyReasoningEffortMetadataChatCompletions(body, req.Metadata, req.Model)
+	if upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata); upstreamModel != "" {
+		body, _ = sjson.SetBytes(body, "model", upstreamModel)
+	}
 	toolsResult := gjson.GetBytes(body, "tools")
 	// I'm addressing the Qwen3 "poisoning" issue, which is caused by the model needing a tool to be defined. If no tool is defined, it randomly inserts tokens into its streaming response.
 	// This will have no real consequences. It's just to scare Qwen3.

internal/translator/antigravity/claude/antigravity_claude_request.go

@@ -123,6 +123,15 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 							functionResponse := client.FunctionResponse{ID: toolCallID, Name: funcName, Response: map[string]interface{}{"result": responseData}}
 							clientContent.Parts = append(clientContent.Parts, client.Part{FunctionResponse: &functionResponse})
 						}
+					} else if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "image" {
+						sourceResult := contentResult.Get("source")
+						if sourceResult.Get("type").String() == "base64" {
+							inlineData := &client.InlineData{
+								MimeType: sourceResult.Get("media_type").String(),
+								Data:     sourceResult.Get("data").String(),
+							}
+							clientContent.Parts = append(clientContent.Parts, client.Part{InlineData: inlineData})
+						}
 					}
 				}
 				contents = append(contents, clientContent)
@@ -180,7 +189,6 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 		if t.Get("type").String() == "enabled" {
 			if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
 				budget := int(b.Int())
-				budget = util.NormalizeThinkingBudget(modelName, budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}

internal/translator/antigravity/claude/antigravity_claude_response.go

@@ -12,6 +12,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	"github.com/tidwall/gjson"
@@ -36,6 +37,9 @@ type Params struct {
 	HasToolUse           bool   // Indicates if tool use was observed in the stream
 }
 
+// toolUseIDCounter provides a process-wide unique counter for tool use identifiers.
+var toolUseIDCounter uint64
+
 // ConvertAntigravityResponseToClaude performs sophisticated streaming response format conversion.
 // This function implements a complex state machine that translates backend client responses
 // into Claude Code-compatible Server-Sent Events (SSE) format. It manages different response types
@@ -216,7 +220,7 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
 
 				// Create the tool use block with unique ID and function details
 				data := fmt.Sprintf(`{"type":"content_block_start","index":%d,"content_block":{"type":"tool_use","id":"","name":"","input":{}}}`, params.ResponseIndex)
-				data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&toolUseIDCounter, 1)))
 				data, _ = sjson.Set(data, "content_block.name", fcName)
 				output = output + fmt.Sprintf("data: %s\n\n\n", data)
 

internal/translator/antigravity/openai/chat-completions/antigravity_openai_request.go

@@ -48,13 +48,13 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -66,15 +66,15 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 
 			if v := tc.Get("thinkingBudget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			} else if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 
@@ -82,7 +82,7 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
-			} else if setBudget && normalized != 0 {
+			} else if setBudget && budget != 0 {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}
 		}
@@ -94,7 +94,7 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 		if t := gjson.GetBytes(rawJSON, "thinking"); t.Exists() && t.IsObject() {
 			if t.Get("type").String() == "enabled" {
 				if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
-					budget := util.NormalizeThinkingBudget(modelName, int(b.Int()))
+					budget := int(b.Int())
 					out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 					out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 				}
@@ -102,15 +102,6 @@ func ConvertOpenAIRequestToAntigravity(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.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/max_tokens
 	if tr := gjson.GetBytes(rawJSON, "temperature"); tr.Exists() && tr.Type == gjson.Number {
 		out, _ = sjson.SetBytes(out, "request.generationConfig.temperature", tr.Num)

internal/translator/antigravity/openai/chat-completions/antigravity_openai_response.go

@@ -11,6 +11,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	. "github.com/router-for-me/CLIProxyAPI/v6/internal/translator/gemini/openai/chat-completions"
@@ -24,6 +25,9 @@ type convertCliResponseToOpenAIChatParams struct {
 	FunctionIndex int
 }
 
+// functionCallIDCounter provides a process-wide unique counter for function call identifiers.
+var functionCallIDCounter uint64
+
 // ConvertAntigravityResponseToOpenAI translates a single chunk of a streaming response from the
 // Gemini CLI API format to the OpenAI Chat Completions streaming format.
 // It processes various Gemini CLI event types and transforms them into OpenAI-compatible JSON responses.
@@ -146,7 +150,7 @@ func ConvertAntigravityResponseToOpenAI(_ context.Context, _ string, originalReq
 
 				functionCallTemplate := `{"id": "","index": 0,"type": "function","function": {"name": "","arguments": ""}}`
 				fcName := functionCallResult.Get("name").String()
-				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&functionCallIDCounter, 1)))
 				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "index", functionCallIndex)
 				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "function.name", fcName)
 				if fcArgsResult := functionCallResult.Get("args"); fcArgsResult.Exists() {

