feat(users): UserManager with per-user SQLCipher, and extract skald-core crate

Two changes developed together in one session; they share the same module
structure (db/mod.rs, the core lib root) and only compile together, so they
land as one commit.

## UserManager + per-user encryption (§9/§11)

New `users::UserManager`: owns the system.db pool plus a map
`userid -> SqlitePool` of unlocked databases. The pool *is* the unlock token —
its connect options carry the DEK as SQLCipher's raw key, so an open pool means
the key is in RAM until restart and dropping it re-locks (§9). Knows nothing
about cookies.

New `crypto` module: envelope encryption. A random 256-bit DEK encrypts
`{userid}.db`; `users.database_password` holds it sealed with AES-256-GCM under
`Argon2id(password, salt)`. The AEAD tag is the password verifier — one
derivation both authenticates and yields the key, so encrypted users store no
second hash. Cleartext users store the Argon2id output directly, compared in
constant time. Argon2 runs in spawn_blocking behind a 2-permit semaphore
(256 MiB per derivation).

- SQLCipher via `libsqlite3-sys` `bundled-sqlcipher-vendored-openssl`, pinned
  <0.38 so it unifies with the one sqlx-sqlite links (a newer copy would apply
  the feature to a SQLite sqlx never uses). OpenSSL is vendored and static, so
  the binary stays self-contained.
- Schema split into `create_registry_tables` (instance-wide, no user key) and
  `create_owner_tables` (one owner's content, identical in every file). No FK in
  the owner bucket may reach the registry — enforced by a standalone test.
  Dropped `chat_history.model_db_id` (write-only, and the only registry-crossing
  key); moved `projects`/`project_tickets` into the owner bucket.
- Provisioning invariant: the file is written before the row, deleted after it,
  so a crash leaves an orphan file, never a user without a database. `open_db`
  never creates: a missing file is an error, not a silent empty database.

Not consumed yet: no login, call sites still use the shared system.db pool.

## Extract crates/skald-core

The headless core moves out of `src/` into its own crate; `skald` (server) and
the coming `skald-setup` are shells around it. Two dependencies on the shell
were inverted rather than dragged along, so the core names neither Tauri nor any
concrete plugin:

- `Plugin::tools(self: Arc<Self>)` — plugins contribute tools through this hook
  (sibling of `http_router`), so the core no longer downcasts to
  `MobileConnectorPlugin`.
- `tools::restart::set_restart_handler` — the desktop shell installs its
  teardown-and-respawn; the core defaults to the supervisor exit code. The core
  loses its `desktop` feature.
- `boot`'s stdout formatter moves to the binary (`src/boot_format.rs`); the core
  only emits tracing events.

