mj_controller/
server.rs

1//! Daemon-owned, phone-oriented control surface for Hel.
2//!
3//! The server deliberately owns no controller business logic. It publishes a
4//! redacted projection of controller state and forwards validated, typed
5//! actions through a channel supplied by the controller.
6
7use std::collections::BTreeMap;
8use std::convert::Infallible;
9use std::net::SocketAddr;
10use std::path::{Component, PathBuf};
11use std::sync::{Arc, Mutex};
12use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
13
14use anyhow::{Context, Result as AnyResult};
15use axum::body::{Body, Bytes, to_bytes};
16use axum::extract::{DefaultBodyLimit, Path, Query, Request, State};
17use axum::http::header::{
18    CACHE_CONTROL, CONTENT_SECURITY_POLICY as CONTENT_SECURITY_POLICY_HEADER, CONTENT_TYPE, COOKIE,
19    HeaderValue, LOCATION, REFERRER_POLICY, SET_COOKIE, X_CONTENT_TYPE_OPTIONS,
20};
21use axum::http::{HeaderMap, Response, StatusCode};
22use axum::middleware::Next;
23use axum::response::IntoResponse;
24use axum::response::sse::{Event, KeepAlive, Sse};
25use axum::routing::{get, post, put};
26use axum::{Json, Router};
27use base64::Engine as _;
28use hmac::{Hmac, KeyInit, Mac};
29use serde::{Deserialize, Serialize};
30use sha2::Sha256;
31use tokio::sync::Semaphore;
32use tokio::sync::{mpsc, watch};
33use tokio_stream::wrappers::ReceiverStream;
34use tokio_util::sync::CancellationToken;
35
36use mj_core::attachment::{AttachmentRef, AttachmentStore, MAX_IMAGE_BYTES, MAX_IMAGES};
37use mj_core::config::{Config, TargetTemplate, project_history_host, validate_id};
38use mj_core::elicitation::{ElicitationRequest, ElicitationResponse, MAX_ELICITATION_BYTES};
39use mj_core::state::{
40    MoveOperation, MovePhase, MovePreparation, MoveSelection, MoveSessionRequest,
41    ProjectSourceIdentity, SessionResourceAllocation, SessionState, SessionTransitionKind,
42    State as AppState,
43};
44
45use crate::targets::AdditionalMount;
46
47use crate::dictation::{
48    DictationError, DictationOperation, DictationRequest, DictationResponse, MAX_AUDIO_BYTES,
49    validate_wav,
50};
51use crate::image::optimize_image;
52
53pub mod api;
54
55pub use api::{
56    ApiFailure, ApiSession, PromptRequest, PromptResponse, SessionListResponse,
57    StartSessionRequest, StartSessionResponse, SubagentBackend, WaitOutcome, WaitRequest,
58    WaitResponse, api_token_path, load_or_create_api_token, map_stop_reason, resolve_wait,
59};
60
61pub use mj_client::web::{
62    BrowserDiffStat, BrowserTranscript, BrowserTranscriptEntry, WebListenerProcess,
63    WebViewerAccess, WebViewerRecovery,
64};
65
66// Keep all control surfaces on the same queue vocabulary. The resume flow
67// used to define a private copy here, which made a move request impossible to
68// pass through the web and daemon boundaries without lossy conversion.
69pub use mj_core::state::ResumeQueueDisposition;
70
71/// Select the process-wide rustls provider before any TLS configuration is built.
72///
73/// Dependency feature unification can enable both rustls providers. Rustls
74/// deliberately refuses to guess in that case, so each executable that links
75/// the controller installs the ring provider at process startup. A provider
76/// installed even earlier is already sufficient and remains in place.
77pub fn install_rustls_crypto_provider() {
78    let _ = rustls::crypto::ring::default_provider().install_default();
79}
80
81pub const COOKIE_NAME: &str = "hel_viewer_session";
82const DEFAULT_SESSION_TTL: Duration = Duration::from_secs(30 * 24 * 60 * 60);
83const EPHEMERAL_SESSION_TTL: Duration = Duration::from_secs(24 * 60 * 60);
84const MAX_BODY_BYTES: usize = 128 * 1024;
85const MAX_CODE_FAILURES: u32 = 5;
86const CODE_LOCKOUT_BASE: Duration = Duration::from_secs(30);
87const CODE_LOCKOUT_CAP: Duration = Duration::from_secs(60 * 60);
88const MAX_TITLE_CHARS: usize = 120;
89const MAX_PROMPT_CHARS: usize = 64 * 1024;
90/// How many repositories one dirty-worktree acknowledgement may name. A bundle
91/// with more repositories than this than has bigger problems than the phone.