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

@@ -165,7 +165,6 @@ func ConvertClaudeRequestToCLI(modelName string, inputRawJSON []byte, _ bool) []
 		if t.Get("type").String() == "enabled" {
 			if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
 				budget := int(b.Int())
-				budget = util.NormalizeThinkingBudget(modelName, budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				out, _ = sjson.Set(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}

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

@@ -12,6 +12,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	"github.com/tidwall/gjson"
@@ -27,6 +28,9 @@ type Params struct {
 	ResponseIndex    int  // Index counter for content blocks in the streaming response
 }
 
+// toolUseIDCounter provides a process-wide unique counter for tool use identifiers.
+var toolUseIDCounter uint64
+
 // ConvertGeminiCLIResponseToClaude performs sophisticated streaming response format conversion.
 // This function implements a complex state machine that translates backend client responses
 // into Claude Code-compatible Server-Sent Events (SSE) format. It manages different response types
@@ -197,7 +201,7 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
 
 				// Create the tool use block with unique ID and function details
 				data := fmt.Sprintf(`{"type":"content_block_start","index":%d,"content_block":{"type":"tool_use","id":"","name":"","input":{}}}`, (*param).(*Params).ResponseIndex)
-				data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&toolUseIDCounter, 1)))
 				data, _ = sjson.Set(data, "content_block.name", fcName)
 				output = output + fmt.Sprintf("data: %s\n\n\n", data)
 

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

@@ -48,13 +48,13 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -66,15 +66,15 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 
 			if v := tc.Get("thinkingBudget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			} else if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 
@@ -82,21 +82,12 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", v.Bool())
-			} else if setBudget && normalized != 0 {
+			} else if setBudget && budget != 0 {
 				out, _ = sjson.SetBytes(out, "request.generationConfig.thinkingConfig.include_thoughts", true)
 			}
 		}
 	}
 
-	// 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)

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

@@ -11,6 +11,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	. "github.com/router-for-me/CLIProxyAPI/v6/internal/translator/gemini/openai/chat-completions"
@@ -24,6 +25,9 @@ type convertCliResponseToOpenAIChatParams struct {
 	FunctionIndex int
 }
 
+// functionCallIDCounter provides a process-wide unique counter for function call identifiers.
+var functionCallIDCounter uint64
+
 // ConvertCliResponseToOpenAI translates a single chunk of a streaming response from the
 // Gemini CLI API format to the OpenAI Chat Completions streaming format.
 // It processes various Gemini CLI event types and transforms them into OpenAI-compatible JSON responses.
@@ -146,7 +150,7 @@ func ConvertCliResponseToOpenAI(_ context.Context, _ string, originalRequestRawJ
 
 				functionCallTemplate := `{"id": "","index": 0,"type": "function","function": {"name": "","arguments": ""}}`
 				fcName := functionCallResult.Get("name").String()
-				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&functionCallIDCounter, 1)))
 				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "index", functionCallIndex)
 				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "function.name", fcName)
 				if fcArgsResult := functionCallResult.Get("args"); fcArgsResult.Exists() {

internal/translator/gemini/claude/gemini_claude_request.go

@@ -158,7 +158,6 @@ func ConvertClaudeRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 		if t.Get("type").String() == "enabled" {
 			if b := t.Get("budget_tokens"); b.Exists() && b.Type == gjson.Number {
 				budget := int(b.Int())
-				budget = util.NormalizeThinkingBudget(modelName, budget)
 				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 			}

internal/translator/gemini/claude/gemini_claude_response.go

@@ -12,6 +12,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	"github.com/tidwall/gjson"
@@ -26,6 +27,9 @@ type Params struct {
 	ResponseIndex    int
 }
 
+// toolUseIDCounter provides a process-wide unique counter for tool use identifiers.
+var toolUseIDCounter uint64
+
 // ConvertGeminiResponseToClaude performs sophisticated streaming response format conversion.
 // This function implements a complex state machine that translates backend client responses
 // into Claude-compatible Server-Sent Events (SSE) format. It manages different response types
@@ -197,7 +201,7 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
 
 				// Create the tool use block with unique ID and function details
 				data := fmt.Sprintf(`{"type":"content_block_start","index":%d,"content_block":{"type":"tool_use","id":"","name":"","input":{}}}`, (*param).(*Params).ResponseIndex)
-				data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				data, _ = sjson.Set(data, "content_block.id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&toolUseIDCounter, 1)))
 				data, _ = sjson.Set(data, "content_block.name", fcName)
 				output = output + fmt.Sprintf("data: %s\n\n\n", data)
 

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

@@ -48,13 +48,13 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -66,15 +66,15 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 
 			if v := tc.Get("thinkingBudget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			} else if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 
@@ -82,7 +82,7 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", v.Bool())
-			} else if setBudget && normalized != 0 {
+			} else if setBudget && budget != 0 {
 				out, _ = sjson.SetBytes(out, "generationConfig.thinkingConfig.include_thoughts", true)
 			}
 		}