All 79 core tests pass; the binary boots and serves in a clean directory, and
the mobile-connector tools still register through the new hook.
This commit is contained in:
2026-07-10 16:48:51 +01:00
parent 38494a85a9
commit 178a38357e
173 changed files with 2650 additions and 1106 deletions
+351
View File
@@ -0,0 +1,351 @@
// The manager only ever handles `Arc<dyn Plugin>`. Naming a concrete plugin crate
// here would make the core depend on every plugin — and, through
// `plugin-transcribe-whisper-local`, on a C build — for no gain: the consumer
// constructs the plugin list and passes it to `Skald::new`.
pub use core_api::plugin::{Plugin, PluginContext, RouterFactory};
use std::collections::HashMap;
use std::sync::{Arc, OnceLock};
use std::time::Duration;
const PLUGIN_START_TIMEOUT_SECS: u64 = 30;
const PLUGIN_STOP_TIMEOUT_SECS: u64 = 5;
use anyhow::Result;
use serde::Serialize;
use serde_json::{Value, json};
use sqlx::SqlitePool;
use tokio::sync::Mutex;
use tokio::time::timeout;
use tracing::{error, info, warn};
use crate::db::plugins as db;
use crate::skald::Skald;
// ── Public plugin info (returned by list_items tool and REST API) ─────────────
#[derive(Debug, Clone, Serialize)]
pub struct PluginInfo {
pub id: String,
pub name: String,
pub description: String,
pub enabled: bool,
pub running: bool,
pub config: Value,
pub config_schema: Value,
pub runtime_status: Option<Value>,
}
// ── PluginManager ─────────────────────────────────────────────────────────────
pub struct PluginManager {
plugins: Vec<Arc<dyn Plugin>>,
db: Arc<SqlitePool>,
skald: OnceLock<Arc<Skald>>,
/// Provided by WebFrontend before start_enabled() is called.
router_factory: OnceLock<RouterFactory>,
/// HTTP port the web server is bound to — provided by WebFrontend before start_enabled().
web_port: OnceLock<u16>,
/// Last known (enabled, config_json) per plugin id — used by the watcher.
known_state: Mutex<HashMap<String, (bool, String)>>,
}
impl PluginManager {
pub fn new(db: Arc<SqlitePool>) -> Self {
Self {
plugins: Vec::new(),
db,
skald: OnceLock::new(),
router_factory: OnceLock::new(),
web_port: OnceLock::new(),
known_state: Mutex::new(HashMap::new()),
}
}
pub fn register(&mut self, plugin: impl Plugin + 'static) {
self.plugins.push(Arc::new(plugin));
}
pub fn register_arc(&mut self, plugin: Arc<dyn Plugin>) {
self.plugins.push(plugin);
}
pub fn set_skald(&self, skald: Arc<Skald>) {
let _ = self.skald.set(skald);
}
/// Called by WebFrontend before start_enabled().
pub fn set_router_factory(&self, factory: RouterFactory) {
let _ = self.router_factory.set(factory);
}
/// Called by WebFrontend before start_enabled().
pub fn set_web_port(&self, port: u16) {
let _ = self.web_port.set(port);
}
fn skald(&self) -> Result<Arc<Skald>> {
self.skald.get().cloned()
.ok_or_else(|| anyhow::anyhow!("PluginManager: skald not initialized"))
}
fn build_context(&self, skald: &Skald) -> Result<PluginContext> {
let router_factory = self.router_factory.get().cloned()
.ok_or_else(|| anyhow::anyhow!("PluginManager: router_factory not set"))?;
let web_port = self.web_port.get().copied()
.ok_or_else(|| anyhow::anyhow!("PluginManager: web_port not set"))?;
Ok(PluginContext {
chat_hub: Arc::clone(skald.chat_hub()) as _,
command: Arc::clone(skald.command_manager()) as _,
approval: Arc::clone(skald.approval()) as _,
inbox: Arc::new(skald.inbox().clone()) as _,
db: Arc::clone(skald.db()),
secrets: Arc::clone(skald.secrets()) as _,
transcribe: Arc::clone(skald.transcribe_manager()) as _,
transcribe_registry: Arc::clone(skald.transcribe_manager()) as _,
image_generate_registry: Arc::clone(skald.image_generator_manager()) as _,
tts_registry: Arc::clone(skald.tts_manager()) as _,
tts_provider: Arc::clone(skald.tts_manager()) as _,
api_provider_registry: Arc::clone(skald.provider_registry()) as _,
location: Arc::clone(skald.location_manager()) as _,
