Files
Termix/src/backend/database/db/schema.pg.ts
T
LukeGus e372b0eda7 feat: remove legacy alert rules, superseded by automations
Alert rules already auto-migrate into automations on boot and the old
engine stands down after. Deletes the rule/firing tables, routes, UI,
and AI tools; keeps notification channels since automations still use
them. Also fixes automations' notify step, which still called the
deleted repository.
2026-09-22 10:45:12 -05:00

2247 lines
78 KiB
TypeScript

// GENERATED FILE — do not edit.
//
// Produced from schema.ts by scripts/generate-dialect-schema.cjs.
// Edit the sqlite schema and re-run `node scripts/generate-dialect-schema.cjs`.
// Target dialect: postgres.
//
// DDL source for drizzle-kit. NOT imported to run queries — repositories use
// schema.ts on every dialect. See the generator header for why that is correct.
import {
pgTable,
text,
varchar,
integer,
serial,
boolean,
doublePrecision,
index,
uniqueIndex,
foreignKey,
type AnyPgColumn,
} from "drizzle-orm/pg-core";
import { sql } from "drizzle-orm";
export const users = pgTable("users", {
id: varchar("id", { length: 255 }).primaryKey(),
username: text("username").notNull(),
passwordHash: text("password_hash").notNull(),
isAdmin: boolean("is_admin").notNull().default(false),
isOidc: boolean("is_oidc").notNull().default(false),
oidcIdentifier: text("oidc_identifier"),
ssoProviderId: integer("sso_provider_id"),
clientId: text("client_id"),
clientSecret: text("client_secret"),
issuerUrl: text("issuer_url"),
authorizationUrl: text("authorization_url"),
tokenUrl: text("token_url"),
identifierPath: text("identifier_path"),
namePath: text("name_path"),
scopes: text().default("openid email profile"),
totpSecret: text("totp_secret"),
totpEnabled: boolean("totp_enabled")
.notNull()
.default(false),
totpBackupCodes: text("totp_backup_codes"),
registeredAt: text("registered_at").notNull().default(sql`CURRENT_TIMESTAMP`),
donationModalDismissed: boolean("donation_modal_dismissed")
.notNull()
.default(false),
});
export const settings = pgTable("settings", {
key: varchar("key", { length: 255 }).primaryKey(),
value: text("value").notNull(),
});
export const ssoProviders = pgTable("sso_providers", {
id: serial("id").primaryKey(),
name: varchar("name", { length: 255 }).notNull(),
type: text("type").notNull(),
enabled: boolean("enabled").notNull().default(true),
displayOrder: integer("display_order").notNull().default(0),
config: text("config").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const sessions = pgTable(
"sessions",
{
id: varchar("id", { length: 255 }).primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
jwtToken: text("jwt_token").notNull(),
deviceType: text("device_type").notNull(),
deviceInfo: text("device_info").notNull(),
oidcSub: text("oidc_sub"),
oidcSid: text("oidc_sid"),
ssoProviderId: integer("sso_provider_id"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
expiresAt: varchar("expires_at", { length: 255 }).notNull(),
lastActiveAt: text("last_active_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Listing a user's devices, and the startup sweep of expired rows.
(table) => [
index("idx_sessions_user_id").on(table.userId),
index("idx_sessions_expires_at").on(table.expiresAt),
],
);
export const trustedDevices = pgTable(
"trusted_devices",
{
id: varchar("id", { length: 255 }).primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
deviceFingerprint: text("device_fingerprint").notNull(),
deviceType: text("device_type").notNull(),
deviceInfo: text("device_info").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
expiresAt: varchar("expires_at", { length: 255 }).notNull(),
lastUsedAt: text("last_used_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_trusted_devices_user_id").on(table.userId)],
);
export const webauthnCredentials = pgTable("webauthn_credentials", {
id: varchar("id", { length: 255 }).primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
credentialId: varchar("credential_id", { length: 255 }).notNull(),
publicKey: text("public_key").notNull(),
counter: integer("counter").notNull().default(0),
deviceType: text("device_type"),
backedUp: boolean("backed_up").notNull().default(false),
transports: text("transports"),
userVerification: text("user_verification").notNull().default("preferred"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
lastUsedAt: text("last_used_at"),
});
export const hosts = pgTable(
"ssh_data",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
connectionType: text("connection_type").notNull().default("ssh"),
name: varchar("name", { length: 255 }),
ip: text("ip").notNull(),
port: integer("port").notNull(),
username: text("username").notNull(),
folder: varchar("folder", { length: 255 }),
// Sub-host nesting: a host acting as an organizational parent for other
// hosts, mutually exclusive with folder (see host route validation).
parentHostId: integer("parent_host_id").references(
(): AnyPgColumn => hosts.id,
{ onDelete: "set null" },
),
tags: text("tags"),
pin: boolean("pin").notNull().default(false),
// Manual drag-to-reorder position within a folder. Null means the host has
// never been manually reordered; falls back to name sort in that case.
sortOrder: integer("sort_order"),
authType: text("auth_type").notNull(),
useWarpgate: boolean("use_warpgate").notNull().default(false),
shareSshAuth: boolean("share_ssh_auth")
.notNull()
.default(false),
forceKeyboardInteractive: text("force_keyboard_interactive"),
password: text("password"),
key: text("key"),
keyPassword: text("key_password"),
keyType: text("key_type"),
sudoPassword: text("sudo_password"),
autostartPassword: text("autostart_password"),
autostartKey: text("autostart_key"),
autostartKeyPassword: text("autostart_key_password"),
credentialId: integer("credential_id").references(() => sshCredentials.id, { onDelete: "set null" }),
overrideCredentialUsername: boolean("override_credential_username"),
// When authType is "vault", the host authenticates via a Vault SSH signer
// profile (shared settings, no secrets). The signing certificate is obtained
// per-user at connect time via an interactive Vault OIDC flow.