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

@@ -11,6 +11,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	"github.com/tidwall/gjson"
@@ -23,6 +24,9 @@ type convertGeminiResponseToOpenAIChatParams struct {
 	FunctionIndex int
 }
 
+// functionCallIDCounter provides a process-wide unique counter for function call identifiers.
+var functionCallIDCounter uint64
+
 // ConvertGeminiResponseToOpenAI translates a single chunk of a streaming response from the
 // Gemini API format to the OpenAI Chat Completions streaming format.
 // It processes various Gemini event types and transforms them into OpenAI-compatible JSON responses.
@@ -148,7 +152,7 @@ func ConvertGeminiResponseToOpenAI(_ context.Context, _ string, originalRequestR
 
 				functionCallTemplate := `{"id": "","index": 0,"type": "function","function": {"name": "","arguments": ""}}`
 				fcName := functionCallResult.Get("name").String()
-				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&functionCallIDCounter, 1)))
 				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "index", functionCallIndex)
 				functionCallTemplate, _ = sjson.Set(functionCallTemplate, "function.name", fcName)
 				if fcArgsResult := functionCallResult.Get("args"); fcArgsResult.Exists() {
@@ -281,7 +285,7 @@ func ConvertGeminiResponseToOpenAINonStream(_ context.Context, _ string, origina
 				}
 				functionCallItemTemplate := `{"id": "","type": "function","function": {"name": "","arguments": ""}}`
 				fcName := functionCallResult.Get("name").String()
-				functionCallItemTemplate, _ = sjson.Set(functionCallItemTemplate, "id", fmt.Sprintf("%s-%d", fcName, time.Now().UnixNano()))
+				functionCallItemTemplate, _ = sjson.Set(functionCallItemTemplate, "id", fmt.Sprintf("%s-%d-%d", fcName, time.Now().UnixNano(), atomic.AddUint64(&functionCallIDCounter, 1)))
 				functionCallItemTemplate, _ = sjson.Set(functionCallItemTemplate, "function.name", fcName)
 				if fcArgsResult := functionCallResult.Get("args"); fcArgsResult.Exists() {
 					functionCallItemTemplate, _ = sjson.Set(functionCallItemTemplate, "function.arguments", fcArgsResult.Raw)

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

@@ -400,16 +400,16 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "minimal":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 1024))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 1024)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "low":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 4096))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 4096)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "medium":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 8192))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 8192)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		case "high":
-			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", util.NormalizeThinkingBudget(modelName, 32768))
+			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", 32768)
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 		default:
 			out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", -1)
@@ -421,32 +421,22 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 	if !hasOfficialThinking && util.ModelSupportsThinking(modelName) {
 		if tc := root.Get("extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
-			var normalized int
+			var budget int
 			if v := tc.Get("thinking_budget"); v.Exists() {
-				normalized = util.NormalizeThinkingBudget(modelName, int(v.Int()))
-				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", normalized)
+				budget = int(v.Int())
+				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", budget)
 				setBudget = true
 			}
 			if v := tc.Get("include_thoughts"); v.Exists() {
 				out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", v.Bool())
 			} else if setBudget {
-				if normalized != 0 {
+				if budget != 0 {
 					out, _ = sjson.Set(out, "generationConfig.thinkingConfig.include_thoughts", true)
 				}
 			}
 		}
 	}
 
-	// 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/gemini/openai/responses/gemini_openai-responses_response.go

@@ -5,6 +5,7 @@ import (
 	"context"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	"github.com/tidwall/gjson"
@@ -37,6 +38,12 @@ type geminiToResponsesState struct {
 	FuncCallIDs map[int]string
 }
 
+// responseIDCounter provides a process-wide unique counter for synthesized response identifiers.
+var responseIDCounter uint64
+
+// funcCallIDCounter provides a process-wide unique counter for function call identifiers.
+var funcCallIDCounter uint64
+
 func emitEvent(event string, payload string) string {
 	return fmt.Sprintf("event: %s\ndata: %s", event, payload)
 }
@@ -205,7 +212,7 @@ func ConvertGeminiResponseToOpenAIResponses(_ context.Context, modelName string,
 					st.FuncArgsBuf[idx] = &strings.Builder{}
 				}
 				if st.FuncCallIDs[idx] == "" {
-					st.FuncCallIDs[idx] = fmt.Sprintf("call_%d", time.Now().UnixNano())
+					st.FuncCallIDs[idx] = fmt.Sprintf("call_%d_%d", time.Now().UnixNano(), atomic.AddUint64(&funcCallIDCounter, 1))
 				}
 				st.FuncNames[idx] = name
 