event_bus: Arc::clone(skald.event_bus()),
system_bus: Arc::clone(skald.system_bus()),
web_port,
remote_slot: Arc::clone(skald.remote()),
router_factory,
})
}
/// Collects the HTTP routers contributed by enabled plugins (plugin.md §12.3).
/// Returns `(plugin_id, router)` pairs; the caller (`WebFrontend::start`)
/// nests each under `/api/plugin/<id>/`. Only plugins with `enabled=true` in
/// the DB and a non-`None` `http_router()` are included.
///
/// Call this AFTER `start_enabled()` so a plugin's router can close over state
/// initialised during `reload`/`start`.
pub async fn collect_plugin_routers(&self) -> Vec<(String, axum::Router)> {
let mut out = Vec::new();
for plugin in &self.plugins {
match db::get(&self.db, plugin.id()).await {
Ok(Some(row)) if row.enabled => {}
Ok(_) => continue,
Err(e) => {
warn!(plugin = plugin.id(), error = %e, "collect_plugin_routers: DB read failed; skipping");
continue;
}
}
if let Some(router) = plugin.http_router() {
info!(plugin = plugin.id(), "plugin contributed an HTTP router → /api/plugin/{}", plugin.id());
out.push((plugin.id().to_string(), router));
}
}
out
}
// ── Startup ───────────────────────────────────────────────────────────────
/// Calls reload() for every plugin that has enabled=true in DB.
/// Plugins without a DB row are skipped (not yet configured).
/// After each successful start, registers the plugin's Memory backend (if any).
/// Must be called after both set_skald() and set_router_factory().
pub async fn start_enabled(&self) -> Result<()> {
let skald = self.skald()?;
// Build a full inventory for the bootstrap report: active (started),
// failed (enabled but errored/timed out), and available (disabled).
let mut active: Vec<String> = Vec::new();
let mut failed: Vec<(String, String)> = Vec::new();
let mut disabled: Vec<String> = Vec::new();
for plugin in &self.plugins {
let row = db::get(&self.db, plugin.id()).await?;
let enabled = row.as_ref().map(|r| r.enabled).unwrap_or(false);
if !enabled {
disabled.push(plugin.id().to_string());
continue;
}
let row = row.expect("enabled implies row present");
let config = serde_json::from_str(&row.config).unwrap_or(json!({}));
let deadline = Duration::from_secs(PLUGIN_START_TIMEOUT_SECS);
let ctx = self.build_context(&skald)?;
match timeout(deadline, plugin.reload(true, config, ctx)).await {
Ok(Ok(())) => {
self.known_state.lock().await
.insert(plugin.id().to_string(), (true, row.config));
info!(plugin = plugin.id(), "plugin started");
if let Some(mem) = plugin.memory() {
skald.memory_manager().register(mem).await;
}
active.push(plugin.id().to_string());
}
Ok(Err(e)) => {
error!(plugin = plugin.id(), error = %e, "plugin failed to start");
failed.push((plugin.id().to_string(), e.to_string()));
}
Err(_) => {
error!(plugin = plugin.id(), secs = PLUGIN_START_TIMEOUT_SECS, "plugin start timed out");
failed.push((plugin.id().to_string(),
format!("start timed out after {PLUGIN_START_TIMEOUT_SECS}s")));
}
}
}
crate::boot::section(format!(
"Plugins — {} active, {} failed, {} available",
active.len(), failed.len(), disabled.len()
));
if !active.is_empty() {
crate::boot::ok(active.join(", "));
}
for (id, reason) in &failed {
crate::boot::fail(format!("{id}{reason}"));
}
if !disabled.is_empty() {
crate::boot::off(disabled.join(", "));
}
Ok(())
}
pub async fn stop_all(&self) {
for plugin in &self.plugins {
if plugin.is_running() {
let deadline = Duration::from_secs(PLUGIN_STOP_TIMEOUT_SECS);
match timeout(deadline, plugin.stop()).await {
Ok(Ok(())) => info!(plugin = plugin.id(), "plugin stopped"),
Ok(Err(e)) => error!(plugin = plugin.id(), error = %e, "plugin stop error"),
Err(_) => warn!(plugin = plugin.id(), secs = PLUGIN_STOP_TIMEOUT_SECS, "plugin stop timed out"),
}
}
}
}
// ── Config update (called by REST API) ────────────────────────────────────
/// Persists the new config to DB, then calls reload() immediately.
