gemini

Documentation

Overview

Package gemini implements engine.Engine for Google Gemini multi-modal understanding.

Gemini supports text generation with optional image/video inputs for analysis. Auth: API key as query param ?key={api_key}, env GEMINI_API_KEY or GOOGLE_API_KEY.

Endpoint: POST /models/{model}:generateContent Default model: gemini-2.0-flash

Index

• Constants
• Variables
• func ConfigSchema() []engine.ConfigField
• func ModelsByCapability() map[string][]string
• type Config
• type Engine
  • func New(cfg Config) *Engine
  • func (e *Engine) Capabilities() engine.Capability
  • func (e *Engine) Execute(ctx context.Context, g workflow.Graph) (engine.Result, error)

Constants

const (
	ModelGemini20Flash     = "gemini-2.0-flash"
	ModelGemini15Pro       = "gemini-1.5-pro"
	ModelGemini20FlashLite = "gemini-2.0-flash-lite"
	ModelGemini15Flash     = "gemini-1.5-flash"
)

Model constants.

Variables

var (
	ErrMissingAPIKey = errors.New("gemini: missing API key (set Config.APIKey or GEMINI_API_KEY / GOOGLE_API_KEY)")
	ErrMissingPrompt = errors.New("gemini: missing prompt in workflow graph")
)

Functions

func ConfigSchema

func ConfigSchema() []engine.ConfigField

ConfigSchema returns the configuration fields for the Gemini engine.

func ModelsByCapability

func ModelsByCapability() map[string][]string

ModelsByCapability returns known Gemini models grouped by capability.

Types

type Config

type Config struct {
	APIKey     string
	BaseURL    string // default: https://generativelanguage.googleapis.com/v1beta
	Model      string // default: gemini-2.0-flash
	HTTPClient *http.Client
}

Config configures the Gemini engine.

type Engine

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

Engine implements engine.Engine for Google Gemini multi-modal understanding.

func New

func New(cfg Config) *Engine

New creates a Gemini engine instance.

func (*Engine) Capabilities

func (e *Engine) Capabilities() engine.Capability

Capabilities implements engine.Describer.

func (*Engine) Execute

func (e *Engine) Execute(ctx context.Context, g workflow.Graph) (engine.Result, error)

Execute sends a multi-modal request to the Gemini API.