vaultProfileId: integer("vault_profile_id").references(
() => vaultProfiles.id,
{ onDelete: "set null" },
),
enableTerminal: boolean("enable_terminal")
.notNull()
.default(true),
enableSessionLogging: boolean("enable_session_logging")
.notNull()
.default(true),
allowSessionSharing: boolean("allow_session_sharing")
.notNull()
.default(true),
enableCommandHistory: boolean("enable_command_history")
.notNull()
.default(true),
enableTunnel: boolean("enable_tunnel")
.notNull()
.default(true),
tunnelConnections: text("tunnel_connections"),
jumpHosts: text("jump_hosts"),
enableFileManager: boolean("enable_file_manager")
.notNull()
.default(true),
scpLegacy: boolean("scp_legacy").notNull().default(false),
enableDocker: boolean("enable_docker")
.notNull()
.default(false),
enableWebUi: boolean("enable_web_ui")
.notNull()
.default(false),
enableTmuxMonitor: boolean("enable_tmux_monitor")
.notNull()
.default(false),
enableTerminalToolbar: boolean("enable_terminal_toolbar")
.notNull()
.default(true),
enableAiAssistant: boolean("enable_ai_assistant")
.notNull()
.default(false),
showTerminalInSidebar: boolean("show_terminal_in_sidebar")
.notNull()
.default(true),
showFileManagerInSidebar: boolean("show_file_manager_in_sidebar")
.notNull()
.default(false),
showTunnelInSidebar: boolean("show_tunnel_in_sidebar")
.notNull()
.default(false),
showDockerInSidebar: boolean("show_docker_in_sidebar")
.notNull()
.default(false),
showServerStatsInSidebar: boolean("show_server_stats_in_sidebar")
.notNull()
.default(false),
defaultPath: text("default_path"),
statsConfig: text("stats_config"),
dockerConfig: text("docker_config"),
webUiConfig: text("web_ui_config"),
enableProxmox: boolean("enable_proxmox")
.notNull()
.default(false),
proxmoxConfig: text("proxmox_config"),
enableProxmoxStats: boolean("enable_proxmox_stats")
.notNull()
.default(false),
proxmoxStatsConfig: text("proxmox_stats_config"),
terminalConfig: text("terminal_config"),
quickActions: text("quick_actions"),
notes: text("notes"),
enableSsh: boolean("enable_ssh").notNull().default(true),
enableRdp: boolean("enable_rdp").notNull().default(false),
enableVnc: boolean("enable_vnc").notNull().default(false),
enableTelnet: boolean("enable_telnet").notNull().default(false),
sshPort: integer("ssh_port").default(22),
rdpPort: integer("rdp_port").default(3389),
vncPort: integer("vnc_port").default(5900),
telnetPort: integer("telnet_port").default(23),
rdpCredentialId: integer("rdp_credential_id").references(() => sshCredentials.id, { onDelete: "set null" }),
rdpUser: text("rdp_user"),
rdpPassword: text("rdp_password"),
rdpDomain: text("rdp_domain"),
rdpSecurity: text("rdp_security"),
rdpIgnoreCert: boolean("rdp_ignore_cert").default(false),
vncCredentialId: integer("vnc_credential_id").references(() => sshCredentials.id, { onDelete: "set null" }),
vncPassword: text("vnc_password"),
vncUser: text("vnc_user"),
telnetUser: text("telnet_user"),
telnetPassword: text("telnet_password"),
telnetCredentialId: integer("telnet_credential_id").references(() => sshCredentials.id, { onDelete: "set null" }),
rdpAuthType: text("rdp_auth_type"),
vncAuthType: text("vnc_auth_type"),
telnetAuthType: text("telnet_auth_type"),
domain: text("domain"),
security: text("security"),
ignoreCert: boolean("ignore_cert").default(false),
guacamoleConfig: text("guacamole_config"),
useSocks5: boolean("use_socks5"),
socks5Host: text("socks5_host"),
socks5Port: integer("socks5_port"),
socks5Username: text("socks5_username"),
socks5Password: text("socks5_password"),
socks5ProxyChain: text("socks5_proxy_chain"),
// null = use the desktop app's global default; "local" | "remote" pins
// this specific host's SSH/Docker-console/Serial connections to originate
// from the embedded local backend or a connected remote sync server.
// Ignored for rdp/vnc/telnet, which always require the remote server.
connectionOrigin: text("connection_origin"),
macAddress: text("mac_address"),
wolBroadcastAddress: text("wol_broadcast_address"),
portKnockSequence: text("port_knock_sequence"),
hostKeyFingerprint: text("host_key_fingerprint"),
hostKeyType: text("host_key_type"),
hostKeyAlgorithm: text("host_key_algorithm").default("sha256"),
hostKeyFirstSeen: text("host_key_first_seen"),
hostKeyLastVerified: text("host_key_last_verified"),
hostKeyChangedCount: integer("host_key_changed_count").default(0),
// Stable identity used to match this row across two independently-seeded
// databases (the embedded backend and a connected remote server) during
// sync -- local autoincrement ids collide across instances.
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Every host read is scoped by owner, so user_id carries the host list.
//
// `folder` is deliberately not indexed: on Postgres/MySQL an indexed text
// column is generated as varchar(255), and folder holds a joined nested path
// with no length cap, so indexing it would truncate deep hierarchies.
(table) => [
index("idx_ssh_data_user_id").on(table.userId),
index("idx_ssh_data_parent_host").on(table.parentHostId),
index("idx_ssh_data_credential").on(table.credentialId),
],
);
export const fileManagerRecent = pgTable(
"file_manager_recent",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
path: text("path").notNull(),
lastOpened: text("last_opened")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Every file manager surface is read for one user on one host at a time.
(table) => [
index("idx_file_manager_recent_user").on(table.userId, table.hostId),
],
);
export const fileManagerPinned = pgTable(
"file_manager_pinned",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
path: text("path").notNull(),
pinnedAt: text("pinned_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_file_manager_pinned_user").on(table.userId, table.hostId),
],
);
export const fileManagerShortcuts = pgTable(
"file_manager_shortcuts",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
path: text("path").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_file_manager_shortcuts_user").on(table.userId, table.hostId),
],
);
export const transferRecent = pgTable(
"transfer_recent",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
sourceHostId: integer("source_host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
destHostId: integer("dest_host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
destPath: text("dest_path").notNull(),
destPathLabel: text("dest_path_label").notNull(),
lastUsed: text("last_used")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_transfer_recent_user").on(table.userId)],
);
export const dismissedAlerts = pgTable(
"dismissed_alerts",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
alertId: text("alert_id").notNull(),
dismissedAt: text("dismissed_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_dismissed_alerts_user_id").on(table.userId)],
);
export const sshCredentials = pgTable(
"ssh_credentials",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
description: text("description"),
folder: varchar("folder", { length: 255 }),
tags: text("tags"),
pin: boolean("pin").notNull().default(false),
// Manual drag-to-reorder position within a folder. Null means the
// credential has never been manually reordered; falls back to name sort
// in that case, same convention as hosts.sortOrder.
