Implement SQLite database persistence for instance management

pkg/config/config.go

@@ -9,6 +9,7 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
+	"time"
 
 	"gopkg.in/yaml.v3"
 )
@@ -42,6 +43,7 @@ type AppConfig struct {
 	Server     ServerConfig          `yaml:"server" json:"server"`
 	Backends   BackendConfig         `yaml:"backends" json:"backends"`
 	Instances  InstancesConfig       `yaml:"instances" json:"instances"`
+	Database   DatabaseConfig        `yaml:"database" json:"database"`
 	Auth       AuthConfig            `yaml:"auth" json:"auth"`
 	LocalNode  string                `yaml:"local_node,omitempty" json:"local_node,omitempty"`
 	Nodes      map[string]NodeConfig `yaml:"nodes,omitempty" json:"nodes,omitempty"`
@@ -71,6 +73,17 @@ type ServerConfig struct {
 	ResponseHeaders map[string]string `yaml:"response_headers,omitempty" json:"response_headers,omitempty"`
 }
 
+// DatabaseConfig contains database configuration settings
+type DatabaseConfig struct {
+	// Database file path (relative to data_dir or absolute)
+	Path string `yaml:"path" json:"path"`
+
+	// Connection settings
+	MaxOpenConnections int           `yaml:"max_open_connections" json:"max_open_connections"`
+	MaxIdleConnections int           `yaml:"max_idle_connections" json:"max_idle_connections"`
+	ConnMaxLifetime    time.Duration `yaml:"connection_max_lifetime" json:"connection_max_lifetime"`
+}
+
 // InstancesConfig contains instance management configuration
 type InstancesConfig struct {
 	// Port range for instances (e.g., 8000,9000)
@@ -204,6 +217,12 @@ func LoadConfig(configPath string) (AppConfig, error) {
 			OnDemandStartTimeout: 120, // 2 minutes
 			TimeoutCheckInterval: 5,   // Check timeouts every 5 minutes
 		},
+		Database: DatabaseConfig{
+			Path:               "llamactl.db", // Relative to data_dir
+			MaxOpenConnections: 25,
+			MaxIdleConnections: 5,
+			ConnMaxLifetime:    5 * time.Minute,
+		},
 		Auth: AuthConfig{
 			RequireInferenceAuth:  true,
 			InferenceKeys:         []string{},
@@ -233,6 +252,11 @@ func LoadConfig(configPath string) (AppConfig, error) {
 		cfg.Instances.LogsDir = filepath.Join(cfg.Instances.DataDir, "logs")
 	}
 
+	// Resolve database path relative to DataDir if it's not absolute
+	if cfg.Database.Path != "" && !filepath.IsAbs(cfg.Database.Path) {
+		cfg.Database.Path = filepath.Join(cfg.Instances.DataDir, cfg.Database.Path)
+	}
+
 	// Validate port range
 	if cfg.Instances.PortRange[0] <= 0 || cfg.Instances.PortRange[1] <= 0 || cfg.Instances.PortRange[0] >= cfg.Instances.PortRange[1] {
 		return AppConfig{}, fmt.Errorf("invalid port range: %v", cfg.Instances.PortRange)
@@ -495,6 +519,26 @@ func loadEnvVars(cfg *AppConfig) {
 	if localNode := os.Getenv("LLAMACTL_LOCAL_NODE"); localNode != "" {
 		cfg.LocalNode = localNode
 	}
+
+	// Database config
+	if dbPath := os.Getenv("LLAMACTL_DATABASE_PATH"); dbPath != "" {
+		cfg.Database.Path = dbPath
+	}
+	if maxOpenConns := os.Getenv("LLAMACTL_DATABASE_MAX_OPEN_CONNECTIONS"); maxOpenConns != "" {
+		if m, err := strconv.Atoi(maxOpenConns); err == nil {
+			cfg.Database.MaxOpenConnections = m
+		}
+	}
+	if maxIdleConns := os.Getenv("LLAMACTL_DATABASE_MAX_IDLE_CONNECTIONS"); maxIdleConns != "" {
+		if m, err := strconv.Atoi(maxIdleConns); err == nil {
+			cfg.Database.MaxIdleConnections = m
+		}
+	}
+	if connMaxLifetime := os.Getenv("LLAMACTL_DATABASE_CONN_MAX_LIFETIME"); connMaxLifetime != "" {
+		if d, err := time.ParseDuration(connMaxLifetime); err == nil {
+			cfg.Database.ConnMaxLifetime = d
+		}
+	}
 }
 
 // ParsePortRange parses port range from string formats like "8000-9000" or "8000,9000"

pkg/database/database.go

@@ -0,0 +1,107 @@
+package database
+
+import (
+	"database/sql"
+	"fmt"
+	"llamactl/pkg/instance"
+	"log"
+	"path/filepath"
+	"time"
+
+	_ "github.com/mattn/go-sqlite3"
+)
+
+// Database defines the interface for instance persistence operations
+type Database interface {
+	Save(inst *instance.Instance) error
+	Delete(name string) error
+	LoadAll() ([]*instance.Instance, error)
+	Close() error
+}
+
+// Config contains database configuration settings
+type Config struct {
+	// Database file path (relative to data_dir or absolute)
+	Path string
+
+	// Connection settings
+	MaxOpenConnections int
+	MaxIdleConnections int
+	ConnMaxLifetime    time.Duration
+}
+
+// DB wraps the database connection with configuration
+type DB struct {
+	*sql.DB
+	config *Config
+}
+
+// Open creates a new database connection with the provided configuration
+func Open(config *Config) (*DB, error) {
+	if config == nil {
+		return nil, fmt.Errorf("database config cannot be nil")
+	}
+
+	if config.Path == "" {
+		return nil, fmt.Errorf("database path cannot be empty")
+	}
+
+	// Ensure the database directory exists
+	dbDir := filepath.Dir(config.Path)
+	if dbDir != "." && dbDir != "/" {
+		// Directory will be created by the manager if auto_create_dirs is enabled
+		log.Printf("Database will be created at: %s", config.Path)
+	}
+
+	// Open SQLite database with proper options
+	// - _journal_mode=WAL: Write-Ahead Logging for better concurrency
+	// - _busy_timeout=5000: Wait up to 5 seconds if database is locked
+	// - _foreign_keys=1: Enable foreign key constraints (for future use)
+	dsn := fmt.Sprintf("file:%s?_journal_mode=WAL&_busy_timeout=5000&_foreign_keys=1", config.Path)
+
+	sqlDB, err := sql.Open("sqlite3", dsn)
+	if err != nil {
+		return nil, fmt.Errorf("failed to open database: %w", err)
+	}
+
+	// Configure connection pool
+	if config.MaxOpenConnections > 0 {
+		sqlDB.SetMaxOpenConns(config.MaxOpenConnections)
+	}
+	if config.MaxIdleConnections > 0 {
+		sqlDB.SetMaxIdleConns(config.MaxIdleConnections)
+	}
+	if config.ConnMaxLifetime > 0 {
+		sqlDB.SetConnMaxLifetime(config.ConnMaxLifetime)
+	}
+
+	// Verify database connection
+	if err := sqlDB.Ping(); err != nil {
+		sqlDB.Close()
+		return nil, fmt.Errorf("failed to ping database: %w", err)
+	}
+
+	log.Printf("Database connection established: %s", config.Path)
+
+	return &DB{
+		DB:     sqlDB,
+		config: config,
+	}, nil
+}
+
+// Close closes the database connection
+func (db *DB) Close() error {
+	if db.DB != nil {
+		log.Println("Closing database connection")
+		return db.DB.Close()
+	}
+	return nil
+}
+
+// HealthCheck verifies the database is accessible
+func (db *DB) HealthCheck() error {
+	if db.DB == nil {
+		return fmt.Errorf("database connection is nil")
+	}
+	return db.DB.Ping()
+}

