backend

type LLMBackend interface {
	// Generate produces a completion for the given prompt.
	// model specifies which model to use (interpretation is backend-specific).
	// maxTokens limits the response length (0 means use backend default).
	Generate(ctx context.Context, prompt string, model string, maxTokens int) (string, error)

	// Name returns a human-readable name for the backend.
	Name() string

	// Close releases any resources held by the backend.
	Close() error
}

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

type LlamaConfig struct {
	// UseSubprocess enables subprocess-based inference for concurrency safety.
	UseSubprocess bool

	// WorkerClient is an optional pre-configured worker client.
	// If nil and UseSubprocess is true, a new worker will be created.
	WorkerClient WorkerClient

	// Default generation parameters
	MaxTokens int
	TopK      int
	TopP      float64
	Temp      float64
}

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

type OpenAIConfig struct {
	APIKey       string
	BaseURL      string // Optional: for Azure or compatible APIs
	DefaultModel string
}

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

type ShellBackend interface {
	// Run executes a shell command and returns combined stdout/stderr.
	Run(ctx context.Context, command string) (string, error)

	// RunWithEnv executes with additional environment variables.
	RunWithEnv(ctx context.Context, command string, env map[string]string) (string, error)
}

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

type ShellConfig struct {
	// Shell is the shell to use (default: "sh")
	Shell string
}

type WorkerClient interface {
	SendRequest(modelSpec, prompt string, maxTokens int) (string, error)
	Close() error
}