feat(gateway): add provider health tracking and dashboard stats API

llm-gateway/cmd/gateway/main.go

@@ -90,6 +90,9 @@ func main() {
 	}
 	log.Printf("Registered %d models", len(cfg.Models))
 
+	// Provider health tracker
+	healthTracker := provider.NewHealthTracker(5 * time.Minute)
+
 	// Auth store (static tokens checked in-memory, not seeded to DB)
 	var staticTokens []auth.StaticToken
 	for _, t := range cfg.Tokens {
@@ -114,12 +117,20 @@ func main() {
 	m := metrics.New()
 
 	// Handlers
-	proxyHandler := proxy.NewHandler(registry, asyncLogger, c, m, cfg)
+	proxyHandler := proxy.NewHandler(registry, asyncLogger, c, m, cfg, healthTracker)
 	modelsHandler := proxy.NewModelsHandler(registry)
 	proxyAuth := proxy.NewAuthMiddleware(authStore)
 	rateLimiter := proxy.NewRateLimiter(db)
 	statsAPI := dashboard.NewStatsAPI(db, authStore)
+	statsAPI.SetHealthTracker(healthTracker)
+	if c != nil {
+		statsAPI.SetCache(c)
+	}
 	dash := dashboard.NewDashboard(authStore, statsAPI)
+	dash.SetRegistry(registry)
+	if c != nil {
+		dash.SetCache(c)
+	}
 
 	// Router
 	r := chi.NewRouter()
@@ -172,6 +183,8 @@ func main() {
 
 		// Dashboard pages (HTMX)
 		r.Get("/dashboard", dash.DashboardPage)
+		r.Get("/logs", dash.LogsPage)
+		r.Get("/models", dash.ModelsPage)
 		r.Get("/tokens", dash.TokensPage)
 		r.Get("/settings", dash.SettingsPage)
 
@@ -205,6 +218,11 @@ func main() {
 		r.Get("/api/stats/providers", statsAPI.Providers)
 		r.Get("/api/stats/tokens", statsAPI.Tokens)
 		r.Get("/api/stats/timeseries", statsAPI.Timeseries)
+		r.Get("/api/stats/logs", statsAPI.Logs)
+		r.Get("/api/stats/latency", statsAPI.Latency)
+		r.Get("/api/stats/cost-breakdown", statsAPI.CostBreakdown)
+		r.Get("/api/stats/provider-health", statsAPI.ProviderHealthHandler)
+		r.Get("/api/stats/cache", statsAPI.CacheStats)
 
 		// Admin-only: user management
 		r.Group(func(r chi.Router) {

llm-gateway/internal/cache/cache.go

@@ -56,6 +56,120 @@ func (c *Cache) Close() error {
 	return c.client.Close()
 }
 
+// CacheStats holds cache statistics from the Valkey/Redis server.
+type CacheStats struct {
+	Hits       int64   `json:"hits"`
+	Misses     int64   `json:"misses"`
+	HitRate    float64 `json:"hit_rate"`
+	MemoryUsed string  `json:"memory_used"`
+	Keys       int64   `json:"keys"`
+	Connected  bool    `json:"connected"`
+}
+
+// Stats returns cache statistics by querying Valkey/Redis INFO.
+func (c *Cache) Stats(ctx context.Context) *CacheStats {
+	stats := &CacheStats{}
+
+	// Check connectivity
+	if err := c.client.Ping(ctx).Err(); err != nil {
+		return stats
+	}
+	stats.Connected = true
+
+	// Parse INFO stats for hits/misses
+	info, err := c.client.Info(ctx, "stats").Result()
+	if err == nil {
+		stats.Hits = parseInfoInt(info, "keyspace_hits")
+		stats.Misses = parseInfoInt(info, "keyspace_misses")
+		total := stats.Hits + stats.Misses
+		if total > 0 {
+			stats.HitRate = float64(stats.Hits) / float64(total)
+		}
+	}
+
+	// Parse INFO memory
+	memInfo, err := c.client.Info(ctx, "memory").Result()
+	if err == nil {
+		stats.MemoryUsed = parseInfoString(memInfo, "used_memory_human")
+	}
+
+	// Parse INFO keyspace
+	ksInfo, err := c.client.Info(ctx, "keyspace").Result()
+	if err == nil {
+		stats.Keys = parseKeyspaceKeys(ksInfo)
+	}
+
+	return stats
+}
+
+func parseInfoInt(info, key string) int64 {
+	prefix := key + ":"
+	for _, line := range splitLines(info) {
+		if len(line) > len(prefix) && line[:len(prefix)] == prefix {
+			var v int64
+			fmt.Sscanf(line[len(prefix):], "%d", &v)
+			return v
+		}
+	}
+	return 0
+}
+
+func parseInfoString(info, key string) string {
+	prefix := key + ":"
+	for _, line := range splitLines(info) {
+		if len(line) > len(prefix) && line[:len(prefix)] == prefix {
+			val := line[len(prefix):]
+			// Trim trailing \r
+			if len(val) > 0 && val[len(val)-1] == '\r' {
+				val = val[:len(val)-1]
+			}
+			return val
+		}
+	}
+	return ""
+}
+
+func parseKeyspaceKeys(info string) int64 {
+	// Format: db0:keys=123,expires=45,avg_ttl=6789
+	for _, line := range splitLines(info) {
+		if len(line) > 3 && line[:2] == "db" {
+			prefix := "keys="
+			idx := -1
+			for i := 0; i <= len(line)-len(prefix); i++ {
+				if line[i:i+len(prefix)] == prefix {
+					idx = i + len(prefix)
+					break
+				}
+			}
+			if idx >= 0 {
+				end := idx
+				for end < len(line) && line[end] >= '0' && line[end] <= '9' {
+					end++
+				}
+				var v int64
+				fmt.Sscanf(line[idx:end], "%d", &v)
+				return v
+			}
+		}
+	}
+	return 0
+}
+
+func splitLines(s string) []string {
+	var lines []string
+	start := 0
+	for i := 0; i < len(s); i++ {
+		if s[i] == '\n' {
+			lines = append(lines, s[start:i])
+			start = i + 1
+		}
+	}
+	if start < len(s) {
+		lines = append(lines, s[start:])
+	}
+	return lines
+}
+
 func (c *Cache) cacheKey(model string, requestBody []byte) string {
 	h := sha256.New()
 	h.Write([]byte(model))

llm-gateway/internal/dashboard/api.go

@@ -3,9 +3,13 @@ package dashboard
 import (
 	"encoding/json"
 	"net/http"
+	"sort"
+	"strconv"
 	"time"
 
 	"llm-gateway/internal/auth"
+	"llm-gateway/internal/cache"
+	"llm-gateway/internal/provider"
 	"llm-gateway/internal/storage"
 )
 
@@ -52,15 +56,70 @@ type TokenUsageStats struct {
 	CostUSD      float64 `json:"cost_usd"`
 }
 
+// RequestLogEntry represents a single request log row.
+type RequestLogEntry struct {
+	Timestamp     int64   `json:"timestamp"`
+	TokenName     string  `json:"token_name"`
+	Model         string  `json:"model"`
+	Provider      string  `json:"provider"`
+	ProviderModel string  `json:"provider_model"`
+	InputTokens   int     `json:"input_tokens"`
+	OutputTokens  int     `json:"output_tokens"`
+	CostUSD       float64 `json:"cost_usd"`
+	LatencyMS     int64   `json:"latency_ms"`
+	Status        string  `json:"status"`
+	ErrorMessage  string  `json:"error_message"`
+	Streaming     bool    `json:"streaming"`
+	Cached        bool    `json:"cached"`
+}
+
+// LogsResult holds paginated logs.
+type LogsResult struct {
+	Logs       []RequestLogEntry `json:"logs"`
+	Page       int               `json:"page"`
+	TotalPages int               `json:"total_pages"`
+	Total      int               `json:"total"`
+}
+
+// LatencyResult holds latency percentiles.
+type LatencyResult struct {
+	P50 float64 `json:"p50"`
+	P95 float64 `json:"p95"`
+	P99 float64 `json:"p99"`
+	Avg float64 `json:"avg"`
+	Min float64 `json:"min"`
+	Max float64 `json:"max"`
+}
+
+// CostBreakdownEntry holds cost data grouped by day and dimension.
+type CostBreakdownEntry struct {
+	Day      string  `json:"day"`
+	GroupBy  string  `json:"group_by"`
+	CostUSD  float64 `json:"cost_usd"`
+	Requests int     `json:"requests"`
+}
+
 type StatsAPI struct {
-	db        *storage.DB
+	db            *storage.DB
-	authStore *auth.Store
+	authStore     *auth.Store
+	healthTracker *provider.HealthTracker
+	cache         *cache.Cache
 }
 
 func NewStatsAPI(db *storage.DB, authStore *auth.Store) *StatsAPI {
 	return &StatsAPI{db: db, authStore: authStore}
 }
 
+// SetHealthTracker sets the provider health tracker.
+func (s *StatsAPI) SetHealthTracker(ht *provider.HealthTracker) {
+	s.healthTracker = ht
+}
+
+// SetCache sets the cache for stats.
+func (s *StatsAPI) SetCache(c *cache.Cache) {
+	s.cache = c
+}
+
 // TokenNamesForUser returns the token names that belong to the user.
 // Admins get nil (no filter), non-admins get their token names.
 func (s *StatsAPI) TokenNamesForUser(user *auth.User) []string {
@@ -227,6 +286,217 @@ func (s *StatsAPI) GetTokenUsage(tokenNames []string) []TokenUsageStats {
 	return results
 }
 