pkg/database/instances.go

@@ -0,0 +1,539 @@
+package database
+
+import (
+	"context"
+	"database/sql"
+	"encoding/json"
+	"fmt"
+	"llamactl/pkg/backends"
+	"llamactl/pkg/instance"
+	"log"
+	"time"
+)
+
+// instanceRow represents a row in the instances table
+type instanceRow struct {
+	ID                int
+	Name              string
+	BackendType       string
+	BackendConfigJSON string
+	Status            string
+	CreatedAt         int64
+	UpdatedAt         int64
+	AutoRestart       int
+	MaxRestarts       int
+	RestartDelay      int
+	OnDemandStart     int
+	IdleTimeout       int
+	DockerEnabled     int
+	CommandOverride   sql.NullString
+	Nodes             sql.NullString
+	Environment       sql.NullString
+	OwnerUserID       sql.NullString
+}
+
+// Create inserts a new instance into the database
+func (db *DB) Create(ctx context.Context, inst *instance.Instance) error {
+	if inst == nil {
+		return fmt.Errorf("instance cannot be nil")
+	}
+
+	opts := inst.GetOptions()
+	if opts == nil {
+		return fmt.Errorf("instance options cannot be nil")
+	}
+
+	// Convert instance to database row
+	row, err := db.instanceToRow(inst)
+	if err != nil {
+		return fmt.Errorf("failed to convert instance to row: %w", err)
+	}
+
+	// Insert into database
+	query := `
+		INSERT INTO instances (
+			name, backend_type, backend_config_json, status,
+			created_at, updated_at,
+			auto_restart, max_restarts, restart_delay,
+			on_demand_start, idle_timeout, docker_enabled,
+			command_override, nodes, environment, owner_user_id
+		) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+	`
+
+	_, err = db.DB.ExecContext(ctx, query,
+		row.Name, row.BackendType, row.BackendConfigJSON, row.Status,
+		row.CreatedAt, row.UpdatedAt,
+		row.AutoRestart, row.MaxRestarts, row.RestartDelay,
+		row.OnDemandStart, row.IdleTimeout, row.DockerEnabled,
+		row.CommandOverride, row.Nodes, row.Environment, row.OwnerUserID,
+	)
+
+	if err != nil {
+		return fmt.Errorf("failed to insert instance: %w", err)
+	}
+
+	return nil
+}
+
+// GetByName retrieves an instance by name
+func (db *DB) GetByName(ctx context.Context, name string) (*instance.Instance, error) {
+	query := `
+		SELECT id, name, backend_type, backend_config_json, status,
+		       created_at, updated_at,
+		       auto_restart, max_restarts, restart_delay,
+		       on_demand_start, idle_timeout, docker_enabled,
+		       command_override, nodes, environment, owner_user_id
+		FROM instances
+		WHERE name = ?
+	`
+
+	var row instanceRow
+	err := db.DB.QueryRowContext(ctx, query, name).Scan(
+		&row.ID, &row.Name, &row.BackendType, &row.BackendConfigJSON, &row.Status,
+		&row.CreatedAt, &row.UpdatedAt,
+		&row.AutoRestart, &row.MaxRestarts, &row.RestartDelay,
+		&row.OnDemandStart, &row.IdleTimeout, &row.DockerEnabled,
+		&row.CommandOverride, &row.Nodes, &row.Environment, &row.OwnerUserID,
+	)
+
+	if err == sql.ErrNoRows {
+		return nil, fmt.Errorf("instance not found: %s", name)
+	}
+	if err != nil {
+		return nil, fmt.Errorf("failed to query instance: %w", err)
+	}
+
+	return db.rowToInstance(&row)
+}
+
+// GetAll retrieves all instances from the database
+func (db *DB) GetAll(ctx context.Context) ([]*instance.Instance, error) {
+	query := `
+		SELECT id, name, backend_type, backend_config_json, status,
+		       created_at, updated_at,
+		       auto_restart, max_restarts, restart_delay,
+		       on_demand_start, idle_timeout, docker_enabled,
+		       command_override, nodes, environment, owner_user_id
+		FROM instances
+		ORDER BY created_at ASC
+	`
+
+	rows, err := db.DB.QueryContext(ctx, query)
+	if err != nil {
+		return nil, fmt.Errorf("failed to query instances: %w", err)
+	}
+	defer rows.Close()
+
+	var instances []*instance.Instance
+	for rows.Next() {
+		var row instanceRow
+		err := rows.Scan(
+			&row.ID, &row.Name, &row.BackendType, &row.BackendConfigJSON, &row.Status,
+			&row.CreatedAt, &row.UpdatedAt,
+			&row.AutoRestart, &row.MaxRestarts, &row.RestartDelay,
+			&row.OnDemandStart, &row.IdleTimeout, &row.DockerEnabled,
+			&row.CommandOverride, &row.Nodes, &row.Environment, &row.OwnerUserID,
+		)
+		if err != nil {
+			log.Printf("Failed to scan instance row: %v", err)
+			continue
+		}
+
+		inst, err := db.rowToInstance(&row)
+		if err != nil {
+			log.Printf("Failed to convert row to instance: %v", err)
+			continue
+		}
+
+		instances = append(instances, inst)
+	}
+
+	if err := rows.Err(); err != nil {
+		return nil, fmt.Errorf("error iterating rows: %w", err)
+	}
+
+	return instances, nil
+}
+
+// Update updates an existing instance
+func (db *DB) Update(ctx context.Context, inst *instance.Instance) error {
+	if inst == nil {
+		return fmt.Errorf("instance cannot be nil")
+	}
+
+	opts := inst.GetOptions()
+	if opts == nil {
+		return fmt.Errorf("instance options cannot be nil")
+	}
+
+	// Convert instance to database row
+	row, err := db.instanceToRow(inst)
+	if err != nil {
+		return fmt.Errorf("failed to convert instance to row: %w", err)
+	}
+
+	// Update in database
+	query := `
+		UPDATE instances SET
+			backend_type = ?, backend_config_json = ?, status = ?,
+			updated_at = ?,
+			auto_restart = ?, max_restarts = ?, restart_delay = ?,
+			on_demand_start = ?, idle_timeout = ?, docker_enabled = ?,
+			command_override = ?, nodes = ?, environment = ?
+		WHERE name = ?
+	`
+
+	result, err := db.DB.ExecContext(ctx, query,
+		row.BackendType, row.BackendConfigJSON, row.Status,