@@ -464,7 +471,7 @@ func ConvertGeminiResponseToOpenAIResponsesNonStream(_ context.Context, _ string
 	// id: prefer provider responseId, otherwise synthesize
 	id := root.Get("responseId").String()
 	if id == "" {
-		id = fmt.Sprintf("resp_%x", time.Now().UnixNano())
+		id = fmt.Sprintf("resp_%x_%d", time.Now().UnixNano(), atomic.AddUint64(&responseIDCounter, 1))
 	}
 	// Normalize to response-style id (prefix resp_ if missing)
 	if !strings.HasPrefix(id, "resp_") {
@@ -575,7 +582,7 @@ func ConvertGeminiResponseToOpenAIResponsesNonStream(_ context.Context, _ string
 			if fc := p.Get("functionCall"); fc.Exists() {
 				name := fc.Get("name").String()
 				args := fc.Get("args")
-				callID := fmt.Sprintf("call_%x", time.Now().UnixNano())
+				callID := fmt.Sprintf("call_%x_%d", time.Now().UnixNano(), atomic.AddUint64(&funcCallIDCounter, 1))
 				outputs = append(outputs, map[string]interface{}{
 					"id":     fmt.Sprintf("fc_%s", callID),
 					"type":   "function_call",

internal/translator/openai/openai/responses/openai_openai-responses_response.go

@@ -5,6 +5,7 @@ import (
 	"context"
 	"fmt"
 	"strings"
+	"sync/atomic"
 	"time"
 
 	"github.com/tidwall/gjson"
@@ -41,6 +42,9 @@ type oaiToResponsesState struct {
 	UsageSeen        bool
 }
 
+// responseIDCounter provides a process-wide unique counter for synthesized response identifiers.
+var responseIDCounter uint64
+
 func emitRespEvent(event string, payload string) string {
 	return fmt.Sprintf("event: %s\ndata: %s", event, payload)
 }
@@ -590,7 +594,7 @@ func ConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream(_ context.Co
 	// id: use provider id if present, otherwise synthesize
 	id := root.Get("id").String()
 	if id == "" {
-		id = fmt.Sprintf("resp_%x", time.Now().UnixNano())
+		id = fmt.Sprintf("resp_%x_%d", time.Now().UnixNano(), atomic.AddUint64(&responseIDCounter, 1))
 	}
 	resp, _ = sjson.Set(resp, "id", id)
 

internal/util/gemini_thinking.go

@@ -1,8 +1,6 @@
 package util
 
 import (
-	"encoding/json"
-	"strconv"
 	"strings"
 
 	"github.com/tidwall/gjson"
@@ -15,80 +13,6 @@ const (
 	GeminiOriginalModelMetadataKey   = "gemini_original_model"
 )
 
-func ParseGeminiThinkingSuffix(model string) (string, *int, *bool, bool) {
-	if model == "" {
-		return model, nil, nil, false
-	}
-	lower := strings.ToLower(model)
-	if !strings.HasPrefix(lower, "gemini-") {
-		return model, nil, nil, false
-	}
-
-	if strings.HasSuffix(lower, "-nothinking") {
-		base := model[:len(model)-len("-nothinking")]
-		budgetValue := 0
-		if strings.HasPrefix(lower, "gemini-2.5-pro") {
-			budgetValue = 128
-		}
-		include := false
-		return base, &budgetValue, &include, true
-	}
-
-	// Handle "-reasoning" suffix: enables thinking with dynamic budget (-1)
-	// Maps: gemini-2.5-flash-reasoning -> gemini-2.5-flash with thinkingBudget=-1
-	if strings.HasSuffix(lower, "-reasoning") {
-		base := model[:len(model)-len("-reasoning")]
-		budgetValue := -1 // Dynamic budget
-		include := true
-		return base, &budgetValue, &include, true
-	}
-
-	idx := strings.LastIndex(lower, "-thinking-")
-	if idx == -1 {
-		return model, nil, nil, false
-	}
-
-	digits := model[idx+len("-thinking-"):]
-	if digits == "" {
-		return model, nil, nil, false
-	}
-	end := len(digits)
-	for i := 0; i < len(digits); i++ {
-		if digits[i] < '0' || digits[i] > '9' {
-			end = i
-			break
-		}
-	}
-	if end == 0 {
-		return model, nil, nil, false
-	}
-	valueStr := digits[:end]
-	value, err := strconv.Atoi(valueStr)
-	if err != nil {
-		return model, nil, nil, false
-	}
-	base := model[:idx]
-	budgetValue := value
-	return base, &budgetValue, nil, true
-}
-
-func NormalizeGeminiThinkingModel(modelName string) (string, map[string]any) {
-	baseModel, budget, include, matched := ParseGeminiThinkingSuffix(modelName)
-	if !matched {
-		return baseModel, nil
-	}
-	metadata := map[string]any{
-		GeminiOriginalModelMetadataKey: modelName,
-	}
-	if budget != nil {
-		metadata[GeminiThinkingBudgetMetadataKey] = *budget
-	}
-	if include != nil {
-		metadata[GeminiIncludeThoughtsMetadataKey] = *include
-	}
-	return baseModel, metadata
-}
-
 func ApplyGeminiThinkingConfig(body []byte, budget *int, includeThoughts *bool) []byte {
 	if budget == nil && includeThoughts == nil {
 		return body
@@ -133,78 +57,45 @@ func ApplyGeminiCLIThinkingConfig(body []byte, budget *int, includeThoughts *boo
 	return updated
 }
 