sortOrder: integer("sort_order"),
authType: text("auth_type").notNull(),
username: text("username"),
password: text("password"),
key: text("key"),
privateKey: text("private_key"),
publicKey: text("public_key"),
keyPassword: text("key_password"),
keyType: text("key_type"),
detectedKeyType: text("detected_key_type"),
certPublicKey: text("cert_public_key"),
usageCount: integer("usage_count").notNull().default(0),
lastUsed: text("last_used"),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_ssh_credentials_user_id").on(table.userId)],
);
export const sshCredentialUsage = pgTable(
"ssh_credential_usage",
{
id: serial("id").primaryKey(),
credentialId: integer("credential_id")
.notNull()
.references(() => sshCredentials.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
usedAt: text("used_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_ssh_credential_usage_credential").on(table.credentialId),
index("idx_ssh_credential_usage_user").on(table.userId),
],
);
export const snippets = pgTable(
"snippets",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
content: text("content").notNull(),
description: text("description"),
folder: varchar("folder", { length: 255 }),
order: integer("order").notNull().default(0),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
hostFilter: text("host_filter"),
isNote: boolean("is_note").notNull().default(false),
},
(table) => [index("idx_snippets_user_id").on(table.userId)],
);
export const snippetFolders = pgTable("snippet_folders", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
color: text("color"),
icon: text("icon"),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const c2sTunnelPresets = pgTable("c2s_tunnel_presets", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
config: text("config").notNull(),
platform: text("platform"),
computerName: text("computer_name"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const snippetAccess = pgTable(
"snippet_access",
{
id: serial("id").primaryKey(),
snippetId: integer("snippet_id")
.notNull()
.references(() => snippets.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 }).references(() => users.id, { onDelete: "cascade" }),
roleId: integer("role_id").references(() => roles.id, {
onDelete: "cascade",
}),
grantedBy: varchar("granted_by", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
permissionLevel: text("permission_level").notNull().default("view"),
expiresAt: varchar("expires_at", { length: 255 }),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Same three lookup shapes as host_access: by grantee, by role, by snippet.
(table) => [
index("idx_snippet_access_user_id").on(table.userId),
index("idx_snippet_access_snippet_id").on(table.snippetId),
index("idx_snippet_access_role_id").on(table.roleId),
],
);
export const sshFolders = pgTable(
"ssh_folders",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
color: text("color"),
icon: text("icon"),
credentialId: integer("credential_id").references(() => sshCredentials.id, {
onDelete: "set null",
}),
// Manual drag-to-reorder position among sibling folders. Null falls back
// to name sort, same convention as hosts.sortOrder.
sortOrder: integer("sort_order"),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_ssh_folders_user_id").on(table.userId)],
);
export const recentActivity = pgTable(
"recent_activity",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
type: text("type").notNull(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
hostName: text("host_name"),
timestamp: varchar("timestamp", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Always read newest-first for one user, so timestamp follows user_id.
(table) => [
index("idx_recent_activity_user_ts").on(table.userId, table.timestamp),
],
);
export const commandHistory = pgTable(
"command_history",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
command: text("command").notNull(),
executedAt: text("executed_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_command_history_user_host").on(table.userId, table.hostId),
],
);
export const networkTopology = pgTable("network_topology", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
topology: text("topology"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const hostAccess = pgTable(
"host_access",
{
id: serial("id").primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.references(() => users.id, { onDelete: "cascade" }),
roleId: integer("role_id")
.references(() => roles.id, { onDelete: "cascade" }),
grantedBy: varchar("granted_by", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
permissionLevel: text("permission_level")
.notNull()
.default("connect"),
expiresAt: varchar("expires_at", { length: 255 }),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
lastAccessedAt: text("last_accessed_at"),
accessCount: integer("access_count").notNull().default(0),
},
// Resolved on every host list request and every permission check, so all
// three lookup shapes (by grantee, by role, by host) need to be indexed.
(table) => [
index("idx_host_access_user_id").on(table.userId),
index("idx_host_access_role_id").on(table.roleId),
index("idx_host_access_host_id").on(table.hostId),
index("idx_host_access_expires_at").on(table.expiresAt),
],
);
export const sharedHostAuthOverrides = pgTable(
"shared_host_auth_overrides",
{
id: serial("id").primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
protocol: varchar("protocol", { length: 255 }).notNull().default("ssh"),
credentialId: integer("credential_id")
.notNull()
.references(() => sshCredentials.id, { onDelete: "cascade" }),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("shared_host_auth_overrides_host_user_protocol_unique").on(
table.hostId,
table.userId,
table.protocol,
),
],
);
export const sharedHostSecrets = pgTable(
"shared_host_secrets",
{
id: serial("id").primaryKey(),
hostAccessId: integer("host_access_id")
.notNull()
.references(() => hostAccess.id, { onDelete: "cascade" }),
targetUserId: varchar("target_user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
protocol: varchar("protocol", { length: 255 }).notNull().default("ssh"),
sourceType: text("source_type").notNull().default("credential"),
originalCredentialId: integer("original_credential_id").references(
() => sshCredentials.id,
{ onDelete: "cascade" },
),
encryptedUsername: text("encrypted_username"),
encryptedAuthType: text("encrypted_auth_type"),
encryptedPassword: text("encrypted_password"),
encryptedKey: text("encrypted_key"),
encryptedKeyPassword: text("encrypted_key_password"),
encryptedKeyType: text("encrypted_key_type"),
encryptedDomain: text("encrypted_domain"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Declared inline in the production DDL as UNIQUE(...), but never here,
// so the generated Postgres and MySQL schemas allowed duplicates the
// SQLite deployment forbids — and the upsert had nothing to conflict on.
(table) => [
uniqueIndex("idx_shared_host_secrets_scope").on(
table.hostAccessId,
table.targetUserId,
table.protocol,
),
],
);
export const roles = pgTable("roles", {
id: serial("id").primaryKey(),
name: varchar("name", { length: 255 }).notNull().unique(),
displayName: text("display_name").notNull(),
description: text("description"),
isSystem: boolean("is_system")
.notNull()
.default(false),
permissions: text("permissions"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const userRoles = pgTable(
"user_roles",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
roleId: integer("role_id")
.notNull()
.references(() => roles.id, { onDelete: "cascade" }),
grantedBy: varchar("granted_by", { length: 255 }).references(() => users.id, {
onDelete: "set null",
}),
grantedAt: text("granted_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Declared inline in the production DDL as UNIQUE(...), but never here,
// so the generated Postgres and MySQL schemas allowed duplicates the
// SQLite deployment forbids — and the upsert had nothing to conflict on.
//
// The unique pair already serves lookups by user, since user_id leads it.
// Listing a role's members starts from role_id, which it cannot serve.
(table) => [
uniqueIndex("idx_user_roles_user_role").on(table.userId, table.roleId),
index("idx_user_roles_role_id").on(table.roleId),
],
);
export const auditLogs = pgTable(
"audit_logs",
{
id: serial("id").primaryKey(),
// Nullable on purpose: the trail outlives the account, and username keeps the
// entry attributable once the reference is gone.
userId: varchar("user_id", { length: 255 }).references(() => users.id, { onDelete: "set null" }),
username: text("username").notNull(),
action: varchar("action", { length: 255 }).notNull(),
resourceType: varchar("resource_type", { length: 255 }).notNull(),
resourceId: text("resource_id"),
resourceName: text("resource_name"),
details: text("details"),
ipAddress: text("ip_address"),
userAgent: text("user_agent"),
success: boolean("success").notNull(),
errorMessage: text("error_message"),
timestamp: varchar("timestamp", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// This table only grows, and is always read newest-first with an optional
// filter. Each composite leads with the filtered column so the same index
// also satisfies the ORDER BY.
