serving

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

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

type CacheState struct {
	// ServerName is the name of the LLM server
	ServerName string
	// IsFlushed indicates whether the cache has been flushed
	IsFlushed bool
	// LastTurnSize is the size of the last turn (in tokens)
	LastTurnSize int
	// LastMemoryPressure is the last known memory pressure (0.0-1.0)
	LastMemoryPressure float64
	// LastWorkflowName is the name of the last workflow that used this cache
	LastWorkflowName string
}

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

type ExecuteTaskOptions struct {
	// MaxTokens is the maximum number of tokens to generate.
	MaxTokens int
	// Stream is whether to stream the response.
	Stream bool
}

type HealthStatus string

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

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

type ServingLayer interface {
	// GetProvider returns a model provider for an agent profile or workflow task
	// If task is nil, uses the agent profile's LLM server configuration
	// If task is not nil and has an LLM server override, uses that instead
	GetProvider(ctx context.Context, profileName string, task *config.WorkflowTask) (model.Provider, error)

	// StartWorkflow initializes a workflow execution
	StartWorkflow(ctx context.Context, workflowName string) (*WorkflowExecution, error)

	// ExecuteTask executes a single workflow task with the given prompt and options
	// Returns the response from the task execution (includes metrics when options.Stream is true)
	ExecuteTask(ctx context.Context, execution *WorkflowExecution, taskIndex int, prompt string, options ExecuteTaskOptions) (*model.Response, error)

	// GetSharedContext returns shared context for a given LLM server
	GetSharedContext(serverName string) *SharedContext

	// GetExecution returns a workflow execution by ID
	GetExecution(executionID string) (*WorkflowExecution, error)
}

type SharedContext struct {
	// ServerName is the name of the LLM server this context belongs to
	ServerName string
	// Messages is the shared conversation history
	Messages []model.Message

	// SyncInterval is how often to synchronize context (in tokens)
	SyncInterval int
	// LastSyncTokenCount is the token count at the last synchronization
	LastSyncTokenCount int
	// contains filtered or unexported fields
}

type WorkflowExecution struct {
	// ExecutionID is the unique identifier for this execution
	ExecutionID string
	// WorkflowName is the name of the workflow being executed
	WorkflowName string
	// CurrentTaskIndex is the index of the current task (0-based)
	CurrentTaskIndex int
	// Tasks is the list of tasks in the workflow
	Tasks []*config.WorkflowTask
	// State is the execution state
	State WorkflowExecutionState
	// Context holds the accumulated messages from previous tasks
	// This is used when tasks have UseContext: true
	Context []model.Message
}

type WorkflowExecutionState string

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