fix(nostr): discovery events expire, heartbeat, and tombstone — stale nodes age out

Presence gets a NIP-40 expiration (48h) and a 12h re-publish heartbeat that
honours the runtime toggle (UI-enabled nodes previously never re-published
at boot). discover() drops pre-TTL events client-side for relays that
ignore NIP-40. Switching discovery off publishes an empty tombstone, and
factory-reset tombstones BEFORE wiping identity — after the wipe the key is
gone and the stale event could never be replaced by anyone.

nostr.discovery-status now also returns the node's own npub (load-only).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

core/archipelago/src/api/rpc/handshake.rs

@@ -21,7 +21,7 @@ use anyhow::{Context, Result};
 use nostr_sdk::FromBech32;
 use serde::{Deserialize, Serialize};
 
-const NOSTR_STATE_FILE: &str = "nostr_discovery_state.json";
+use crate::nostr_handshake::DISCOVERY_STATE_FILE as NOSTR_STATE_FILE;
 
 /// Runtime override for `Config::nostr_discovery_enabled`. The OS-level
 /// config file is read once at boot and is OFF by default; this state file
@@ -55,10 +55,16 @@ async fn save_discovery_state(
 }
 
 impl RpcHandler {
-    /// Read the current runtime discoverability flag.
+    /// Read the current runtime discoverability flag. Also returns the npub
+    /// this node publishes as (the discoverability UI shows it — that npub,
+    /// not the onion, is what's actually visible on the relays). Load-only:
+    /// null until discovery keys exist.
     pub(super) async fn handle_nostr_discovery_status(&self) -> Result<serde_json::Value> {
         let state = load_discovery_state(&self.config.data_dir).await;
-        Ok(serde_json::json!({ "enabled": state.enabled }))
+        let npub = nostr_handshake::own_npub(&self.config.data_dir.join("identity"))
+            .await
+            .unwrap_or(None);
+        Ok(serde_json::json!({ "enabled": state.enabled, "npub": npub }))
     }
 
     /// Set the runtime discoverability flag. If turning ON, publish presence
@@ -101,6 +107,24 @@ impl RpcHandler {
                     tracing::warn!("Initial presence publish failed: {}", e);
                 }
             });
+        } else if !enabled {
+            // Switching off: overwrite our presence with an empty tombstone so
+            // the node disappears from other nodes' discovery lists now, not
+            // at the next TTL expiry.
+            let identity_dir = self.config.data_dir.join("identity");
+            let relays = self.handshake_relays().await;
+            let tor_proxy = self.config.nostr_tor_proxy.clone();
+            tokio::spawn(async move {
+                if let Err(e) = nostr_handshake::publish_tombstone(
+                    &identity_dir,
+                    &relays,
+                    tor_proxy.as_deref(),
+                )
+                .await
+                {
+                    tracing::warn!("Presence tombstone publish failed: {}", e);
+                }
+            });
         }
 
         Ok(serde_json::json!({ "enabled": enabled }))

core/archipelago/src/api/rpc/system/handlers.rs

@@ -921,6 +921,29 @@ impl RpcHandler {
             return Err(anyhow::anyhow!("Password Incorrect"));
         }
 
+        // Overwrite our Nostr presence with a tombstone BEFORE the wipe: the
+        // discovery keys die with the identity dir, and once they're gone the
+        // stale presence event can never be replaced by anyone — it would
+        // list this dead install to the whole network until relays expire it.
+        // Best-effort with a hard cap so a dead relay can't stall the reset.
+        {
+            let identity_dir = self.config.data_dir.join("identity");
+            let relays = crate::nostr_relays::merged_relay_list(
+                &self.config.data_dir,
+                &self.config.nostr_relays,
+            )
+            .await;
+            let _ = tokio::time::timeout(
+                std::time::Duration::from_secs(15),
+                crate::nostr_handshake::publish_tombstone(
+                    &identity_dir,
+                    &relays,
+                    self.config.nostr_tor_proxy.as_deref(),
+                ),
+            )
+            .await;
+        }
+
         tracing::warn!("Factory reset initiated — wiping ALL user data and containers");
 
         let data_dir = &self.config.data_dir;

core/archipelago/src/nostr_handshake.rs

@@ -35,6 +35,37 @@ use tracing::warn;
 
 const NOSTR_SECRET_FILE: &str = "nostr_secret";
 