(table) => [
index("idx_audit_logs_timestamp").on(table.timestamp),
index("idx_audit_logs_user_ts").on(table.userId, table.timestamp),
index("idx_audit_logs_action_ts").on(table.action, table.timestamp),
index("idx_audit_logs_resource_ts").on(table.resourceType, table.timestamp),
],
);
export const sessionRecordings = pgTable(
"session_recordings",
{
id: serial("id").primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
// Nullable on purpose: a recording is evidence about the host as much as the
// person, so it outlives the account. username keeps it attributable.
userId: varchar("user_id", { length: 255 }).references(() => users.id, { onDelete: "set null" }),
username: text("username"),
accessId: integer("access_id").references(() => hostAccess.id, {
onDelete: "set null",
}),
startedAt: varchar("started_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
endedAt: text("ended_at"),
duration: integer("duration"),
commands: text("commands"),
dangerousActions: text("dangerous_actions"),
recordingPath: text("recording_path"),
protocol: varchar("protocol", { length: 255 }).notNull().default("ssh"),
format: text("format").notNull().default("text"),
terminatedByOwner: boolean("terminated_by_owner").default(false),
terminationReason: text("termination_reason"),
},
// Listed newest-first per user, and audited per host.
(table) => [
index("idx_session_recordings_user_started").on(
table.userId,
table.startedAt,
),
index("idx_session_recordings_host").on(table.hostId),
],
);
export const sessionShares = pgTable(
"session_shares",
{
id: varchar("id", { length: 255 }).primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
ownerUserId: varchar("owner_user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
protocol: varchar("protocol", { length: 255 }).notNull(),
// Live-session binding: TerminalSessionManager's session.id for SSH, or
// guacd's own guacamoleConnectionId for rdp/vnc/telnet. Neither is a DB
// row (process-local, in-memory) so this intentionally has no FK.
sessionId: varchar("session_id", { length: 255 }).notNull(),
tabInstanceId: text("tab_instance_id"),
shareType: text("share_type").notNull(), // "link" | "user"
targetUserId: varchar("target_user_id", { length: 255 }).references(() => users.id, {
onDelete: "cascade",
}),
linkToken: varchar("link_token", { length: 255 }).unique(),
permissionLevel: text("permission_level").notNull().default("read-only"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
expiresAt: varchar("expires_at", { length: 255 }).notNull(),
revokedAt: text("revoked_at"),
lastJoinedAt: text("last_joined_at"),
joinCount: integer("join_count").notNull().default(0),
},
// Resolved from the live session on join, and listed per host.
(table) => [
index("idx_session_shares_session_id").on(table.sessionId),
index("idx_session_shares_host_id").on(table.hostId),
],
);
export const sessionShareParticipants = pgTable(
"session_share_participants",
{
id: serial("id").primaryKey(),
shareId: varchar("share_id", { length: 255 })
.notNull()
.references(() => sessionShares.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 }).references(() => users.id, {
onDelete: "cascade",
}),
guestLabel: text("guest_label"),
joinedAt: text("joined_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
leftAt: text("left_at"),
},
);
export const opksshTokens = pgTable(
"opkssh_tokens",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
sshCert: text("ssh_cert").notNull(),
privateKey: text("private_key").notNull(),
email: text("email"),
sub: text("sub"),
issuer: text("issuer"),
audience: text("audience"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
expiresAt: varchar("expires_at", { length: 255 }).notNull(),
lastUsed: text("last_used"),
},
// Declared inline in the production DDL as UNIQUE(...), but never here,
// so the generated Postgres and MySQL schemas allowed duplicates the
// SQLite deployment forbids — and the upsert had nothing to conflict on.
(table) => [uniqueIndex("idx_opkssh_tokens_user_host").on(table.userId, table.hostId)],
);
// Vault SSH signer profiles. These hold ONLY non-secret connection settings and
// are intended to be shared across users (shared === true makes a profile
// visible to every user on the server). Each user authenticates to Vault via an
// interactive OIDC flow at connect time; no tokens or keys are stored here.
export const vaultProfiles = pgTable("vault_profiles", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
description: text("description"),
folder: varchar("folder", { length: 255 }),
tags: text("tags"),
// Vault server connection (non-secret)
vaultAddr: text("vault_addr").notNull(),
vaultNamespace: text("vault_namespace"),
// OIDC auth method mount + role used to obtain a Vault token interactively
oidcMount: text("oidc_mount"),
oidcRole: text("oidc_role"),
// SSH secrets engine mount + signer role used to sign the ephemeral key
sshMount: text("ssh_mount"),
sshRole: text("ssh_role").notNull(),
validPrincipals: text("valid_principals"),
// Ephemeral keypair algorithm to generate per connection
keyType: text("key_type"),
// When true the profile is visible/usable by all users on the server
shared: boolean("shared").notNull().default(false),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// Per-user cache of the ephemeral SSH private key + Vault-signed certificate.
// Transient: rows live only until the certificate expires. Secret fields are
// encrypted under the user's data-encryption key (see field-crypto.ts).
export const vaultTokens = pgTable(
"vault_tokens",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
profileId: integer("profile_id")
.notNull()
.references(() => vaultProfiles.id, { onDelete: "cascade" }),
sshCert: text("ssh_cert").notNull(),
privateKey: text("private_key").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
expiresAt: varchar("expires_at", { length: 255 }).notNull(),
lastUsed: text("last_used"),
},
// Declared inline in the production DDL as UNIQUE(...), but never here,
// so the generated Postgres and MySQL schemas allowed duplicates the
// SQLite deployment forbids — and the upsert had nothing to conflict on.
(table) => [uniqueIndex("idx_vault_tokens_user_profile").on(table.userId, table.profileId)],
);
export const apiKeys = pgTable(
"api_keys",
{
id: varchar("id", { length: 255 }).primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
tokenHash: text("token_hash").notNull(),
tokenPrefix: text("token_prefix").notNull(),
createdAt: varchar("created_at", { length: 255 }).notNull().default(sql`CURRENT_TIMESTAMP`),
expiresAt: varchar("expires_at", { length: 255 }),
lastUsedAt: text("last_used_at"),
isActive: boolean("is_active").notNull().default(true),
},
(table) => [index("idx_api_keys_user_id").on(table.userId)],
);
export const userOpenTabs = pgTable(
"user_open_tabs",
{
id: varchar("id", { length: 255 }).primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
tabType: text("tab_type").notNull(),
hostId: integer("host_id").references(() => hosts.id, {
onDelete: "cascade",
}),
label: varchar("label", { length: 255 }).notNull(),
tabOrder: integer("tab_order").notNull().default(0),
backendSessionId: text("backend_session_id"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_user_open_tabs_user_id").on(table.userId)],
);
export const userPreferences = pgTable("user_preferences", {
userId: varchar("user_id", { length: 255 })
.primaryKey()
.references(() => users.id, { onDelete: "cascade" }),
reopenTabsOnLogin: boolean("reopen_tabs_on_login")
.notNull()
.default(false),
theme: text("theme"),
fontSize: text("font_size"),
accentColor: text("accent_color"),
language: text("language"),
storageMode: text("storage_mode"),
commandAutocomplete: boolean("command_autocomplete"),
commandPaletteEnabled: boolean("command_palette_enabled"),
showHostTags: boolean("show_host_tags"),
hostTrayOnClick: boolean("host_tray_on_click"),
pinAppRail: boolean("pin_app_rail"),
expandAppRailOnHover: boolean("expand_app_rail_on_hover"),
showPinAppRailButton: boolean("show_pin_app_rail_button"),
foldersCollapsed: boolean("folders_collapsed"),
confirmSnippetExecution: boolean("confirm_snippet_execution"),
disableUpdateCheck: boolean("disable_update_check"),
confirmTabClose: boolean("confirm_tab_close"),
hiddenRailTabs: text("hidden_rail_tabs"),
// null means the user has not been asked yet; the assistant stays hidden
// until this is explicitly true and the admin global is on.
