Startup and Configuration

July 15, 2026 · View on GitHub

This guide explains Mostrix’s boot sequence and configuration surfaces.

Overview

  • Entry: src/main.rs:98
  • Initializes database, derives identity keys, initializes settings, then logger, terminal (raw mode), shared state, Nostr client, and background tasks.
  • Shows an animated startup splash (Mostro wordmark + loading dots) while post-terminal init runs; see Startup splash below.
  • Enters the main event loop to handle UI updates and user input.

Startup splash

After the terminal enters alternate screen mode, Mostrix draws a full-screen splash until background boot work finishes (src/startup.rs, src/ui/startup_splash.rs).

  • Wordmark: multi-line logo from the project art (same style as the desktop logo.txt export).
  • Loading dots: reusable glyph <> repeated 1–4 times on the last logo row, each prefixed by a space, cycled every ~400 ms while the splash tick runs (~150 ms redraw interval).
  • Phase text: short status under the art (Starting…, Connecting to relays…, Loading market data…, Restoring chats…, Almost ready…) updated as init steps complete.
  • Minimum display: splash stays visible for at least ~800 ms so fast boots do not flash the screen.
  • Narrow terminals: if the terminal is narrower than the padded logo width, a one-line mostro is loading + dots + phase is shown instead.
  • Background: full-screen fill via fill_splash_background (solid BACKGROUND_COLOR block, same pattern as the Exit tab) so ASCII art and status text do not show mismatched per-span backgrounds.
  • CI / scripts: set MOSTRIX_NO_SPLASH=1 to skip the splash loop and run init directly.

Initialization Sequence

1. Database Initialization

The database is initialized at startup to ensure the schema is ready.

Source: src/db.rs

  • Creates the SQLite database file at ~/.mostrix/mostrix.db.
  • Ensures tables exist (orders, users).
  • If the users table is empty, User::new() generates a new 12-word BIP-39 mnemonic and persists it in the users table (this mnemonic is the root for user identity/trade key derivation).
  • For existing databases, runs migrations automatically to keep the schema up to date.

2. Settings Initialization

Mostrix uses centralized settings management in src/settings.rs.

Source: src/settings.rs

pub fn init_settings(identity_keys: Option<Keys>)
    -> Result<InitSettingsResult, anyhow::Error>
  • On first run, settings.toml is generated from an embedded template compiled into the binary (rather than copying from the repo root).
  • If identity_keys is provided (derived from the DB identity/index-0 key), Mostrix derives the nsec_privkey for settings.toml so DB keys and settings keys match.
  • The returned InitSettingsResult.did_generate_new_settings_file indicates whether this process generated a brand-new settings.toml.
  • When did_generate_new_settings_file is true, main.rs shows the BackupNewKeys popup overlay immediately on the current initial tab, prompting the user to save the generated 12-word mnemonic.

Error Handling: Startup failures in init_settings() are propagated as anyhow::Error (causing a clean process exit with an error message). If settings are accessed later at runtime before initialization (via the SETTINGS global), those failures are surfaced as user-friendly messages using OperationResult::Error instead of panicking. This ensures graceful degradation and clear feedback to users in both cases.

3. Logger Setup

Logging is configured via setup_logger in src/main.rs.

Source: src/main.rs:41

fn setup_logger(level: &str) -> Result<(), fern::InitError> {
    let log_level = match level.to_lowercase().as_str() {
        // ... level mapping ...
    };
    Dispatch::new()
        .format(|out, message, record| {
            out.finish(format_args!(
                "[{}] [{}] - {}",
                Local::now().format("%Y-%m-%d %H:%M:%S"),
                record.level(),
                message
            ))
        })
        .level(log_level)
        .chain(fern::log_file("app.log")?) // Writes to app.log
        .apply()?;
    Ok(())
}
  • Sets the log level based on the log_level field in settings.toml.
  • Outputs log messages to app.log.

4. TUI Initialization

The TUI uses ratatui with the crossterm backend.

Source: src/main.rs:104

    enable_raw_mode()?;
    let mut out = stdout();
    execute!(
        out,
        EnterAlternateScreen,
        crossterm::event::EnableMouseCapture
    )?;
    let backend = CrosstermBackend::new(out);
    let mut terminal = Terminal::new(backend)?;
  • Enables terminal raw mode.
  • Enters the alternate screen and enables mouse capture.

Configuration Structure

The Settings struct defines all available configuration options.