92const MAX_DIRTY_ACKNOWLEDGEMENTS: usize = 32;
93/// The largest draft a phone may store. A composer is for a prompt, and a
94/// prompt this size has other problems; the bound exists so one viewer cannot
95/// fill the daemon's database with text it never sent.
96const MAX_DRAFT_BYTES: usize = 64 * 1024;
97/// How many prompt-history matches one search returns. Public because the
98/// controller loop performs the search and must use the same bound the phone
99/// was promised.
100pub const MAX_HISTORY_MATCHES: usize = 40;
101/// Image prompts need far more room than any other phone request. Browser
102/// uploads are base64-encoded, so two ordinary photographs already exceed the
103/// general body limit even when each one fits it. The larger bound therefore
104/// stays scoped to the action route that carries prompts.
105const MAX_PROMPT_BODY_BYTES: usize = 32 * 1024 * 1024;
106/// A browser uploads one source image at a time. The image optimizer has its
107/// own decoded-allocation bound; this is the HTTP envelope bound before that
108/// work starts.
109const MAX_ATTACHMENT_UPLOAD_BYTES: usize = 64 * 1024 * 1024;
110/// Keep two browser uploads/transcriptions in flight. The permit is acquired
111/// before reading the request body, so an overloaded client is rejected
112/// without accepting megabytes that cannot be processed yet.
113const MAX_CONCURRENT_DICTATIONS: usize = 2;
114/// Keep this in sync with the prompt admission bound and the browser composer.
115pub const MAX_PROMPT_IMAGES: usize = MAX_IMAGES;
116const COOKIE_KEY_BYTES: usize = 32;
117const COOKIE_KEY_FILE: &str = "phone-cookie-key";
118
119/// How long stored viewer state outlives its last use.
120///
121/// It matches the session cookie's own lifetime: state keyed to an identity
122/// that can no longer authenticate has nothing left to belong to.
123pub const fn default_session_ttl() -> Duration {
124    DEFAULT_SESSION_TTL
125}
126
127pub fn cookie_key_path() -> PathBuf {
128    mj_core::config::data_dir().join(COOKIE_KEY_FILE)
129}
130
131/// Load the phone cookie signing key, creating it on first use.
132///
133/// Session cookies are stateless, so this file is the only thing that keeps a
134/// signed-in phone signed in across daemon restarts. Deleting it is
135/// therefore the explicit sign-everyone-out gesture: the next start writes a
136/// new key and every outstanding cookie stops validating. A missing file is
137/// ordinary first use; an unreadable or too-short one is replaced loudly,
138/// because refusing to start would be a worse answer than asking phones to
139/// enter the viewer code again.
140pub fn load_or_create_cookie_key(path: &std::path::Path) -> AnyResult<Vec<u8>> {
141    match std::fs::read(path) {
142        Ok(key) if key.len() >= COOKIE_KEY_BYTES => return Ok(key),
143        Ok(key) => tracing::warn!(
144            path = %path.display(),
145            bytes = key.len(),
146            "phone cookie key is shorter than {COOKIE_KEY_BYTES} bytes; generating a new key signs every phone out"
147        ),
148        Err(error) if error.kind() == std::io::ErrorKind::NotFound => {}
149        Err(error) => tracing::warn!(
150            path = %path.display(),
151            "could not read the phone cookie key ({error}); generating a new key signs every phone out"
152        ),
153    }
154    let key = generate_cookie_key()?;
155    mj_core::config::atomic_write(path, &key)
156        .with_context(|| format!("persist Mjolnir phone cookie key {}", path.display()))?;
157    Ok(key.to_vec())
158}
159
160/// Options for the daemon's phone service.