-func GeminiThinkingFromMetadata(metadata map[string]any) (*int, *bool, bool) {
-	if len(metadata) == 0 {
-		return nil, nil, false
+// modelsWithDefaultThinking lists models that should have thinking enabled by default
+// when no explicit thinkingConfig is provided.
+var modelsWithDefaultThinking = map[string]bool{
+	"gemini-3-pro-preview": true,
+}
+
+// ModelHasDefaultThinking returns true if the model should have thinking enabled by default.
+func ModelHasDefaultThinking(model string) bool {
+	return modelsWithDefaultThinking[model]
+}
+
+// ApplyDefaultThinkingIfNeeded injects default thinkingConfig for models that require it.
+// For standard Gemini API format (generationConfig.thinkingConfig path).
+// Returns the modified body if thinkingConfig was added, otherwise returns the original.
+func ApplyDefaultThinkingIfNeeded(model string, body []byte) []byte {
+	if !ModelHasDefaultThinking(model) {
+		return body
 	}
-	var (
-		budgetPtr  *int
-		includePtr *bool
-		matched    bool
-	)
-	if rawBudget, ok := metadata[GeminiThinkingBudgetMetadataKey]; ok {
-		switch v := rawBudget.(type) {
-		case int:
-			budget := v
-			budgetPtr = &budget
-			matched = true
-		case int32:
-			budget := int(v)
-			budgetPtr = &budget
-			matched = true
-		case int64:
-			budget := int(v)
-			budgetPtr = &budget
-			matched = true
-		case float64:
-			budget := int(v)
-			budgetPtr = &budget
-			matched = true
-		case json.Number:
-			if val, err := v.Int64(); err == nil {
-				budget := int(val)
-				budgetPtr = &budget
-				matched = true
-			}
-		}
+	if gjson.GetBytes(body, "generationConfig.thinkingConfig").Exists() {
+		return body
 	}
-	if rawInclude, ok := metadata[GeminiIncludeThoughtsMetadataKey]; ok {
-		switch v := rawInclude.(type) {
-		case bool:
-			include := v
-			includePtr = &include
-			matched = true
-		case string:
-			if parsed, err := strconv.ParseBool(v); err == nil {
-				include := parsed
-				includePtr = &include
-				matched = true
-			}
-		case json.Number:
-			if val, err := v.Int64(); err == nil {
-				include := val != 0
-				includePtr = &include
-				matched = true
-			}
-		case int:
-			include := v != 0
-			includePtr = &include
-			matched = true
-		case int32:
-			include := v != 0
-			includePtr = &include
-			matched = true
-		case int64:
-			include := v != 0
-			includePtr = &include
-			matched = true
-		case float64:
-			include := v != 0
-			includePtr = &include
-			matched = true
-		}
+	updated, _ := sjson.SetBytes(body, "generationConfig.thinkingConfig.thinkingBudget", -1)
+	updated, _ = sjson.SetBytes(updated, "generationConfig.thinkingConfig.include_thoughts", true)
+	return updated
+}
+
+// ApplyDefaultThinkingIfNeededCLI injects default thinkingConfig for models that require it.
+// For Gemini CLI API format (request.generationConfig.thinkingConfig path).
+// Returns the modified body if thinkingConfig was added, otherwise returns the original.
+func ApplyDefaultThinkingIfNeededCLI(model string, body []byte) []byte {
+	if !ModelHasDefaultThinking(model) {
+		return body
 	}
-	return budgetPtr, includePtr, matched
+	if gjson.GetBytes(body, "request.generationConfig.thinkingConfig").Exists() {
+		return body
+	}
+	updated, _ := sjson.SetBytes(body, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
+	updated, _ = sjson.SetBytes(updated, "request.generationConfig.thinkingConfig.include_thoughts", true)
+	return updated
 }
 
 // StripThinkingConfigIfUnsupported removes thinkingConfig from the request body
@@ -223,6 +114,32 @@ func StripThinkingConfigIfUnsupported(model string, body []byte) []byte {
 	return updated
 }
 