+		row.UpdatedAt,
+		row.AutoRestart, row.MaxRestarts, row.RestartDelay,
+		row.OnDemandStart, row.IdleTimeout, row.DockerEnabled,
+		row.CommandOverride, row.Nodes, row.Environment,
+		row.Name,
+	)
+
+	if err != nil {
+		return fmt.Errorf("failed to update instance: %w", err)
+	}
+
+	rowsAffected, err := result.RowsAffected()
+	if err != nil {
+		return fmt.Errorf("failed to get rows affected: %w", err)
+	}
+
+	if rowsAffected == 0 {
+		return fmt.Errorf("instance not found: %s", inst.Name)
+	}
+
+	return nil
+}
+
+// UpdateStatus updates only the status of an instance (optimized operation)
+func (db *DB) UpdateStatus(ctx context.Context, name string, status instance.Status) error {
+	// Convert status to string
+	statusJSON, err := status.MarshalJSON()
+	if err != nil {
+		return fmt.Errorf("failed to marshal status: %w", err)
+	}
+	var statusStr string
+	if err := json.Unmarshal(statusJSON, &statusStr); err != nil {
+		return fmt.Errorf("failed to unmarshal status string: %w", err)
+	}
+
+	query := `
+		UPDATE instances SET
+			status = ?,
+			updated_at = ?
+		WHERE name = ?
+	`
+
+	result, err := db.DB.ExecContext(ctx, query, statusStr, time.Now().Unix(), name)
+	if err != nil {
+		return fmt.Errorf("failed to update instance status: %w", err)
+	}
+
+	rowsAffected, err := result.RowsAffected()
+	if err != nil {
+		return fmt.Errorf("failed to get rows affected: %w", err)
+	}
+
+	if rowsAffected == 0 {
+		return fmt.Errorf("instance not found: %s", name)
+	}
+
+	return nil
+}
+
+// DeleteInstance removes an instance from the database
+func (db *DB) DeleteInstance(ctx context.Context, name string) error {
+	query := `DELETE FROM instances WHERE name = ?`
+
+	result, err := db.DB.ExecContext(ctx, query, name)
+	if err != nil {
+		return fmt.Errorf("failed to delete instance: %w", err)
+	}
+
+	rowsAffected, err := result.RowsAffected()
+	if err != nil {
+		return fmt.Errorf("failed to get rows affected: %w", err)
+	}
+
+	if rowsAffected == 0 {
+		return fmt.Errorf("instance not found: %s", name)
+	}
+
+	return nil
+}
+
+// instanceToRow converts an Instance to a database row
+func (db *DB) instanceToRow(inst *instance.Instance) (*instanceRow, error) {
+	opts := inst.GetOptions()
+	if opts == nil {
+		return nil, fmt.Errorf("instance options cannot be nil")
+	}
+
+	// Marshal backend options to JSON (this uses the MarshalJSON method which handles typed backends)
+	backendJSON, err := json.Marshal(&opts.BackendOptions)
+	if err != nil {
+		return nil, fmt.Errorf("failed to marshal backend options: %w", err)
+	}
+
+	// Extract just the backend_options field from the marshaled JSON
+	var backendWrapper struct {
+		BackendOptions map[string]any `json:"backend_options"`
+	}
+	if err := json.Unmarshal(backendJSON, &backendWrapper); err != nil {
+		return nil, fmt.Errorf("failed to unmarshal backend wrapper: %w", err)
+	}
+
+	backendConfigJSON, err := json.Marshal(backendWrapper.BackendOptions)
+	if err != nil {
+		return nil, fmt.Errorf("failed to marshal backend config: %w", err)
+	}
+
+	// Convert nodes map to JSON array
+	var nodesJSON sql.NullString
+	if len(opts.Nodes) > 0 {
+		nodesList := make([]string, 0, len(opts.Nodes))
+		for node := range opts.Nodes {
+			nodesList = append(nodesList, node)
+		}
+		nodesBytes, err := json.Marshal(nodesList)
+		if err != nil {
+			return nil, fmt.Errorf("failed to marshal nodes: %w", err)
+		}
+		nodesJSON = sql.NullString{String: string(nodesBytes), Valid: true}
+	}
+
+	// Convert environment map to JSON
+	var envJSON sql.NullString
+	if len(opts.Environment) > 0 {
+		envBytes, err := json.Marshal(opts.Environment)
+		if err != nil {
+			return nil, fmt.Errorf("failed to marshal environment: %w", err)
+		}
+		envJSON = sql.NullString{String: string(envBytes), Valid: true}
+	}
+
+	// Convert command override
+	var cmdOverride sql.NullString
+	if opts.CommandOverride != "" {
+		cmdOverride = sql.NullString{String: opts.CommandOverride, Valid: true}
+	}
+
+	// Convert boolean pointers to integers (0 or 1)
+	autoRestart := 0
+	if opts.AutoRestart != nil && *opts.AutoRestart {
+		autoRestart = 1
+	}
+
+	maxRestarts := -1
+	if opts.MaxRestarts != nil {
+		maxRestarts = *opts.MaxRestarts
+	}
+
+	restartDelay := 0
+	if opts.RestartDelay != nil {
+		restartDelay = *opts.RestartDelay
+	}
+
+	onDemandStart := 0
+	if opts.OnDemandStart != nil && *opts.OnDemandStart {
+		onDemandStart = 1
+	}
+
+	idleTimeout := 0
+	if opts.IdleTimeout != nil {
+		idleTimeout = *opts.IdleTimeout
+	}
+
+	dockerEnabled := 0
+	if opts.DockerEnabled != nil && *opts.DockerEnabled {
+		dockerEnabled = 1
+	}
+
+	now := time.Now().Unix()
+
+	// Convert status to string
+	statusJSON, err := inst.GetStatus().MarshalJSON()
+	if err != nil {
+		return nil, fmt.Errorf("failed to marshal status: %w", err)
+	}
+	var statusStr string
+	if err := json.Unmarshal(statusJSON, &statusStr); err != nil {
+		return nil, fmt.Errorf("failed to unmarshal status string: %w", err)
+	}
+
+	return &instanceRow{
+		Name:              inst.Name,
+		BackendType:       string(opts.BackendOptions.BackendType),
+		BackendConfigJSON: string(backendConfigJSON),
+		Status:            statusStr,
+		CreatedAt:         inst.Created,
+		UpdatedAt:         now,
+		AutoRestart:       autoRestart,
+		MaxRestarts:       maxRestarts,
+		RestartDelay:      restartDelay,
+		OnDemandStart:     onDemandStart,
+		IdleTimeout:       idleTimeout,
+		DockerEnabled:     dockerEnabled,
