enact-config 0.0.2

//! Configuration Types - Core configuration structure
//!
//! This module defines the configuration structure that matches the TypeScript schema.
//! It's kept in sync with packages/enact-schemas/src/config.schemas.ts

use anyhow::{Context, Result};
use serde::{Deserialize, Serialize};
use std::fmt;
use std::path::Path;
use std::str::FromStr;

/// Runtime mode
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Default)]
pub enum RuntimeMode {
    #[default]
    #[serde(rename = "local")]
    Local,
    #[serde(rename = "airgapped")]
    AirGapped,
    #[serde(rename = "cloud")]
    Cloud,
}

impl FromStr for RuntimeMode {
    type Err = std::convert::Infallible;

    fn from_str(value: &str) -> Result<Self, Self::Err> {
        Ok(match value.to_ascii_lowercase().as_str() {
            "airgapped" => RuntimeMode::AirGapped,
            "cloud" => RuntimeMode::Cloud,
            _ => RuntimeMode::Local,
        })
    }
}

impl fmt::Display for RuntimeMode {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        match self {
            RuntimeMode::Local => write!(f, "local"),
            RuntimeMode::AirGapped => write!(f, "airgapped"),
            RuntimeMode::Cloud => write!(f, "cloud"),
        }
    }
}

/// Provider configuration
#[derive(Debug, Clone, Serialize, Deserialize, Default, PartialEq)]
pub struct Providers {
    pub azure: Option<AzureProvider>,
    pub anthropic: Option<AnthropicProvider>,
    pub openai: Option<OpenAIProvider>,
    pub ollama: Option<OllamaProvider>,
    pub google: Option<GoogleProvider>,
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct AzureProvider {
    pub endpoint: Option<String>,
    pub api_key: Option<String>,
    pub deployment_name: Option<String>,
    #[serde(default = "default_api_version")]
    pub api_version: String,
}

fn default_api_version() -> String {
    "2024-02-15-preview".to_string()
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct AnthropicProvider {
    pub api_key: Option<String>,
    #[serde(default = "default_anthropic_base_url")]
    pub base_url: String,
}

fn default_anthropic_base_url() -> String {
    "https://api.anthropic.com".to_string()
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct OpenAIProvider {
    pub api_key: Option<String>,
    #[serde(default = "default_openai_base_url")]
    pub base_url: String,
    pub organization: Option<String>,
}

fn default_openai_base_url() -> String {
    "https://api.openai.com/v1".to_string()
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct OllamaProvider {
    #[serde(default = "default_ollama_base_url")]
    pub base_url: String,
}

fn default_ollama_base_url() -> String {
    "http://localhost:11434".to_string()
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct GoogleProvider {
    pub api_key: Option<String>,
    #[serde(default = "default_google_base_url")]
    pub base_url: String,
}

fn default_google_base_url() -> String {
    "https://generativelanguage.googleapis.com/v1".to_string()
}

/// Runtime configuration
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct Runtime {
    #[serde(default)]
    pub mode: RuntimeMode,
    #[serde(default = "default_max_concurrent")]
    pub max_concurrent_executions: u32,
    #[serde(default = "default_timeout")]
    pub default_timeout: u64,
    #[serde(default = "default_true")]
    pub enable_telemetry: bool,
    #[serde(default = "default_true")]
    pub allow_network: bool,
}

fn default_max_concurrent() -> u32 {
    10
}

fn default_timeout() -> u64 {
    30000
}

fn default_true() -> bool {
    true
}

impl Default for Runtime {
    fn default() -> Self {
        Self {
            mode: RuntimeMode::Local,
            max_concurrent_executions: 10,
            default_timeout: 30000,
            enable_telemetry: true,
            allow_network: true,
        }
    }
}

/// Storage configuration
#[derive(Debug, Clone, Serialize, Deserialize, Default, PartialEq)]
pub struct Storage {
    #[serde(default = "default_event_store")]
    pub event_store: EventStore,
    #[serde(default = "default_state_store")]
    pub state_store: StateStore,
    #[serde(default = "default_filesystem_store")]
    pub artifact_store: ArtifactStore,
    #[serde(default = "default_sqlite_vector_store")]
    pub vector_store: VectorStore,
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct EventStore {
    #[serde(default = "default_sqlite")]
    pub r#type: String,
    pub path: Option<String>,
    pub dsn: Option<String>,
}