+// NormalizeGeminiThinkingBudget normalizes the thinkingBudget value in a standard Gemini
+// request body (generationConfig.thinkingConfig.thinkingBudget path).
+func NormalizeGeminiThinkingBudget(model string, body []byte) []byte {
+	const budgetPath = "generationConfig.thinkingConfig.thinkingBudget"
+	budget := gjson.GetBytes(body, budgetPath)
+	if !budget.Exists() {
+		return body
+	}
+	normalized := NormalizeThinkingBudget(model, int(budget.Int()))
+	updated, _ := sjson.SetBytes(body, budgetPath, normalized)
+	return updated
+}
+
+// NormalizeGeminiCLIThinkingBudget normalizes the thinkingBudget value in a Gemini CLI
+// request body (request.generationConfig.thinkingConfig.thinkingBudget path).
+func NormalizeGeminiCLIThinkingBudget(model string, body []byte) []byte {
+	const budgetPath = "request.generationConfig.thinkingConfig.thinkingBudget"
+	budget := gjson.GetBytes(body, budgetPath)
+	if !budget.Exists() {
+		return body
+	}
+	normalized := NormalizeThinkingBudget(model, int(budget.Int()))
+	updated, _ := sjson.SetBytes(body, budgetPath, normalized)
+	return updated
+}
+
 // ConvertThinkingLevelToBudget checks for "generationConfig.thinkingConfig.thinkingLevel"
 // and converts it to "thinkingBudget".
 // "high" -> 32768

