feat(temporal): anti-drift injection for all provider requests

cmd/server/main.go

@@ -505,6 +505,7 @@ func main() {
 				misc.StartAntigravityVersionUpdater(context.Background())
 				if !localModel {
 					registry.StartModelsUpdater(context.Background())
+					registry.StartOpenRouterEnrichment(context.Background())
 				}
 				hook := tui.NewLogHook(2000)
 				hook.SetFormatter(&logging.LogFormatter{})
@@ -581,6 +582,7 @@ func main() {
 			misc.StartAntigravityVersionUpdater(context.Background())
 			if !localModel {
 				registry.StartModelsUpdater(context.Background())
+				registry.StartOpenRouterEnrichment(context.Background())
 			}
 			cmd.StartService(cfg, configFilePath, password)
 		}

internal/api/handlers/management/model_definitions.go

@@ -3,6 +3,7 @@ package management
 import (
 	"net/http"
 	"strings"
+	"time"
 
 	"github.com/gin-gonic/gin"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
@@ -31,3 +32,93 @@ func (h *Handler) GetStaticModelDefinitions(c *gin.Context) {
 		"models":  models,
 	})
 }
+
+// GetModelsHealth returns comprehensive health information for all registered models.
+func (h *Handler) GetModelsHealth(c *gin.Context) {
+	globalRegistry := registry.GetGlobalRegistry()
+
+	// Get ALL registered models including suspended/unhealthy for full operator visibility
+	models := globalRegistry.GetAllRegisteredModels("openai")
+	// Build enhanced model health with suspension and provider info
+	enhancedModels := make([]map[string]any, 0, len(models))
+	for _, model := range models {
+		modelID, _ := model["id"].(string)
+		if modelID == "" {
+			continue
+		}
+		// Get health details for this model
+		providers, suspendedClients, count := globalRegistry.GetModelHealthDetails(modelID)
+		// Build suspension summary
+		suspensionSummary := make(map[string]any)
+		if len(suspendedClients) > 0 {
+			suspensionSummary["count"] = len(suspendedClients)
+			suspensionSummary["clients"] = suspendedClients
+			// Extract unique reasons
+			reasons := make(map[string]bool)
+			for _, reason := range suspendedClients {
+				if reason == "" {
+					reasons["unknown"] = true
+				} else {
+					reasons[strings.ToLower(reason)] = true
+				}
+			}
+			reasonList := make([]string, 0, len(reasons))
+			for reason := range reasons {
+				reasonList = append(reasonList, reason)
+			}
+			suspensionSummary["reasons"] = reasonList
+		} else {
+			suspensionSummary["count"] = 0
+			suspensionSummary["clients"] = map[string]string{}
+			suspensionSummary["reasons"] = []string{}
+		}
+
+		// Build provider list
+		providerList := make([]string, 0, len(providers))
+		for provider := range providers {
+			providerList = append(providerList, provider)
+		}
+		// Create enhanced model entry
+		enhancedModel := make(map[string]any)
+		for k, v := range model {
+			enhancedModel[k] = v
+		}
+		enhancedModel["total_clients"] = count
+		enhancedModel["providers"] = providerList
+		enhancedModel["provider_counts"] = providers
+		enhancedModel["suspension"] = suspensionSummary
+		enhancedModels = append(enhancedModels, enhancedModel)
+	}
+	// Build the sources map indicating where context_length came from
+	sources := registry.BuildModelSources(globalRegistry)
+	// Get last refresh timestamp from OpenRouter enrichment
+	lastRefresh := registry.GetOpenRouterLastRefresh()
+	lastRefreshStr := ""
+	if !lastRefresh.IsZero() {
+		lastRefreshStr = lastRefresh.Format(time.RFC3339)
+	}
+	response := gin.H{
+		"models":           enhancedModels,
+		"sources":          sources,
+		"last_refresh":     lastRefreshStr,
+		"refresh_interval": "24h",
+	}
+	c.JSON(http.StatusOK, response)
+}
+
+// RefreshModels triggers an immediate refresh of model metadata from OpenRouter and the enrichment cache.
+// This re-fetches context_length from OpenRouter's public /api/v1/models endpoint and enriches registered models that lack this data.
+func (h *Handler) RefreshModels(c *gin.Context) {
+	count := registry.TriggerOpenRouterRefresh(c.Request.Context())
+	if count == 0 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "no models registered"})
+		return
+	}
+	lastRefresh := registry.GetOpenRouterLastRefresh()
+	c.JSON(http.StatusOK, gin.H{
+		"status":         "refreshed",
+		"enriched_count": count,
+		"last_refresh":   lastRefresh.Format(time.RFC3339),
+		"total_models":   len(registry.GetGlobalRegistry().GetAvailableModels("openai")),
+	})
+}