+// GetLogs returns paginated request logs with filters.
+func (s *StatsAPI) GetLogs(tokenNames []string, page int, model, token, status string) *LogsResult {
+	if page < 1 {
+		page = 1
+	}
+	limit := 50
+	offset := (page - 1) * limit
+
+	tokenFilter, filterArgs := buildTokenFilter(tokenNames)
+
+	where := "WHERE 1=1" + tokenFilter
+	args := make([]any, 0)
+	args = append(args, filterArgs...)
+
+	if model != "" {
+		where += " AND model = ?"
+		args = append(args, model)
+	}
+	if token != "" {
+		where += " AND token_name = ?"
+		args = append(args, token)
+	}
+	if status != "" {
+		where += " AND status = ?"
+		args = append(args, status)
+	}
+
+	// Get total count
+	var total int
+	countArgs := make([]any, len(args))
+	copy(countArgs, args)
+	s.db.QueryRow("SELECT COUNT(*) FROM request_logs "+where, countArgs...).Scan(&total)
+
+	totalPages := (total + limit - 1) / limit
+	if totalPages < 1 {
+		totalPages = 1
+	}
+
+	// Get page
+	query := `SELECT timestamp, token_name, model, provider, provider_model,
+		input_tokens, output_tokens, cost_usd, latency_ms, status,
+		COALESCE(error_message, ''), streaming, cached
+		FROM request_logs ` + where + ` ORDER BY timestamp DESC LIMIT ? OFFSET ?`
+	args = append(args, limit, offset)
+
+	rows, err := s.db.Query(query, args...)
+	if err != nil {
+		return &LogsResult{Logs: []RequestLogEntry{}, Page: page, TotalPages: totalPages, Total: total}
+	}
+	defer rows.Close()
+
+	var logs []RequestLogEntry
+	for rows.Next() {
+		var l RequestLogEntry
+		var streaming, cached int
+		rows.Scan(&l.Timestamp, &l.TokenName, &l.Model, &l.Provider, &l.ProviderModel,
+			&l.InputTokens, &l.OutputTokens, &l.CostUSD, &l.LatencyMS, &l.Status,
+			&l.ErrorMessage, &streaming, &cached)
+		l.Streaming = streaming == 1
+		l.Cached = cached == 1
+		logs = append(logs, l)
+	}
+	if logs == nil {
+		logs = []RequestLogEntry{}
+	}
+
+	return &LogsResult{
+		Logs:       logs,
+		Page:       page,
+		TotalPages: totalPages,
+		Total:      total,
+	}
+}
+
+// GetDistinctModels returns distinct model names from logs.
+func (s *StatsAPI) GetDistinctModels() []string {
+	rows, err := s.db.Query("SELECT DISTINCT model FROM request_logs ORDER BY model")
+	if err != nil {
+		return nil
+	}
+	defer rows.Close()
+	var models []string
+	for rows.Next() {
+		var m string
+		rows.Scan(&m)
+		models = append(models, m)
+	}
+	return models
+}
+
+// GetDistinctTokens returns distinct token names from logs.
+func (s *StatsAPI) GetDistinctTokens() []string {
+	rows, err := s.db.Query("SELECT DISTINCT token_name FROM request_logs ORDER BY token_name")
+	if err != nil {
+		return nil
+	}
+	defer rows.Close()
+	var tokens []string
+	for rows.Next() {
+		var t string
+		rows.Scan(&t)
+		tokens = append(tokens, t)
+	}
+	return tokens
+}
+
+// GetLatency computes latency percentiles from request_logs.
+func (s *StatsAPI) GetLatency(tokenNames []string, period, model, providerName string) *LatencyResult {
+	var since int64
+	switch period {
+	case "7d":
+		since = time.Now().AddDate(0, 0, -7).Unix()
+	case "30d":
+		since = time.Now().AddDate(0, -1, 0).Unix()
+	default:
+		since = time.Now().Add(-24 * time.Hour).Unix()
+	}
+
+	tokenFilter, filterArgs := buildTokenFilter(tokenNames)
+
+	where := "WHERE timestamp >= ? AND status = 'success'" + tokenFilter
+	args := []any{since}
+	args = append(args, filterArgs...)
+
+	if model != "" {
+		where += " AND model = ?"
+		args = append(args, model)
+	}
+	if providerName != "" {
+		where += " AND provider = ?"
+		args = append(args, providerName)
+	}
+
+	rows, err := s.db.Query("SELECT latency_ms FROM request_logs "+where+" ORDER BY latency_ms", args...)
+	if err != nil {
+		return &LatencyResult{}
+	}
+	defer rows.Close()
+
+	var latencies []float64
+	for rows.Next() {
+		var l float64
+		rows.Scan(&l)
+		latencies = append(latencies, l)
+	}
+
+	if len(latencies) == 0 {
+		return &LatencyResult{}
+	}
+
+	sort.Float64s(latencies)
+	n := len(latencies)
+	var sum float64
+	for _, l := range latencies {
+		sum += l
+	}
+
+	return &LatencyResult{
+		P50: latencies[n*50/100],
+		P95: latencies[n*95/100],
+		P99: latencies[min(n*99/100, n-1)],
+		Avg: sum / float64(n),
+		Min: latencies[0],
+		Max: latencies[n-1],
+	}
+}
+
+// GetCostBreakdown returns cost data grouped by day and dimension.
+func (s *StatsAPI) GetCostBreakdown(tokenNames []string, period, groupBy string) []CostBreakdownEntry {
+	var since int64
+	switch period {
+	case "30d":
+		since = time.Now().AddDate(0, -1, 0).Unix()
+	case "7d":
+		since = time.Now().AddDate(0, 0, -7).Unix()
+	default:
+		since = time.Now().Add(-24 * time.Hour).Unix()
+	}
+
+	tokenFilter, filterArgs := buildTokenFilter(tokenNames)
+
+	groupCol := "model"
+	if groupBy == "token" {
+		groupCol = "token_name"
+	} else if groupBy == "provider" {
+		groupCol = "provider"
+	}
+
+	args := []any{since}
+	args = append(args, filterArgs...)
+
+	query := `SELECT date(timestamp, 'unixepoch') as day, ` + groupCol + `,
+		COALESCE(SUM(cost_usd), 0), COUNT(*)
+		FROM request_logs WHERE timestamp >= ?` + tokenFilter + `
+		GROUP BY day, ` + groupCol + ` ORDER BY day, ` + groupCol
+
+	rows, err := s.db.Query(query, args...)
+	if err != nil {
+		return nil
+	}
+	defer rows.Close()
+
+	var results []CostBreakdownEntry
+	for rows.Next() {
+		var e CostBreakdownEntry
+		rows.Scan(&e.Day, &e.GroupBy, &e.CostUSD, &e.Requests)
+		results = append(results, e)
+	}
+	return results
+}
+
 // JSON HTTP handlers (thin wrappers).
 
 func (s *StatsAPI) Summary(w http.ResponseWriter, r *http.Request) {
@@ -302,6 +572,58 @@ func (s *StatsAPI) Timeseries(w http.ResponseWriter, r *http.Request) {
 	writeJSON(w, results)
 }
 
+// Logs serves the paginated logs API.
+func (s *StatsAPI) Logs(w http.ResponseWriter, r *http.Request) {
+	tokenNames := s.tokenNamesForUser(r)
+	page, _ := strconv.Atoi(r.URL.Query().Get("page"))
+	model := r.URL.Query().Get("model")
+	token := r.URL.Query().Get("token")
+	status := r.URL.Query().Get("status")
+	result := s.GetLogs(tokenNames, page, model, token, status)
+	writeJSON(w, result)
+}
+
+// Latency serves latency percentiles API.
+func (s *StatsAPI) Latency(w http.ResponseWriter, r *http.Request) {
+	tokenNames := s.tokenNamesForUser(r)
+	period := r.URL.Query().Get("period")
+	model := r.URL.Query().Get("model")
+	providerName := r.URL.Query().Get("provider")
+	result := s.GetLatency(tokenNames, period, model, providerName)
+	writeJSON(w, result)
+}
+
+// CostBreakdown serves cost breakdown API.
+func (s *StatsAPI) CostBreakdown(w http.ResponseWriter, r *http.Request) {
+	tokenNames := s.tokenNamesForUser(r)
+	period := r.URL.Query().Get("period")
+	groupBy := r.URL.Query().Get("group_by")
+	if groupBy == "" {
+		groupBy = "model"
+	}
+	result := s.GetCostBreakdown(tokenNames, period, groupBy)
+	writeJSON(w, result)
+}
+
+// ProviderHealthHandler serves provider health status API.
+func (s *StatsAPI) ProviderHealthHandler(w http.ResponseWriter, r *http.Request) {
+	if s.healthTracker == nil {
+		writeJSON(w, []provider.ProviderHealth{})
+		return
+	}
+	writeJSON(w, s.healthTracker.Status())
+}
+
+// CacheStats serves cache statistics API.
+func (s *StatsAPI) CacheStats(w http.ResponseWriter, r *http.Request) {
+	if s.cache == nil {
+		writeJSON(w, map[string]any{"enabled": false})
+		return
+	}
+	stats := s.cache.Stats(r.Context())
+	writeJSON(w, stats)
+}
+
 func writeJSON(w http.ResponseWriter, v any) {
 	w.Header().Set("Content-Type", "application/json")
 	json.NewEncoder(w).Encode(v)

llm-gateway/internal/dashboard/handler.go

@@ -5,9 +5,12 @@ import (
 	"fmt"
 	"html/template"
 	"net/http"
+	"strconv"
 	"time"
 
 	"llm-gateway/internal/auth"
+	"llm-gateway/internal/cache"
+	"llm-gateway/internal/provider"
 )
 
