config

type AudioConfig struct {
	Backend        string `toml:"backend"`
	DeviceID       string `toml:"device_id"`
	OutputDeviceID string `toml:"output_device_id"`
	SampleRate     int    `toml:"sample_rate"`
	Channels       int    `toml:"channels"`
	FrameSizeMs    int    `toml:"frame_size_ms"`
	LatencyHint    string `toml:"latency_hint"`
}

type Config struct {
	General        GeneralConfig        `toml:"general"`
	Audio          AudioConfig          `toml:"audio"`
	UI             UIConfig             `toml:"ui"`
	Vocabulary     VocabularyConfig     `toml:"vocabulary"`
	Shortcuts      ShortcutsConfig      `toml:"shortcuts"`
	ModelSelection ModelSelectionConfig `toml:"model_selection"`

	// ServerConnection points the device/local-target at a remote SpeechKit
	// Server-Target. Only consulted when at least one mode in ModelSelection
	// has mode_source = "server". Disabled by default; the desktop app runs
	// fully self-contained until a user opts a mode into server-side
	// execution (typically via onboarding or settings).
	ServerConnection ServerConnectionConfig `toml:"server_connection"`

	Local       LocalConfig       `toml:"local"`
	LocalLLM    LocalLLMConfig    `toml:"local_llm"`
	VPS         VPSConfig         `toml:"vps"`
	HuggingFace HuggingFaceConfig `toml:"huggingface"`
	Routing     RoutingConfig     `toml:"routing"`
	Feedback    FeedbackConfig    `toml:"feedback"` // legacy compat; prefer Store
	Store       StoreConfig       `toml:"store"`
	Providers   ProvidersConfig   `toml:"providers"`
	TTS         TTSConfig         `toml:"tts"`
	VoiceAgent  VoiceAgentConfig  `toml:"voice_agent"`

	// Server configures the standalone Linux server binary (cmd/speechkit-server).
	// All fields are optional; the desktop app (cmd/speechkit) ignores them entirely.
	Server    ServerConfig     `toml:"server"`
	Personas  []PersonaConfig  `toml:"personas"`
	Roles     []RoleConfig     `toml:"roles"`
	Sequences []SequenceConfig `toml:"sequences"`
}

type FeedbackConfig struct {
	SaveAudio          bool   `toml:"save_audio"`
	AudioRetentionDays int    `toml:"audio_retention_days"`
	DBPath             string `toml:"db_path"`
	MaxAudioStorageMB  int    `toml:"max_audio_storage_mb"`
}

type GeneralConfig struct {
	Language                 string `toml:"language"`
	Hotkey                   string `toml:"hotkey"` // Deprecated: legacy single-hotkey field kept for config file compat. Use DictateHotkey.
	DictateHotkey            string `toml:"dictate_hotkey"`
	AssistHotkey             string `toml:"assist_hotkey"`
	VoiceAgentHotkey         string `toml:"voice_agent_hotkey"`
	DictateHotkeyBehavior    string `toml:"dictate_hotkey_behavior"`
	AssistHotkeyBehavior     string `toml:"assist_hotkey_behavior"`
	VoiceAgentHotkeyBehavior string `toml:"voice_agent_hotkey_behavior"`
	DictateEnabled           bool   `toml:"dictate_enabled"`
	AssistEnabled            bool   `toml:"assist_enabled"`
	VoiceAgentEnabled        bool   `toml:"voice_agent_enabled"`
	AutoStartOnLaunch        bool   `toml:"auto_start_on_launch"`
	AgentHotkey              string `toml:"agent_hotkey"`
	AgentMode                string `toml:"agent_mode"`  // "assist" or "voice_agent" — determines what agent_hotkey triggers
	ActiveMode               string `toml:"active_mode"` // legacy compat
	HotkeyMode               string `toml:"hotkey_mode"` // legacy compat for single behavior setting
	AutoStopSilenceMs        int    `toml:"auto_stop_silence_ms"`
	FastModeSilenceMs        int    `toml:"fast_mode_silence_ms"` // silence threshold for Quick Capture auto-stop
	ModelDownloadDir         string `toml:"model_download_dir"`   // Default directory for downloaded local model files
}

type GoogleProviderConfig struct {
	Enabled      bool   `toml:"enabled"`
	APIKeyEnv    string `toml:"api_key_env"`
	STTModel     string `toml:"stt_model"`
	UtilityModel string `toml:"utility_model"`
	AssistModel  string `toml:"assist_model"`
	AgentModel   string `toml:"agent_model"`
}

type GroqProviderConfig struct {
	Enabled      bool   `toml:"enabled"`
	APIKeyEnv    string `toml:"api_key_env"`
	STTModel     string `toml:"stt_model"`
	UtilityModel string `toml:"utility_model"`
	AssistModel  string `toml:"assist_model"`
	AgentModel   string `toml:"agent_model"`
}

type HuggingFaceConfig struct {
	Enabled      bool   `toml:"enabled"`
	Model        string `toml:"model"`
	UtilityModel string `toml:"utility_model"`
	AssistModel  string `toml:"assist_model"`
	AgentModel   string `toml:"agent_model"`
	TokenEnv     string `toml:"token_env"`
}

type InstallMode string

type InstallState struct {
	Mode      InstallMode `toml:"mode"`
	SetupDone bool        `toml:"setup_done"`
	DeviceID  string      `toml:"device_id"`
}

type LocalConfig struct {
	Enabled   bool   `toml:"enabled"`
	Model     string `toml:"model"`
	ModelPath string `toml:"model_path"`
	Port      int    `toml:"port"`
	GPU       string `toml:"gpu"`
}

type LocalLLMConfig struct {
	Enabled      bool   `toml:"enabled"`
	BaseURL      string `toml:"base_url"`
	Model        string `toml:"model"`
	ModelPath    string `toml:"model_path"`
	Port         int    `toml:"port"`
	GPU          string `toml:"gpu"`
	UtilityModel string `toml:"utility_model"`
	AssistModel  string `toml:"assist_model"`
	AgentModel   string `toml:"agent_model"`
}

type ModeModelSelection struct {
	PrimaryProfileID  string `toml:"primary_profile_id"`
	FallbackProfileID string `toml:"fallback_profile_id"`

	// ModeSource selects whether this mode runs locally (Framework kernel
	// in-process, default) or against a remote SpeechKit Server-Target
	// configured under [server_connection]. Empty string is treated as
	// ModeSourceLocal so existing configs keep behaving as before.
	ModeSource string `toml:"mode_source"`
}

type ModelSelectionConfig struct {
	Dictate    ModeModelSelection `toml:"dictate"`
	Assist     ModeModelSelection `toml:"assist"`
	VoiceAgent ModeModelSelection `toml:"voice_agent"`
}

type OllamaProviderConfig struct {
	Enabled      bool   `toml:"enabled"`
	BaseURL      string `toml:"base_url"`
	STTModel     string `toml:"stt_model"`
	UtilityModel string `toml:"utility_model"`
	AssistModel  string `toml:"assist_model"`
	AgentModel   string `toml:"agent_model"`
}

type OpenAIProviderConfig struct {
	Enabled       bool   `toml:"enabled"`
	APIKeyEnv     string `toml:"api_key_env"`
	STTModel      string `toml:"stt_model"`
	UtilityModel  string `toml:"utility_model"`
	AssistModel   string `toml:"assist_model"`
	AgentModel    string `toml:"agent_model"`
	TTSModel      string `toml:"tts_model"`
	TTSVoice      string `toml:"tts_voice"`
	RealtimeModel string `toml:"realtime_model"`
}

type OpenRouterProviderConfig struct {
	Enabled      bool   `toml:"enabled"`
	APIKeyEnv    string `toml:"api_key_env"`
	UtilityModel string `toml:"utility_model"`
	AssistModel  string `toml:"assist_model"`
	AgentModel   string `toml:"agent_model"`
}

type OverlayFreePosition struct {
	X int `toml:"x"`
	Y int `toml:"y"`
}

type PersonaConfig struct {
	ID          string            `toml:"id"`
	DisplayName string            `toml:"display_name"`
	Description string            `toml:"description"`
	Voice       string            `toml:"voice"`
	Locale      string            `toml:"locale"`
	DefaultRole string            `toml:"default_role"`
	Tags        []string          `toml:"tags"`
	Metadata    map[string]string `toml:"metadata"`
}