internal/util/thinking_suffix.go

@@ -0,0 +1,327 @@
+package util
+
+import (
+	"encoding/json"
+	"strconv"
+	"strings"
+)
+
+const (
+	ThinkingBudgetMetadataKey          = "thinking_budget"
+	ThinkingIncludeThoughtsMetadataKey = "thinking_include_thoughts"
+	ReasoningEffortMetadataKey         = "reasoning_effort"
+	ThinkingOriginalModelMetadataKey   = "thinking_original_model"
+)
+
+// NormalizeThinkingModel parses dynamic thinking suffixes on model names and returns
+// the normalized base model with extracted metadata. Supported patterns:
+//   - "-thinking-<number>" extracts a numeric budget
+//   - "-thinking-<level>" extracts a reasoning effort level (minimal/low/medium/high/xhigh/auto/none)
+//   - "-thinking" maps to a default reasoning effort of "medium"
+//   - "-reasoning" maps to dynamic budget (-1) and include_thoughts=true
+//   - "-nothinking" maps to budget=0 and include_thoughts=false
+func NormalizeThinkingModel(modelName string) (string, map[string]any) {
+	if modelName == "" {
+		return modelName, nil
+	}
+
+	lower := strings.ToLower(modelName)
+	baseModel := modelName
+
+	var (
+		budgetOverride  *int
+		includeThoughts *bool
+		reasoningEffort *string
+		matched         bool
+	)
+
+	switch {
+	case strings.HasSuffix(lower, "-nothinking"):
+		baseModel = modelName[:len(modelName)-len("-nothinking")]
+		budget := 0
+		include := false
+		budgetOverride = &budget
+		includeThoughts = &include
+		matched = true
+	case strings.HasSuffix(lower, "-reasoning"):
+		baseModel = modelName[:len(modelName)-len("-reasoning")]
+		budget := -1
+		include := true
+		budgetOverride = &budget
+		includeThoughts = &include
+		matched = true
+	default:
+		if idx := strings.LastIndex(lower, "-thinking-"); idx != -1 {
+			value := modelName[idx+len("-thinking-"):]
+			if value != "" {
+				if parsed, ok := parseIntPrefix(value); ok {
+					baseModel = modelName[:idx]
+					budgetOverride = &parsed
+					matched = true
+				} else if effort, okEffort := normalizeReasoningEffort(value); okEffort {
+					baseModel = modelName[:idx]
+					reasoningEffort = &effort
+					matched = true
+				}
+			}
+		} else if strings.HasSuffix(lower, "-thinking") {
+			baseModel = modelName[:len(modelName)-len("-thinking")]
+			effort := "medium"
+			reasoningEffort = &effort
+			matched = true
+		}
+	}
+
+	if !matched {
+		return baseModel, nil
+	}
+
+	metadata := map[string]any{
+		ThinkingOriginalModelMetadataKey: modelName,
+	}
+	if budgetOverride != nil {
+		metadata[ThinkingBudgetMetadataKey] = *budgetOverride
+	}
+	if includeThoughts != nil {
+		metadata[ThinkingIncludeThoughtsMetadataKey] = *includeThoughts
+	}
+	if reasoningEffort != nil {
+		metadata[ReasoningEffortMetadataKey] = *reasoningEffort
+	}
+	return baseModel, metadata
+}
+
+// ThinkingFromMetadata extracts thinking overrides from metadata produced by NormalizeThinkingModel.
+// It accepts both the new generic keys and legacy Gemini-specific keys.
+func ThinkingFromMetadata(metadata map[string]any) (*int, *bool, *string, bool) {
+	if len(metadata) == 0 {
+		return nil, nil, nil, false
+	}
+
+	var (
+		budgetPtr  *int
+		includePtr *bool
+		effortPtr  *string
+		matched    bool
+	)
+
+	readBudget := func(key string) {
+		if budgetPtr != nil {
+			return
+		}
+		if raw, ok := metadata[key]; ok {
+			if v, okNumber := parseNumberToInt(raw); okNumber {
+				budget := v
+				budgetPtr = &budget
+				matched = true
+			}
+		}
+	}
+
+	readInclude := func(key string) {
+		if includePtr != nil {
+			return
+		}
+		if raw, ok := metadata[key]; ok {
+			switch v := raw.(type) {
+			case bool:
+				val := v
+				includePtr = &val
+				matched = true
+			case *bool:
+				if v != nil {
+					val := *v
+					includePtr = &val
+					matched = true
+				}
+			}
+		}
+	}
+
+	readEffort := func(key string) {
+		if effortPtr != nil {
+			return
+		}
+		if raw, ok := metadata[key]; ok {
+			if val, okStr := raw.(string); okStr && strings.TrimSpace(val) != "" {
+				normalized := strings.ToLower(strings.TrimSpace(val))
+				effortPtr = &normalized
+				matched = true
+			}
+		}
+	}
+
+	readBudget(ThinkingBudgetMetadataKey)
+	readBudget(GeminiThinkingBudgetMetadataKey)
+	readInclude(ThinkingIncludeThoughtsMetadataKey)
+	readInclude(GeminiIncludeThoughtsMetadataKey)
+	readEffort(ReasoningEffortMetadataKey)
+	readEffort("reasoning.effort")
+
+	return budgetPtr, includePtr, effortPtr, matched
+}
+
+// ResolveThinkingConfigFromMetadata derives thinking budget/include overrides,
+// converting reasoning effort strings into budgets when possible.
+func ResolveThinkingConfigFromMetadata(model string, metadata map[string]any) (*int, *bool, bool) {
+	budget, include, effort, matched := ThinkingFromMetadata(metadata)
+	if !matched {
+		return nil, nil, false
+	}
+
+	if budget == nil && effort != nil {
+		if derived, ok := ThinkingEffortToBudget(model, *effort); ok {
+			budget = &derived
+		}
+	}
+	return budget, include, budget != nil || include != nil || effort != nil
+}
+
+// ReasoningEffortFromMetadata resolves a reasoning effort string from metadata,
+// inferring "auto" and "none" when budgets request dynamic or disabled thinking.
+func ReasoningEffortFromMetadata(metadata map[string]any) (string, bool) {
+	budget, include, effort, matched := ThinkingFromMetadata(metadata)
+	if !matched {
+		return "", false
+	}
+	if effort != nil && *effort != "" {
+		return *effort, true
+	}
+	if budget != nil {
+		switch *budget {
+		case -1:
+			return "auto", true
+		case 0:
+			return "none", true
+		}
+	}
+	if include != nil && !*include {
+		return "none", true
+	}
+	return "", true
+}
+
+// ThinkingEffortToBudget maps reasoning effort levels to approximate budgets,
+// clamping the result to the model's supported range.
+func ThinkingEffortToBudget(model, effort string) (int, bool) {
+	if effort == "" {
+		return 0, false
+	}
+	switch strings.ToLower(effort) {
+	case "none":
+		return 0, true
+	case "auto":
+		return NormalizeThinkingBudget(model, -1), true
+	case "minimal":
+		return NormalizeThinkingBudget(model, 512), true
+	case "low":
+		return NormalizeThinkingBudget(model, 1024), true
+	case "medium":
+		return NormalizeThinkingBudget(model, 8192), true
+	case "high":
+		return NormalizeThinkingBudget(model, 24576), true
+	case "xhigh":
+		return NormalizeThinkingBudget(model, 32768), true
+	default:
+		return 0, false
+	}
+}
+
+// ResolveOriginalModel returns the original model name stored in metadata (if present),
+// otherwise falls back to the provided model.
+func ResolveOriginalModel(model string, metadata map[string]any) string {
+	normalize := func(name string) string {
+		if name == "" {
+			return ""
+		}
+		if base, _ := NormalizeThinkingModel(name); base != "" {
+			return base
+		}
+		return strings.TrimSpace(name)
+	}
+
+	if metadata != nil {
+		if v, ok := metadata[ThinkingOriginalModelMetadataKey]; ok {
+			if s, okStr := v.(string); okStr && strings.TrimSpace(s) != "" {
+				if base := normalize(s); base != "" {
+					return base
+				}
+			}
+		}
+		if v, ok := metadata[GeminiOriginalModelMetadataKey]; ok {
+			if s, okStr := v.(string); okStr && strings.TrimSpace(s) != "" {
+				if base := normalize(s); base != "" {
+					return base
+				}
+			}
+		}
+	}
+	// Fallback: try to re-normalize the model name when metadata was dropped.
+	if base := normalize(model); base != "" {
+		return base
+	}
+	return model
+}
+
+func parseIntPrefix(value string) (int, bool) {
+	if value == "" {
+		return 0, false
+	}
+	digits := strings.TrimLeft(value, "-")
+	if digits == "" {
+		return 0, false
+	}
+	end := len(digits)
+	for i := 0; i < len(digits); i++ {
+		if digits[i] < '0' || digits[i] > '9' {
+			end = i
+			break
+		}
+	}
+	if end == 0 {
+		return 0, false
+	}
+	val, err := strconv.Atoi(digits[:end])
+	if err != nil {
+		return 0, false
+	}
+	return val, true
+}
+
+func parseNumberToInt(raw any) (int, bool) {
+	switch v := raw.(type) {
+	case int:
+		return v, true
+	case int32:
+		return int(v), true
+	case int64:
+		return int(v), true
+	case float64:
+		return int(v), true
+	case json.Number:
+		if val, err := v.Int64(); err == nil {
+			return int(val), true
+		}
+	case string:
+		if strings.TrimSpace(v) == "" {
+			return 0, false
+		}
+		if parsed, err := strconv.Atoi(strings.TrimSpace(v)); err == nil {
+			return parsed, true
+		}
+	}
+	return 0, false
+}
+
+func normalizeReasoningEffort(value string) (string, bool) {
+	if value == "" {
+		return "", false
+	}
+	effort := strings.ToLower(strings.TrimSpace(value))
+	switch effort {
+	case "minimal", "low", "medium", "high", "xhigh", "auto", "none":
+		return effort, true
+	default:
+		return "", false
+	}
+}