 //go:embed templates/*.html templates/partials/*.html
@@ -20,9 +23,18 @@ var templateFuncs = template.FuncMap{
 		}
 		return time.Unix(ts, 0).Format("2006-01-02")
 	},
+	"formatTimeDetail": func(ts int64) string {
+		if ts == 0 {
+			return "never"
+		}
+		return time.Unix(ts, 0).Format("2006-01-02 15:04:05")
+	},
 	"addInt": func(a, b int) int {
 		return a + b
 	},
+	"subInt": func(a, b int) int {
+		return a - b
+	},
 	"formatCost": func(v float64) string {
 		if v == 0 {
 			return "$0.00"
@@ -32,19 +44,87 @@ var templateFuncs = template.FuncMap{
 		}
 		return fmt.Sprintf("$%.4f", v)
 	},
+	"formatPrice": func(v float64) string {
+		if v == 0 {
+			return "-"
+		}
+		return fmt.Sprintf("$%.2f", v)
+	},
+	"formatPct": func(v float64) string {
+		return fmt.Sprintf("%.1f%%", v*100)
+	},
+	"budgetPct": func(spend, budget float64) float64 {
+		if budget <= 0 {
+			return 0
+		}
+		return spend / budget * 100
+	},
+	"budgetColor": func(pct float64) string {
+		if pct >= 80 {
+			return "#f87171"
+		}
+		if pct >= 50 {
+			return "#fbbf24"
+		}
+		return "#4ade80"
+	},
+	"seq": func(start, end int) []int {
+		var s []int
+		for i := start; i <= end; i++ {
+			s = append(s, i)
+		}
+		return s
+	},
+	"paginationStart": func(page, totalPages int) int {
+		start := page - 2
+		if start < 1 {
+			start = 1
+		}
+		if totalPages-start < 4 && totalPages > 4 {
+			start = totalPages - 4
+		}
+		return start
+	},
+	"paginationEnd": func(page, totalPages int) int {
+		start := page - 2
+		if start < 1 {
+			start = 1
+		}
+		end := start + 4
+		if end > totalPages {
+			end = totalPages
+		}
+		return end
+	},
 }
 
 // PageData is the common data passed to all templates.
 type PageData struct {
 	ActivePage string
 	User       *auth.User
-	// Page-specific data
+	// Dashboard data
-	Summary    *SummaryResult
+	Summary        *SummaryResult
-	Models     []ModelStats
+	Models         []ModelStats
-	Providers  []ProviderStats
+	Providers      []ProviderStats
-	TokenStats []TokenUsageStats
+	TokenStats     []TokenUsageStats
+	ProviderHealth []provider.ProviderHealth
+	Latency        *LatencyResult
+	CacheEnabled   bool
+	CacheInfo      *cache.CacheStats
+	// Tokens page data
 	Tokens     []auth.APIToken
-	Users      []auth.User
+	TokenSpend map[string]float64
+	// Users page data
+	Users []auth.User
+	// Logs page data
+	LogsResult   *LogsResult
+	LogModels    []string
+	LogTokens    []string
+	FilterModel  string
+	FilterToken  string
+	FilterStatus string
+	// Models routing page data
+	ModelRoutes []provider.ModelRouteInfo
 }
 
 // Dashboard serves the HTMX-based dashboard pages.
@@ -52,6 +132,8 @@ type Dashboard struct {
 	templates *template.Template
 	authStore *auth.Store
 	statsAPI  *StatsAPI
+	registry  *provider.Registry
+	cache     *cache.Cache
 }
 
 // NewDashboard creates a new Dashboard handler.
@@ -70,6 +152,16 @@ func NewDashboard(authStore *auth.Store, statsAPI *StatsAPI) *Dashboard {
 	}
 }
 
+// SetRegistry sets the provider registry for model routing display.
+func (d *Dashboard) SetRegistry(r *provider.Registry) {
+	d.registry = r
+}
+
+// SetCache sets the cache reference for cache stats display.
+func (d *Dashboard) SetCache(c *cache.Cache) {
+	d.cache = c
+}
+
 // LoginPage serves the login page.
 func (d *Dashboard) LoginPage(w http.ResponseWriter, r *http.Request) {
 	if !d.authStore.HasAnyUser() {
@@ -106,11 +198,66 @@ func (d *Dashboard) DashboardPage(w http.ResponseWriter, r *http.Request) {
 		Models:     d.statsAPI.GetModels(tokenNames),
 		Providers:  d.statsAPI.GetProviders(tokenNames),
 		TokenStats: d.statsAPI.GetTokenUsage(tokenNames),
+		Latency:    d.statsAPI.GetLatency(tokenNames, "24h", "", ""),
+	}
+
+	// Provider health
+	if d.statsAPI.healthTracker != nil {
+		data.ProviderHealth = d.statsAPI.healthTracker.Status()
+	}
+
+	// Cache stats
+	if d.cache != nil {
+		data.CacheEnabled = true
+		data.CacheInfo = d.cache.Stats(r.Context())
 	}
 
 	d.renderDashboardPage(w, r, "partials/dashboard.html", data)
 }
 
+// LogsPage serves the request logs view.
+func (d *Dashboard) LogsPage(w http.ResponseWriter, r *http.Request) {
+	user := auth.UserFromContext(r.Context())
+	tokenNames := d.statsAPI.TokenNamesForUser(user)
+
+	page, _ := strconv.Atoi(r.URL.Query().Get("page"))
+	if page < 1 {
+		page = 1
+	}
+	model := r.URL.Query().Get("model")
+	token := r.URL.Query().Get("token")
+	status := r.URL.Query().Get("status")
+
+	data := PageData{
+		ActivePage:   "logs",
+		User:         user,
+		LogsResult:   d.statsAPI.GetLogs(tokenNames, page, model, token, status),
+		LogModels:    d.statsAPI.GetDistinctModels(),
+		LogTokens:    d.statsAPI.GetDistinctTokens(),
+		FilterModel:  model,
+		FilterToken:  token,
+		FilterStatus: status,
+	}
+
+	d.renderDashboardPage(w, r, "partials/logs.html", data)
+}
+
+// ModelsPage serves the model routing table view.
+func (d *Dashboard) ModelsPage(w http.ResponseWriter, r *http.Request) {
+	user := auth.UserFromContext(r.Context())
+
+	data := PageData{
+		ActivePage: "models",
+		User:       user,
+	}
+
+	if d.registry != nil {
+		data.ModelRoutes = d.registry.AllRoutes()
+	}
+
+	d.renderDashboardPage(w, r, "partials/models-page.html", data)
+}
+
 // TokensPage serves the tokens management view.
 func (d *Dashboard) TokensPage(w http.ResponseWriter, r *http.Request) {
 	user := auth.UserFromContext(r.Context())
@@ -125,10 +272,17 @@ func (d *Dashboard) TokensPage(w http.ResponseWriter, r *http.Request) {
 		tokens = []auth.APIToken{}
 	}
 
+	// Get today's spend for budget display
+	spend, _ := d.statsAPI.db.TodaySpendAll()
+	if spend == nil {
+		spend = make(map[string]float64)
+	}
+
 	d.renderDashboardPage(w, r, "partials/tokens.html", PageData{
 		ActivePage: "tokens",
 		User:       user,
 		Tokens:     tokens,
+		TokenSpend: spend,
 	})
 }
 

llm-gateway/internal/dashboard/templates/layout.html

@@ -5,93 +5,228 @@
 <meta charset="UTF-8">
 <meta name="viewport" content="width=device-width, initial-scale=1.0">
 <title>LLM Gateway</title>
+<script>
+// Prevent flash of wrong theme
+(function() {
+  var pref = localStorage.getItem('theme') || 'auto';
+  var effective = pref;
+  if (pref === 'auto') effective = window.matchMedia('(prefers-color-scheme: light)').matches ? 'light' : 'dark';
+  if (effective === 'light') document.documentElement.setAttribute('data-theme', 'light');
+})();
+</script>
 <script src="https://unpkg.com/htmx.org@2.0.4"></script>
 <script src="https://unpkg.com/htmx-ext-json-enc@2.0.3/json-enc.js"></script>
 <script src="https://unpkg.com/htmx-ext-sse@2.2.2/sse.js"></script>
 <script src="https://cdn.jsdelivr.net/npm/chart.js@4"></script>
 <style>
+  :root {
+    --bg-primary: #0f172a;
+    --bg-secondary: #1e293b;
+    --bg-tertiary: #334155;
+    --border-color: #334155;
+    --text-primary: #e2e8f0;
+    --text-secondary: #94a3b8;
+    --text-muted: #64748b;
+    --text-heading: #f8fafc;
+    --text-subheading: #cbd5e1;
+    --accent-blue: #3b82f6;
+    --accent-blue-hover: #2563eb;
+    --accent-blue-bg: #3b82f620;
+    --accent-green: #4ade80;
+    --accent-green-bg: #4ade8020;
+    --accent-red: #f87171;
+    --accent-red-bg: #7f1d1d40;
+    --accent-red-border: #991b1b;
+    --accent-red-text: #fca5a5;
+    --accent-yellow: #fbbf24;
+    --accent-yellow-bg: #92400e40;
+    --accent-purple: #a78bfa;
+    --accent-purple-bg: #a78bfa20;
+    --success-bg: #14532d40;
+    --success-border: #166534;
+    --success-text: #86efac;
+    --modal-overlay: #00000080;
+    --chart-grid: #1e293b;
+  }
+  [data-theme="light"] {
+    --bg-primary: #f8fafc;
+    --bg-secondary: #ffffff;
+    --bg-tertiary: #e2e8f0;
+    --border-color: #cbd5e1;
+    --text-primary: #1e293b;
+    --text-secondary: #475569;
+    --text-muted: #94a3b8;
+    --text-heading: #0f172a;
+    --text-subheading: #334155;
+    --accent-blue: #2563eb;
+    --accent-blue-hover: #1d4ed8;
+    --accent-blue-bg: #dbeafe;
+    --accent-green: #16a34a;
+    --accent-green-bg: #dcfce7;
+    --accent-red: #dc2626;
+    --accent-red-bg: #fef2f2;
+    --accent-red-border: #fca5a5;
+    --accent-red-text: #991b1b;
+    --accent-yellow: #d97706;
+    --accent-yellow-bg: #fef3c7;
+    --accent-purple: #7c3aed;
+    --accent-purple-bg: #ede9fe;
+    --success-bg: #f0fdf4;
+    --success-border: #86efac;
+    --success-text: #166534;
+    --modal-overlay: #00000040;
+    --chart-grid: #e2e8f0;
+  }
+
   * { margin: 0; padding: 0; box-sizing: border-box; }
-  body { font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; background: #0f172a; color: #e2e8f0; min-height: 100vh; display: flex; }
+  body { font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; background: var(--bg-primary); color: var(--text-primary); min-height: 100vh; display: flex; }
 
