llm

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

type ContainerGGUF struct {
	ByteOrder binary.ByteOrder

	Version uint32

	V1 struct {
		NumTensor uint32
		NumKV     uint32
	}

	V2 struct {
		NumTensor uint64
		NumKV     uint64
	}

	V3 struct {
		NumTensor uint64
		NumKV     uint64
	}
}

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 []float64 `json:"embedding"`
}

type GGML struct {
	Size int64
	// contains filtered or unexported fields
}

type GGUFModel struct {
	*ContainerGGUF

	KV
	Tensors []Tensor
	// contains filtered or unexported fields
}

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

type KV map[string]any

type LLM interface {
	Predict(context.Context, PredictOpts, func(PredictResult)) error
	Embedding(context.Context, string) ([]float64, error)
	Encode(context.Context, string) ([]int, error)
	Decode(context.Context, []int) (string, error)
	Close()
}

type ModelGGLA struct {
	*ContainerGGLA
	// contains filtered or unexported fields
}

type PredictOpts struct {
	Prompt  string
	Format  string
	Images  []ImageData
	Options api.Options
}

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

type Tensor struct {
	Name   string
	Kind   uint32
	Offset uint64

	// shape is the number of elements in each dimension
	Shape []uint64

	FileName      string
	OffsetPadding uint64
	FileOffsets   []uint64
}

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

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