llmrepo

package
v0.0.53 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Aug 20, 2025 License: Apache-2.0 Imports: 8 Imported by: 0

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func NewModelManager added in v0.0.53

func NewModelManager(runtime *runtimestate.State, tokenizer ollamatokenizer.Tokenizer, config ModelManagerConfig) (*modelManager, error)

Types

type EmbedRequest added in v0.0.53

type EmbedRequest struct {
	ModelName    string
	ProviderType string
	Tracker      activitytracker.ActivityTracker // Now exported
}

type ModelConfig added in v0.0.53

type ModelConfig struct {
	Name     string
	Provider string
}

type ModelManagerConfig added in v0.0.53

type ModelManagerConfig struct {
	DefaultPromptModel    ModelConfig
	DefaultEmbeddingModel ModelConfig
	DefaultChatModel      ModelConfig
}

type ModelRepo

type ModelRepo interface {
	GetTokenizer(ctx context.Context, modelName string) (Tokenizer, error)
	PromptExecute(
		ctx context.Context,
		req Request,
	) (libmodelprovider.LLMPromptExecClient, error)
	Chat(
		ctx context.Context,
		req Request,
	) (libmodelprovider.LLMChatClient, string, error)
	Embed(
		ctx context.Context,
		embedReq EmbedRequest,
	) (libmodelprovider.LLMEmbedClient, error)
	Stream(
		ctx context.Context,
		req Request,
	) (libmodelprovider.LLMStreamClient, error)
}

type Request added in v0.0.53

type Request struct {
	ProviderTypes []string // Optional: if empty, uses all default providers
	ModelNames    []string // Optional: if empty, any model is considered
	ContextLength int      // Minimum required context length
	Tracker       activitytracker.ActivityTracker
}

Unified Request type for all operations

type Tokenizer

type Tokenizer interface {
	Tokenize(ctx context.Context, prompt string) ([]int, error)
	CountTokens(ctx context.Context, prompt string) (int, error)
}

Directories

Path Synopsis
internal

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL