llm

type CompletionRequest struct {
	Prompt  string
	Format  json.RawMessage
	Images  []ImageData
	Options *api.Options
}

type CompletionResponse struct {
	Content            string
	DoneReason         string
	Done               bool
	PromptEvalCount    int
	PromptEvalDuration time.Duration
	EvalCount          int
	EvalDuration       time.Duration
}

type DetokenizeRequest struct {
	Tokens []int `json:"tokens"`
}

type DetokenizeResponse struct {
	Content string `json:"content"`
}

type EmbeddingRequest struct {
	Content string `json:"content"`
}

type EmbeddingResponse struct {
	Embedding []float32 `json:"embedding"`
}

type ImageData struct {
	Data          []byte `json:"data"`
	ID            int    `json:"id"`
	AspectRatioID int    `json:"aspect_ratio_id"`
}

type LlamaServer interface {
	Ping(ctx context.Context) error
	WaitUntilRunning(ctx context.Context) error
	Completion(ctx context.Context, req CompletionRequest, fn func(CompletionResponse)) error
	Embedding(ctx context.Context, input string) ([]float32, error)
	Tokenize(ctx context.Context, content string) ([]int, error)
	Detokenize(ctx context.Context, tokens []int) (string, error)
	Close() error
	EstimatedVRAM() uint64 // Total VRAM across all GPUs
	EstimatedTotal() uint64
	EstimatedVRAMByGPU(gpuID string) uint64
}

type MemoryEstimate struct {
	// How many layers we predict we can load
	Layers int

	// The size of the graph which occupies the main GPU
	Graph uint64

	// How much VRAM will be allocated given the number of layers we predict
	VRAMSize uint64

	// The total size of the model if loaded into VRAM.  If all layers are loaded, VRAMSize == TotalSize
	TotalSize uint64

	// For multi-GPU scenarios, this provides the tensor split parameter
	TensorSplit string

	// For multi-GPU scenarios, this is the size in bytes per GPU
	GPUSizes []uint64
	// contains filtered or unexported fields
}

type ServerStatus int

type ServerStatusResp struct {
	Status          string  `json:"status"`
	SlotsIdle       int     `json:"slots_idle"`
	SlotsProcessing int     `json:"slots_processing"`
	Error           string  `json:"error"`
	Progress        float32 `json:"progress"`
}

type StatusWriter struct {
	LastErrMsg string
	// contains filtered or unexported fields
}

type TokenizeRequest struct {
	Content string `json:"content"`
}

type TokenizeResponse struct {
	Tokens []int `json:"tokens"`
}