161///
162/// `ServerOptions::new` generates both the six-digit viewer code and an
163/// ephemeral cookie key. A caller that wants cookies to survive server
164/// restarts installs a persisted key with `set_cookie_key`, which
165/// `load_or_create_cookie_key` reads from its private Hel data directory. The
166/// key and viewer code are intentionally omitted from `Debug` output.
167#[derive(Clone)]
168pub struct ServerOptions {
169    pub bind: SocketAddr,
170    pub snapshot_rx: watch::Receiver<ViewerSnapshot>,
171    pub conversation_rx: watch::Receiver<BTreeMap<String, BrowserTranscript>>,
172    pub action_tx: mpsc::Sender<ControllerRequest>,
173    pub bundle_tx: mpsc::Sender<BundleRequest>,
174    pub receipt_tx: mpsc::Sender<ReadReceiptRequest>,
175    pub preflight_tx: mpsc::Sender<PreflightRequest>,
176    pub move_preparation_tx: mpsc::Sender<MovePreparationRequest>,
177    pub client_state_tx: mpsc::Sender<ClientStateRequest>,
178    pub dictation_tx: mpsc::Sender<DictationRequest>,
179    /// Dedicated bounded path for stopping one live background task. This is
180    /// deliberately separate from [`ControllerAction`]: stopping a provider
181    /// task does not occupy the controller's action admission slot.
182    background_task_stop_tx: mpsc::Sender<BackgroundTaskStopRequest>,
183    pub shutdown: CancellationToken,
184    pub session_ttl: Duration,
185    /// Keep this enabled for direct HTTPS or an HTTPS reverse proxy. It may be
186    /// disabled only for an explicitly trusted HTTP development endpoint.
187    pub secure_cookie: bool,
188    tls_config: Option<axum_server::tls_rustls::RustlsConfig>,
189    viewer_code: String,
190    login_token: String,
191    cookie_key: Vec<u8>,
192    api_token: String,
193    subagent: Option<Arc<dyn api::SubagentBackend>>,
194}
195
196/// Typed request channels served by the authenticated HTTP surface.
197pub struct ServerRequests {
198    pub action_tx: mpsc::Sender<ControllerRequest>,
199    pub bundle_tx: mpsc::Sender<BundleRequest>,
200    pub receipt_tx: mpsc::Sender<ReadReceiptRequest>,
201    pub preflight_tx: mpsc::Sender<PreflightRequest>,
202    pub move_preparation_tx: mpsc::Sender<MovePreparationRequest>,
203    pub client_state_tx: mpsc::Sender<ClientStateRequest>,
204    pub dictation_tx: mpsc::Sender<DictationRequest>,
205}
206
207impl ServerOptions {
208    pub fn new(
209        bind: SocketAddr,
210        snapshot_rx: watch::Receiver<ViewerSnapshot>,
211        conversation_rx: watch::Receiver<BTreeMap<String, BrowserTranscript>>,
212        requests: ServerRequests,
213    ) -> AnyResult<Self> {
214        Ok(Self {
215            bind,
216            snapshot_rx,
217            conversation_rx,
218            action_tx: requests.action_tx,
219            bundle_tx: requests.bundle_tx,
220            receipt_tx: requests.receipt_tx,
221            preflight_tx: requests.preflight_tx,
222            move_preparation_tx: requests.move_preparation_tx,
223            client_state_tx: requests.client_state_tx,
224            dictation_tx: requests.dictation_tx,
225            background_task_stop_tx: mpsc::channel(1).0,
226            shutdown: CancellationToken::new(),
227            session_ttl: DEFAULT_SESSION_TTL,
228            secure_cookie: true,
229            tls_config: None,
230            viewer_code: generate_viewer_code()?,
231            login_token: generate_login_token()?,
232            cookie_key: generate_cookie_key()?.to_vec(),
233            // An empty token authenticates nothing: the daemon installs the
234            // persisted one, and a server without it serves the viewer only.
235            api_token: String::new(),
236            subagent: None,
237        })
238    }
239
240    pub fn viewer_code(&self) -> &str {
241        &self.viewer_code
242    }
243
244    pub fn login_token(&self) -> &str {
245        &self.login_token
246    }
247
248    /// Serve HTTPS directly using the supplied Rustls configuration. Hel's
249    /// CLI can load its persisted certificate (including a Tailscale-issued
250    /// certificate) and pass it here without coupling this module to disk.