+		CommandOverride:   cmdOverride,
+		Nodes:             nodesJSON,
+		Environment:       envJSON,
+	}, nil
+}
+
+// rowToInstance converts a database row to an Instance
+func (db *DB) rowToInstance(row *instanceRow) (*instance.Instance, error) {
+	// Unmarshal backend config
+	var backendConfig map[string]any
+	if err := json.Unmarshal([]byte(row.BackendConfigJSON), &backendConfig); err != nil {
+		return nil, fmt.Errorf("failed to unmarshal backend config: %w", err)
+	}
+
+	// Create backends.Options by marshaling and unmarshaling to trigger the UnmarshalJSON logic
+	// This ensures the typed backend fields (LlamaServerOptions, VllmServerOptions, etc.) are populated
+	var backendOptions backends.Options
+	backendJSON, err := json.Marshal(map[string]any{
+		"backend_type":    row.BackendType,
+		"backend_options": backendConfig,
+	})
+	if err != nil {
+		return nil, fmt.Errorf("failed to marshal backend for unmarshaling: %w", err)
+	}
+
+	if err := json.Unmarshal(backendJSON, &backendOptions); err != nil {
+		return nil, fmt.Errorf("failed to unmarshal backend options: %w", err)
+	}
+
+	// Unmarshal nodes
+	var nodes map[string]struct{}
+	if row.Nodes.Valid && row.Nodes.String != "" {
+		var nodesList []string
+		if err := json.Unmarshal([]byte(row.Nodes.String), &nodesList); err != nil {
+			return nil, fmt.Errorf("failed to unmarshal nodes: %w", err)
+		}
+		nodes = make(map[string]struct{}, len(nodesList))
+		for _, node := range nodesList {
+			nodes[node] = struct{}{}
+		}
+	}
+
+	// Unmarshal environment
+	var environment map[string]string
+	if row.Environment.Valid && row.Environment.String != "" {
+		if err := json.Unmarshal([]byte(row.Environment.String), &environment); err != nil {
+			return nil, fmt.Errorf("failed to unmarshal environment: %w", err)
+		}
+	}
+
+	// Convert integers to boolean pointers
+	autoRestart := row.AutoRestart == 1
+	maxRestarts := row.MaxRestarts
+	restartDelay := row.RestartDelay
+	onDemandStart := row.OnDemandStart == 1
+	idleTimeout := row.IdleTimeout
+	dockerEnabled := row.DockerEnabled == 1
+
+	// Create instance options
+	opts := &instance.Options{
+		AutoRestart:     &autoRestart,
+		MaxRestarts:     &maxRestarts,
+		RestartDelay:    &restartDelay,
+		OnDemandStart:   &onDemandStart,
+		IdleTimeout:     &idleTimeout,
+		DockerEnabled:   &dockerEnabled,
+		CommandOverride: row.CommandOverride.String,
+		Nodes:           nodes,
+		Environment:     environment,
+		BackendOptions:  backendOptions,
+	}
+
+	// Create instance struct and manually unmarshal fields
+	// We do this manually because BackendOptions and Nodes have json:"-" tags
+	// and would be lost if we used the marshal/unmarshal cycle
+	inst := &instance.Instance{
+		Name:    row.Name,
+		Created: row.CreatedAt,
+	}
+
+	// Create a temporary struct for unmarshaling the status and simple fields
+	type instanceAux struct {
+		Name    string `json:"name"`
+		Created int64  `json:"created"`
+		Status  string `json:"status"`
+		Options struct {
+			AutoRestart     *bool             `json:"auto_restart,omitempty"`
+			MaxRestarts     *int              `json:"max_restarts,omitempty"`
+			RestartDelay    *int              `json:"restart_delay,omitempty"`
+			OnDemandStart   *bool             `json:"on_demand_start,omitempty"`
+			IdleTimeout     *int              `json:"idle_timeout,omitempty"`
+			DockerEnabled   *bool             `json:"docker_enabled,omitempty"`
+			CommandOverride string            `json:"command_override,omitempty"`
+			Environment     map[string]string `json:"environment,omitempty"`
+		} `json:"options"`
+	}
+
+	aux := instanceAux{
+		Name:    row.Name,
+		Created: row.CreatedAt,
+		Status:  row.Status,
+	}
+	aux.Options.AutoRestart = opts.AutoRestart
+	aux.Options.MaxRestarts = opts.MaxRestarts
+	aux.Options.RestartDelay = opts.RestartDelay
+	aux.Options.OnDemandStart = opts.OnDemandStart
+	aux.Options.IdleTimeout = opts.IdleTimeout
+	aux.Options.DockerEnabled = opts.DockerEnabled
+	aux.Options.CommandOverride = opts.CommandOverride
+	aux.Options.Environment = opts.Environment
+
+	instJSON, err := json.Marshal(aux)
+	if err != nil {
+		return nil, fmt.Errorf("failed to marshal instance: %w", err)
+	}
+
+	if err := json.Unmarshal(instJSON, inst); err != nil {
+		return nil, fmt.Errorf("failed to unmarshal instance: %w", err)
+	}
+
+	// Manually set the fields that have json:"-" tags by using SetOptions
+	// We need to set the whole options object because GetOptions returns a copy
+	// and we need to ensure BackendOptions and Nodes (which have json:"-") are set
+	inst.SetOptions(opts)
+
+	return inst, nil
+}
+
+// Database interface implementation
+
+// Save saves an instance to the database (insert or update)
+func (db *DB) Save(inst *instance.Instance) error {
+	ctx := context.Background()
+
+	// Try to get existing instance
+	existing, err := db.GetByName(ctx, inst.Name)
+	if err != nil {
+		// Instance doesn't exist, create it
+		return db.Create(ctx, inst)
+	}
+
+	// Instance exists, update it
+	if existing != nil {
+		return db.Update(ctx, inst)
+	}
+
+	return db.Create(ctx, inst)
+}
+
+// Delete removes an instance from the database
+func (db *DB) Delete(name string) error {
+	ctx := context.Background()
+	return db.DeleteInstance(ctx, name)
+}
+
+// LoadAll loads all instances from the database
+func (db *DB) LoadAll() ([]*instance.Instance, error) {
+	ctx := context.Background()
+	return db.GetAll(ctx)
+}