internal/api/server.go

@@ -627,6 +627,8 @@ func (s *Server) registerManagementRoutes() {
 		mgmt.GET("/auth-files", s.mgmt.ListAuthFiles)
 		mgmt.GET("/auth-files/models", s.mgmt.GetAuthFileModels)
 		mgmt.GET("/model-definitions/:channel", s.mgmt.GetStaticModelDefinitions)
+		mgmt.GET("/models/health", s.mgmt.GetModelsHealth)
+		mgmt.POST("/models/refresh", s.mgmt.RefreshModels)
 		mgmt.GET("/auth-files/download", s.mgmt.DownloadAuthFile)
 		mgmt.POST("/auth-files", s.mgmt.UploadAuthFile)
 		mgmt.DELETE("/auth-files", s.mgmt.DeleteAuthFile)
@@ -773,8 +775,8 @@ func (s *Server) unifiedModelsHandler(openaiHandler *openai.OpenAIAPIHandler, cl
 	return func(c *gin.Context) {
 		userAgent := c.GetHeader("User-Agent")
 
-		// Route to Claude handler if User-Agent starts with "claude-cli"
-		if strings.HasPrefix(userAgent, "claude-cli") {
+		// Route to Claude handler if User-Agent starts with "claude-cli" or "claude-code"
+		if strings.HasPrefix(userAgent, "claude-cli") || strings.HasPrefix(userAgent, "claude-code") || strings.HasPrefix(userAgent, "claude/") {
 			// log.Debugf("Routing /v1/models to Claude handler for User-Agent: %s", userAgent)
 			claudeHandler.ClaudeModels(c)
 		} else {

internal/config/config.go

@@ -14,9 +14,10 @@ import (
 	"syscall"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
-	log "github.com/sirupsen/logrus"
-	"golang.org/x/crypto/bcrypt"
-	"gopkg.in/yaml.v3"
+	temporal "github.com/router-for-me/CLIProxyAPI/v6/internal/temporal"
+ 	log "github.com/sirupsen/logrus"
+ 	"golang.org/x/crypto/bcrypt"
+ 	"gopkg.in/yaml.v3"
 )
 
 const (
@@ -126,6 +127,10 @@ type Config struct {
 	// Payload defines default and override rules for provider payload parameters.
 	Payload PayloadConfig `yaml:"payload" json:"payload"`
 
+	// Temporal controls temporal anti-drift injection into LLM request payloads.
+	// nil means use defaults (enabled, every request).
+	Temporal *temporal.Config `yaml:"temporal" json:"temporal"`
+
 	legacyMigrationPending bool `yaml:"-" json:"-"`
 }
 