Source: src/settings.rs

#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct Settings {
    pub mostro_pubkey: String,
    pub nsec_privkey: String,
    pub admin_privkey: String,
    pub relays: Vec<String>,
    pub log_level: String,
    pub currencies_filter: Vec<String>,
    #[serde(default = "default_user_mode")]
    pub user_mode: String, // "user" or "admin", default "user"
    #[serde(default)]
    pub ln_address: String, // Lightning address for buyer receive; empty = unset
    #[serde(default)]
    pub blossom_servers: Vec<String>, // Blossom upload hosts; empty = built-in defaults
}

Fields:

  • mostro_pubkey: The public key of the Mostro instance to interact with.
  • nsec_privkey: The user's Nostr private key (nsec format).
  • admin_privkey: The admin's private key, required for solving disputes when in admin mode.
  • relays: A list of Nostr relay URLs to connect to.
  • log_level: The verbosity of logging (e.g., "debug", "info", "warn", "error").
  • currencies_filter: Optional list of fiat currency filters (ISO codes).
    • When empty, all currencies published by the Mostro instance are shown.
    • When non-empty (e.g. ["USD"], ["USD", "EUR"]), only orders whose fiat code is in this list are displayed.
  • user_mode: Either "user" or "admin". Controls the UI and available actions.
  • ln_address: Optional Lightning address (user@domain.com) used when the local user acts as buyer (receive via LNURL-pay). The embedded template includes ln_address = "". Older settings.toml files without this key still load (#[serde(default)] yields an empty string). Saving from the Settings tab runs an async check that the LNURL metadata URL returns JSON with tag: "payRequest" before writing disk (spawn_verify_and_save_ln_address_task in src/ui/key_handler/async_tasks.rs, helper in src/util/ln_address.rs). The spawned task reports on ln_address_result_tx (LnAddressVerifyResult), not on order_result_tx, so settings verification does not share the order/dispute result queue. Clear removes the value without a network call.
  • blossom_servers: Optional list of HTTPS Blossom bases for My Trades attachment upload (Ctrl+O send). When empty, Mostrix uses DEFAULT_BLOSSOM_SERVERS in src/util/blossom.rs (same defaults as Mostro Mobile). Example in repo settings.toml: commented # blossom_servers = ["https://blossom.primal.net", …]. Resolved at send time via blossom_servers_from_settings in src/util/send_attachment.rs (main loop reloads settings from disk when draining the send queue).

Proof-of-work for published events is taken from the Mostro instance status event (kind 38385, tag pow), not from settings.toml.

Mostro instance info (kind 38385)

Background and manual refresh (Mostro Info tab → Enter) fetch the daemon status event and update UI state:

  • AppState.mostro_info: parsed tags (pow, bond_enabled, protocol_version, LND metadata, …) — see mostro_info_from_tags.
  • AppState.transport: resolved wire transport for protocol DMs via transport_from_instance. Updated through AppState.set_mostro_info (startup await, main loop MostroInfoFetchResult, reconnect, invalid-pubkey clear).
  • Startup: when relays are reachable, run_post_terminal_startup awaits fetch_mostro_instance_info before spawning the DM listener so the first subscription uses the correct transport (v1 GiftWrap or v2 kind 14). On fetch failure or offline boot, transport defaults to GiftWrap.

Displayed on the Mostro Info tab: protocol version (1 / 2 / unknown) and wire transport label (GiftWrap vs NIP-44 direct).

Nostr & Background Tasks

Nostr Client Connection

Mostrix initializes a nostr_sdk::Client with the user's keys, adds configured relays, and connects using a panic-safe wrapper (connect_client_safely).

Current startup behavior:

  • Trims relay strings and skips empty entries before adding.
  • Computes relays_reachable with any_relay_reachable for offline UI behavior.
  • Calls connect_client_safely(&client) (instead of raw client.connect().await) to prevent background panic crashes when connectivity is unstable.
  • Logs a warning if no configured relays are reachable at boot.