aiAssistantEnabled: boolean("ai_assistant_enabled"),
// Opt-in to letting the assistant run allowlisted read-only diagnostics
// without a per-command approval click.
aiReadOnlyCommands: boolean("ai_read_only_commands"),
compactHostView: boolean("compact_host_view"),
statusColorScheme: text("status_color_scheme"),
customThemes: text("custom_themes"),
customKeybindings: text("custom_keybindings"),
terminalDefaults: text("terminal_defaults"),
rdpDefaults: text("rdp_defaults"),
terminalMacros: text("terminal_macros"),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const hostMetricsPreferences = pgTable(
"host_metrics_preferences",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
// JSON-encoded HostMetricsLayout. Layout has no secrets, so it is stored as
// plain JSON (no field-level encryption).
layout: text("layout").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// One layout per user per host. Enforced in production since the inline DDL
// creates it, but it was never declared here, so the generated Postgres and
// MySQL schemas lacked it — and the upsert has nothing to conflict on.
(table) => [
uniqueIndex("idx_host_metrics_prefs_user_host").on(table.userId, table.hostId),
],
);
export const proxmoxStatsPreferences = pgTable(
"proxmox_stats_preferences",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 }).notNull().references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id").notNull().references(() => hosts.id, { onDelete: "cascade" }),
// JSON-encoded ProxmoxStatsLayout. Layout has no secrets, so it is stored as
// plain JSON (no field-level encryption), same convention as hostMetricsPreferences.layout.
layout: text("layout").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("idx_proxmox_stats_prefs_user_host").on(table.userId, table.hostId),
],
);
export const hostSidebarPreferences = pgTable("host_sidebar_preferences", {
userId: varchar("user_id", { length: 255 })
.primaryKey()
.references(() => users.id, { onDelete: "cascade" }),
// JSON-encoded HostSidebarPreferences. No secrets in this blob, stored as
// plain JSON like hostMetricsPreferences.layout.
data: text("data").notNull(),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const credentialSidebarPreferences = pgTable(
"credential_sidebar_preferences",
{
userId: varchar("user_id", { length: 255 })
.primaryKey()
.references(() => users.id, { onDelete: "cascade" }),
// JSON-encoded CredentialSidebarPreferences. No secrets in this blob,
// same convention as hostSidebarPreferences.data.
data: text("data").notNull(),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
);
export const uiPreferences = pgTable("ui_preferences", {
userId: varchar("user_id", { length: 255 })
.primaryKey()
.references(() => users.id, { onDelete: "cascade" }),
// JSON-encoded UiPreferences (preset + per-area overrides + onboarding
// state). No secrets in this blob, same convention as
// hostSidebarPreferences.data.
data: text("data").notNull(),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const hostHealthChecks = pgTable(
"host_health_checks",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
// JSON array of { id, name, type: "tcp"|"http", target, port, path }
checks: text("checks").notNull(),
intervalSeconds: integer("interval_seconds").notNull().default(300),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// Same as above: one set of checks per user per host.
(table) => [
uniqueIndex("idx_host_health_checks_user_host").on(table.userId, table.hostId),
],
);
export const hostHealthHistory = pgTable("host_health_history", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
checkId: text("check_id").notNull(),
ts: text("ts").notNull().default(sql`CURRENT_TIMESTAMP`),
ok: boolean("ok").notNull(),
latencyMs: integer("latency_ms"),
detail: text("detail"),
});
export const dashboardServiceLinks = pgTable("dashboard_service_links", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
label: varchar("label", { length: 255 }).notNull(),
url: text("url").notNull(),
order: integer("order").notNull().default(0),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// --- termix-id begin ---
// A user claims a unique public handle. Their published SSH public keys are
// served at an unauthenticated resolver endpoint in authorized_keys format,
// so any server can be provisioned with `curl <host>/termix-id/u/<handle> >> ~/.ssh/authorized_keys`.
export const termixIdentities = pgTable("termix_identities", {
id: serial("id").primaryKey(),
// One Termix ID per user — enforced in schema, not just in code.
userId: varchar("user_id", { length: 255 })
.notNull()
.unique()
.references(() => users.id, { onDelete: "cascade" }),
handle: varchar("handle", { length: 255 }).notNull().unique(),
description: text("description"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const termixIdentityKeys = pgTable("termix_identity_keys", {
id: serial("id").primaryKey(),
identityId: integer("identity_id")
.notNull()
.references(() => termixIdentities.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// Public keys are non-secret, so they are stored in plaintext (no field-level
// encryption). This is what lets the unauthenticated resolver serve them.
publicKey: text("public_key").notNull(),
// Raw algorithm token (e.g. "ssh-ed25519"), and a normalized group used for
// the /<ALGO> resolver filter (RSA / ED25519 / ECDSA / ...).
keyType: text("key_type").notNull(),
algorithm: text("algorithm").notNull(),
label: varchar("label", { length: 255 }),
comment: text("comment"),
// "manual" (pasted) or "credential" (imported from an ssh_credentials entry).
source: text("source").notNull().default("manual"),
credentialId: integer("credential_id").references(() => sshCredentials.id, {
onDelete: "set null",
}),
enabled: boolean("enabled").notNull().default(true),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// Per-identity certificate authority. Servers that trust this CA (via
// TrustedUserCAKeys / @cert-authority) accept any user certificate it signs,
// giving central revocation (rotate the CA) and expiry (cert validity).
export const termixIdentityCa = pgTable("termix_identity_ca", {
id: serial("id").primaryKey(),
identityId: integer("identity_id")
.notNull()
.unique()
.references(() => termixIdentities.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// CA public key (plaintext — it is published); CA private key is field-encrypted.
publicKey: text("public_key").notNull(),
privateKey: text("private_key").notNull(),
validityDays: integer("validity_days").notNull().default(90),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// --- termix-id end ---
// --- tmux-monitor begin ---
export const tmuxSessionTags = pgTable("tmux_session_tags", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
sessionName: text("session_name").notNull(),
tag: text("tag").notNull(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// --- tmux-monitor end ---
// --- metrics-history begin ---
export const hostMetricsHistory = pgTable("host_metrics_history", {
id: serial("id").primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
ts: text("ts")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
cpuPercent: doublePrecision("cpu_percent"),
memPercent: doublePrecision("mem_percent"),
diskPercent: doublePrecision("disk_percent"),
netRxBytes: integer("net_rx_bytes"),
netTxBytes: integer("net_tx_bytes"),
});
// --- metrics-history end ---
// --- proxmox-node-history begin ---
export const proxmoxNodeHistory = pgTable("proxmox_node_history", {
id: serial("id").primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
ts: text("ts")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
cpuPercent: doublePrecision("cpu_percent"),
memPercent: doublePrecision("mem_percent"),
diskPercent: doublePrecision("disk_percent"),
netRxBytes: integer("net_rx_bytes"),
netTxBytes: integer("net_tx_bytes"),
});
// --- proxmox-node-history end ---
// --- alerts begin ---
export const notificationChannels = pgTable("notification_channels", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
type: text("type").notNull(),
config: text("config").notNull(),
enabled: boolean("enabled").notNull().default(true),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// --- alerts end ---
// --- automations begin ---
export const automations = pgTable(
"automations",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
description: text("description"),
enabled: boolean("enabled").notNull().default(true),
// The whole trigger + steps graph, shaped by AutomationDefinition. Read and
// written as a unit, never queried by its inner structure.
definition: text("definition").notNull(),
definitionVersion: integer("definition_version").notNull().default(1),
concurrencyPolicy: text("concurrency_policy").notNull().default("skip"),
maxRunSeconds: integer("max_run_seconds").notNull().default(300),
dryRun: boolean("dry_run").notNull().default(false),
lastRunAt: text("last_run_at"),
lastRunStatus: text("last_run_status"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// The scheduler sweeps enabled automations for one user at a time.
(table) => [index("idx_automations_user").on(table.userId, table.enabled)],
);
/**
* Durable per-target trigger state. state_key scopes a trigger to what it is
* actually watching ("<hostId>", "<hostId>:/data", "<hostId>:<container>"), so
* a sustained-breach window can track one filesystem rather than a whole host.
* Living in the database rather than memory means cooldowns and dwell windows
* survive a restart.
*/
export const automationTriggerState = pgTable(
"automation_trigger_state",
{
id: serial("id").primaryKey(),
automationId: integer("automation_id")
.notNull()
.references(() => automations.id, { onDelete: "cascade" }),
stateKey: varchar("state_key", { length: 255 }).notNull(),
breachStartedAt: text("breach_started_at"),
lastFiredAt: text("last_fired_at"),
lastValue: doublePrecision("last_value"),
lastObservedState: text("last_observed_state"),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("idx_automation_trigger_state_key").on(
table.automationId,
table.stateKey,
),
],
);
export const automationSchedules = pgTable(
"automation_schedules",
{
id: serial("id").primaryKey(),
automationId: integer("automation_id")
.notNull()
.references(() => automations.id, { onDelete: "cascade" }),
cron: text("cron"),
intervalSeconds: integer("interval_seconds"),
timezone: text("timezone"),
nextDueAt: varchar("next_due_at", { length: 255 }),
lastTickAt: text("last_tick_at"),
},
(table) => [
uniqueIndex("idx_automation_schedules_automation").on(table.automationId),
index("idx_automation_schedules_due").on(table.nextDueAt),
],
);
export const automationRuns = pgTable(
"automation_runs",
{
id: serial("id").primaryKey(),
automationId: integer("automation_id")
.notNull()
.references(() => automations.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
triggerType: text("trigger_type").notNull(),
triggerContext: text("trigger_context"),
status: varchar("status", { length: 255 }).notNull(),
startedAt: varchar("started_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
finishedAt: text("finished_at"),
durationMs: integer("duration_ms"),
error: text("error"),
dryRun: boolean("dry_run").notNull().default(false),
// Set when one automation invoked another, so a chain can be traced.
parentRunId: integer("parent_run_id"),
},
(table) => [
index("idx_automation_runs_automation").on(
table.automationId,
table.startedAt,
),
index("idx_automation_runs_user").on(table.userId, table.startedAt),
],
);
export const automationRunSteps = pgTable(
"automation_run_steps",
{
id: serial("id").primaryKey(),
runId: integer("run_id")
.notNull()
.references(() => automationRuns.id, { onDelete: "cascade" }),
stepIndex: integer("step_index").notNull(),
stepId: text("step_id").notNull(),
stepType: text("step_type").notNull(),
status: varchar("status", { length: 255 }).notNull(),
startedAt: varchar("started_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
finishedAt: text("finished_at"),
output: text("output"),
error: text("error"),
truncated: boolean("truncated")
.notNull()
.default(false),
},
(table) => [
index("idx_automation_run_steps_run").on(table.runId, table.stepIndex),
],
);
export const automationChannels = pgTable(
"automation_channels",
{
id: serial("id").primaryKey(),
automationId: integer("automation_id")
.notNull()
.references(() => automations.id, { onDelete: "cascade" }),
channelId: integer("channel_id")
.notNull()
.references(() => notificationChannels.id, { onDelete: "cascade" }),
},
(table) => [
uniqueIndex("idx_automation_channels_pair").on(
table.automationId,
table.channelId,
),
],
);
// --- automations end ---
// --- homepage begin ---
export const homepageItems = pgTable(
"homepage_items",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
typeId: text("type_id").notNull(),
title: text("title"),
config: text("config").notNull().default("{}"),
folderId: integer("folder_id"),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_homepage_items_user_id").on(table.userId)],
);
export const homepageLayouts = pgTable("homepage_layouts", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.unique()
.references(() => users.id, { onDelete: "cascade" }),
// JSON: { entries: HomepageLayoutEntry[], pan: {x,y}, zoom: number }
layout: text("layout").notNull().default("{}"),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// --- homepage end ---
// --- fleets begin ---
export const fleets = pgTable("fleets", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
description: text("description"),
color: text("color"),
icon: text("icon"),
// JSON array of { tag: string } rules, unioned with static fleetMembers at
// resolution time. Kept to tag-equality matching for v1.
tagRules: text("tag_rules"),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
export const fleetMembers = pgTable(
"fleet_members",
{
id: serial("id").primaryKey(),
fleetId: integer("fleet_id")
.notNull()
.references(() => fleets.id, { onDelete: "cascade" }),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
addedAt: text("added_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// fleet_id leads the unique pair, so listing a fleet's hosts is already
// served. Finding the fleets a host belongs to starts from host_id.
(table) => [
uniqueIndex("idx_fleet_members_fleet_host").on(table.fleetId, table.hostId),
index("idx_fleet_members_host").on(table.hostId),
],
);
// Latest-only inventory snapshot per host, overwritten on each refresh - no
// historical log, matching the "latest snapshot only" scope decision.
export const fleetInventory = pgTable(
"fleet_inventory",
{
id: serial("id").primaryKey(),
hostId: integer("host_id")
.notNull()
.references(() => hosts.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
osPrettyName: text("os_pretty_name"),
kernel: text("kernel"),
architecture: text("architecture"),
hostname: text("hostname"),
uptimeSeconds: integer("uptime_seconds"),
ip: text("ip"),
packageManager: text("package_manager"),
collectedAt: text("collected_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
// host_id leads the unique pair; a user's whole inventory is read by user_id.
(table) => [
uniqueIndex("idx_fleet_inventory_host").on(table.hostId, table.userId),
index("idx_fleet_inventory_user").on(table.userId),
],
);
// --- fleets end ---
// --- workspaces begin ---
export const userWorkspaces = pgTable(
"user_workspaces",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
color: text("color"),
icon: text("icon"),
// "manual" | "last_session" - exactly one last_session row per user.
kind: text("kind").notNull().default("manual"),
isDefault: boolean("is_default")
.notNull()
.default(false),
// JSON-encoded WorkspacePayload: tabs, splitMode, paneTabIds, rowSizes, rowColSizes
payload: text("payload").notNull().default("{}"),
syncId: varchar("sync_id", { length: 255 }).unique(),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
lastUsedAt: text("last_used_at"),
},
(table) => [index("idx_user_workspaces_user_id").on(table.userId)],
);
// --- workspaces end ---
// --- sync begin ---
// Records a delete for a synced entity type so the other side of a sync
// pair (embedded desktop backend <-> connected remote server) learns about
// the deletion instead of re-creating the row on its next pull.
export const syncTombstones = pgTable("sync_tombstones", {
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
entityType: text("entity_type").notNull(),
syncId: varchar("sync_id", { length: 255 }).notNull(),
deletedAt: text("deleted_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
});
// --- sync end ---
// --- ai begin ---
/**
* A user's connection to one AI provider. api_key is encrypted at rest via
* FieldCrypto; it is never returned to the frontend, which only ever sees
* api_key_prefix for display.
*/
export const aiProviders = pgTable(
"ai_providers",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// ollama | anthropic | openai | gemini | openai_compatible
providerType: text("provider_type").notNull(),
label: varchar("label", { length: 255 }).notNull(),
// Required for ollama and openai_compatible, optional elsewhere.
baseUrl: text("base_url"),
apiKey: text("api_key"),
// First few characters, kept in the clear so the UI can identify a key.
apiKeyPrefix: text("api_key_prefix"),
defaultModel: text("default_model"),
enabled: boolean("enabled").notNull().default(true),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("idx_ai_providers_user_label").on(table.userId, table.label),
],
);
export const aiConversations = pgTable(
"ai_conversations",
{
id: serial("id").primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
title: text("title"),
providerId: integer("provider_id"),
model: text("model"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_ai_conversations_user").on(table.userId, table.updatedAt),
],
);
export const aiMessages = pgTable(
"ai_messages",
{
id: serial("id").primaryKey(),
conversationId: integer("conversation_id")
.notNull()
.references(() => aiConversations.id, { onDelete: "cascade" }),
// user | assistant | tool
role: text("role").notNull(),
content: text("content").notNull().default(""),
// Serialized tool calls and their results for this turn.
toolCalls: text("tool_calls"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_ai_messages_conversation").on(
table.conversationId,
table.createdAt,
),
],
);
/**
* A change the assistant wants to make. Nothing here has been applied: the
* payload is re-validated against the tool schema at apply time and only then
* dispatched through the same repository logic a human action uses.
*/
export const aiProposals = pgTable(
"ai_proposals",
{
id: serial("id").primaryKey(),
conversationId: integer("conversation_id")
.notNull()
.references(() => aiConversations.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// The propose_* tool name that produced this.
kind: text("kind").notNull(),
summary: text("summary"),
payload: text("payload").notNull().default("{}"),
// pending | applied | rejected | expired
status: varchar("status", { length: 255 }).notNull().default("pending"),
appliedAt: text("applied_at"),
resultSummary: text("result_summary"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_ai_proposals_user").on(table.userId, table.status),
index("idx_ai_proposals_conversation").on(table.conversationId),
],
);
// --- ai end ---
// --- collab rooms ---
/**
* A collaboration room: a group of users watching one "stage" - the live
* session the current presenter is showing. The stage points at a
* shareType="room" row in session_shares, so transport, gating, recording and
* expiry all reuse the session-sharing machinery.
*/
export const collabRooms = pgTable(
"collab_rooms",
{
id: varchar("id", { length: 255 }).primaryKey(),
name: varchar("name", { length: 255 }).notNull(),
ownerUserId: varchar("owner_user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// Persistent rooms survive being emptied and can be re-used; one-off
// rooms are ended explicitly and never listed again.
persistent: boolean("persistent")
.notNull()
.default(false),
presenterUserId: varchar("presenter_user_id", { length: 255 }).references(() => users.id, {
onDelete: "set null",
}),
stageProtocol: text("stage_protocol"),
stageHostId: integer("stage_host_id").references(() => hosts.id, {
onDelete: "set null",
}),
stageShareId: varchar("stage_share_id", { length: 255 }).references(() => sessionShares.id, {
onDelete: "set null",
}),
// Set = anonymous guests may watch the stage through this token.
guestLinkToken: varchar("guest_link_token", { length: 255 }),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
endedAt: text("ended_at"),
},
(table) => [
index("idx_collab_rooms_owner").on(table.ownerUserId),
uniqueIndex("idx_collab_rooms_guest_token").on(table.guestLinkToken),
],
);
export const collabRoomMembers = pgTable(
"collab_room_members",
{
id: serial("id").primaryKey(),
roomId: varchar("room_id", { length: 255 })
.notNull()
.references(() => collabRooms.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// "host" runs the room: invites, force-switches the presenter, ends it.
roomRole: text("room_role").notNull().default("member"),
addedBy: varchar("added_by", { length: 255 }).references(() => users.id, {
onDelete: "set null",
}),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("idx_collab_room_members_room_user").on(
table.roomId,
table.userId,
),
index("idx_collab_room_members_user").on(table.userId),
],
);
// --- secret sources ---
/**
* An external password manager Termix pulls secrets from at connect time,
* instead of storing them. Only the access token is secret; it is encrypted
* with the owner's data key under the row id. Hosts and credentials refer to
* entries by reference ("op://vault/item/field") in their secret fields.
*/
export const secretSources = pgTable(
"secret_sources",
{
id: varchar("id", { length: 255 }).primaryKey(),
userId: varchar("user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
name: varchar("name", { length: 255 }).notNull(),
// "onepassword-connect" for now; the reference syntax is per kind.
kind: text("kind").notNull().default("onepassword-connect"),
baseUrl: text("base_url").notNull(),
token: text("token").notNull(),
// Visible to every user; secrets still decrypt with the owner's key.
shared: boolean("shared").notNull().default(false),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [index("idx_secret_sources_user").on(table.userId)],
);
// --- credential sharing ---
/**
* Who may use or manage someone else's credential. Same shape as
* snippet_access; "use" attaches it to hosts and connects, "manage" also
* edits and re-shares it.
*/
export const credentialAccess = pgTable(
"credential_access",
{
id: serial("id").primaryKey(),
credentialId: integer("credential_id")
.notNull()
.references(() => sshCredentials.id, { onDelete: "cascade" }),
userId: varchar("user_id", { length: 255 }).references(() => users.id, { onDelete: "cascade" }),
roleId: integer("role_id").references(() => roles.id, {
onDelete: "cascade",
}),
grantedBy: varchar("granted_by", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
permissionLevel: text("permission_level").notNull().default("use"),
expiresAt: varchar("expires_at", { length: 255 }),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_credential_access_user_id").on(table.userId),
index("idx_credential_access_role_id").on(table.roleId),
index("idx_credential_access_credential_id").on(table.credentialId),
],
);
/**
* A recipient's copy of a shared credential's secrets, re-encrypted under
* the recipient's data key (the owner's key cannot be used by anyone else).
* Rebuilt whenever the owner edits the credential; one row per grant and
* recipient, like shared_host_secrets.
*/
export const sharedCredentialSecrets = pgTable(
"shared_credential_secrets",
{
id: serial("id").primaryKey(),
credentialAccessId: integer("credential_access_id").notNull(),
targetUserId: varchar("target_user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
credentialId: integer("credential_id")
.notNull()
.references(() => sshCredentials.id, { onDelete: "cascade" }),
encryptedUsername: text("encrypted_username"),
authType: text("auth_type").notNull().default("password"),
encryptedPassword: text("encrypted_password"),
encryptedKey: text("encrypted_key"),
encryptedKeyPassword: text("encrypted_key_password"),
keyType: text("key_type"),
publicKey: text("public_key"),
certPublicKey: text("cert_public_key"),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
foreignKey({
columns: [table.credentialAccessId],
foreignColumns: [credentialAccess.id],
name: "shared_cred_secrets_access_id_fk",
}).onDelete("cascade"),
uniqueIndex("idx_shared_credential_secrets_scope").on(
table.credentialAccessId,
table.targetUserId,
),
index("idx_shared_credential_secrets_target").on(
table.targetUserId,
table.credentialId,
),
],
);
// --- folder access rules ---
/**
* A standing share on a host folder. Sharing a folder fans out host_access
* grants to the hosts in it today; this row is what makes hosts created in
* or moved into the folder later inherit the same access.
*/
export const folderAccess = pgTable(
"folder_access",
{
id: serial("id").primaryKey(),
ownerUserId: varchar("owner_user_id", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
// The folder path as stored on hosts ("Parent / Child"); subfolders inherit.
folder: varchar("folder", { length: 255 }).notNull(),
userId: varchar("user_id", { length: 255 }).references(() => users.id, { onDelete: "cascade" }),
roleId: integer("role_id").references(() => roles.id, {
onDelete: "cascade",
}),
grantedBy: varchar("granted_by", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
permissionLevel: text("permission_level").notNull().default("connect"),
expiresAt: varchar("expires_at", { length: 255 }),
createdAt: varchar("created_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
index("idx_folder_access_owner_folder").on(table.ownerUserId, table.folder),
],
);
// --- plugins begin ---
/**
* An installed plugin. id matches the manifest's own id (not autoincrement),
* so a plugin can be looked up the same way the manifest and registry refer
* to it. manifest_json is the full manifest as it was at install time, kept
* for audit/rollback even after a registry updates or removes the entry.
*/
export const plugins = pgTable(
"plugins",
{
id: varchar("id", { length: 255 }).primaryKey(),
name: varchar("name", { length: 255 }).notNull(),
version: text("version").notNull(),
tier: text("tier").notNull().default("available"),
source: text("source").notNull().default("community"),
registryId: varchar("registry_id", { length: 255 }),
state: text("state").notNull().default("disabled"),
installedAt: text("installed_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
autoUpdate: boolean("auto_update")
.notNull()
.default(false),
manifestJson: text("manifest_json").notNull(),
},
(table) => [index("idx_plugins_registry_id").on(table.registryId)],
);
export const pluginPermissionGrants = pgTable(
"plugin_permission_grants",
{
id: serial("id").primaryKey(),
pluginId: varchar("plugin_id", { length: 255 })
.notNull()
.references(() => plugins.id, { onDelete: "cascade" }),
capability: varchar("capability", { length: 255 }).notNull(),
grantedAt: text("granted_at")
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
grantedBy: varchar("granted_by", { length: 255 })
.notNull()
.references(() => users.id, { onDelete: "cascade" }),
},
// A plugin's grants are always read together, and re-granting the same
// capability should update the existing row rather than duplicate it.
(table) => [
uniqueIndex("idx_plugin_permission_grants_plugin_capability").on(
table.pluginId,
table.capability,
),
],
);
export const pluginRegistries = pgTable("plugin_registries", {
id: varchar("id", { length: 255 }).primaryKey(),
name: varchar("name", { length: 255 }).notNull(),
url: text("url").notNull(),
kind: text("kind").notNull().default("community"),
enabled: boolean("enabled").notNull().default(true),
signingKey: text("signing_key"),
lastCheckedAt: text("last_checked_at"),
lastIndexHash: text("last_index_hash"),
});
/**
* Install counts populated by a background job (GitHub release download
* counts, aggregated telemetry, or a manual override) rather than by the
* install/uninstall actions themselves — kept separate from `plugins` so
* that job can overwrite counts without touching install state.
*/
export const pluginInstallCounts = pgTable(
"plugin_install_counts",
{
id: serial("id").primaryKey(),
pluginId: varchar("plugin_id", { length: 255 }).notNull(),
registryId: varchar("registry_id", { length: 255 }).notNull(),
count: integer("count").notNull().default(0),
source: text("source").notNull().default("aggregate-telemetry"),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("idx_plugin_install_counts_plugin_registry").on(
table.pluginId,
table.registryId,
),
],
);
/**
* Per-plugin key/value state, written through ctx.storage. Rows are always
* scoped to the calling plugin by the broker, so a plugin cannot name another
* plugin's scope. Cascades with the plugin so uninstalling leaves nothing.
*/
export const pluginStorage = pgTable(
"plugin_storage",
{
id: serial("id").primaryKey(),
pluginId: varchar("plugin_id", { length: 255 })
.notNull()
.references(() => plugins.id, { onDelete: "cascade" }),
key: varchar("key", { length: 255 }).notNull(),
value: text("value").notNull(),
updatedAt: varchar("updated_at", { length: 255 })
.notNull()
.default(sql`CURRENT_TIMESTAMP`),
},
(table) => [
uniqueIndex("idx_plugin_storage_plugin_key").on(table.pluginId, table.key),
],
);
// --- plugins end ---