sdk/api/handlers/gemini/gemini_handlers.go

@@ -48,8 +48,24 @@ func (h *GeminiAPIHandler) Models() []map[string]any {
 // GeminiModels handles the Gemini models listing endpoint.
 // It returns a JSON response containing available Gemini models and their specifications.
 func (h *GeminiAPIHandler) GeminiModels(c *gin.Context) {
+	rawModels := h.Models()
+	normalizedModels := make([]map[string]any, 0, len(rawModels))
+	defaultMethods := []string{"generateContent"}
+	for _, model := range rawModels {
+		normalizedModel := make(map[string]any, len(model))
+		for k, v := range model {
+			normalizedModel[k] = v
+		}
+		if name, ok := normalizedModel["name"].(string); ok && name != "" && !strings.HasPrefix(name, "models/") {
+			normalizedModel["name"] = "models/" + name
+		}
+		if _, ok := normalizedModel["supportedGenerationMethods"]; !ok {
+			normalizedModel["supportedGenerationMethods"] = defaultMethods
+		}
+		normalizedModels = append(normalizedModels, normalizedModel)
+	}
 	c.JSON(http.StatusOK, gin.H{
-		"models": h.Models(),
+		"models": normalizedModels,
 	})
 }
 

sdk/api/handlers/handlers.go

@@ -323,18 +323,32 @@ func (h *BaseAPIHandler) getRequestDetails(modelName string) (providers []string
 
 	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(resolvedModelName)
+	targetModelName := resolvedModelName
+	if isDynamic {
+		targetModelName = extractedModelName
+	}
+
+	// Normalize the model name to handle dynamic thinking suffixes before determining the provider.
+	normalizedModel, metadata = normalizeModelMetadata(targetModelName)
 
 	if isDynamic {
 		providers = []string{providerName}
-		// For dynamic models, the extractedModelName is already normalized by parseDynamicModel
-		// so we use it as the final normalizedModel.
-		normalizedModel = extractedModelName
 	} else {
 		// For non-dynamic models, use the normalizedModel to get the provider name.
 		providers = util.GetProviderName(normalizedModel)
+		if len(providers) == 0 && metadata != nil {
+			if originalRaw, ok := metadata[util.ThinkingOriginalModelMetadataKey]; ok {
+				if originalModel, okStr := originalRaw.(string); okStr {
+					originalModel = strings.TrimSpace(originalModel)
+					if originalModel != "" && !strings.EqualFold(originalModel, normalizedModel) {
+						if altProviders := util.GetProviderName(originalModel); len(altProviders) > 0 {
+							providers = altProviders
+							normalizedModel = originalModel
+						}
+					}
+				}
+			}
+		}
 	}
 
 	if len(providers) == 0 {
@@ -382,7 +396,7 @@ func cloneBytes(src []byte) []byte {
 }
 
 func normalizeModelMetadata(modelName string) (string, map[string]any) {
-	return util.NormalizeGeminiThinkingModel(modelName)
+	return util.NormalizeThinkingModel(modelName)
 }
 
 func cloneMetadata(src map[string]any) map[string]any {

sdk/cliproxy/service.go

@@ -498,7 +498,7 @@ func (s *Service) Run(ctx context.Context) error {
 	}()
 
 	time.Sleep(100 * time.Millisecond)
-	fmt.Printf("API server started successfully on: %d\n", s.cfg.Port)
+	fmt.Printf("API server started successfully on: %s:%d\n", s.cfg.Host, s.cfg.Port)
 
 	if s.hooks.OnAfterStart != nil {
 		s.hooks.OnAfterStart(s)
@@ -779,7 +779,7 @@ func (s *Service) registerModelsForAuth(a *coreauth.Auth) {
 							Created:     time.Now().Unix(),
 							OwnedBy:     compat.Name,
 							Type:        "openai-compatibility",
-							DisplayName: m.Name,
+							DisplayName: modelID,
 						})
 					}
 					// Register and return