Documentation
¶
Overview ¶
Package ai provides a unified interface for working with multiple LLM providers. It offers a simple, opinionated abstraction layer designed for tool-building workflows.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
Types ¶
type CacheConfig ¶
type CacheConfig struct {
// CacheSystemPrompt enables caching of the system prompt.
// This is beneficial when reusing the same system prompt across multiple requests.
CacheSystemPrompt bool
// TTL is the cache time-to-live ("5m" or "1h").
// Defaults to "5m" (5 minutes).
TTL string
}
CacheConfig controls prompt caching behavior.
type Client ¶
type Client interface {
// Complete sends a completion request and returns the response.
Complete(ctx context.Context, req CompletionRequest) (*CompletionResponse, error)
// Provider returns the provider name (e.g., "anthropic", "openai").
Provider() string
}
Client represents an LLM provider client. Implementations include Anthropic Claude, OpenAI, and others.
type CompletionRequest ¶
type CompletionRequest struct {
// SystemPrompt defines the AI's role and behavior.
// This is typically cached when the same prompt is reused.
SystemPrompt string
// Messages is the conversation history.
Messages []Message
// MaxTokens is the maximum number of tokens to generate.
MaxTokens int
// Temperature controls randomness (0.0 = deterministic, 1.0 = creative).
Temperature float64
// Model is an optional model override.
// If empty, the client's default model is used.
Model string
// CacheConfig controls prompt caching behavior.
CacheConfig *CacheConfig
}
CompletionRequest is a provider-agnostic completion request.
type CompletionResponse ¶
type CompletionResponse struct {
// Content is the generated text response.
Content string
// StopReason indicates why generation stopped (e.g., "end_turn", "max_tokens").
StopReason string
// InputTokens is the number of input tokens processed.
InputTokens int
// OutputTokens is the number of output tokens generated.
OutputTokens int
// CacheReadTokens is the number of tokens read from cache (if caching enabled).
CacheReadTokens int
// CacheCreationTokens is the number of tokens written to cache (if caching enabled).
CacheCreationTokens int
}
CompletionResponse is a provider-agnostic completion response.
type Error ¶
type Error struct {
// Type categorizes the error.
Type ErrorType
// Message provides a human-readable error description.
Message string
// Status is the HTTP status code (if applicable).
Status int
// Err is the underlying error (if any).
Err error
}
Error represents an AI-related error with type information.
func NewAuthError ¶
NewAuthError creates an authentication error.
func NewInvalidRequestError ¶
NewInvalidRequestError creates an invalid request error.
func NewNetworkError ¶
NewNetworkError creates a network error.
func NewRateLimitError ¶
NewRateLimitError creates a rate limit error.
func NewServerError ¶
NewServerError creates a server error.
type ErrorType ¶
type ErrorType string
ErrorType categorizes AI-related errors.
const ( // ErrTypeAuth indicates authentication failure (invalid API key, etc.). ErrTypeAuth ErrorType = "auth" // ErrTypeRateLimit indicates rate limiting by the provider. ErrTypeRateLimit ErrorType = "rate_limit" // ErrTypeInvalidReq indicates an invalid request (malformed, missing fields). ErrTypeInvalidReq ErrorType = "invalid_request" // ErrTypeServer indicates a server error from the provider. ErrTypeServer ErrorType = "server_error" // ErrTypeNetwork indicates a network or connection error. ErrTypeNetwork ErrorType = "network" )
type Message ¶
type Message struct {
// Role is the message role (user or assistant).
Role MessageRole
// Content is the message text.
Content string
}
Message represents a single message in a conversation.
type MessageRole ¶
type MessageRole string
MessageRole identifies the speaker of a message.
const ( // RoleUser represents a user message. RoleUser MessageRole = "user" // RoleAssistant represents an assistant message. RoleAssistant MessageRole = "assistant" )
Source Files
Directories