fn default_sqlite() -> String {
    "sqlite".to_string()
}

impl Default for EventStore {
    fn default() -> Self {
        Self {
            r#type: "jsonl".to_string(),
            path: Some("events".to_string()),
            dsn: None,
        }
    }
}

fn default_event_store() -> EventStore {
    EventStore::default()
}

impl Default for StateStore {
    fn default() -> Self {
        Self {
            r#type: "jsonl".to_string(),
            path: Some("state".to_string()),
            dsn: None,
        }
    }
}

fn default_state_store() -> StateStore {
    StateStore::default()
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct StateStore {
    #[serde(default = "default_sqlite")]
    pub r#type: String,
    pub path: Option<String>,
    pub dsn: Option<String>,
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct ArtifactStore {
    #[serde(default = "default_filesystem")]
    pub r#type: String,
    pub path: Option<String>,
    #[serde(default = "default_zstd")]
    pub compression: String,
}

fn default_filesystem() -> String {
    "filesystem".to_string()
}

fn default_zstd() -> String {
    "zstd".to_string()
}

impl Default for ArtifactStore {
    fn default() -> Self {
        Self {
            r#type: "filesystem".to_string(),
            path: None,
            compression: "zstd".to_string(),
        }
    }
}

fn default_filesystem_store() -> ArtifactStore {
    ArtifactStore::default()
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct VectorStore {
    #[serde(default = "default_sqlite")]
    pub r#type: String,
    pub url: Option<String>,
    pub collection: Option<String>,
    pub path: Option<String>,
    pub dsn: Option<String>,
}

impl Default for VectorStore {
    fn default() -> Self {
        Self {
            r#type: "sqlite".to_string(),
            url: None,
            collection: None,
            path: None,
            dsn: None,
        }
    }
}

fn default_sqlite_vector_store() -> VectorStore {
    VectorStore::default()
}

/// Tools configuration
#[derive(Debug, Clone, Serialize, Deserialize, Default, PartialEq)]
pub struct Tools {
    #[serde(default)]
    pub ingestion: IngestionTools,
}

#[derive(Debug, Clone, Serialize, Deserialize, Default, PartialEq)]
pub struct IngestionTools {
    #[serde(default = "default_pdf")]
    pub pdf: PdfIngestion,
    #[serde(default = "default_ocr")]
    pub ocr: OcrIngestion,
    #[serde(default = "default_embeddings")]
    pub embeddings: EmbeddingsIngestion,
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct PdfIngestion {
    #[serde(default = "default_pdfium")]
    pub engine: String,
}

impl Default for PdfIngestion {
    fn default() -> Self {
        Self {
            engine: "pdfium".to_string(),
        }
    }
}

fn default_pdfium() -> String {
    "pdfium".to_string()
}

fn default_pdf() -> PdfIngestion {
    PdfIngestion {
        engine: "pdfium".to_string(),
    }
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct OcrIngestion {
    #[serde(default = "default_tesseract")]
    pub engine: String,
    #[serde(default = "default_languages")]
    pub languages: Vec<String>,
}

impl Default for OcrIngestion {
    fn default() -> Self {
        Self {
            engine: "tesseract".to_string(),
            languages: vec!["eng".to_string()],
        }
    }
}

fn default_tesseract() -> String {
    "tesseract".to_string()
}

fn default_languages() -> Vec<String> {
    vec!["eng".to_string()]
}

fn default_ocr() -> OcrIngestion {
    OcrIngestion {
        engine: "tesseract".to_string(),
        languages: vec!["eng".to_string()],
    }
}

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct EmbeddingsIngestion {
    #[serde(default = "default_fastembed")]
    pub engine: String,
    pub model: Option<String>,
}

impl Default for EmbeddingsIngestion {
    fn default() -> Self {
        Self {
            engine: "fastembed".to_string(),
            model: None,
        }
    }
}

fn default_fastembed() -> String {
    "fastembed".to_string()
}

fn default_embeddings() -> EmbeddingsIngestion {
    EmbeddingsIngestion {
        engine: "fastembed".to_string(),
        model: None,
    }
}

/// Cloud configuration
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct Cloud {
    pub api_url: Option<String>,
    pub tenant_id: Option<String>,
    #[serde(default)]
    pub auto_sync: bool,
}

/// Human-in-the-loop approval configuration
///
/// Configures when and how tool calls require human approval before execution.
///
/// ## Policies
///
/// - `always_approve` - Auto-approve all tool calls (default)
/// - `always_deny` - Block all tool calls
/// - `ask` or `always_require` - Prompt user for every tool call
/// - `ask_once` - Prompt once per tool, remember decisions for session
/// - `pattern` - Only prompt for tools matching `require_patterns` regexes
///
/// ## Example YAML
///
/// ```yaml
/// approval:
///   enabled: true
///   policy: pattern
///   require_patterns:
///     - "Edit|Write|Bash"   # Prompt for file-modifying tools
///     - "mcp__.*"           # Prompt for all MCP tools
///   timeout_seconds: 300
///   tool_overrides:
///     Read: always_approve  # Auto-approve Read tool
///     Glob: always_approve  # Auto-approve Glob tool
/// ```
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct ApprovalConfig {
    /// Whether approval is enabled
    #[serde(default)]
    pub enabled: bool,

    /// Policy type: "always_approve", "always_deny", "ask", "ask_once", "pattern"
    #[serde(default = "default_approval_policy")]
    pub policy: String,

    /// Maximum steps before requiring approval (for threshold policy)
    pub max_steps: Option<usize>,

    /// Patterns that require approval (for pattern policy)
    /// e.g., ["Edit|Write|Bash", "mcp__.*"]
    pub require_patterns: Option<Vec<String>>,

    /// Timeout for waiting for approval (in seconds)
    #[serde(default = "default_approval_timeout")]
    pub timeout_seconds: u64,

    /// Per-tool policy overrides (tool_name -> policy)
    /// e.g., {"Read": "always_approve", "Bash": "ask"}
    #[serde(default)]
    pub tool_overrides: Option<std::collections::HashMap<String, String>>,
}

fn default_approval_policy() -> String {
    "always_approve".to_string()
}

fn default_approval_timeout() -> u64 {
    300 // 5 minutes
}

impl Default for ApprovalConfig {
    fn default() -> Self {
        Self {
            enabled: false,
            policy: "always_approve".to_string(),
            max_steps: None,
            require_patterns: None,
            timeout_seconds: 300,
            tool_overrides: None,
        }
    }
}

/// Episodic memory configuration
///
/// Configures daily logs, session snapshots, and consolidation rules
/// for episodic (short-term) memory storage.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct MemoryConfig {
    /// Memory backend: "sqlite", "markdown", "none"
    #[serde(default = "default_memory_backend")]
    pub backend: String,

    /// Directory for daily log files (relative to workspace)
    #[serde(default = "default_daily_logs_dir")]
    pub daily_logs_dir: String,

    /// Maximum entries per daily log before rolling
    #[serde(default = "default_max_daily_entries")]
    pub max_daily_entries: usize,

    /// Whether to automatically consolidate to semantic memory
    #[serde(default)]
    pub auto_consolidate: bool,

    /// Time of day to run consolidation (HH:MM format)
    pub consolidation_time: Option<String>,

    /// Number of days to retain episodic logs (None = forever)
    pub retention_days: Option<u32>,

    /// Whether to include timestamps in entries
    #[serde(default = "default_true")]
    pub include_timestamps: bool,
}

fn default_memory_backend() -> String {
    "markdown".to_string()
}

fn default_daily_logs_dir() -> String {
    "memory".to_string()
}

fn default_max_daily_entries() -> usize {
    100
}

impl Default for MemoryConfig {
    fn default() -> Self {
        Self {
            backend: "markdown".to_string(),
            daily_logs_dir: "memory".to_string(),
            max_daily_entries: 100,
            auto_consolidate: true,
            consolidation_time: Some("03:00".to_string()),
            retention_days: Some(30),
            include_timestamps: true,
        }
    }
}

/// Session/conversation management configuration
///
/// Controls context rotation, idle timeouts, and cleanup behavior.
/// Can be overridden at agent or channel level using `SessionConfigOverride`.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct SessionConfig {
    /// Maximum conversation turns before context rotation.
    #[serde(default = "default_max_turns")]
    pub max_turns: usize,

    /// Rotate context when estimated token usage exceeds this percentage of model max.
    #[serde(default = "default_rotation_threshold_pct")]
    pub rotation_threshold_pct: u8,

    /// Rotate context after this many seconds of inactivity.
    #[serde(default = "default_idle_timeout_secs")]
    pub idle_timeout_secs: u64,

    /// Interval for cleanup task to remove stale conversations (seconds).
    #[serde(default = "default_cleanup_interval_secs")]
    pub cleanup_interval_secs: u64,

    /// Remove conversations idle longer than this (seconds).
    #[serde(default = "default_cleanup_idle_threshold_secs")]
    pub cleanup_idle_threshold_secs: u64,
}

fn default_max_turns() -> usize {
    20
}

fn default_rotation_threshold_pct() -> u8 {
    80
}

fn default_idle_timeout_secs() -> u64 {
    1800 // 30 minutes
}

fn default_cleanup_interval_secs() -> u64 {
    300 // 5 minutes
}

fn default_cleanup_idle_threshold_secs() -> u64 {
    3600 // 1 hour
}

impl Default for SessionConfig {
    fn default() -> Self {
        Self {
            max_turns: default_max_turns(),
            rotation_threshold_pct: default_rotation_threshold_pct(),
            idle_timeout_secs: default_idle_timeout_secs(),
            cleanup_interval_secs: default_cleanup_interval_secs(),
            cleanup_idle_threshold_secs: default_cleanup_idle_threshold_secs(),
        }
    }
}

/// Optional session config overrides for agent.yaml or channels.yaml.
///
/// Fields set to `None` inherit from the parent configuration level.
/// Override hierarchy: Channel > Agent > Global > Defaults
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Default)]
pub struct SessionConfigOverride {
    pub max_turns: Option<usize>,
    pub rotation_threshold_pct: Option<u8>,
    pub idle_timeout_secs: Option<u64>,
    pub cleanup_interval_secs: Option<u64>,
    pub cleanup_idle_threshold_secs: Option<u64>,
}

impl SessionConfigOverride {
    /// Apply this override onto a base SessionConfig.
    /// Fields that are `None` inherit from the base.
    pub fn apply_to(&self, base: &SessionConfig) -> SessionConfig {
        SessionConfig {
            max_turns: self.max_turns.unwrap_or(base.max_turns),
            rotation_threshold_pct: self
                .rotation_threshold_pct
                .unwrap_or(base.rotation_threshold_pct),
            idle_timeout_secs: self.idle_timeout_secs.unwrap_or(base.idle_timeout_secs),
            cleanup_interval_secs: self
                .cleanup_interval_secs
                .unwrap_or(base.cleanup_interval_secs),
            cleanup_idle_threshold_secs: self
                .cleanup_idle_threshold_secs
                .unwrap_or(base.cleanup_idle_threshold_secs),
        }
    }
}

/// HTTP/gRPC server configuration
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct ServerConfig {
    /// Port for the HTTP gateway (default: 8080)
    #[serde(default = "default_server_port")]
    pub port: u16,
    /// Host address to bind to (default: "0.0.0.0")
    #[serde(default = "default_server_host")]
    pub host: String,
    /// gRPC port (default: 50051)
    #[serde(default = "default_grpc_port")]
    pub grpc_port: u16,
    /// Documentation server port (default: 1111)
    #[serde(default = "default_docs_port")]
    pub docs_port: u16,
    /// Base URL for the docs server (e.g. "http://127.0.0.1:1111"). When unset, derived from docs_port.
    /// Used to rewrite asset URLs in HTML so CSS/JS load correctly when serving pre-built docs locally.
    #[serde(default)]
    pub docs_base_url: Option<String>,
}

fn default_server_port() -> u16 {
    8080
}

fn default_server_host() -> String {
    "0.0.0.0".to_string()
}

fn default_grpc_port() -> u16 {
    50051
}

fn default_docs_port() -> u16 {
    1111
}

impl Default for ServerConfig {
    fn default() -> Self {
        Self {
            port: default_server_port(),
            host: default_server_host(),
            grpc_port: default_grpc_port(),
            docs_port: default_docs_port(),
            docs_base_url: None,
        }
    }
}

/// Logging paths (relative to ENACT_HOME)
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct LoggingConfig {
    /// Path relative to ENACT_HOME for daemon log file
    #[serde(default = "default_daemon_log")]
    pub daemon_log: String,
    /// Path relative to ENACT_HOME for serve log file
    #[serde(default = "default_serve_log")]
    pub serve_log: String,
}

/// Observability configuration
///
/// Controls tracing, logging, and metrics for LLM calls, token usage,
/// memory access, guardrails, and context windows.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct ObservabilityConfig {
    /// Enable LLM call tracing (start/end/failed events)
    #[serde(default = "default_true")]
    pub trace_llm_calls: bool,

    /// Log full prompts (expensive - use for debugging only)
    #[serde(default)]
    pub log_full_prompts: bool,

    /// Log full responses (expensive - use for debugging only)
    #[serde(default)]
    pub log_full_responses: bool,

    /// Track token usage and emit token.usage events
    #[serde(default = "default_true")]
    pub track_token_usage: bool,

    /// Trace memory access (recall/store events)
    #[serde(default = "default_true")]
    pub trace_memory_access: bool,

    /// Enable context window snapshots
    #[serde(default)]
    pub enable_context_snapshots: bool,

    /// Enable reasoning trace capture
    #[serde(default)]
    pub capture_reasoning_traces: bool,

    /// Maximum content length for logged prompts/responses (truncation limit)
    #[serde(default = "default_max_content_length")]
    pub max_content_length: usize,
}

fn default_max_content_length() -> usize {
    1000
}

impl Default for ObservabilityConfig {
    fn default() -> Self {
        Self {
            trace_llm_calls: true,
            log_full_prompts: false,
            log_full_responses: false,
            track_token_usage: true,
            trace_memory_access: true,
            enable_context_snapshots: false,
            capture_reasoning_traces: false,
            max_content_length: 1000,
        }
    }
}

fn default_daemon_log() -> String {
    "logs/daemon.log".to_string()
}

fn default_serve_log() -> String {
    "logs/serve.log".to_string()
}

impl Default for LoggingConfig {
    fn default() -> Self {
        Self {
            daemon_log: default_daemon_log(),
            serve_log: default_serve_log(),
        }
    }
}

/// Main configuration structure
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Default)]
#[serde(deny_unknown_fields)]
pub struct Config {
    /// Default model id for run/serve (e.g. azure_foundry:text:gpt-4-1-mini). Read from ~/.enact/config.yaml.
    #[serde(default)]
    pub default_model_id: Option<String>,
    #[serde(default)]
    pub providers: Providers,
    #[serde(default)]
    pub runtime: Runtime,
    #[serde(default)]
    pub storage: Storage,
    #[serde(default)]
    pub tools: Tools,
    pub cloud: Option<Cloud>,
    /// HTTP/gRPC server configuration
    #[serde(default)]
    pub server: ServerConfig,
    /// Human-in-the-loop approval configuration
    #[serde(default)]
    pub approval: ApprovalConfig,
    /// Memory configuration
    #[serde(default)]
    pub memory: MemoryConfig,
    /// Session/conversation management configuration
    #[serde(default)]
    pub session: SessionConfig,
    /// Log file paths (relative to ENACT_HOME)
    #[serde(default)]
    pub logging: LoggingConfig,
    /// Observability configuration (LLM tracing, token usage, memory access, etc.)
    #[serde(default)]
    pub observability: ObservabilityConfig,
}

impl Config {
    /// Load global config from ENACT_HOME/config.yaml.
    /// If the file does not exist, returns default config.
    pub fn load_from_home() -> Result<Self> {
        let path = crate::home::enact_home().join("config.yaml");
        Self::load_from_yaml_path(&path)
    }

    /// Write default config to `path` if the file does not exist.
    /// Used by install/doctor/serve to ensure config.yaml is present.
    pub fn ensure_default_at(path: &Path) -> Result<()> {
        if path.exists() {
            return Ok(());
        }
        if let Some(parent) = path.parent() {
            std::fs::create_dir_all(parent).context("Failed to create config directory")?;
        }
        Self::default().save_to_yaml_path(path)
    }

    /// Save global config to ENACT_HOME/config.yaml.
    /// Creates a rolling backup of all config files before writing.
    pub fn save_to_home(&self) -> Result<()> {
        crate::home::create_config_backup()?;
        let path = crate::home::enact_home().join("config.yaml");
        self.save_to_yaml_path(&path)
    }

    /// Load config from a YAML file path.
    pub fn load_from_yaml_path(path: &Path) -> Result<Self> {
        if !path.exists() {
            return Ok(Self::default());
        }
        let s = std::fs::read_to_string(path).context("Failed to read config file")?;
        let config: Config = serde_yaml::from_str(&s).context("Failed to parse config YAML")?;
        Ok(config)
    }

    /// Save config to a YAML file path.
    pub fn save_to_yaml_path(&self, path: &Path) -> Result<()> {
        if let Some(parent) = path.parent() {
            std::fs::create_dir_all(parent).context("Failed to create config directory")?;
        }
        let s = serde_yaml::to_string(self).context("Failed to serialize config to YAML")?;
        std::fs::write(path, s).context("Failed to write config file")?;
        Ok(())
    }
}