   /* Sidebar */
-  .sidebar { width: 220px; background: #1e293b; border-right: 1px solid #334155; min-height: 100vh; display: flex; flex-direction: column; position: fixed; top: 0; left: 0; }
+  .sidebar { width: 220px; background: var(--bg-secondary); border-right: 1px solid var(--border-color); min-height: 100vh; display: flex; flex-direction: column; position: fixed; top: 0; left: 0; }
-  .sidebar-brand { padding: 20px 16px; font-size: 1.1rem; font-weight: 700; color: #f8fafc; border-bottom: 1px solid #334155; }
+  .sidebar-brand { padding: 20px 16px; font-size: 1.1rem; font-weight: 700; color: var(--text-heading); border-bottom: 1px solid var(--border-color); }
   .sidebar-nav { flex: 1; padding: 12px 0; }
-  .sidebar-nav a { display: block; padding: 10px 20px; color: #94a3b8; text-decoration: none; font-size: 0.9rem; transition: all 0.15s; }
+  .sidebar-nav a { display: block; padding: 10px 20px; color: var(--text-secondary); text-decoration: none; font-size: 0.9rem; transition: all 0.15s; }
-  .sidebar-nav a:hover { background: #334155; color: #e2e8f0; }
+  .sidebar-nav a:hover { background: var(--bg-tertiary); color: var(--text-primary); }
-  .sidebar-nav a.active { background: #3b82f620; color: #3b82f6; border-right: 3px solid #3b82f6; }
+  .sidebar-nav a.active { background: var(--accent-blue-bg); color: var(--accent-blue); border-right: 3px solid var(--accent-blue); }
-  .sidebar-footer { padding: 16px; border-top: 1px solid #334155; }
+  .sidebar-footer { padding: 16px; border-top: 1px solid var(--border-color); }
-  .sidebar-footer .user-info { font-size: 0.85rem; color: #94a3b8; margin-bottom: 8px; }
+  .sidebar-footer .user-info { font-size: 0.85rem; color: var(--text-secondary); margin-bottom: 8px; }
-  .sidebar-footer a { display: block; padding: 6px 0; color: #94a3b8; text-decoration: none; font-size: 0.85rem; }
+  .sidebar-footer a { display: block; padding: 6px 0; color: var(--text-secondary); text-decoration: none; font-size: 0.85rem; }
-  .sidebar-footer a:hover { color: #f87171; }
+  .sidebar-footer a:hover { color: var(--accent-red); }
+  .theme-toggle { cursor: pointer; background: var(--bg-tertiary); border: 1px solid var(--border-color); color: var(--text-secondary); padding: 6px 12px; border-radius: 6px; font-size: 0.8rem; width: 100%; margin-bottom: 8px; }
+  .theme-toggle:hover { color: var(--text-primary); }
 
   /* Main content */
   .main { flex: 1; margin-left: 220px; padding: 24px; min-height: 100vh; max-width: calc(100vw - 220px); overflow-x: hidden; }
 
   /* Cards & tables */
   .cards { display: grid; grid-template-columns: repeat(auto-fit, minmax(160px, 1fr)); gap: 12px; margin-bottom: 24px; max-width: 100%; }
-  .card { background: #1e293b; border-radius: 8px; padding: 16px; }
+  .card { background: var(--bg-secondary); border-radius: 8px; padding: 16px; }
-  .card .label { font-size: 0.75rem; color: #94a3b8; text-transform: uppercase; letter-spacing: 0.05em; }
+  .card .label { font-size: 0.75rem; color: var(--text-secondary); text-transform: uppercase; letter-spacing: 0.05em; }
   .card .value { font-size: 1.5rem; font-weight: 700; margin-top: 4px; }
-  .card .sub { font-size: 0.75rem; color: #64748b; margin-top: 2px; }
+  .card .sub { font-size: 0.75rem; color: var(--text-muted); margin-top: 2px; }
-  .section { background: #1e293b; border-radius: 8px; padding: 16px; margin-bottom: 16px; overflow-x: auto; }
+  .section { background: var(--bg-secondary); border-radius: 8px; padding: 16px; margin-bottom: 16px; overflow-x: auto; }
-  .section h2 { font-size: 1.1rem; margin-bottom: 12px; color: #cbd5e1; }
+  .section h2 { font-size: 1.1rem; margin-bottom: 12px; color: var(--text-subheading); }
   .tabs { display: flex; gap: 8px; margin-bottom: 16px; }
-  .tabs button { background: #1e293b; border: 1px solid #334155; color: #94a3b8; padding: 6px 14px; border-radius: 6px; cursor: pointer; font-size: 0.8rem; }
+  .tabs button { background: var(--bg-secondary); border: 1px solid var(--border-color); color: var(--text-secondary); padding: 6px 14px; border-radius: 6px; cursor: pointer; font-size: 0.8rem; }
-  .tabs button.active { background: #3b82f6; border-color: #3b82f6; color: #fff; }
+  .tabs button.active { background: var(--accent-blue); border-color: var(--accent-blue); color: #fff; }
   table { width: 100%; border-collapse: collapse; font-size: 0.85rem; }
-  th { text-align: left; padding: 8px; color: #94a3b8; border-bottom: 1px solid #334155; font-weight: 500; }
+  th { text-align: left; padding: 8px; color: var(--text-secondary); border-bottom: 1px solid var(--border-color); font-weight: 500; }
-  td { padding: 8px; border-bottom: 1px solid #334155; }
+  td { padding: 8px; border-bottom: 1px solid var(--border-color); }
-  .green { color: #4ade80; }
+  .green { color: var(--accent-green); }
-  .red { color: #f87171; }
+  .red { color: var(--accent-red); }
-  .blue { color: #60a5fa; }
+  .blue { color: var(--accent-blue); }
+  .yellow { color: var(--accent-yellow); }
 
   /* Buttons */
   .btn { display: inline-block; padding: 10px 20px; border-radius: 6px; border: none; cursor: pointer; font-size: 0.9rem; font-weight: 500; text-decoration: none; }
-  .btn-primary { background: #3b82f6; color: #fff; }
+  .btn-primary { background: var(--accent-blue); color: #fff; }
-  .btn-primary:hover { background: #2563eb; }
+  .btn-primary:hover { background: var(--accent-blue-hover); }
   .btn-danger { background: #ef4444; color: #fff; }
   .btn-danger:hover { background: #dc2626; }
   .btn-sm { padding: 6px 12px; font-size: 0.8rem; }
-  .btn-outline { background: transparent; border: 1px solid #334155; color: #94a3b8; }
+  .btn-outline { background: transparent; border: 1px solid var(--border-color); color: var(--text-secondary); }
-  .btn-outline:hover { border-color: #64748b; color: #e2e8f0; }
+  .btn-outline:hover { border-color: var(--text-muted); color: var(--text-primary); }
 
   /* Forms */
   .form-group { margin-bottom: 16px; }
-  .form-group label { display: block; font-size: 0.85rem; color: #94a3b8; margin-bottom: 4px; }
+  .form-group label { display: block; font-size: 0.85rem; color: var(--text-secondary); margin-bottom: 4px; }
-  .form-group input, .form-group select { width: 100%; padding: 10px 12px; background: #0f172a; border: 1px solid #334155; border-radius: 6px; color: #e2e8f0; font-size: 0.95rem; }
+  .form-group input, .form-group select { width: 100%; padding: 10px 12px; background: var(--bg-primary); border: 1px solid var(--border-color); border-radius: 6px; color: var(--text-primary); font-size: 0.95rem; }
-  .form-group input:focus { outline: none; border-color: #3b82f6; }
+  .form-group input:focus, .form-group select:focus { outline: none; border-color: var(--accent-blue); }
-  .error-msg { background: #7f1d1d40; border: 1px solid #991b1b; color: #fca5a5; padding: 10px; border-radius: 6px; margin-bottom: 16px; font-size: 0.85rem; }
+  .error-msg { background: var(--accent-red-bg); border: 1px solid var(--accent-red-border); color: var(--accent-red-text); padding: 10px; border-radius: 6px; margin-bottom: 16px; font-size: 0.85rem; }
-  .success-msg { background: #14532d40; border: 1px solid #166534; color: #86efac; padding: 10px; border-radius: 6px; margin-bottom: 16px; font-size: 0.85rem; }
+  .success-msg { background: var(--success-bg); border: 1px solid var(--success-border); color: var(--success-text); padding: 10px; border-radius: 6px; margin-bottom: 16px; font-size: 0.85rem; }
 
