loadtest

type Metrics struct {
	TotalCalls     atomic.Int64
	SuccessCount   atomic.Int64
	ErrorCount     atomic.Int64
	TotalLatencyNs atomic.Int64 // sum of all call latencies in nanoseconds
	MinLatencyNs   atomic.Int64
	MaxLatencyNs   atomic.Int64
}

type MetricsSnapshot struct {
	TotalCalls   int64
	SuccessCount int64
	ErrorCount   int64
	AvgLatency   time.Duration
	MinLatency   time.Duration
	MaxLatency   time.Duration
}

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

type ProviderConfig struct {
	// Name is the provider name returned by Name().
	Name string

	// Model is the model identifier returned by Model().
	Model string

	// BaseLatency is the minimum latency per Chat call.
	// Simulates LLM time-to-first-token.
	BaseLatency time.Duration

	// LatencyJitter adds random jitter up to this duration on top of BaseLatency.
	// Total latency per call = BaseLatency + rand(0, LatencyJitter).
	LatencyJitter time.Duration

	// ErrorRate is the probability [0.0, 1.0] that a Chat call returns an error.
	// Simulates 429s, timeouts, and transient failures.
	ErrorRate float64

	// ErrorMessage is the error returned when error injection fires.
	// Defaults to "loadtest: simulated provider error" if empty.
	ErrorMessage string

	// Response is the canned response returned on success.
	// If nil, a default response is generated.
	Response *types.LLMResponse

	// ResponseFunc generates a dynamic response based on the input messages.
	// Takes precedence over Response if set.
	ResponseFunc func(messages []types.Message, tools []shuttle.Tool) *types.LLMResponse

	// StreamChunkSize is the number of characters per streaming chunk.
	// Only used when the provider is accessed via ChatStream.
	// Defaults to 10.
	StreamChunkSize int

	// StreamChunkDelay is the delay between streaming chunks.
	// Defaults to 5ms.
	StreamChunkDelay time.Duration

	// InputTokensPerMessage is the simulated input token count per message.
	// Defaults to 50.
	InputTokensPerMessage int

	// OutputTokens is the simulated output token count.
	// Defaults to 100.
	OutputTokens int

	// CostPerCall is the simulated cost in USD per call.
	// Defaults to 0.001.
	CostPerCall float64
}