pkg/database/migrations.go

@@ -0,0 +1,78 @@
+package database
+
+import (
+	"embed"
+	"fmt"
+	"log"
+
+	"github.com/golang-migrate/migrate/v4"
+	"github.com/golang-migrate/migrate/v4/database/sqlite3"
+	"github.com/golang-migrate/migrate/v4/source/iofs"
+)
+
+//go:embed migrations/*.sql
+var migrationFiles embed.FS
+
+// RunMigrations applies all pending database migrations
+func RunMigrations(db *DB) error {
+	if db == nil || db.DB == nil {
+		return fmt.Errorf("database connection is nil")
+	}
+
+	// Create migration source from embedded files
+	sourceDriver, err := iofs.New(migrationFiles, "migrations")
+	if err != nil {
+		return fmt.Errorf("failed to create migration source: %w", err)
+	}
+
+	// Create database driver for migrations
+	dbDriver, err := sqlite3.WithInstance(db.DB, &sqlite3.Config{})
+	if err != nil {
+		return fmt.Errorf("failed to create database driver: %w", err)
+	}
+
+	// Create migrator
+	migrator, err := migrate.NewWithInstance("iofs", sourceDriver, "sqlite3", dbDriver)
+	if err != nil {
+		return fmt.Errorf("failed to create migrator: %w", err)
+	}
+
+	// Get current version
+	currentVersion, dirty, err := migrator.Version()
+	if err != nil && err != migrate.ErrNilVersion {
+		return fmt.Errorf("failed to get current migration version: %w", err)
+	}
+
+	if dirty {
+		return fmt.Errorf("database is in dirty state at version %d - manual intervention required", currentVersion)
+	}
+
+	// Run migrations
+	log.Printf("Running database migrations (current version: %v)", currentVersionString(currentVersion, err))
+
+	if err := migrator.Up(); err != nil {
+		if err == migrate.ErrNoChange {
+			log.Println("Database schema is up to date")
+			return nil
+		}
+		return fmt.Errorf("failed to run migrations: %w", err)
+	}
+
+	// Get new version
+	newVersion, _, err := migrator.Version()
+	if err != nil {
+		log.Printf("Migrations completed (unable to determine new version: %v)", err)
+	} else {
+		log.Printf("Migrations completed successfully (new version: %d)", newVersion)
+	}
+
+	return nil
+}
+
+// currentVersionString returns a string representation of the current version
+func currentVersionString(version uint, err error) string {
+	if err == migrate.ErrNilVersion {
+		return "none"
+	}
+	return fmt.Sprintf("%d", version)
+}

pkg/database/migrations/001_initial_schema.down.sql

@@ -0,0 +1,7 @@
+-- Drop indexes first
+DROP INDEX IF EXISTS idx_instances_backend_type;
+DROP INDEX IF EXISTS idx_instances_status;
+DROP INDEX IF EXISTS idx_instances_name;
+
+-- Drop tables
+DROP TABLE IF EXISTS instances;

pkg/database/migrations/001_initial_schema.up.sql

@@ -0,0 +1,43 @@
+-- -----------------------------------------------------------------------------
+-- Instances Table: Central configuration and state for LLM backends
+-- -----------------------------------------------------------------------------
+CREATE TABLE IF NOT EXISTS instances (
+    -- Primary identification
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    name TEXT NOT NULL UNIQUE,
+
+    -- Backend configuration
+    backend_type TEXT NOT NULL CHECK(backend_type IN ('llama_cpp', 'mlx_lm', 'vllm')),
+    backend_config_json TEXT NOT NULL, -- Backend-specific options (150+ fields for llama_cpp, etc.)
+
+    -- Instance state
+    status TEXT NOT NULL CHECK(status IN ('stopped', 'running', 'failed', 'restarting', 'shutting_down')) DEFAULT 'stopped',
+
+    -- Timestamps (created_at stored as Unix timestamp for compatibility with existing JSON format)
+    created_at INTEGER NOT NULL,
+    updated_at INTEGER NOT NULL,
+
+    -- Common instance options (extracted from Instance.Options)
+    -- NOT NULL with defaults to match config behavior (nil pointers use these defaults)
+    auto_restart INTEGER NOT NULL DEFAULT 0,        -- Boolean: Enable automatic restart on failure
+    max_restarts INTEGER NOT NULL DEFAULT -1,       -- Maximum restart attempts (-1 = unlimited)
+    restart_delay INTEGER NOT NULL DEFAULT 0,       -- Delay between restarts in seconds
+    on_demand_start INTEGER NOT NULL DEFAULT 0,     -- Boolean: Enable on-demand instance start
+    idle_timeout INTEGER NOT NULL DEFAULT 0,        -- Idle timeout in minutes before auto-stop
+    docker_enabled INTEGER NOT NULL DEFAULT 0,      -- Boolean: Run instance in Docker container
+    command_override TEXT,                          -- Custom command to override default backend command (nullable)
+
+    -- JSON fields for complex structures (nullable - empty when not set)
+    nodes TEXT,                            -- JSON array of node names for remote instances
+    environment TEXT,                      -- JSON map of environment variables
+
+    -- Future extensibility hook
+    owner_user_id TEXT NULL                -- Future: OIDC user ID for ownership
+);
+
+-- -----------------------------------------------------------------------------
+-- Indexes for performance
+-- -----------------------------------------------------------------------------
+CREATE UNIQUE INDEX IF NOT EXISTS idx_instances_name ON instances(name);
+CREATE INDEX IF NOT EXISTS idx_instances_status ON instances(status);
+CREATE INDEX IF NOT EXISTS idx_instances_backend_type ON instances(backend_type);

pkg/manager/manager.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"llamactl/pkg/config"
+	"llamactl/pkg/database"
 	"llamactl/pkg/instance"
 	"log"
 	"sync"
@@ -28,11 +29,11 @@ type InstanceManager interface {
 
 type instanceManager struct {
 	// Components (each with own synchronization)
-	registry    *instanceRegistry
-	ports       *portAllocator
-	persistence *instancePersister
-	remote      *remoteManager
-	lifecycle   *lifecycleManager
+	registry     *instanceRegistry
+	ports        *portAllocator
+	db           database.Database
+	remote       *remoteManager
+	lifecycle    *lifecycleManager
 