   /* Modal */
-  .modal-overlay { position: fixed; top: 0; left: 0; right: 0; bottom: 0; background: #00000080; display: none; align-items: center; justify-content: center; z-index: 100; }
+  .modal-overlay { position: fixed; top: 0; left: 0; right: 0; bottom: 0; background: var(--modal-overlay); display: none; align-items: center; justify-content: center; z-index: 100; }
   .modal-overlay.show { display: flex; }
-  .modal { background: #1e293b; border-radius: 12px; padding: 24px; width: 100%; max-width: 440px; }
+  .modal { background: var(--bg-secondary); border-radius: 12px; padding: 24px; width: 100%; max-width: 440px; }
-  .modal h2 { margin-bottom: 16px; color: #cbd5e1; }
+  .modal h2 { margin-bottom: 16px; color: var(--text-subheading); }
   .modal-actions { display: flex; gap: 8px; justify-content: flex-end; margin-top: 16px; }
 
   /* Token display */
-  .token-key { background: #0f172a; padding: 8px 12px; border-radius: 6px; font-family: monospace; font-size: 0.85rem; word-break: break-all; margin: 8px 0; display: flex; align-items: center; gap: 8px; }
+  .token-key { background: var(--bg-primary); padding: 8px 12px; border-radius: 6px; font-family: monospace; font-size: 0.85rem; word-break: break-all; margin: 8px 0; display: flex; align-items: center; gap: 8px; }
   .token-key code { flex: 1; }
-  .copy-btn { background: #334155; border: none; color: #94a3b8; padding: 4px 8px; border-radius: 4px; cursor: pointer; font-size: 0.75rem; }
+  .copy-btn { background: var(--bg-tertiary); border: none; color: var(--text-secondary); padding: 4px 8px; border-radius: 4px; cursor: pointer; font-size: 0.75rem; }
-  .copy-btn:hover { color: #e2e8f0; }
+  .copy-btn:hover { color: var(--text-primary); }
 
   /* Badge */
   .badge { display: inline-block; padding: 2px 8px; border-radius: 12px; font-size: 0.7rem; font-weight: 600; }
-  .badge-admin { background: #3b82f620; color: #60a5fa; }
+  .badge-admin { background: var(--accent-blue-bg); color: var(--accent-blue); }
-  .badge-user { background: #4ade8020; color: #4ade80; }
+  .badge-user { background: var(--accent-green-bg); color: var(--accent-green); }
-  .badge-totp { background: #a78bfa20; color: #a78bfa; }
+  .badge-totp { background: var(--accent-purple-bg); color: var(--accent-purple); }
+  .badge-healthy { background: var(--accent-green-bg); color: var(--accent-green); }
+  .badge-degraded { background: var(--accent-yellow-bg); color: var(--accent-yellow); }
+  .badge-down { background: var(--accent-red-bg); color: var(--accent-red); }
+  .badge-success { background: var(--accent-green-bg); color: var(--accent-green); }
+  .badge-error { background: var(--accent-red-bg); color: var(--accent-red); }
+  .badge-cached { background: var(--accent-blue-bg); color: var(--accent-blue); }
+  .badge-priority { background: var(--bg-tertiary); color: var(--text-secondary); }
 
   .page-header { display: flex; align-items: center; gap: 12px; margin-bottom: 20px; }
-  .page-header h1 { font-size: 1.3rem; color: #f8fafc; }
+  .page-header h1 { font-size: 1.3rem; color: var(--text-heading); }
+
+  /* Filter bar */
+  .filter-bar { display: flex; gap: 12px; align-items: center; margin-bottom: 16px; flex-wrap: wrap; }
+  .filter-bar select { padding: 6px 10px; background: var(--bg-primary); border: 1px solid var(--border-color); border-radius: 6px; color: var(--text-primary); font-size: 0.85rem; }
+  .filter-bar select:focus { outline: none; border-color: var(--accent-blue); }
+
+  /* Pagination */
+  .pagination { display: flex; gap: 4px; align-items: center; margin-top: 16px; justify-content: center; }
+  .pagination button, .pagination span { padding: 6px 12px; border-radius: 6px; font-size: 0.8rem; border: 1px solid var(--border-color); background: var(--bg-secondary); color: var(--text-secondary); cursor: pointer; }
+  .pagination button:hover { background: var(--bg-tertiary); color: var(--text-primary); }
+  .pagination button.active { background: var(--accent-blue); border-color: var(--accent-blue); color: #fff; }
+  .pagination button:disabled { opacity: 0.5; cursor: not-allowed; }
+  .pagination .page-info { border: none; background: none; cursor: default; color: var(--text-muted); font-size: 0.8rem; }
+
+  /* Progress bar */
+  .progress-bar { width: 100%; height: 8px; background: var(--bg-tertiary); border-radius: 4px; overflow: hidden; }
+  .progress-bar-fill { height: 100%; border-radius: 4px; transition: width 0.3s; }
+  .budget-info { font-size: 0.75rem; color: var(--text-muted); margin-top: 2px; }
+
+  /* Expandable row */
+  .expandable { cursor: pointer; }
+  .expand-content { display: none; padding: 8px 12px; background: var(--bg-primary); border-radius: 6px; margin: 4px 0; font-size: 0.8rem; font-family: monospace; white-space: pre-wrap; word-break: break-all; }
+  .expand-content.show { display: block; }
+
+  /* Health status row */
+  .health-row { display: flex; gap: 12px; flex-wrap: wrap; margin-bottom: 16px; }
+  .health-item { display: flex; align-items: center; gap: 8px; padding: 8px 16px; background: var(--bg-secondary); border-radius: 8px; }
+  .health-item .provider-name { font-weight: 600; font-size: 0.85rem; }
 </style>
 </head>
 <body>
+<script>
+function getThemePref() { return localStorage.getItem('theme') || 'auto'; }
+function applyTheme(pref) {
+  var effective = pref;
+  if (pref === 'auto') effective = window.matchMedia('(prefers-color-scheme: light)').matches ? 'light' : 'dark';
+  if (effective === 'light') document.documentElement.setAttribute('data-theme', 'light');
+  else document.documentElement.removeAttribute('data-theme');
+}
+function themeLabel(pref) {
+  if (pref === 'light') return 'Light';
+  if (pref === 'dark') return 'Dark';
+  return 'Auto';
+}
+function toggleTheme() {
+  var order = ['dark', 'light', 'auto'];
+  var cur = getThemePref();
+  var next = order[(order.indexOf(cur) + 1) % order.length];
+  localStorage.setItem('theme', next);
+  applyTheme(next);
+  var btn = document.getElementById('theme-btn');
+  if (btn) btn.textContent = 'Theme: ' + themeLabel(next);
+}
+// Follow system changes when set to auto
+window.matchMedia('(prefers-color-scheme: light)').addEventListener('change', function() {
+  if (getThemePref() === 'auto') applyTheme('auto');
+});
+</script>
 <div class="sidebar">
   <div class="sidebar-brand">LLM Gateway</div>
   <nav class="sidebar-nav">
     <a href="/dashboard" hx-get="/dashboard" hx-target="#content" hx-push-url="true" {{if eq .ActivePage "dashboard"}}class="active"{{end}}>Dashboard</a>
+    <a href="/logs" hx-get="/logs" hx-target="#content" hx-push-url="true" {{if eq .ActivePage "logs"}}class="active"{{end}}>Logs</a>
+    <a href="/models" hx-get="/models" hx-target="#content" hx-push-url="true" {{if eq .ActivePage "models"}}class="active"{{end}}>Models</a>
     <a href="/tokens" hx-get="/tokens" hx-target="#content" hx-push-url="true" {{if eq .ActivePage "tokens"}}class="active"{{end}}>API Tokens</a>
     {{if .User.IsAdmin}}
     <a href="/users" hx-get="/users" hx-target="#content" hx-push-url="true" {{if eq .ActivePage "users"}}class="active"{{end}}>Users</a>
@@ -99,6 +234,7 @@
     <a href="/settings" hx-get="/settings" hx-target="#content" hx-push-url="true" {{if eq .ActivePage "settings"}}class="active"{{end}}>Settings</a>
   </nav>
   <div class="sidebar-footer">
+    <button class="theme-toggle" onclick="toggleTheme()" id="theme-btn">Switch Theme</button>
     <div class="user-info">{{.User.Username}}</div>
     <a href="#" hx-post="/api/auth/logout" hx-swap="none" onclick="setTimeout(()=>window.location='/login',100)">Logout</a>
   </div>
@@ -115,6 +251,11 @@ document.body.addEventListener('htmx:pushedIntoHistory', function(e) {
     a.classList.toggle('active', a.getAttribute('href') === window.location.pathname);
   });
 });
+// Set initial theme button label
+(function() {
+  var btn = document.getElementById('theme-btn');
+  if (btn) btn.textContent = 'Theme: ' + themeLabel(getThemePref());
+})();
 </script>
 </body>
 </html>

llm-gateway/internal/dashboard/templates/partials/dashboard.html

@@ -17,6 +17,38 @@
   {{end}}
 </div>
 
+{{if .ProviderHealth}}
+<div class="section">
+  <h2>Provider Health</h2>
+  <div class="health-row">
+    {{range .ProviderHealth}}
+    <div class="health-item">
+      <span class="provider-name">{{.Provider}}</span>
+      <span class="badge badge-{{.Status}}">{{.Status}}</span>
+      <span style="font-size:0.75rem;color:var(--text-muted)">{{printf "%.0f" .AvgLatency}}ms avg | {{formatPct .ErrorRate}} errors</span>
+    </div>
+    {{end}}
+  </div>
+</div>
+{{end}}
+
+{{if .Latency}}{{if gt .Latency.Max 0.0}}
+<div class="cards">
+  <div class="card"><div class="label">P50 Latency</div><div class="value">{{printf "%.0f" .Latency.P50}}ms</div></div>
+  <div class="card"><div class="label">P95 Latency</div><div class="value yellow">{{printf "%.0f" .Latency.P95}}ms</div></div>
+  <div class="card"><div class="label">P99 Latency</div><div class="value red">{{printf "%.0f" .Latency.P99}}ms</div></div>
+  <div class="card"><div class="label">Avg Latency</div><div class="value">{{printf "%.0f" .Latency.Avg}}ms</div></div>
+</div>
+{{end}}{{end}}
+
+{{if .CacheEnabled}}{{if .CacheInfo}}{{if .CacheInfo.Connected}}
+<div class="cards">
+  <div class="card"><div class="label">Cache Hit Rate</div><div class="value green">{{formatPct .CacheInfo.HitRate}}</div><div class="sub">{{.CacheInfo.Hits}} hits / {{.CacheInfo.Misses}} misses</div></div>
+  <div class="card"><div class="label">Cache Memory</div><div class="value">{{.CacheInfo.MemoryUsed}}</div></div>
+  <div class="card"><div class="label">Cached Keys</div><div class="value">{{.CacheInfo.Keys}}</div></div>
+</div>
+{{end}}{{end}}{{end}}
+
 <div class="tabs">
   <button class="active" onclick="loadTimeseries('24h', this)">24h</button>
   <button onclick="loadTimeseries('7d', this)">7d</button>
@@ -27,6 +59,16 @@
   <canvas id="chart" height="200"></canvas>
 </div>
 
