model

type Chunk struct {
	// Delta is the incremental content in this chunk.
	Delta Part `json:"delta"`
	// Usage is the cumulative usage (only set in the final chunk; nil otherwise).
	Usage *Usage `json:"usage,omitempty"`
}

type Message struct {
	// Role indicates who sent this message.
	Role Role `json:"role"`
	// Content is the list of content parts in this message.
	Content []Part `json:"content"`
}

type Part struct {
	// Kind indicates the type of content in this part.
	Kind PartKind `json:"kind"`
	// Text contains plain text content (used with PartKindText).
	Text string `json:"text,omitempty"`
	// ImageURL contains a URL to an image (used with PartKindImage).
	ImageURL string `json:"image_url,omitempty"`
	// ImageData contains base64-encoded image data (used with PartKindImage).
	ImageData string `json:"image_data,omitempty"`
	// ImageMIME contains the MIME type of the image (used with PartKindImage).
	ImageMIME string `json:"image_mime,omitempty"`
	// ToolUseID is the unique identifier of a tool invocation (used with PartKindToolUse).
	ToolUseID string `json:"tool_use_id,omitempty"`
	// ToolName is the name of the tool being invoked (used with PartKindToolUse).
	ToolName string `json:"tool_name,omitempty"`
	// ToolArgs contains the JSON arguments for the tool invocation (used with PartKindToolUse).
	ToolArgs json.RawMessage `json:"tool_args,omitempty"`
	// ToolResultID is the ID of the tool invocation this result corresponds to (used with PartKindToolResult).
	ToolResultID string `json:"tool_result_id,omitempty"`
	// ToolResultContent contains the result content from the tool (used with PartKindToolResult).
	ToolResultContent string `json:"tool_result_content,omitempty"`
	// IsError indicates whether this tool result is an error (used with PartKindToolResult).
	IsError bool `json:"is_error,omitempty"`
}

type PartKind string

type PricePerMillionTokens struct {
	// Input is the cost per million input tokens.
	Input float64
	// Output is the cost per million output tokens.
	Output float64
	// CacheRead is the cost per million cache-read tokens.
	CacheRead float64
}

type Request struct {
	// Model is the model identifier (e.g., "claude-3-5-sonnet-20241022", "gpt-4o").
	Model string `json:"model"`
	// System is optional system/instruction context prepended to the messages.
	System string `json:"system,omitempty"`
	// Messages is the conversation history.
	Messages []Message `json:"messages"`
	// Tools is the list of tools the model can invoke.
	Tools []ToolSchema `json:"tools,omitempty"`
	// Temperature controls randomness in model outputs (0 to 1, where 1 is max randomness).
	Temperature *float64 `json:"temperature,omitempty"`
	// MaxTokens is the maximum number of tokens to generate in the response.
	MaxTokens int `json:"max_tokens,omitempty"`
	// StopSeqs are sequences that will cause generation to stop early.
	StopSeqs []string `json:"stop_seqs,omitempty"`
}

type Response struct {
	// Message is the model's response message.
	Message Message `json:"message"`
	// StopReason indicates why generation stopped.
	StopReason StopReason `json:"stop_reason"`
	// Usage is the token usage and cost information.
	Usage Usage `json:"usage"`
}

type Role string

type StopReason string

type Table map[string]PricePerMillionTokens

type ToolSchema struct {
	// Name is the unique identifier of the tool.
	Name string `json:"name"`
	// Description explains what the tool does.
	Description string `json:"description"`
	// InputSchema is the JSON schema describing the tool's input parameters.
	InputSchema json.RawMessage `json:"input_schema"`
}

type Usage struct {
	// InputTokens is the number of input tokens processed.
	InputTokens int `json:"input_tokens"`
	// OutputTokens is the number of output tokens generated.
	OutputTokens int `json:"output_tokens"`
	// CacheReadTokens is the number of tokens read from cache (omitted if zero).
	CacheReadTokens int `json:"cache_read_tokens,omitempty"`
	// CostUSD is the cost of the request in USD (omitted if zero).
	CostUSD float64 `json:"cost_usd,omitempty"`
}