 	// Configuration
 	globalConfig *config.AppConfig
@@ -42,8 +43,8 @@ type instanceManager struct {
 	shutdownOnce  sync.Once
 }
 
-// New creates a new instance of InstanceManager.
-func New(globalConfig *config.AppConfig) InstanceManager {
+// New creates a new instance of InstanceManager with dependency injection.
+func New(globalConfig *config.AppConfig, db database.Database) InstanceManager {
 
 	if globalConfig.Instances.TimeoutCheckInterval <= 0 {
 		globalConfig.Instances.TimeoutCheckInterval = 5 // Default to 5 minutes if not set
@@ -56,9 +57,6 @@ func New(globalConfig *config.AppConfig) InstanceManager {
 	portRange := globalConfig.Instances.PortRange
 	ports := newPortAllocator(portRange[0], portRange[1])
 
-	// Initialize persistence
-	persistence := newInstancePersister(globalConfig.Instances.InstancesDir)
-
 	// Initialize remote manager
 	remote := newRemoteManager(globalConfig.Nodes, 30*time.Second)
 
@@ -66,7 +64,7 @@ func New(globalConfig *config.AppConfig) InstanceManager {
 	im := &instanceManager{
 		registry:     registry,
 		ports:        ports,
-		persistence:  persistence,
+		db:           db,
 		remote:       remote,
 		globalConfig: globalConfig,
 	}
@@ -86,9 +84,9 @@ func New(globalConfig *config.AppConfig) InstanceManager {
 	return im
 }
 
-// persistInstance saves an instance using the persistence component
+// persistInstance saves an instance using the persistence layer
 func (im *instanceManager) persistInstance(inst *instance.Instance) error {
-	return im.persistence.save(inst)
+	return im.db.Save(inst)
 }
 
 func (im *instanceManager) Shutdown() {
@@ -116,13 +114,18 @@ func (im *instanceManager) Shutdown() {
 		}
 		wg.Wait()
 		fmt.Println("All instances stopped.")
+
+		// 4. Close database connection
+		if err := im.db.Close(); err != nil {
+			log.Printf("Error closing database: %v\n", err)
+		}
 	})
 }
 
-// loadInstances restores all instances from disk using the persistence component
+// loadInstances restores all instances from the persistence layer
 func (im *instanceManager) loadInstances() error {
 	// Load all instances from persistence
-	instances, err := im.persistence.loadAll()
+	instances, err := im.db.LoadAll()
 	if err != nil {
 		return fmt.Errorf("failed to load instances: %w", err)
 	}

pkg/manager/manager_test.go

@@ -4,20 +4,34 @@ import (
 	"fmt"
 	"llamactl/pkg/backends"
 	"llamactl/pkg/config"
+	"llamactl/pkg/database"
 	"llamactl/pkg/instance"
 	"llamactl/pkg/manager"
-	"os"
-	"path/filepath"
 	"sync"
 	"testing"
+	"time"
 )
 
 func TestManager_PersistsAndLoadsInstances(t *testing.T) {
 	tempDir := t.TempDir()
 	appConfig := createTestAppConfig(tempDir)
+	// Use file-based database for this test since we need to persist across connections
+	appConfig.Database.Path = tempDir + "/test.db"
 
-	// Create instance and check file was created
-	manager1 := manager.New(appConfig)
+	// Create instance and check database was created
+	db1, err := database.Open(&database.Config{
+		Path:               appConfig.Database.Path,
+		MaxOpenConnections: appConfig.Database.MaxOpenConnections,
+		MaxIdleConnections: appConfig.Database.MaxIdleConnections,
+		ConnMaxLifetime:    appConfig.Database.ConnMaxLifetime,
+	})
+	if err != nil {
+		t.Fatalf("Failed to open database: %v", err)
+	}
+	if err := database.RunMigrations(db1); err != nil {
+		t.Fatalf("Failed to run migrations: %v", err)
+	}
+	manager1 := manager.New(appConfig, db1)
 	options := &instance.Options{
 		BackendOptions: backends.Options{
 			BackendType: backends.BackendTypeLlamaCpp,
@@ -28,18 +42,28 @@ func TestManager_PersistsAndLoadsInstances(t *testing.T) {
 		},
 	}
 
-	_, err := manager1.CreateInstance("test-instance", options)
+	_, err = manager1.CreateInstance("test-instance", options)
 	if err != nil {
 		t.Fatalf("CreateInstance failed: %v", err)
 	}
 
-	expectedPath := filepath.Join(tempDir, "test-instance.json")
-	if _, err := os.Stat(expectedPath); os.IsNotExist(err) {
-		t.Errorf("Expected persistence file %s to exist", expectedPath)
-	}
+	// Shutdown first manager to close database connection
+	manager1.Shutdown()
 
-	// Load instances from disk
-	manager2 := manager.New(appConfig)
+	// Load instances from database
+	db2, err := database.Open(&database.Config{
+		Path:               appConfig.Database.Path,
+		MaxOpenConnections: appConfig.Database.MaxOpenConnections,
+		MaxIdleConnections: appConfig.Database.MaxIdleConnections,
+		ConnMaxLifetime:    appConfig.Database.ConnMaxLifetime,
+	})
+	if err != nil {
+		t.Fatalf("Failed to open database: %v", err)
+	}
+	if err := database.RunMigrations(db2); err != nil {
+		t.Fatalf("Failed to run migrations: %v", err)
+	}
+	manager2 := manager.New(appConfig, db2)
 	instances, err := manager2.ListInstances()
 	if err != nil {
 		t.Fatalf("ListInstances failed: %v", err)
@@ -50,13 +74,29 @@ func TestManager_PersistsAndLoadsInstances(t *testing.T) {
 	if instances[0].Name != "test-instance" {
 		t.Errorf("Expected loaded instance name 'test-instance', got %q", instances[0].Name)
 	}
+
+	manager2.Shutdown()
 }
 
-func TestDeleteInstance_RemovesPersistenceFile(t *testing.T) {
+func TestDeleteInstance_RemovesFromDatabase(t *testing.T) {
 	tempDir := t.TempDir()
 	appConfig := createTestAppConfig(tempDir)
 