+<div class="section">
+  <h2>Cost Breakdown</h2>
+  <div class="tabs" id="cost-tabs">
+    <button class="active" onclick="loadCostBreakdown('model', this)">By Model</button>
+    <button onclick="loadCostBreakdown('token', this)">By Token</button>
+    <button onclick="loadCostBreakdown('provider', this)">By Provider</button>
+  </div>
+  <canvas id="cost-chart" height="200"></canvas>
+</div>
+
 {{if .Models}}
 <div class="section">
   <h2>Models</h2>
@@ -88,7 +130,22 @@
 {{end}}
 
 <script>
-var _chart;
+var _chart, _costChart;
+function chartColors() {
+  var isLight = document.documentElement.hasAttribute('data-theme');
+  return {
+    text: isLight ? '#475569' : '#94a3b8',
+    grid: isLight ? '#e2e8f020' : '#334155',
+    green: '#4ade80',
+    legend: isLight ? '#1e293b' : '#e2e8f0'
+  };
+}
+function formatCostTick(v) {
+  if (v === 0) return '$0';
+  if (v < 0.001) return '$' + v.toFixed(6);
+  if (v < 0.01) return '$' + v.toFixed(4);
+  return '$' + v.toFixed(2);
+}
 function loadTimeseries(period, btn) {
   document.querySelectorAll('.tabs button').forEach(function(b) { b.classList.remove('active'); });
   if (btn) btn.classList.add('active');
@@ -96,6 +153,7 @@ function loadTimeseries(period, btn) {
   fetch('/api/stats/timeseries?period=' + period, {credentials: 'same-origin'})
     .then(function(r) { return r.json(); })
     .then(function(data) {
+      var c = chartColors();
       var labels = (data||[]).map(function(d) { return d.bucket; });
       var requests = (data||[]).map(function(d) { return d.requests; });
       var costs = (data||[]).map(function(d) { return d.cost_usd; });
@@ -105,24 +163,68 @@ function loadTimeseries(period, btn) {
         data: {
           labels: labels,
           datasets: [
-            { label: 'Requests', data: requests, backgroundColor: '#3b82f680', yAxisID: 'y' },
+            { label: 'Requests', data: requests, backgroundColor: 'rgba(59,130,246,0.5)', yAxisID: 'y' },
-            { label: 'Cost ($)', data: costs, type: 'line', borderColor: '#4ade80', backgroundColor: '#4ade8020', yAxisID: 'y1', tension: 0.3 }
+            { label: 'Cost ($)', data: costs, type: 'line', borderColor: c.green, backgroundColor: 'rgba(74,222,128,0.1)', yAxisID: 'y1', tension: 0.3, pointRadius: 3 }
           ]
         },
         options: {
           responsive: true,
           interaction: { mode: 'index', intersect: false },
           scales: {
-            y: { position: 'left', ticks: { color: '#94a3b8' }, grid: { color: '#1e293b' } },
+            y: { position: 'left', beginAtZero: true, ticks: { color: c.text, precision: 0 }, grid: { color: c.grid } },
-            y1: { position: 'right', ticks: { color: '#4ade80' }, grid: { display: false } },
+            y1: { position: 'right', beginAtZero: true, ticks: { color: c.green, callback: formatCostTick }, grid: { display: false } },
-            x: { ticks: { color: '#94a3b8', maxRotation: 45 }, grid: { color: '#1e293b' } }
+            x: { ticks: { color: c.text, maxRotation: 45 }, grid: { color: c.grid } }
           },
-          plugins: { legend: { labels: { color: '#e2e8f0' } } }
+          plugins: { legend: { labels: { color: c.legend } } }
         }
       });
     }).catch(function(){});
 }
 loadTimeseries('24h');
+
+function loadCostBreakdown(groupBy, btn) {
+  document.querySelectorAll('#cost-tabs button').forEach(function(b) { b.classList.remove('active'); });
+  if (btn) btn.classList.add('active');
+  fetch('/api/stats/cost-breakdown?period=7d&group_by=' + groupBy, {credentials: 'same-origin'})
+    .then(function(r) { return r.json(); })
+    .then(function(data) {
+      if (!data || data.length === 0) {
+        if (_costChart) _costChart.destroy();
+        return;
+      }
+      var c = chartColors();
+      var days = [], groups = {};
+      (data||[]).forEach(function(d) {
+        if (days.indexOf(d.day) === -1) days.push(d.day);
+        if (!groups[d.group_by]) groups[d.group_by] = {};
+        groups[d.group_by][d.day] = d.cost_usd;
+      });
+      var palette = ['#3b82f6','#4ade80','#f87171','#fbbf24','#a78bfa','#f472b6','#22d3ee','#fb923c'];
+      var datasets = [], ci = 0;
+      for (var g in groups) {
+        datasets.push({
+          label: g,
+          data: days.map(function(day) { return groups[g][day] || 0; }),
+          backgroundColor: palette[ci % palette.length] + '80'
+        });
+        ci++;
+      }
+      if (_costChart) _costChart.destroy();
+      _costChart = new Chart(document.getElementById('cost-chart'), {
+        type: 'bar',
+        data: { labels: days, datasets: datasets },
+        options: {
+          responsive: true,
+          scales: {
+            x: { stacked: true, ticks: { color: c.text }, grid: { color: c.grid } },
+            y: { stacked: true, beginAtZero: true, ticks: { color: c.text, callback: formatCostTick }, grid: { color: c.grid } }
+          },
+          plugins: { legend: { labels: { color: c.legend } } }
+        }
+      });
+    }).catch(function(){});
+}
+loadCostBreakdown('model');
 </script>
 </div>
 {{end}}