pub async fn update_config(&self, id: &str, enabled: bool, config: Value) -> Result<()> {
let plugin = self.find(id)?;
let config_json = serde_json::to_string(&config)?;
db::upsert(&self.db, id, enabled, &config_json).await?;
let skald = self.skald()?;
plugin.reload(enabled, config, self.build_context(&skald)?).await?;
self.known_state.lock().await
.insert(id.to_string(), (enabled, config_json));
info!(plugin = id, enabled, "plugin config updated");
Ok(())
}
/// Toggle only the enabled flag, keeping existing config.
pub async fn toggle(&self, id: &str, enabled: bool) -> Result<()> {
let row = db::get(&self.db, id).await?
.unwrap_or_else(|| crate::db::plugins::PluginRow {
id: id.to_string(),
enabled,
config: "{}".to_string(),
});
let config: Value = serde_json::from_str(&row.config).unwrap_or(json!({}));
self.update_config(id, enabled, config).await
}
// ── Background config watcher ─────────────────────────────────────────────
/// Spawns a Tokio task that polls the DB every 30 s and calls reload()
/// on any plugin whose (enabled, config) has changed since last check.
/// This is the fallback path; normal updates go through update_config().
pub fn start_config_watcher(self: &Arc<Self>, shutdown: tokio_util::sync::CancellationToken) {
let this = Arc::clone(self);
tokio::spawn(async move {
let mut interval = tokio::time::interval(Duration::from_secs(30));
interval.tick().await; // skip immediate first tick
loop {
tokio::select! {
_ = shutdown.cancelled() => { break; }
_ = interval.tick() => {
if let Err(e) = this.check_and_reload().await {
error!(error = %e, "plugin config watcher error");
}
}
}
}
});
}
async fn check_and_reload(&self) -> Result<()> {
let rows = db::list(&self.db).await?;
let skald = self.skald()?;
// Collect what needs reloading while holding the lock briefly.
let to_reload: Vec<_> = {
let known = self.known_state.lock().await;
rows.into_iter()
.filter(|row| {
known.get(&row.id)
.map_or(true, |(e, c)| *e != row.enabled || c != &row.config)
})
.collect()
};
for row in to_reload {
let Ok(plugin) = self.find(&row.id) else { continue };
let config = serde_json::from_str(&row.config).unwrap_or(json!({}));
let ctx = self.build_context(&skald)?;
match plugin.reload(row.enabled, config, ctx).await {
Ok(()) => {
self.known_state.lock().await
.insert(row.id.clone(), (row.enabled, row.config));
info!(plugin = row.id, "plugin reloaded by config watcher");
if row.enabled {
if let Some(mem) = plugin.memory() {
skald.memory_manager().register(mem).await;
}
}
}
Err(e) => error!(plugin = row.id, error = %e, "plugin reload failed"),
}
}
Ok(())
}
// ── Queries ───────────────────────────────────────────────────────────────
pub async fn list(&self) -> Result<Vec<PluginInfo>> {
let mut out = Vec::new();
for plugin in &self.plugins {
let row = db::get(&self.db, plugin.id()).await?;
let (enabled, config_json) = row
.map(|r| (r.enabled, r.config))
.unwrap_or((false, "{}".to_string()));
out.push(PluginInfo {
id: plugin.id().to_string(),
name: plugin.name().to_string(),
description: plugin.description().to_string(),
enabled,
running: plugin.is_running(),
config: serde_json::from_str(&config_json).unwrap_or(json!({})),
config_schema: plugin.config_schema(),
runtime_status: plugin.runtime_status(),
});
}
Ok(out)
}
/// Every registered plugin, enabled or not. Lets the core ask each one for
/// its contributions (`Plugin::tools`, `Plugin::http_router`) without ever
/// naming a concrete plugin type.
pub fn all(&self) -> &[Arc<dyn Plugin>] {
&self.plugins
}
pub fn get_plugin_typed<T: Plugin + 'static>(&self, id: &str) -> Option<Arc<T>> {
self.plugins.iter()
.find(|p| p.id() == id)
.and_then(|p| Arc::clone(p).as_arc_any().downcast::<T>().ok())
}
fn find(&self, id: &str) -> Result<Arc<dyn Plugin>> {
self.plugins.iter()
.find(|p| p.id() == id)
.cloned()
.ok_or_else(|| anyhow::anyhow!("plugin not found: {id}"))
}
}