Background Tasks

Several background tasks are spawned to keep the UI and data in sync:

  1. Order Refresh: Periodically fetches pending orders from Mostro.
  2. Relay order DB reconcile (startup + ~30s orders updater): run_relay_order_db_reconcile_once (bulk terminal sync from nostr order events) and run_targeted_relay_order_db_reconcile_tick (round-robin per-order fetch for local non-terminal trades with keys). See relay_order_db_reconcile.rs and MESSAGE_FLOW_AND_PROTOCOL.md (Relay → SQLite section).
  3. Trade Message Listener: Listens for new messages related to active orders.
  4. Network Status Monitor:
    • spawn_network_status_monitor runs every 5 seconds.
    • Re-checks relay reachability from disk settings and emits NetworkStatus::Offline/Online.
    • On Offline, startup overlay text indicates automatic retry.
    • On Online, main.rs triggers reload_runtime_session_after_reconnect(...) to reconnect and reload runtime background tasks.
  5. Shared-key chat subscription router (listen_for_chat_messages in src/util/chat_listener.rs):
    • A single long-lived task (spawned in src/startup.rs, respawned on reconnect/key reload) maintains one relay subscription whose filter batches all active shared-key pubkeys (kind: 1059, #p: [shared pubkeys]) — the same model as Mostro Mobile's SubscriptionManager. No timed polling.
    • Track/untrack via the global command channel (ChatRouterCmd, published by set_chat_router_cmd_tx). Helpers: track_order_chat / untrack_order_chat / track_dispute_chat / untrack_dispute_chat.
    • On TrackChatKey, the router hydrates history once (fetch_gift_wraps_for_shared_key, filtered by the last-seen cursor). Buffered track/untrack commands are drained and applied before a single batched resubscribe (idempotent re-tracks are no-ops). Subscribe failures retry with short backoff.
    • Live kind: 1059 gift wraps are routed by the event's p tag to the owning shared key, decrypted with unwrap_giftwrap_with_shared_key, and emitted on the existing admin_chat_updates / user_order_chat_updates channels.
    • Application is unchanged: ui::helpers::apply_admin_chat_updates / apply_user_order_chat_updates (in src/ui/helpers/startup.rs) still merge updates, persist cursors (buyer_chat_last_seen / seller_chat_last_seen, order transcripts), and dedupe.
    • Startup track set (option B): track_startup_chats emits tracks for Order::get_startup_active_orders rows (active + success) with a resolvable shared key (User), and each InProgress dispute's buyer/seller shared key (Admin).
    • Untrack when an order reaches TERMINAL_DM_STATUSES or its row is removed/reverted (DM router hooks), or a dispute leaves InProgress. Chat is not untracked on success; on-disk transcripts preserve history for untracked terminal orders.

Source: src/util/chat_listener.rs (router), src/startup.rs + src/ui/helpers/startup.rs (track_startup_chats), src/util/dm_utils/mod.rs (track/untrack hooks), src/ui/helpers/startup.rs (apply_admin_chat_updates)

  1. DM Router Wiring (trade messages):
    • App channel creation includes dm_subscription_tx / dm_subscription_rx.
    • set_dm_router_cmd_tx(dm_subscription_tx.clone()) publishes the sender globally for wait_for_dm (returns Result; startup fails fast if the mutex is poisoned).
    • Before spawning the listener, hydrate_startup_active_order_dm_state loads non-terminal orders from SQLite and returns active_order_trade_indices plus order_last_seen_dm_ts cursors; main.rs seeds the shared active-order map.
    • listen_for_order_messages(client, mostro_pubkey, transport, pool, …, order_last_seen_dm_ts, …, dm_subscription_rx) runs as the single router loop consuming:
      • TrackOrder commands for long-lived trade subscriptions.
      • RegisterWaiter commands for one-shot request/response waits.
    • After bootstrapping per-order protocol-DM subscriptions (ensure_order_dm_subscription), the listener performs a fetch_events replay (fetch_and_replay_startup_trade_dms) so the Messages UI is populated from relay history (in-memory messages are not stored in the DB). Replay uses notify: false to avoid duplicate popups/badge noise.
    • Startup transport: startup.rs awaits instance info when relays are reachable, then spawns the listener with resolved app.transport.
    • Reload / reconnect transport: dm_transport_for_mostro re-fetches instance info and updates app.transport before respawning the listener (key reload, fetch-scheduler reload, network reconnect).
    • This unifies in-flight response handling and background trade notifications on top of one notification stream.

See DM_LISTENER_FLOW.md for DmSubscriptionMode (StartupCatchUp, StartupSince, LiveOnly), filter_protocol_dm_from_mostro, waiter vs TrackOrder ordering, and replay details.

