//
//  Message.swift
//  Zyquo Local
//
//  Author: Simon-Pierre Boucher
//  Mail: contact@spboucher.ai
//

import Foundation

/// A single chat message. `thinking` holds the extracted `<think>…</think>`
/// content for reasoning models; `stats` is present on assistant messages.
struct Message: Identifiable, Codable, Hashable, Sendable {
    enum Role: String, Codable, Sendable {
        case system
        case user
        case assistant
    }

    var id: UUID
    var role: Role
    var content: String
    var thinking: String?
    var stats: MessageStats?
    var date: Date

    init(
        id: UUID = UUID(),
        role: Role,
        content: String,
        thinking: String? = nil,
        stats: MessageStats? = nil,
        date: Date = Date()
    ) {
        self.id = id
        self.role = role
        self.content = content
        self.thinking = thinking
        self.stats = stats
        self.date = date
    }
}

/// Per-response performance statistics (first-class per the spec).
struct MessageStats: Codable, Hashable, Sendable {
    /// Seconds from send to the first streamed token.
    var timeToFirstToken: TimeInterval
    /// Generation-phase tokens per second.
    var tokensPerSecond: Double
    var promptTokenCount: Int
    var generationTokenCount: Int
    /// Peak MLX GPU memory during the response, in bytes.
    var peakMemoryBytes: Int
}