llm-gateway/internal/dashboard/templates/partials/logs.html

@@ -0,0 +1,120 @@
+{{define "content"}}
+<div class="page-header">
+  <h1>Request Logs</h1>
+  <span style="font-size:0.85rem;color:var(--text-muted)">{{.LogsResult.Total}} total</span>
+</div>
+
+<div class="filter-bar">
+  <select id="filter-model" onchange="applyLogsFilter()">
+    <option value="">All Models</option>
+    {{range .LogModels}}<option value="{{.}}" {{if eq . $.FilterModel}}selected{{end}}>{{.}}</option>{{end}}
+  </select>
+  <select id="filter-token" onchange="applyLogsFilter()">
+    <option value="">All Tokens</option>
+    {{range .LogTokens}}<option value="{{.}}" {{if eq . $.FilterToken}}selected{{end}}>{{.}}</option>{{end}}
+  </select>
+  <select id="filter-status" onchange="applyLogsFilter()">
+    <option value="">All Status</option>
+    <option value="success" {{if eq .FilterStatus "success"}}selected{{end}}>Success</option>
+    <option value="error" {{if eq .FilterStatus "error"}}selected{{end}}>Errors Only</option>
+    <option value="cached" {{if eq .FilterStatus "cached"}}selected{{end}}>Cached</option>
+  </select>
+  <button class="btn btn-sm btn-outline" onclick="clearLogsFilter()">Clear</button>
+</div>
+
+<div class="section">
+  <table>
+    <thead>
+      <tr>
+        <th>Time</th>
+        <th>Token</th>
+        <th>Model</th>
+        <th>Provider</th>
+        <th>Status</th>
+        <th>Latency</th>
+        <th>Tokens</th>
+        <th>Cost</th>
+      </tr>
+    </thead>
+    <tbody>
+    {{range $i, $log := .LogsResult.Logs}}
+    <tr class="{{if $log.ErrorMessage}}expandable{{end}}" {{if $log.ErrorMessage}}onclick="toggleExpand('expand-{{$i}}')"{{end}}>
+      <td>{{formatTimeDetail $log.Timestamp}}</td>
+      <td>{{$log.TokenName}}</td>
+      <td>{{$log.Model}}</td>
+      <td>{{$log.Provider}}</td>
+      <td>
+        {{if eq $log.Status "success"}}<span class="badge badge-success">success</span>
+        {{else if eq $log.Status "error"}}<span class="badge badge-error">error</span>
+        {{else if eq $log.Status "cached"}}<span class="badge badge-cached">cached</span>
+        {{else}}<span class="badge">{{$log.Status}}</span>{{end}}
+        {{if $log.Streaming}} <span class="badge badge-totp">stream</span>{{end}}
+      </td>
+      <td>{{$log.LatencyMS}}ms</td>
+      <td>{{$log.InputTokens}} / {{$log.OutputTokens}}</td>
+      <td class="green">{{formatCost $log.CostUSD}}</td>
+    </tr>
+    {{if $log.ErrorMessage}}
+    <tr>
+      <td colspan="8" style="padding:0;">
+        <div id="expand-{{$i}}" class="expand-content {{if eq $.FilterStatus "error"}}show{{end}}">{{$log.ErrorMessage}}</div>
+      </td>
+    </tr>
+    {{end}}
+    {{end}}
+    {{if not .LogsResult.Logs}}
+    <tr><td colspan="8" style="text-align:center;color:var(--text-muted);padding:24px;">No logs found</td></tr>
+    {{end}}
+    </tbody>
+  </table>
+
+  {{if gt .LogsResult.TotalPages 1}}
+  <div class="pagination">
+    <button {{if le .LogsResult.Page 1}}disabled{{end}} onclick="goToLogsPage(1)">First</button>
+    <button {{if le .LogsResult.Page 1}}disabled{{end}} onclick="goToLogsPage({{subInt .LogsResult.Page 1}})">Prev</button>
+    {{$page := .LogsResult.Page}}
+    {{$total := .LogsResult.TotalPages}}
+    {{range seq (paginationStart $page $total) (paginationEnd $page $total)}}
+    <button class="{{if eq . $page}}active{{end}}" onclick="goToLogsPage({{.}})">{{.}}</button>
+    {{end}}
+    <button {{if ge .LogsResult.Page .LogsResult.TotalPages}}disabled{{end}} onclick="goToLogsPage({{addInt .LogsResult.Page 1}})">Next</button>
+    <button {{if ge .LogsResult.Page .LogsResult.TotalPages}}disabled{{end}} onclick="goToLogsPage({{.LogsResult.TotalPages}})">Last</button>
+    <span class="page-info">Page {{.LogsResult.Page}} of {{.LogsResult.TotalPages}}</span>
+  </div>
+  {{end}}
+</div>
+
+<script>
+function buildLogsURL(page) {
+  var params = [];
+  var model = document.getElementById('filter-model').value;
+  var token = document.getElementById('filter-token').value;
+  var status = document.getElementById('filter-status').value;
+  if (model) params.push('model=' + encodeURIComponent(model));
+  if (token) params.push('token=' + encodeURIComponent(token));
+  if (status) params.push('status=' + encodeURIComponent(status));
+  if (page > 1) params.push('page=' + page);
+  return '/logs' + (params.length ? '?' + params.join('&') : '');
+}
+function applyLogsFilter() {
+  var url = buildLogsURL(1);
+  htmx.ajax('GET', url, {target: '#content', swap: 'innerHTML'});
+  history.pushState({}, '', url);
+}
+function goToLogsPage(page) {
+  var url = buildLogsURL(page);
+  htmx.ajax('GET', url, {target: '#content', swap: 'innerHTML'});
+  history.pushState({}, '', url);
+}
+function clearLogsFilter() {
+  document.getElementById('filter-model').value = '';
+  document.getElementById('filter-token').value = '';
+  document.getElementById('filter-status').value = '';
+  applyLogsFilter();
+}
+function toggleExpand(id) {
+  var el = document.getElementById(id);
+  if (el) el.classList.toggle('show');
+}
+</script>
+{{end}}

llm-gateway/internal/dashboard/templates/partials/models-page.html

@@ -0,0 +1,39 @@
+{{define "content"}}
+<div class="page-header">
+  <h1>Model Routing</h1>
+</div>
+
+{{if .ModelRoutes}}
+{{range .ModelRoutes}}
+<div class="section">
+  <h2>{{.Name}}</h2>
+  <table>
+    <thead>
+      <tr>
+        <th>Provider</th>
+        <th>Provider Model</th>
+        <th>Priority</th>
+        <th>Input Price (per 1M)</th>
+        <th>Output Price (per 1M)</th>
+      </tr>
+    </thead>
+    <tbody>
+    {{range .Routes}}
+    <tr>
+      <td>{{.ProviderName}}</td>
+      <td><code>{{.ProviderModel}}</code></td>
+      <td><span class="badge badge-priority">{{.Priority}}</span></td>
+      <td>{{formatPrice .InputPrice}}</td>
+      <td>{{formatPrice .OutputPrice}}</td>
+    </tr>
+    {{end}}
+    </tbody>
+  </table>
+</div>
+{{end}}
+{{else}}
+<div class="section" style="text-align:center;color:var(--text-muted);padding:24px;">
+  No models configured
+</div>
+{{end}}
+{{end}}

llm-gateway/internal/dashboard/templates/partials/tokens.html

@@ -10,9 +10,9 @@
 </div>
 
 <div class="section">
-  <h2>Static Tokens <span style="font-size:0.75rem;color:#64748b;font-weight:400;">(from config, managed via environment variables)</span></h2>
+  <h2>Static Tokens <span style="font-size:0.75rem;color:var(--text-muted);font-weight:400;">(from config, managed via environment variables)</span></h2>
   <table>
-    <thead><tr><th>Name</th><th>Prefix</th><th>Rate Limit</th><th>Budget</th><th></th></tr></thead>
+    <thead><tr><th>Name</th><th>Prefix</th><th>Rate Limit</th><th>Budget</th><th>Today's Spend</th><th></th></tr></thead>
     <tbody>
     {{range .Tokens}}{{if lt .ID 0}}
     <tr>
@@ -20,6 +20,18 @@
       <td><code>{{.KeyPrefix}}...</code></td>
       <td>{{if eq .RateLimitRPM 0}}unlimited{{else}}{{.RateLimitRPM}} rpm{{end}}</td>
       <td>{{if gt .DailyBudgetUSD 0.0}}${{printf "%.2f" .DailyBudgetUSD}}{{else}}unlimited{{end}}</td>
+      <td>
+        {{$spend := index $.TokenSpend .Name}}
+        {{if gt .DailyBudgetUSD 0.0}}
+          {{$pct := budgetPct $spend .DailyBudgetUSD}}
+          <div style="min-width:120px;">
+            <div class="progress-bar"><div class="progress-bar-fill" style="width:{{if gt $pct 100.0}}100{{else}}{{printf "%.0f" $pct}}{{end}}%;background:{{budgetColor $pct}};"></div></div>
+            <div class="budget-info">${{printf "%.4f" $spend}} / ${{printf "%.2f" .DailyBudgetUSD}} ({{printf "%.1f" $pct}}%)</div>
+          </div>
+        {{else}}
+          {{if gt $spend 0.0}}{{formatCost $spend}}{{else}}-{{end}}
+        {{end}}
+      </td>
       <td><span class="badge badge-totp">config</span></td>
     </tr>
     {{end}}{{end}}
@@ -28,17 +40,28 @@
 </div>
 
 <div class="section">
-  <h2>Dynamic Tokens <span style="font-size:0.75rem;color:#64748b;font-weight:400;">(created via dashboard)</span></h2>
+  <h2>Dynamic Tokens <span style="font-size:0.75rem;color:var(--text-muted);font-weight:400;">(created via dashboard)</span></h2>
   <table>
-    <thead><tr><th>Name</th><th>Prefix</th><th>Rate Limit</th><th>Budget</th><th>Created</th><th>Last Used</th><th></th></tr></thead>
+    <thead><tr><th>Name</th><th>Prefix</th><th>Rate Limit</th><th>Budget</th><th>Today's Spend</th><th>Created</th><th>Last Used</th><th></th></tr></thead>
     <tbody id="tokens-tbody">
-    {{$hasDynamic := false}}
     {{range .Tokens}}{{if gt .ID 0}}
     <tr>
       <td>{{.Name}}</td>
       <td><code>{{.KeyPrefix}}...</code></td>
       <td>{{if eq .RateLimitRPM 0}}unlimited{{else}}{{.RateLimitRPM}} rpm{{end}}</td>
       <td>{{if gt .DailyBudgetUSD 0.0}}${{printf "%.2f" .DailyBudgetUSD}}{{else}}unlimited{{end}}</td>
+      <td>
+        {{$spend := index $.TokenSpend .Name}}
+        {{if gt .DailyBudgetUSD 0.0}}
+          {{$pct := budgetPct $spend .DailyBudgetUSD}}
+          <div style="min-width:120px;">
+            <div class="progress-bar"><div class="progress-bar-fill" style="width:{{if gt $pct 100.0}}100{{else}}{{printf "%.0f" $pct}}{{end}}%;background:{{budgetColor $pct}};"></div></div>
+            <div class="budget-info">${{printf "%.4f" $spend}} / ${{printf "%.2f" .DailyBudgetUSD}} ({{printf "%.1f" $pct}}%)</div>
+          </div>
+        {{else}}
+          {{if gt $spend 0.0}}{{formatCost $spend}}{{else}}-{{end}}
+        {{end}}
+      </td>
       <td>{{formatTime .CreatedAt}}</td>
       <td>{{if gt .LastUsedAt 0}}{{formatTime .LastUsedAt}}{{else}}never{{end}}</td>
       <td><button class="btn btn-sm btn-danger" onclick="deleteToken({{.ID}})">Revoke</button></td>