+/// Runtime discoverability override written by the `nostr.set-discovery` RPC.
+/// Lives here (not api/rpc) so the server's heartbeat can honour the same
+/// state the toggle writes.
+pub const DISCOVERY_STATE_FILE: &str = "nostr_discovery_state.json";
+
+/// How long a presence event stays valid. Published as a NIP-40 expiration
+/// tag AND enforced client-side in `discover` (relay NIP-40 support varies).
+/// Must be comfortably longer than the re-publish heartbeat (12h in
+/// server.rs) so a node that misses one heartbeat doesn't vanish: 48h
+/// tolerates three misses.
+pub const PRESENCE_TTL_SECS: u64 = 48 * 3600;
+
+/// Read the runtime discovery override. `None` means the toggle has never
+/// been used on this node — callers fall back to the config flag.
+pub async fn discovery_enabled_override(data_dir: &Path) -> Option<bool> {
+    let raw = fs::read_to_string(data_dir.join(DISCOVERY_STATE_FILE))
+        .await
+        .ok()?;
+    let v: serde_json::Value = serde_json::from_str(&raw).ok()?;
+    v.get("enabled").and_then(|e| e.as_bool())
+}
+
+/// This node's own published npub (bech32), if discovery keys exist.
+/// Load-only: never mints keys on a read.
+pub async fn own_npub(identity_dir: &Path) -> Result<Option<String>> {
+    Ok(load_nostr_keys(identity_dir)
+        .await?
+        .map(|k| k.public_key().to_bech32().unwrap_or_default())
+        .filter(|s| !s.is_empty()))
+}
+
 /// Message types exchanged inside NIP-44 encrypted DMs (kind 4).
 ///
 /// Note: NONE of these variants carry an onion address. The onion is only
@@ -164,8 +195,13 @@ pub async fn publish_presence(
         warn!("Nostr relay connection timed out after 10s, continuing anyway");
     }
 
-    let builder =
+    // NIP-40 expiration: relays that honour it garbage-collect the event if
-        EventBuilder::new(Kind::Custom(30078), content).tag(Tag::identifier("archipelago-node"));
+    // this node stops heartbeating (reinstall, decommission, long outage).
+    // `discover` enforces the same window client-side for relays that don't.
+    let expires = Timestamp::from(Timestamp::now().as_u64() + PRESENCE_TTL_SECS);
+    let builder = EventBuilder::new(Kind::Custom(30078), content)
+        .tag(Tag::identifier("archipelago-node"))
+        .tag(Tag::expiration(expires));
     let _ = client.send_event_builder(builder).await;
     client.disconnect().await;
 
@@ -176,6 +212,43 @@ pub async fn publish_presence(
     Ok(())
 }
 
+/// Overwrite this node's presence with an empty tombstone (NIP-33: same
+/// author + kind + d-tag replaces). Called when discovery is switched off
+/// and — critically — during factory-reset BEFORE the keys are wiped: once
+/// the secret is gone, nothing can ever replace the stale event.
+pub async fn publish_tombstone(
+    identity_dir: &Path,
+    relays: &[String],
+    tor_proxy: Option<&str>,
+) -> Result<()> {
+    if relays.is_empty() {
+        return Ok(());
+    }
+    let Some(keys) = load_nostr_keys(identity_dir).await? else {
+        return Ok(()); // never published — nothing to tombstone
+    };
+    let client = build_client(keys, tor_proxy)?;
+    for url in relays {
+        let _ = client.add_relay(url).await;
+    }
+    if tokio::time::timeout(Duration::from_secs(10), client.connect())
+        .await
+        .is_err()
+    {
+        warn!("Nostr relay connection timed out after 10s, continuing anyway");
+    }
+    // Tombstone also expires: after TTL the relay may drop it entirely,
+    // which is the desired end state (nothing left to list).
+    let expires = Timestamp::from(Timestamp::now().as_u64() + PRESENCE_TTL_SECS);
+    let builder = EventBuilder::new(Kind::Custom(30078), "{}")
+        .tag(Tag::identifier("archipelago-node"))
+        .tag(Tag::expiration(expires));
+    let _ = client.send_event_builder(builder).await;
+    client.disconnect().await;
+    tracing::info!("🔒 Published presence tombstone to {} relays", relays.len());
+    Ok(())
+}
+
 /// Discover other Archipelago nodes (presence-only — no onion addresses).
 /// Returns Nostr pubkeys and DIDs of discoverable nodes.
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -221,7 +294,17 @@ pub async fn discover_nodes(
     client.disconnect().await;
 
