loadbalance

type HealthMonitor struct {
	// contains filtered or unexported fields
}

type HealthMonitorConfig struct {
	// ConsecutiveErrorThreshold is the number of consecutive errors before marking unhealthy
	ConsecutiveErrorThreshold int `json:"consecutive_error_threshold" yaml:"consecutive_error_threshold"`
	// RecoveryTimeoutSeconds is the time in seconds before auto-recovery
	RecoveryTimeoutSeconds int `json:"recovery_timeout_seconds" yaml:"recovery_timeout_seconds"`
	// ProbeEnabled enables health check probing before marking service healthy
	ProbeEnabled bool `json:"probe_enabled" yaml:"probe_enabled"`
}

type HealthProbeFunc func(serviceID string) bool

type HealthStatus int

type Service struct {
	Provider      string       `yaml:"provider" json:"provider"`                                 // Provider name / uuid
	Model         string       `yaml:"model" json:"model"`                                       // Model name
	Weight        int          `yaml:"weight" json:"weight"`                                     // Weight for load balancing
	Active        bool         `yaml:"active" json:"active"`                                     // Whether this service is active
	TimeWindow    int          `yaml:"time_window" json:"time_window"`                           // Statistics time window in seconds
	ModelCapacity *int         `yaml:"model_capacity,omitempty" json:"model_capacity,omitempty"` // ModelCapacity overrides the provider's default_model_capacity for this specific model
	Stats         ServiceStats `yaml:"-" json:"-"`                                               // Service usage statistics (stored in SQLite, not in config)
}

type ServiceHealth struct {
	ServiceID         string        // Unique service identifier (provider:model)
	Status            HealthStatus  // Current health status
	LastError         error         // Last error that caused unhealthy state
	LastErrorTime     time.Time     // When the error occurred
	ConsecutiveErrors int           // Count of consecutive errors
	RateLimited       bool          // True if last error was 429
	AuthError         bool          // True if last error was 401/403
	LastHealthCheck   time.Time     // Last time health was checked
	RecoveryTimeout   time.Duration // Time before auto-recovery
	// contains filtered or unexported fields
}

type ServiceID struct {
	ProviderUUID string `json:"provider_uuid"`
	Model        string `json:"model"`
}

type ServiceStats struct {
	ServiceID            string    `json:"service_id"`             // Unique service identifier
	RequestCount         int64     `json:"request_count"`          // Total request count
	LastUsed             time.Time `json:"last_used"`              // Last usage timestamp
	WindowStart          time.Time `json:"window_start"`           // Current time window start
	WindowRequestCount   int64     `json:"window_request_count"`   // Requests in current window
	WindowTokensConsumed int64     `json:"window_tokens_consumed"` // Tokens consumed in current window (input + output)
	WindowInputTokens    int64     `json:"window_input_tokens"`    // Input tokens in current window
	WindowOutputTokens   int64     `json:"window_output_tokens"`   // Output tokens in current window
	TimeWindow           int       `json:"time_window"`            // Copy of service's time window

	// Latency tracking fields
	LatencySamples    []int64   `json:"-"`                   // Rolling window of latency samples (in ms)
	AvgLatencyMs      float64   `json:"avg_latency_ms"`      // Average latency in current window
	P50LatencyMs      float64   `json:"p50_latency_ms"`      // 50th percentile latency
	P95LatencyMs      float64   `json:"p95_latency_ms"`      // 95th percentile latency
	P99LatencyMs      float64   `json:"p99_latency_ms"`      // 99th percentile latency
	LastLatencyUpdate time.Time `json:"last_latency_update"` // When latency was last updated

	// Token speed tracking fields (tokens per second)
	SpeedSamples    []float64 `json:"-"`                 // Rolling window of token speed samples
	AvgTokenSpeed   float64   `json:"avg_token_speed"`   // Average tokens per second
	LastSpeedUpdate time.Time `json:"last_speed_update"` // When speed was last updated

	// TTFT (Time To First Token) tracking fields
	TTFTSamples    []int64   `json:"-"`                // Rolling window of TTFT samples (in ms)
	AvgTTFTMs      float64   `json:"avg_ttft_ms"`      // Average TTFT in milliseconds
	P50TTFTMs      float64   `json:"p50_ttft_ms"`      // 50th percentile TTFT
	P95TTFTMs      float64   `json:"p95_ttft_ms"`      // 95th percentile TTFT
	P99TTFTMs      float64   `json:"p99_ttft_ms"`      // 99th percentile TTFT
	LastTTFTUpdate time.Time `json:"last_ttft_update"` // When TTFT was last updated

	// Cache tracking fields
	WindowCacheHits   int64   `json:"window_cache_hits"`   // Cache hits in current window
	WindowCacheMisses int64   `json:"window_cache_misses"` // Cache misses in current window
	CacheHitRate      float64 `json:"cache_hit_rate"`      // Cache hit rate (hits / total)

	// Cost tracking fields (token-based)
	WindowCostTokens int64 `json:"window_cost_tokens"` // Total tokens as cost proxy in current window
	// contains filtered or unexported fields
}

type TacticType int