ollama

README

agent-api ollama-provider

The Ollama provider for the agent-api

---

👷🏗️ The Ollama provider is a work in progress and the API may change unexpectedly.

Documentation

Index

• type OllamaProvider
  • func NewOllamaProvider(logger logr.Logger, baseURL string, port int, model string) *OllamaProvider
  • func (p *OllamaProvider) EstimateTokens(ctx context.Context, message string) (int, error)
  • func (p *OllamaProvider) GenerateResponse(ctx context.Context, messages []message.Message) (*message.Message, error)
  • func (p *OllamaProvider) GenerateStream(ctx context.Context, messages []message.Message, opts *core.InferenceOptions) (<-chan *message.Message, <-chan error)
  • func (p *OllamaProvider) GenerateStreamWithTools(ctx context.Context, messages []message.Message, tools []tool.Tool, ...) (<-chan *message.Message, <-chan error)
  • func (p *OllamaProvider) GenerateWithTools(ctx context.Context, messages []message.Message, tools []tool.Tool) (*message.Message, error)
  • func (p *OllamaProvider) GetCapabilities(ctx context.Context) (*core.ProviderCapabilities, error)
  • func (p *OllamaProvider) GetModelList(ctx context.Context) ([]string, error)
  • func (p *OllamaProvider) ValidatePrompt(ctx context.Context, messages []message.Message) error

Types

type OllamaProvider

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

OllamaProvider implements the LLMProvider interface for Ollama

func NewOllamaProvider

func NewOllamaProvider(logger logr.Logger, baseURL string, port int, model string) *OllamaProvider

NewOllamaProvider creates a new Ollama provider

TODO: - need to handle base URL better (with trailing slashes, etc.) - need to construct actual URL using baseURL, port, etc.

func (*OllamaProvider) EstimateTokens

func (p *OllamaProvider) EstimateTokens(ctx context.Context, message string) (int, error)

EstimateTokens estimates the number of tokens in a message

func (*OllamaProvider) GenerateResponse

func (p *OllamaProvider) GenerateResponse(ctx context.Context, messages []message.Message) (*message.Message, error)

GenerateResponse implements the LLMProvider interface for basic responses

func (*OllamaProvider) GenerateStream

func (p *OllamaProvider) GenerateStream(ctx context.Context, messages []message.Message, opts *core.InferenceOptions) (<-chan *message.Message, <-chan error)

GenerateStream streams the response token by token

func (*OllamaProvider) GenerateStreamWithTools

func (p *OllamaProvider) GenerateStreamWithTools(ctx context.Context, messages []message.Message, tools []tool.Tool, opts *core.InferenceOptions) (<-chan *message.Message, <-chan error)

GenerateStreamWithTools streams a response with tools token by token

func (*OllamaProvider) GenerateWithTools

func (p *OllamaProvider) GenerateWithTools(ctx context.Context, messages []message.Message, tools []tool.Tool) (*message.Message, error)

GenerateWithTools implements the LLMProvider interface for tool-using responses

TODO: - handle automatically generating a system message descriptor

func (*OllamaProvider) GetCapabilities

func (p *OllamaProvider) GetCapabilities(ctx context.Context) (*core.ProviderCapabilities, error)

func (*OllamaProvider) GetModelList

func (p *OllamaProvider) GetModelList(ctx context.Context) ([]string, error)

GetModelList returns available models for this provider

func (*OllamaProvider) ValidatePrompt

func (p *OllamaProvider) ValidatePrompt(ctx context.Context, messages []message.Message) error

ValidatePrompt checks if a prompt is valid for the provider