251    pub fn set_tls_config(&mut self, config: axum_server::tls_rustls::RustlsConfig) {
252        self.tls_config = Some(config);
253        self.secure_cookie = true;
254    }
255
256    /// Install a persisted signing key. Rotating this value signs every phone
257    /// out without maintaining a server-side session database.
258    pub fn set_cookie_key(&mut self, key: Vec<u8>) -> AnyResult<()> {
259        anyhow::ensure!(
260            key.len() >= COOKIE_KEY_BYTES,
261            "cookie signing key must be at least {COOKIE_KEY_BYTES} bytes"
262        );
263        self.cookie_key = key;
264        Ok(())
265    }
266
267    /// Install the controller's bounded background-task stop path.
268    pub fn set_background_task_stop_tx(&mut self, tx: mpsc::Sender<BackgroundTaskStopRequest>) {
269        self.background_task_stop_tx = tx;
270    }
271
272    /// Install the persisted bearer token for the `/api/v1` routes. Rotating
273    /// it revokes every client that still holds the old one.
274    pub fn set_api_token(&mut self, token: String) {
275        self.api_token = token;
276    }
277
278    /// Install the daemon-side backend the `/api/v1` routes drive sessions
279    /// through. Without it those routes answer 503.
280    pub fn set_subagent_backend(&mut self, backend: Arc<dyn api::SubagentBackend>) {
281        self.subagent = Some(backend);
282    }
283
284    #[cfg(test)]
285    fn with_test_credentials(mut self, code: &str, key: &[u8]) -> Self {
286        self.viewer_code = code.to_string();
287        self.login_token = "test-login-token".into();
288        self.cookie_key = key.to_vec();
289        self.secure_cookie = false;
290        self.api_token = "test-api-token".into();
291        self
292    }
293}
294
295/// Run the phone server until its shutdown token is cancelled.
296///
297/// This binds only the requested listener. It does not daemonize, provision a
298/// target, or keep sessions alive: controller availability is required, just
299/// like MJ's explicit remote-viewer model.
300pub async fn run_server(options: ServerOptions) -> AnyResult<()> {
301    let listener = tokio::net::TcpListener::bind(options.bind)
302        .await
303        .with_context(|| format!("bind web viewer to {}", options.bind))?;
304    run_server_on_listener(options, listener).await
305}
306
307/// Serve a reserved socket so readiness and advertised ports reflect a real listener.
308pub async fn run_server_on_listener(
309    options: ServerOptions,
310    listener: tokio::net::TcpListener,
311) -> AnyResult<()> {
312    let mut options = options;
313    let bind = listener.local_addr().context("read web viewer address")?;
314    let shutdown = options.shutdown.clone();
315    let viewer_code = options.viewer_code.clone();
316    let tls_config = options.tls_config.take();
317    let app = router(options);
318    println!("Mjolnir viewer code: {viewer_code}");
319    let listener = listener.into_std().context("prepare web viewer listener")?;
320    let handle = axum_server::Handle::new();
321    let shutdown_handle = handle.clone();
322    let serve = async move {
323        if let Some(tls_config) = tls_config {
324            axum_server::from_tcp_rustls(listener, tls_config)
325                .handle(handle)
326                .serve(app.into_make_service())
327                .await
328        } else {
329            axum_server::from_tcp(listener)
330                .handle(handle)
331                .serve(app.into_make_service())
332                .await
333        }
334    };
335    tokio::pin!(serve);
336    tokio::select! {
337        result = &mut serve => result,
338        _ = shutdown.cancelled() => {
339            shutdown_handle.graceful_shutdown(Some(Duration::from_secs(2)));
340            serve.await
341        }
342    }
343    .with_context(|| format!("serve web viewer on {bind}"))
344}
345
346mod viewer_types;
347pub use viewer_types::*;
348mod actions;
349pub use actions::*;
350mod routes;
351use routes::*;
352mod handlers;
353use handlers::*;
354mod validation;
355use validation::*;
356mod errors;
357use errors::*;
358mod auth;
359pub use auth::*;
360mod assets;
361use assets::*;
362mod config_view;
363pub use config_view::*;
364
365#[cfg(test)]
366mod tests;
mj_controller/server.rs

mj_controller/
server.rs
