package ai

import (
	"context"
	"time"

	"ai-operator/internal/media"
)

type VoiceProvider interface {
	StartSession(ctx context.Context, config VoiceSessionConfig) error
	SendAudio(ctx context.Context, chunk media.AudioChunk) error
	SendToolResult(ctx context.Context, result ToolResult) error
	Close(ctx context.Context) error
	Events() <-chan VoiceEvent
	Stats() VoiceProviderStats
}

type VoiceSessionConfig struct {
	CallID            string
	ProviderSessionID string
	Language          string
	RegionCode        string
	SystemPrompt      string
	InputAudioFormat  string
	OutputAudioFormat string
	InputSampleRate   int
	OutputSampleRate  int
	Metadata          map[string]string
}

type VoiceEventType string

const (
	VoiceEventSessionStarted           VoiceEventType = "session.started"
	VoiceEventSessionUpdated           VoiceEventType = "session.updated"
	VoiceEventUserTranscriptDelta      VoiceEventType = "user.transcript.delta"
	VoiceEventUserTranscriptDone       VoiceEventType = "user.transcript.done"
	VoiceEventAssistantTranscriptDelta VoiceEventType = "assistant.transcript.delta"
	VoiceEventAssistantTranscriptDone  VoiceEventType = "assistant.transcript.done"
	VoiceEventAssistantAudioDelta      VoiceEventType = "assistant.audio.delta"
	VoiceEventAssistantAudioDone       VoiceEventType = "assistant.audio.done"
	VoiceEventToolCall                 VoiceEventType = "tool.call"
	VoiceEventInterruption             VoiceEventType = "interruption"
	VoiceEventRateLimitsUpdated        VoiceEventType = "rate_limits.updated"
	VoiceEventError                    VoiceEventType = "error"
	VoiceEventClosed                   VoiceEventType = "closed"
)

type VoiceEvent struct {
	Type     VoiceEventType
	CallID   string
	Text     string
	Audio    []byte
	ToolCall *ToolCall
	Error    string
	Metadata map[string]any
	At       time.Time
}

type ToolCall struct {
	ID           string
	Name         string
	Arguments    map[string]any
	RawArguments string
}
type ToolResult struct {
	CallID     string
	ToolCallID string
	Result     any
	Error      string
}

type VoiceProviderStats struct {
	InputAudioFrames  int64
	InputAudioBytes   int64
	OutputAudioFrames int64
	OutputAudioBytes  int64
	EventsReceived    int64
	EventsSent        int64
	ToolCallsReceived int64
	Errors            int64
	StartedAt         *time.Time
	ClosedAt          *time.Time
}