Admin Chat Restore at Startup

In addition to the background scheduler, Mostrix restores admin chat state during startup:

  • All persisted admin disputes are loaded from the admin_disputes table.
  • For disputes in InProgress state, ui::helpers::recover_admin_chat_from_files:
    • Reads chat transcripts from ~/.mostrix/disputes_chat/<dispute_id>.txt (if present).
    • Reconstructs AppState.admin_dispute_chats so the "Disputes in Progress" tab immediately shows prior messages.
    • Updates in‑memory admin_chat_last_seen entries for Buyer and Seller based on file timestamps.
  • Subsequent background NIP‑59 fetches use the stored buyer_chat_last_seen / seller_chat_last_seen values as cursors, ensuring:
    • Instant UI restore after restart.
    • Incremental network sync without replaying the full chat history from relays.

User order chat restore at startup (My Trades)

For User role, Mostrix restores peer-to-peer order chat alongside trade DMs:

  • Cached transcripts live under ~/.mostrix/orders_chat/<order_id>.txt and are loaded into AppState.order_chats by load_user_order_chats_at_startup.
  • Attachment rows in transcripts are stored as JSON (image_encrypted / file_encrypted via serialize_attachment_for_transcript) so Ctrl+S and file counts work immediately after restart; legacy [Image: … - Ctrl+S to save] lines are hydrated in memory when relay returns the same attachment at the same timestamp.
  • Disk restore via load_user_order_chats_at_startup seeds AppState.order_chats and order_chat_last_seen. Relay history is hydrated once by the shared-key chat subscription router when track_startup_chats emits TrackChatKey — no separate startup poll and no timed polling.
  • apply_user_order_chat_updates skips relay echoes of the local trade pubkey; peer dedup is scoped to existing Peer rows so optimistic You sends are not mirrored as Peer and do not suppress unrelated peer text at the same timestamp. See MESSAGE_FLOW_AND_PROTOCOL.md — "User order chat local cache".

Main Event Loop

The TUI runs in a tokio::select! loop that handles (among others):

  1. Fatal errors: fatal_error_rx — aborts background work and shows an error popup.
  2. Network status: network_status_rx — offline overlay vs reconnect + runtime reload.
  3. Order / dispute / attachment / observer async results: order_result_rx — OperationResult; includes dispute-list refresh side effects for certain Info messages and My Trades DB resync for OrderHistoryDeleted.
  4. Lightning address verify-and-save (settings): ln_address_result_rx — LnAddressVerifyResult; mapped to OperationResult::Info / Error and passed to handle_operation_result so UI behavior matches other operation-result popups without mixing traffic into order_result_rx.
  5. Key rotation / seed words / message notifications / admin & user chat fetches / Mostro instance info / user input / periodic ticks: see src/main.rs (create_app_channels in src/ui/key_handler/async_tasks.rs lists all paired senders and receivers, including save_attachment_tx/rx for Ctrl+S downloads and send_order_attachment_tx/rx for outbound My Trades uploads via SendOrderAttachmentJob). User order chat results arrive on user_order_chat_updates_rx and are applied via apply_user_order_chat_updates.

Source: src/main.rs (outer loop + tokio::select! + terminal.draw).

// Simplified shape (not exhaustive — see src/main.rs for full select!)
loop {
    tokio::select! {
        // fatal_error_rx, network_status_rx, ...
        result = order_result_rx.recv() => { apply_order_result(...) }
        ln_address_verify = ln_address_result_rx.recv() => { /* map LnAddressVerifyResult → OperationResult */ }
        // key_rotation_rx, seed_words_rx, message_notification_rx, ...
        maybe_event = events.next() => { /* handle_key_event, paste, mouse */ }
        _ = refresh_interval.tick() => { /* 150 ms — redraw even without input */ }
    }
    // Before every frame (not only on keypress):
    drain_save_attachment_queue(...)        // start Blossom downloads queued by Ctrl+S popups
    drain_send_order_attachment_queue(...)  // Ctrl+O / Ctrl+Shift+O jobs: encrypt → Blossom → DM
    drain_order_result_queue(...)           // apply OperationResult (e.g. "Saved to …", "Attachment sent: …")
    expire_attachment_toast(&mut app);
    terminal.draw(|f| ui_draw(f, &app, &orders, Some(&status_line)))?;
}

Why drain before draw: My Trades Enter on the save-attachment popup may enqueue the download asynchronously (DB lookup for decryption key). Outbound sends enqueue on send_order_attachment_rx the same way (FromPath or RetryPrepared). Without draining save_attachment_rx, send_order_attachment_rx, and order_result_rx on each frame, success/error popups (including upload-ok/send-failed) could appear only after an unrelated keypress. The 150 ms refresh_interval tick plus this drain keeps attachment save and send feedback timely.