type ProvidersConfig struct {
	OpenAI     OpenAIProviderConfig     `toml:"openai"`
	Groq       GroqProviderConfig       `toml:"groq"`
	Google     GoogleProviderConfig     `toml:"google"`
	Ollama     OllamaProviderConfig     `toml:"ollama"`
	OpenRouter OpenRouterProviderConfig `toml:"openrouter"`
}

type RoleConfig struct {
	ID                          string   `toml:"id"`
	DisplayName                 string   `toml:"display_name"`
	SystemPrompt                string   `toml:"system_prompt"`
	RefinementPrompt            string   `toml:"refinement_prompt"`
	Locale                      string   `toml:"locale"`
	VocabularyHint              string   `toml:"vocabulary_hint"`
	ToolAllowlist               []string `toml:"tool_allowlist"`
	Temperature                 float64  `toml:"temperature"`
	ThinkingEnabled             bool     `toml:"thinking_enabled"`
	ThinkingLevel               string   `toml:"thinking_level"`
	IncludeThoughts             bool     `toml:"include_thoughts"`
	ThinkingBudget              int      `toml:"thinking_budget"`
	AutomaticActivityDetection  bool     `toml:"automatic_activity_detection"`
	VADStartSensitivity         string   `toml:"vad_start_sensitivity"`
	VADEndSensitivity           string   `toml:"vad_end_sensitivity"`
	VADPrefixPaddingMs          int      `toml:"vad_prefix_padding_ms"`
	VADSilenceDurationMs        int      `toml:"vad_silence_duration_ms"`
	ActivityHandling            string   `toml:"activity_handling"`
	TurnCoverage                string   `toml:"turn_coverage"`
	ContextCompressionEnabled   bool     `toml:"context_compression_enabled"`
	ContextCompressionTriggerTk int64    `toml:"context_compression_trigger_tokens"`
	ContextCompressionTargetTk  int64    `toml:"context_compression_target_tokens"`
	EnableAffectiveDialog       bool     `toml:"enable_affective_dialog"`
}

type RoutingConfig struct {
	Strategy                string  `toml:"strategy"`
	PreferLocalUnderSeconds float64 `toml:"prefer_local_under_seconds"`
	ParallelCloud           bool    `toml:"parallel_cloud"`
	ReplaceOnBetter         bool    `toml:"replace_on_better"`
}

type SequenceConfig struct {
	ID          string               `toml:"id"`
	DisplayName string               `toml:"display_name"`
	Description string               `toml:"description"`
	Completion  string               `toml:"completion"` // "all_steps" | "explicit_close" | "max_turns"
	MaxTurns    int                  `toml:"max_turns"`
	Steps       []SequenceStepConfig `toml:"steps"`
}

type SequenceStepConfig struct {
	ID           string   `toml:"id"`
	Instruction  string   `toml:"instruction"`
	ExitCriteria string   `toml:"exit_criteria"`
	RequireTools []string `toml:"require_tools"`
	MaxTurns     int      `toml:"max_turns"`
}

type ServerConfig struct {
	ListenAddr            string   `toml:"listen_addr"`          // e.g. ":8080"
	Modes                 []string `toml:"modes"`                // subset of ["dictation","assist","voiceagent"]; empty = all
	AuthMode              string   `toml:"auth_mode"`            // "bearer" | "edge_hmac" | "bearer_or_edge"
	BearerTokenEnv        string   `toml:"bearer_token_env"`     // env var name holding the bearer token
	EdgeAuthSecretEnv     string   `toml:"edge_auth_secret_env"` // env var name holding the HMAC secret
	CORSAllowedOrigins    []string `toml:"cors_allowed_origins"`
	RateLimitRPS          float64  `toml:"rate_limit_rps"`
	RateLimitBurst        int      `toml:"rate_limit_burst"`
	MaxUploadMB           int      `toml:"max_upload_mb"`
	MaxVoiceAgentSessions int      `toml:"max_voiceagent_sessions"` // global cap
	MaxSessionsPerUser    int      `toml:"max_sessions_per_user"`
	TicketTTLSec          int      `toml:"ticket_ttl_sec"` // Voice Agent WS ticket TTL
	// VoiceAgentIdleTimeoutSec terminates a Voice Agent WebSocket session
	// after N seconds without any client- or provider-side activity.
	// Defaults to 900 (15 min). Set to 0 to disable the server-side idle
	// timeout (kernel-level idle handling stays in effect either way).
	VoiceAgentIdleTimeoutSec int    `toml:"voiceagent_idle_timeout_sec"`
	WhisperBinary            string `toml:"whisper_binary"` // absolute path inside container
	WhisperPort              int    `toml:"whisper_port"`   // loopback port for whisper.cpp server
	ModelDir                 string `toml:"model_dir"`      // persistent volume, e.g. /var/lib/speechkit/models
	LogFormat                string `toml:"log_format"`     // "json" | "text"
	LogLevel                 string `toml:"log_level"`      // "debug" | "info" | "warn" | "error"
}

type ServerConnectionConfig struct {
	// Enabled gates the entire server connection. When false, every mode is
	// forced to run locally regardless of its mode_source. Lets users keep
	// their server URL in config but temporarily flip back to fully local.
	Enabled bool `toml:"enabled"`

	// URL is the base URL of the speechkit-server, e.g.
	// "https://speechkit.example.com" or "http://localhost:8080".
	URL string `toml:"url"`

	// BearerTokenEnv names the env var that holds the bearer token sent in
	// the Authorization header. Defaults to SPEECHKIT_SERVER_TOKEN. The
	// value is never read from the TOML file itself — only the env var name
	// is configured here.
	BearerTokenEnv string `toml:"bearer_token_env"`

	// FallbackToLocal makes the device app fall back to the in-process
	// Framework kernel if a server call fails or the server is unreachable.
	// Useful for laptop deployments that may be offline; should be false
	// for kiosks that must never silently downgrade to local processing.
	FallbackToLocal bool `toml:"fallback_to_local"`

	// RequestTimeoutSec caps non-streaming HTTP calls (Dictation, Assist).
	// 0 means no explicit timeout (the underlying http.Client default
	// applies). Voice Agent WebSocket sessions are not affected.
	RequestTimeoutSec int `toml:"request_timeout_sec"`
}

type ShortcutLocaleConfig struct {
	LeadingFillers []string `toml:"leading_fillers"`
	CopyLast       []string `toml:"copy_last"`
	InsertLast     []string `toml:"insert_last"`
	Summarize      []string `toml:"summarize"`
	QuickNote      []string `toml:"quick_note"`
}

type ShortcutsConfig struct {
	Locale map[string]ShortcutLocaleConfig `toml:"locale"`
}

type StoreConfig struct {
	Backend            string `toml:"backend"` // "sqlite" | "postgres" | registered name
	SQLitePath         string `toml:"sqlite_path"`
	PostgresDSN        string `toml:"postgres_dsn"`
	SaveAudio          bool   `toml:"save_audio"`
	AudioRetentionDays int    `toml:"audio_retention_days"`
	MaxAudioStorageMB  int    `toml:"max_audio_storage_mb"`
}

type TTSConfig struct {
	Enabled     bool           `toml:"enabled"`
	Strategy    string         `toml:"strategy"` // "cloud-first", "local-first", "cloud-only", "local-only"
	Voice       string         `toml:"voice"`    // Global default voice override
	Speed       float64        `toml:"speed"`    // Global speed 0.25-4.0, default 1.0
	Format      string         `toml:"format"`   // "mp3", "wav", "opus", "pcm"
	OpenAI      TTSOpenAI      `toml:"openai"`
	Google      TTSGoogle      `toml:"google"`
	HuggingFace TTSHuggingFace `toml:"huggingface"`
	Local       TTSLocal       `toml:"local"`
}

type TTSGoogle struct {
	Enabled bool   `toml:"enabled"`
	Voice   string `toml:"voice"` // e.g. "de-DE-Neural2-B"
}

type TTSHuggingFace struct {
	Enabled bool   `toml:"enabled"`
	Model   string `toml:"model"` // e.g. "parler-tts/parler-tts-mini-multilingual-v1.1"
}

type TTSLocal struct {
	Enabled   bool   `toml:"enabled"`
	Model     string `toml:"model"`
	ModelPath string `toml:"model_path"`
	Port      int    `toml:"port"`
}

type TTSOpenAI struct {
	Enabled bool   `toml:"enabled"`
	Model   string `toml:"model"` // "tts-1" or "tts-1-hd"
	Voice   string `toml:"voice"` // alloy, echo, fable, onyx, nova, shimmer
}

type UIConfig struct {
	OverlayEnabled          bool                           `toml:"overlay_enabled"`
	OverlayPosition         string                         `toml:"overlay_position"` // "top", "bottom", "left", "right"
	OverlayMovable          bool                           `toml:"overlay_movable"`
	OverlayFreeX            int                            `toml:"overlay_free_x"`
	OverlayFreeY            int                            `toml:"overlay_free_y"`
	OverlayMonitorPositions map[string]OverlayFreePosition `toml:"overlay_monitor_positions"`
	Visualizer              string                         `toml:"visualizer"`
	Design                  string                         `toml:"design"`
	AssistOverlayMode       string                         `toml:"assist_overlay_mode"`
	VoiceAgentOverlayMode   string                         `toml:"voice_agent_overlay_mode"`
}

type VPSConfig struct {
	Enabled   bool   `toml:"enabled"`
	URL       string `toml:"url"`
	APIKeyEnv string `toml:"api_key_env"`
}

type VocabularyConfig struct {
	Dictionary string `toml:"dictionary"`
}

type VoiceAgentConfig struct {
	Enabled bool `toml:"enabled"`
	// Provider selects the backend that drives a Voice Agent session.
	// Supported values:
	//   ""          (default) — same as "gemini"
	//   "gemini"    — Google Gemini Live (cloud, GOOGLE_AI_API_KEY required)
	//   "cascaded"  — self-hosted whisper.cpp → Genkit agent LLM → TTS pipeline
	//                 (CPU-capable; no external realtime dependency)
	//   "moshi"     — self-hosted Kyutai Moshi Rust server (GPU required, M9b)
	//
	// The Server-Target reads this field via cmd/speechkit-server; the Device-
	// Target currently always uses "gemini" and ignores it.
	Provider                        string `toml:"provider"`
	Model                           string `toml:"model"`             // Real-time model ID (e.g. "gemini-3.1-flash-live-preview")
	FallbackModel                   string `toml:"fallback_model"`    // Fallback real-time model
	Voice                           string `toml:"voice"`             // Voice name for real-time model
	FrameworkPrompt                 string `toml:"framework_prompt"`  // Durable host/framework instruction that defines the Voice Agent behavior
	RefinementPrompt                string `toml:"refinement_prompt"` // User-specific refinement appended to the framework prompt
	Instruction                     string `toml:"instruction"`       // Legacy alias for FrameworkPrompt
	AutoStartOnLaunch               bool   `toml:"auto_start_on_launch"`
	CloseBehavior                   string `toml:"close_behavior"` // "continue" keeps the conversation window in the taskbar; "new_chat" ends the current chat on close
	ReminderAfterIdleSec            int    `toml:"reminder_after_idle_sec"`
	DeactivateAfterIdleSec          int    `toml:"deactivate_after_idle_sec"`
	PipelineFallback                bool   `toml:"pipeline_fallback"` // Use STT -> Agent LLM -> optional TTS when the selected Voice Agent profile is not native realtime.
	ShowPrompter                    bool   `toml:"show_prompter"`     // Show live transcript prompter window
	EnableSessionSummary            bool   `toml:"enable_session_summary"`
	EnableInputTranscript           bool   `toml:"enable_input_transcript"`
	EnableOutputTranscript          bool   `toml:"enable_output_transcript"`
	EnableAffectiveDialog           bool   `toml:"enable_affective_dialog"`
	ThinkingEnabled                 bool   `toml:"thinking_enabled"`
	IncludeThoughts                 bool   `toml:"include_thoughts"`
	ThinkingBudget                  int    `toml:"thinking_budget"`
	ThinkingLevel                   string `toml:"thinking_level"`
	ContextCompressionEnabled       bool   `toml:"context_compression_enabled"`
	ContextCompressionTriggerTokens int64  `toml:"context_compression_trigger_tokens"`
	ContextCompressionTargetTokens  int64  `toml:"context_compression_target_tokens"`
	AutomaticActivityDetection      bool   `toml:"automatic_activity_detection"`
	ActivityHandling                string `toml:"activity_handling"`
	TurnCoverage                    string `toml:"turn_coverage"`
	VADStartSensitivity             string `toml:"vad_start_sensitivity"`
	VADEndSensitivity               string `toml:"vad_end_sensitivity"`
	VADPrefixPaddingMs              int    `toml:"vad_prefix_padding_ms"`
	VADSilenceDurationMs            int    `toml:"vad_silence_duration_ms"`
}