cmd

Documentation

Index

• func Execute()
• type ChatSession
  • func NewChatSession(api *server.APIClient, model string, cfg *config.Config, ...) *ChatSession
  • func (s *ChatSession) Run(prompt string) error
  • func (s *ChatSession) SetSamplingOptions(temp, topP, minP, repeatPenalty, presencePenalty, frequencyPenalty float64, ...)
  • func (s *ChatSession) SetSystemPrompt(prompt string)
• type ModelInfo

Functions

func Execute

func Execute()

Types

type ChatSession

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

ChatSession handles one-shot chat completions for CLI prompts.

func NewChatSession

func NewChatSession(api *server.APIClient, model string, cfg *config.Config, persona *config.Persona, preset *presets.Preset) *ChatSession

NewChatSession creates a new chat session.

func (*ChatSession) Run

func (s *ChatSession) Run(prompt string) error

Run sends the prompt to the model and streams the response.

func (*ChatSession) SetSamplingOptions

func (s *ChatSession) SetSamplingOptions(temp, topP, minP, repeatPenalty, presencePenalty, frequencyPenalty float64, topK, maxTokens int)

SetSamplingOptions sets the sampling parameters for generation.

func (*ChatSession) SetSystemPrompt

func (s *ChatSession) SetSystemPrompt(prompt string)

SetSystemPrompt sets the system prompt for the session.

type ModelInfo

type ModelInfo struct {
	User     string
	Repo     string
	Quant    string
	Backend  string
	Path     string
	Size     int64
	LastUsed time.Time
}

ModelInfo represents a locally downloaded model. Backend / Path mirror hf.LocalModel so cmd code can delete by the inventory's authoritative answer rather than re-deriving paths per backend.