internal/registry/model_registry.go

@@ -243,6 +243,14 @@ func (r *ModelRegistry) RegisterClient(clientID, clientProvider string, models [
 		if model == nil || model.ID == "" {
 			continue
 		}
+		if staticInfo := LookupStaticModelInfo(model.ID); staticInfo != nil {
+			if model.ContextLength == 0 && staticInfo.ContextLength > 0 {
+				model.ContextLength = staticInfo.ContextLength
+			}
+			if model.MaxCompletionTokens == 0 && staticInfo.MaxCompletionTokens > 0 {
+				model.MaxCompletionTokens = staticInfo.MaxCompletionTokens
+			}
+		}
 		rawModelIDs = append(rawModelIDs, model.ID)
 		newCounts[model.ID]++
 		if _, exists := newModels[model.ID]; exists {
@@ -875,6 +883,25 @@ func cloneModelMapValue(value any) any {
 //
 // Returns:
 //   - []*ModelInfo: List of available models for the provider
+// GetAllRegisteredModels returns metadata for every registered model regardless of
+// suspension or quota state. Use this for administrative/health endpoints that need
+// visibility into ALL models, including unhealthy ones.
+func (r *ModelRegistry) GetAllRegisteredModels(handlerType string) []map[string]any {
+	r.mutex.RLock()
+	defer r.mutex.RUnlock()
+	models := make([]map[string]any, 0, len(r.models))
+	for _, registration := range r.models {
+		if registration.Info == nil {
+			continue
+		}
+		model := r.convertModelToMap(registration.Info, handlerType)
+		if model != nil {
+			models = append(models, model)
+		}
+	}
+	return models
+}
+
 func (r *ModelRegistry) GetAvailableModelsByProvider(provider string) []*ModelInfo {
 	provider = strings.ToLower(strings.TrimSpace(provider))
 	if provider == "" {
@@ -1158,6 +1185,14 @@ func (r *ModelRegistry) convertModelToMap(model *ModelInfo, handlerType string)
 		if model.DisplayName != "" {
 			result["display_name"] = model.DisplayName
 		}
+		if model.ContextLength > 0 {
+			result["context_length"] = model.ContextLength
+			result["context_window_size"] = model.ContextLength
+		}
+		if model.MaxCompletionTokens > 0 {
+			result["max_completion_tokens"] = model.MaxCompletionTokens
+			result["max_output_tokens"] = model.MaxCompletionTokens
+		}
 		return result
 
 	case "gemini":
@@ -1315,3 +1350,42 @@ func (r *ModelRegistry) GetModelsForClient(clientID string) []*ModelInfo {
 	}
 	return result
 }
+
+// GetModelHealthDetails returns detailed health information for a model including
+// providers, suspended clients with reasons, and availability counts.
+func (r *ModelRegistry) GetModelHealthDetails(modelID string) (providers map[string]int, suspendedClients map[string]string, count int) {
+	r.mutex.RLock()
+	defer r.mutex.RUnlock()
+
+	if registration, exists := r.models[modelID]; exists && registration != nil {
+		// Clone providers map
+		if len(registration.Providers) > 0 {
+			providers = make(map[string]int, len(registration.Providers))
+			for k, v := range registration.Providers {
+				providers[k] = v
+			}
+		}
+		// Clone suspended clients map
+		if len(registration.SuspendedClients) > 0 {
+			suspendedClients = make(map[string]string, len(registration.SuspendedClients))
+			for k, v := range registration.SuspendedClients {
+				suspendedClients[k] = v
+			}
+		}
+		count = registration.Count
+	}
+	return
+}
+
+// SetModelContextLength updates the context_length on the live model registration.
+// This writes through to the actual stored ModelInfo, not a clone.
+func (r *ModelRegistry) SetModelContextLength(modelID string, contextLength int) bool {
+	r.mutex.Lock()
+	defer r.mutex.Unlock()
+	reg, ok := r.models[modelID]
+	if !ok || reg == nil || reg.Info == nil {
+		return false
+	}
+	reg.Info.ContextLength = contextLength
+	return true
+}

internal/registry/models/models.json

@@ -2708,6 +2708,7 @@
       "display_name": "Gemini 3.1 Flash Image",
       "name": "gemini-3.1-flash-image",
       "description": "Gemini 3.1 Flash Image",
+      "context_length": 1048576,
       "thinking": {
         "min": 128,
         "max": 32768,