llms

type CallOption func(*CallOptions)

type CallOptions struct {
	// Model specifies the model to use (overrides default).
	Model string `json:"model"`
	// Temperature controls randomness in generation (0.0 to 2.0).
	Temperature float64 `json:"temperature"`
	// MaxTokens limits the maximum tokens in the response.
	MaxTokens int `json:"max_tokens"`
	// StopWords specifies sequences where generation should stop.
	StopWords []string `json:"stop_words"`
	// TopP controls diversity via nucleus sampling (0.0 to 1.0).
	TopP float64 `json:"top_p"`
	// TopK limits sampling to top K tokens.
	TopK int `json:"top_k"`
	// Seed sets a deterministic seed for reproducible outputs.
	Seed int `json:"seed"`
	// Metadata contains additional provider-specific options.
	Metadata map[string]any `json:"metadata,omitempty"`
	// StreamingFunc is called for each chunk when streaming is enabled.
	StreamingFunc func(ctx context.Context, chunk []byte) error `json:"-"`
}

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

type LLMRerankerOption func(*LLMReranker)

type Model interface {
	// GenerateContent generates a response from the LLM given a conversation history.
	// Use this for multi-turn conversations or when you need access to full response metadata.
	GenerateContent(ctx context.Context, messages []schema.MessageContent, options ...CallOption) (*schema.ContentResponse, error)
	// Call is a convenience method for single-turn prompts.
	// It returns the generated text directly.
	Call(ctx context.Context, prompt string, options ...CallOption) (string, error)
}

type Tokenizer interface {
	// CountTokens returns the number of tokens in the text.
	CountTokens(ctx context.Context, text string) (int, error)
}