     let mut nodes = Vec::new();
+    let stale_cutoff = Timestamp::from(Timestamp::now().as_u64().saturating_sub(PRESENCE_TTL_SECS));
     for event in events {
+        // Client-side staleness enforcement: pre-TTL events (and events from
+        // relays that ignore NIP-40) would otherwise list dead installs
+        // forever — every reinstall mints a new key, so the old author can
+        // never replace its own event.
+        if event.created_at < stale_cutoff {
+            continue;
+        }
+        // A tombstone ("{}" content) parses but yields no pubkey — the
+        // nostr_pubkey.is_empty() guard below already drops it.
         if let Ok(content) = serde_json::from_str::<serde_json::Value>(&event.content) {
             let nostr_pubkey = content
                 .get("nostr_pubkey")

core/archipelago/src/server.rs

@@ -212,7 +212,15 @@ impl Server {
 
         // Publish presence-only to Nostr (DID + Nostr pubkey, NO onion address).
         // Onion addresses are exchanged privately via NIP-44 encrypted DMs.
-        if config.nostr_discovery_enabled && !config.nostr_relays.is_empty() {
+        //
+        // This is a heartbeat, not a one-shot: presence events carry a NIP-40
+        // expiration of PRESENCE_TTL_SECS, so a node that stops re-publishing
+        // ages out of discovery instead of lingering forever. First tick runs
+        // immediately (preserving the old startup-publish behaviour); the
+        // runtime toggle (nostr.set-discovery) is re-read every tick, so a
+        // node switched on via the UI heartbeats too — not just ones with the
+        // config flag baked in.
+        {
             let identity_dir = config.data_dir.join("identity");
             let did =
                 identity::did_key_from_pubkey_hex(&data.server_info.pubkey).unwrap_or_default();
@@ -221,11 +229,22 @@ impl Server {
             // where handshake peers actually read (2026-07-22 unification).
             let data_dir_for_relays = config.data_dir.clone();
             let config_relays = config.nostr_relays.clone();
+            let config_flag = config.nostr_discovery_enabled;
             let tor_proxy = config.nostr_tor_proxy.clone();
             tokio::spawn(async move {
-                let relays =
+                const HEARTBEAT_SECS: u64 = 12 * 3600; // < PRESENCE_TTL_SECS/3
-                    crate::nostr_relays::merged_relay_list(&data_dir_for_relays, &config_relays)
+                loop {
+                    let enabled =
+                        nostr_handshake::discovery_enabled_override(&data_dir_for_relays)
+                            .await
+                            .unwrap_or(config_flag);
+                    if enabled {
+                        let relays = crate::nostr_relays::merged_relay_list(
+                            &data_dir_for_relays,
+                            &config_relays,
+                        )
                         .await;
+                        if !relays.is_empty() {
                             if let Err(e) = nostr_handshake::publish_presence(
                                 &identity_dir,
                                 &did,
@@ -237,6 +256,10 @@ impl Server {
                             {
                                 tracing::debug!("Nostr presence publish (non-fatal): {}", e);
                             }
+                        }
+                    }
+                    tokio::time::sleep(std::time::Duration::from_secs(HEARTBEAT_SECS)).await;
+                }
             });
         }
         info!(