-	mgr := manager.New(appConfig)
+	db, err := database.Open(&database.Config{
+		Path:               appConfig.Database.Path,
+		MaxOpenConnections: appConfig.Database.MaxOpenConnections,
+		MaxIdleConnections: appConfig.Database.MaxIdleConnections,
+		ConnMaxLifetime:    appConfig.Database.ConnMaxLifetime,
+	})
+	if err != nil {
+		t.Fatalf("Failed to open database: %v", err)
+	}
+	if err := database.RunMigrations(db); err != nil {
+		t.Fatalf("Failed to run migrations: %v", err)
+	}
+	mgr := manager.New(appConfig, db)
+	defer mgr.Shutdown()
+
 	options := &instance.Options{
 		BackendOptions: backends.Options{
 			BackendType: backends.BackendTypeLlamaCpp,
@@ -67,20 +107,33 @@ func TestDeleteInstance_RemovesPersistenceFile(t *testing.T) {
 		},
 	}
 
-	_, err := mgr.CreateInstance("test-instance", options)
+	_, err = mgr.CreateInstance("test-instance", options)
 	if err != nil {
 		t.Fatalf("CreateInstance failed: %v", err)
 	}
 
-	expectedPath := filepath.Join(tempDir, "test-instance.json")
+	// Verify instance exists
+	instances, err := mgr.ListInstances()
+	if err != nil {
+		t.Fatalf("ListInstances failed: %v", err)
+	}
+	if len(instances) != 1 {
+		t.Fatalf("Expected 1 instance, got %d", len(instances))
+	}
 
+	// Delete instance
 	err = mgr.DeleteInstance("test-instance")
 	if err != nil {
 		t.Fatalf("DeleteInstance failed: %v", err)
 	}
 
-	if _, err := os.Stat(expectedPath); !os.IsNotExist(err) {
-		t.Error("Expected persistence file to be deleted")
+	// Verify instance was deleted from database
+	instances, err = mgr.ListInstances()
+	if err != nil {
+		t.Fatalf("ListInstances failed: %v", err)
+	}
+	if len(instances) != 0 {
+		t.Errorf("Expected 0 instances after deletion, got %d", len(instances))
 	}
 }
 
@@ -158,6 +211,12 @@ func createTestAppConfig(instancesDir string) *config.AppConfig {
 			DefaultRestartDelay:  5,
 			TimeoutCheckInterval: 5,
 		},
+		Database: config.DatabaseConfig{
+			Path:               ":memory:",
+			MaxOpenConnections: 25,
+			MaxIdleConnections: 5,
+			ConnMaxLifetime:    5 * time.Minute,
+		},
 		LocalNode: "main",
 		Nodes:     map[string]config.NodeConfig{},
 	}
@@ -166,5 +225,17 @@ func createTestAppConfig(instancesDir string) *config.AppConfig {
 func createTestManager(t *testing.T) manager.InstanceManager {
 	tempDir := t.TempDir()
 	appConfig := createTestAppConfig(tempDir)
-	return manager.New(appConfig)
+	db, err := database.Open(&database.Config{
+		Path:               appConfig.Database.Path,
+		MaxOpenConnections: appConfig.Database.MaxOpenConnections,
+		MaxIdleConnections: appConfig.Database.MaxIdleConnections,
+		ConnMaxLifetime:    appConfig.Database.ConnMaxLifetime,
+	})
+	if err != nil {
+		t.Fatalf("Failed to open database: %v", err)
+	}
+	if err := database.RunMigrations(db); err != nil {
+		t.Fatalf("Failed to run migrations: %v", err)
+	}
+	return manager.New(appConfig, db)
 }

pkg/manager/operations.go

@@ -317,9 +317,9 @@ func (im *instanceManager) DeleteInstance(name string) error {
 		im.remote.removeInstance(name)
 		im.registry.remove(name)
 
-		// Delete the instance's persistence file
-		if err := im.persistence.delete(name); err != nil {
-			return fmt.Errorf("failed to delete config file for remote instance %s: %w", name, err)
+		// Delete the instance's persistence
+		if err := im.db.Delete(name); err != nil {
+			return fmt.Errorf("failed to delete remote instance %s: %w", name, err)
 		}
 
 		return nil
@@ -343,9 +343,9 @@ func (im *instanceManager) DeleteInstance(name string) error {
 		return fmt.Errorf("failed to remove instance from registry: %w", err)
 	}
 
-	// Delete persistence file
-	if err := im.persistence.delete(name); err != nil {
-		return fmt.Errorf("failed to delete config file for instance %s: %w", name, err)
+	// Delete from persistence
+	if err := im.db.Delete(name); err != nil {
+		return fmt.Errorf("failed to delete instance from persistence %s: %w", name, err)
 	}
 
 	return nil

pkg/manager/operations_test.go

@@ -3,10 +3,12 @@ package manager_test
 import (
 	"llamactl/pkg/backends"
 	"llamactl/pkg/config"
+	"llamactl/pkg/database"
 	"llamactl/pkg/instance"
 	"llamactl/pkg/manager"
 	"strings"
 	"testing"
+	"time"
 )
 
 func TestCreateInstance_FailsWithDuplicateName(t *testing.T) {
@@ -49,10 +51,28 @@ func TestCreateInstance_FailsWhenMaxInstancesReached(t *testing.T) {
 			MaxInstances:         1, // Very low limit for testing
 			TimeoutCheckInterval: 5,
 		},
+		Database: config.DatabaseConfig{
+			Path:               ":memory:",
+			MaxOpenConnections: 25,
+			MaxIdleConnections: 5,
+			ConnMaxLifetime:    5 * time.Minute,
+		},
 		LocalNode: "main",
 		Nodes:     map[string]config.NodeConfig{},
 	}
-	limitedManager := manager.New(appConfig)
+	db, err := database.Open(&database.Config{
+		Path:               appConfig.Database.Path,
+		MaxOpenConnections: appConfig.Database.MaxOpenConnections,
+		MaxIdleConnections: appConfig.Database.MaxIdleConnections,
+		ConnMaxLifetime:    appConfig.Database.ConnMaxLifetime,
+	})
+	if err != nil {
+		t.Fatalf("Failed to open database: %v", err)
+	}
+	if err := database.RunMigrations(db); err != nil {
+		t.Fatalf("Failed to run migrations: %v", err)
+	}
+	limitedManager := manager.New(appConfig, db)
 
 	options := &instance.Options{
 		BackendOptions: backends.Options{
@@ -63,7 +83,7 @@ func TestCreateInstance_FailsWhenMaxInstancesReached(t *testing.T) {
 		},
 	}
 