llm-gateway/internal/provider/health.go

@@ -0,0 +1,121 @@
+package provider
+
+import (
+	"sync"
+	"time"
+)
+
+// HealthEvent represents a single request outcome for a provider.
+type HealthEvent struct {
+	Timestamp time.Time
+	LatencyMS int64
+	IsError   bool
+	ErrorMsg  string
+}
+
+// ProviderHealth is the computed health status for a provider.
+type ProviderHealth struct {
+	Provider   string  `json:"provider"`
+	Status     string  `json:"status"` // healthy, degraded, down
+	ErrorRate  float64 `json:"error_rate"`
+	AvgLatency float64 `json:"avg_latency_ms"`
+	Total      int     `json:"total"`
+	Errors     int     `json:"errors"`
+}
+
+// HealthTracker tracks per-provider health using a sliding window.
+type HealthTracker struct {
+	mu       sync.RWMutex
+	windows  map[string][]HealthEvent
+	windowDu time.Duration
+}
+
+// NewHealthTracker creates a health tracker with the given window duration.
+func NewHealthTracker(window time.Duration) *HealthTracker {
+	if window == 0 {
+		window = 5 * time.Minute
+	}
+	return &HealthTracker{
+		windows:  make(map[string][]HealthEvent),
+		windowDu: window,
+	}
+}
+
+// Record adds a health event for a provider.
+func (h *HealthTracker) Record(provider string, latencyMS int64, err error) {
+	event := HealthEvent{
+		Timestamp: time.Now(),
+		LatencyMS: latencyMS,
+		IsError:   err != nil,
+	}
+	if err != nil {
+		event.ErrorMsg = err.Error()
+	}
+
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
+	h.windows[provider] = append(h.windows[provider], event)
+	h.prune(provider)
+}
+
+// Status returns computed health for all tracked providers.
+func (h *HealthTracker) Status() []ProviderHealth {
+	h.mu.RLock()
+	defer h.mu.RUnlock()
+
+	cutoff := time.Now().Add(-h.windowDu)
+	var results []ProviderHealth
+
+	for provider, events := range h.windows {
+		var total, errors int
+		var totalLatency int64
+
+		for _, e := range events {
+			if e.Timestamp.Before(cutoff) {
+				continue
+			}
+			total++
+			totalLatency += e.LatencyMS
+			if e.IsError {
+				errors++
+			}
+		}
+
+		if total == 0 {
+			continue
+		}
+
+		errorRate := float64(errors) / float64(total)
+		status := "healthy"
+		if errorRate >= 0.5 {
+			status = "down"
+		} else if errorRate >= 0.1 {
+			status = "degraded"
+		}
+
+		results = append(results, ProviderHealth{
+			Provider:   provider,
+			Status:     status,
+			ErrorRate:  errorRate,
+			AvgLatency: float64(totalLatency) / float64(total),
+			Total:      total,
+			Errors:     errors,
+		})
+	}
+
+	return results
+}
+
+// prune removes events outside the window. Must be called with lock held.
+func (h *HealthTracker) prune(provider string) {
+	cutoff := time.Now().Add(-h.windowDu)
+	events := h.windows[provider]
+	i := 0
+	for i < len(events) && events[i].Timestamp.Before(cutoff) {
+		i++
+	}
+	if i > 0 {
+		h.windows[provider] = events[i:]
+	}
+}

llm-gateway/internal/provider/registry.go

@@ -70,3 +70,38 @@ func (r *Registry) Lookup(model string) ([]Route, bool) {
 func (r *Registry) ModelNames() []string {
 	return r.order
 }
+
+// RouteInfo exposes route details for dashboard display.
+type RouteInfo struct {
+	ProviderName  string  `json:"provider_name"`
+	ProviderModel string  `json:"provider_model"`
+	Priority      int     `json:"priority"`
+	InputPrice    float64 `json:"input_price"`
+	OutputPrice   float64 `json:"output_price"`
+}
+
+// ModelRouteInfo exposes a model and its routes for dashboard display.
+type ModelRouteInfo struct {
+	Name   string      `json:"name"`
+	Routes []RouteInfo `json:"routes"`
+}
+
+// AllRoutes returns all models and their routes in config order.
+func (r *Registry) AllRoutes() []ModelRouteInfo {
+	results := make([]ModelRouteInfo, 0, len(r.order))
+	for _, name := range r.order {
+		routes := r.routes[name]
+		info := ModelRouteInfo{Name: name}
+		for _, rt := range routes {
+			info.Routes = append(info.Routes, RouteInfo{
+				ProviderName:  rt.Provider.Name(),
+				ProviderModel: rt.ProviderModel,
+				Priority:      rt.Priority,
+				InputPrice:    rt.InputPrice,
+				OutputPrice:   rt.OutputPrice,
+			})
+		}
+		results = append(results, info)
+	}
+	return results
+}

llm-gateway/internal/proxy/handler.go

@@ -41,20 +41,22 @@ func getAPIToken(ctx context.Context) *auth.APIToken {
 }
 
 type Handler struct {
-	registry *provider.Registry
+	registry      *provider.Registry
-	logger   *storage.AsyncLogger
+	logger        *storage.AsyncLogger
-	cache    *cache.Cache
+	cache         *cache.Cache
-	metrics  *metrics.Metrics
+	metrics       *metrics.Metrics
-	cfg      *config.Config
+	cfg           *config.Config
+	healthTracker *provider.HealthTracker
 }
 
-func NewHandler(registry *provider.Registry, logger *storage.AsyncLogger, c *cache.Cache, m *metrics.Metrics, cfg *config.Config) *Handler {
+func NewHandler(registry *provider.Registry, logger *storage.AsyncLogger, c *cache.Cache, m *metrics.Metrics, cfg *config.Config, ht *provider.HealthTracker) *Handler {
 	return &Handler{
-		registry: registry,
+		registry:      registry,
-		logger:   logger,
+		logger:        logger,
-		cache:    c,
+		cache:         c,
-		metrics:  m,
+		metrics:       m,
-		cfg:      cfg,
+		cfg:           cfg,
+		healthTracker: ht,
 	}
 }
 
@@ -117,6 +119,9 @@ func (h *Handler) handleNonStream(w http.ResponseWriter, r *http.Request, req *p
 				// Client error — don't retry
 				h.metrics.RecordRequest(req.Model, route.Provider.Name(), tokenName, "error", latency, 0, 0, 0)
 				h.logRequest(tokenName, req.Model, route.Provider.Name(), route.ProviderModel, 0, 0, 0, latency, "error", err.Error(), false, false)
+				if h.healthTracker != nil {
+					h.healthTracker.Record(route.Provider.Name(), latency, err)
+				}
 				writeErrorRaw(w, pe.StatusCode, pe.Body)
 				return
 			}
@@ -124,9 +129,16 @@ func (h *Handler) handleNonStream(w http.ResponseWriter, r *http.Request, req *p
 			log.Printf("Provider %s failed for %s: %v", route.Provider.Name(), req.Model, err)
 			h.metrics.RecordRequest(req.Model, route.Provider.Name(), tokenName, "error", latency, 0, 0, 0)
 			h.logRequest(tokenName, req.Model, route.Provider.Name(), route.ProviderModel, 0, 0, 0, latency, "error", err.Error(), false, false)
+			if h.healthTracker != nil {
+				h.healthTracker.Record(route.Provider.Name(), latency, err)
+			}
 			continue
 		}
 
+		if h.healthTracker != nil {
+			h.healthTracker.Record(route.Provider.Name(), latency, nil)
+		}
+
 		// Compute cost
 		inputTokens, outputTokens := 0, 0
 		if resp.Usage != nil {

llm-gateway/internal/proxy/stream.go

@@ -31,12 +31,19 @@ func (h *Handler) handleStream(w http.ResponseWriter, r *http.Request, req *prov
 				latency := time.Since(start).Milliseconds()
 				h.metrics.RecordRequest(req.Model, route.Provider.Name(), tokenName, "error", latency, 0, 0, 0)
 				h.logRequest(tokenName, req.Model, route.Provider.Name(), route.ProviderModel, 0, 0, 0, latency, "error", err.Error(), true, false)
+				if h.healthTracker != nil {
+					h.healthTracker.Record(route.Provider.Name(), latency, err)
+				}
 				writeErrorRaw(w, pe.StatusCode, pe.Body)
 				return
 			}
 			lastErr = err
+			latency := time.Since(start).Milliseconds()
 			log.Printf("Provider %s stream failed for %s: %v", route.Provider.Name(), req.Model, err)
-			h.logRequest(tokenName, req.Model, route.Provider.Name(), route.ProviderModel, 0, 0, 0, time.Since(start).Milliseconds(), "error", err.Error(), true, false)
+			h.logRequest(tokenName, req.Model, route.Provider.Name(), route.ProviderModel, 0, 0, 0, latency, "error", err.Error(), true, false)
+			if h.healthTracker != nil {
+				h.healthTracker.Record(route.Provider.Name(), latency, err)
+			}
 			continue
 		}
 
@@ -84,6 +91,9 @@ func (h *Handler) handleStream(w http.ResponseWriter, r *http.Request, req *prov
 		cost := computeCost(inputTokens, outputTokens, route.InputPrice, route.OutputPrice)
 		h.metrics.RecordRequest(req.Model, route.Provider.Name(), tokenName, "success", latency, inputTokens, outputTokens, cost)
 		h.logRequest(tokenName, req.Model, route.Provider.Name(), route.ProviderModel, inputTokens, outputTokens, cost, latency, "success", "", true, false)
+		if h.healthTracker != nil {
+			h.healthTracker.Record(route.Provider.Name(), latency, nil)
+		}
 		return
 	}
 

llm-gateway/internal/storage/db.go

@@ -101,3 +101,27 @@ func (db *DB) TodaySpend(tokenName string) (float64, error) {
 	}
 	return total.Float64, nil
 }
+
+// TodaySpendAll returns today's spend for all tokens as a map.
+func (db *DB) TodaySpendAll() (map[string]float64, error) {
+	startOfDay := time.Now().Truncate(24 * time.Hour).Unix()
+	rows, err := db.Query(
+		"SELECT token_name, SUM(cost_usd) FROM request_logs WHERE timestamp >= ? GROUP BY token_name",
+		startOfDay,
+	)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	result := make(map[string]float64)
+	for rows.Next() {
+		var name string
+		var total float64
+		if err := rows.Scan(&name, &total); err != nil {
+			continue
+		}
+		result[name] = total
+	}
+	return result, nil
+}