-	_, err := limitedManager.CreateInstance("instance1", options)
+	_, err = limitedManager.CreateInstance("instance1", options)
 	if err != nil {
 		t.Fatalf("CreateInstance 1 failed: %v", err)
 	}

pkg/manager/persistence.go

@@ -1,198 +0,0 @@
-package manager
-
-import (
-	"encoding/json"
-	"fmt"
-	"llamactl/pkg/instance"
-	"log"
-	"os"
-	"path/filepath"
-	"strings"
-	"sync"
-)
-
-// instancePersister provides atomic file-based persistence with durability guarantees.
-type instancePersister struct {
-	mu           sync.Mutex
-	instancesDir string
-}
-
-// newInstancePersister creates a new instance persister.
-// If instancesDir is empty, persistence is disabled.
-func newInstancePersister(instancesDir string) *instancePersister {
-	return &instancePersister{
-		instancesDir: instancesDir,
-	}
-}
-
-// Save persists an instance to disk with atomic write
-func (p *instancePersister) save(inst *instance.Instance) error {
-	if inst == nil {
-		return fmt.Errorf("cannot save nil instance")
-	}
-
-	// Validate instance name to prevent path traversal
-	validatedName, err := p.validateInstanceName(inst.Name)
-	if err != nil {
-		return err
-	}
-
-	p.mu.Lock()
-	defer p.mu.Unlock()
-
-	instancePath := filepath.Join(p.instancesDir, validatedName+".json")
-	tempPath := instancePath + ".tmp"
-
-	// Serialize instance to JSON
-	jsonData, err := json.MarshalIndent(inst, "", "  ")
-	if err != nil {
-		return fmt.Errorf("failed to marshal instance %s: %w", inst.Name, err)
-	}
-
-	// Create temporary file
-	tempFile, err := os.OpenFile(tempPath, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0644)
-	if err != nil {
-		return fmt.Errorf("failed to create temp file for instance %s: %w", inst.Name, err)
-	}
-
-	// Write data to temporary file
-	if _, err := tempFile.Write(jsonData); err != nil {
-		tempFile.Close()
-		os.Remove(tempPath)
-		return fmt.Errorf("failed to write temp file for instance %s: %w", inst.Name, err)
-	}
-
-	// Sync to disk before rename to ensure durability
-	if err := tempFile.Sync(); err != nil {
-		tempFile.Close()
-		os.Remove(tempPath)
-		return fmt.Errorf("failed to sync temp file for instance %s: %w", inst.Name, err)
-	}
-
-	// Close the file
-	if err := tempFile.Close(); err != nil {
-		os.Remove(tempPath)
-		return fmt.Errorf("failed to close temp file for instance %s: %w", inst.Name, err)
-	}
-
-	// Atomic rename (this is atomic on POSIX systems)
-	if err := os.Rename(tempPath, instancePath); err != nil {
-		os.Remove(tempPath)
-		return fmt.Errorf("failed to rename temp file for instance %s: %w", inst.Name, err)
-	}
-
-	return nil
-}
-
-// Delete removes an instance's persistence file from disk.
-func (p *instancePersister) delete(name string) error {
-	validatedName, err := p.validateInstanceName(name)
-	if err != nil {
-		return err
-	}
-
-	p.mu.Lock()
-	defer p.mu.Unlock()
-
-	instancePath := filepath.Join(p.instancesDir, validatedName+".json")
-
-	if err := os.Remove(instancePath); err != nil {
-		if os.IsNotExist(err) {
-			// Not an error if file doesn't exist
-			return nil
-		}
-		return fmt.Errorf("failed to delete instance file for %s: %w", name, err)
-	}
-
-	return nil
-}
-
-// LoadAll loads all persisted instances from disk.
-// Returns a slice of instances and any errors encountered during loading.
-func (p *instancePersister) loadAll() ([]*instance.Instance, error) {
-	p.mu.Lock()
-	defer p.mu.Unlock()
-
-	// Check if instances directory exists
-	if _, err := os.Stat(p.instancesDir); os.IsNotExist(err) {
-		return nil, nil // No instances directory, return empty list
-	}
-
-	// Read all JSON files from instances directory
-	files, err := os.ReadDir(p.instancesDir)
-	if err != nil {
-		return nil, fmt.Errorf("failed to read instances directory: %w", err)
-	}
-
-	instances := make([]*instance.Instance, 0)
-	var loadErrors []string
-
-	for _, file := range files {
-		if file.IsDir() || !strings.HasSuffix(file.Name(), ".json") {
-			continue
-		}
-
-		instanceName := strings.TrimSuffix(file.Name(), ".json")
-		instancePath := filepath.Join(p.instancesDir, file.Name())
-
-		inst, err := p.loadInstanceFile(instanceName, instancePath)
-		if err != nil {
-			log.Printf("Failed to load instance %s: %v", instanceName, err)
-			loadErrors = append(loadErrors, fmt.Sprintf("%s: %v", instanceName, err))
-			continue
-		}
-
-		instances = append(instances, inst)
-	}
-
-	if len(loadErrors) > 0 {
-		log.Printf("Loaded %d instances with %d errors", len(instances), len(loadErrors))
-	} else if len(instances) > 0 {
-		log.Printf("Loaded %d instances from persistence", len(instances))
-	}
-
-	return instances, nil
-}
-
-// loadInstanceFile is an internal helper that loads a single instance file.
-// Note: This assumes the mutex is already held by the caller.
-func (p *instancePersister) loadInstanceFile(name, path string) (*instance.Instance, error) {
-	data, err := os.ReadFile(path)
-	if err != nil {
-		return nil, fmt.Errorf("failed to read instance file: %w", err)
-	}
-
-	var inst instance.Instance
-	if err := json.Unmarshal(data, &inst); err != nil {
-		return nil, fmt.Errorf("failed to unmarshal instance: %w", err)
-	}
-
-	// Validate the instance name matches the filename
-	if inst.Name != name {
-		return nil, fmt.Errorf("instance name mismatch: file=%s, instance.Name=%s", name, inst.Name)
-	}
-
-	return &inst, nil
-}
-
-// validateInstanceName ensures the instance name is safe for filesystem operations.
-// Returns the validated name if valid, or an error if invalid.
-func (p *instancePersister) validateInstanceName(name string) (string, error) {
-	if name == "" {
-		return "", fmt.Errorf("instance name cannot be empty")
-	}
-
-	// Check for path separators and parent directory references
-	// This prevents path traversal attacks
-	if strings.Contains(name, "/") || strings.Contains(name, "\\") || strings.Contains(name, "..") {
-		return "", fmt.Errorf("invalid instance name: %s (cannot contain path separators or '..')", name)
-	}
-
-	// Additional check: ensure the name doesn't start with a dot (hidden files)
-	// or contain any other suspicious characters
-	if strings.HasPrefix(name, ".") {
-		return "", fmt.Errorf("invalid instance name: %s (cannot start with '.')", name)
-	}
-
-	return name, nil
-}