huskies/server/src/chat/transport/matrix/bot/context.rs

//! Matrix bot context — shared state for the Matrix bot (rooms, history, permissions).
use crate::chat::ChatTransport;
use crate::service::gateway::config::ProjectEntry;
use crate::service::timer::TimerStore;
use crate::services::Services;
use matrix_sdk::ruma::{OwnedEventId, OwnedRoomId, OwnedUserId};
use std::collections::{BTreeMap, HashSet, VecDeque};
use std::sync::Arc;
use std::sync::atomic::AtomicI64;
use tokio::sync::Mutex as TokioMutex;
use tokio::sync::RwLock;

use super::history::ConversationHistory;

/// Maximum number of incoming event IDs retained for deduplication.
///
/// FIFO ring (VecDeque + HashSet): when full, the oldest entry is evicted.
/// Matrix sync replays are temporally clustered — any event replayed more than
/// 1 000 events later was long since processed, so FIFO eviction is correct.
/// 1 000 × ~80 B ≈ 80 KB, negligible memory cost.
pub const SEEN_EVENT_IDS_CAP: usize = 1_000;

/// Bounded FIFO set for deduplicating incoming Matrix event IDs.
pub struct SeenEventIds {
    deque: VecDeque<OwnedEventId>,
    set: HashSet<OwnedEventId>,
    cap: usize,
}

impl SeenEventIds {
    /// Create a new set with the given capacity cap.
    pub fn new(cap: usize) -> Self {
        Self {
            deque: VecDeque::with_capacity(cap),
            set: HashSet::with_capacity(cap),
            cap,
        }
    }

    /// Insert an event ID. Returns `true` if the ID was new (never seen),
    /// `false` if already present (duplicate — caller should skip processing).
    /// When the set is full, the oldest entry is evicted before inserting.
    pub fn insert(&mut self, id: OwnedEventId) -> bool {
        if self.set.contains(&id) {
            return false;
        }
        if self.deque.len() == self.cap
            && let Some(old) = self.deque.pop_front()
        {
            self.set.remove(&old);
        }
        self.deque.push_back(id.clone());
        self.set.insert(id);
        true
    }
}

/// Shared context injected into Matrix event handlers.
#[derive(Clone)]
pub struct BotContext {
    /// Shared services bundle (project root, agent pool, bot identity, permissions).
    pub services: Arc<Services>,
    /// Matrix-specific parsed user ID (e.g. `@timmy:homeserver.local`).
    /// Transport-specific — kept separate from `services.bot_user_id` (String)
    /// because Matrix SDK APIs require `OwnedUserId` for comparisons and
    /// `.localpart()` extraction.
    pub matrix_user_id: OwnedUserId,
    /// All room IDs the bot listens in.
    pub target_room_ids: Vec<OwnedRoomId>,
    pub allowed_users: Vec<String>,
    /// Shared, per-room rolling conversation history.
    pub history: ConversationHistory,
    /// Maximum number of entries to keep per room before trimming the oldest.
    pub history_size: usize,
    /// Event IDs of messages the bot has sent.  Used to detect replies to the
    /// bot so it can continue a conversation thread without requiring an
    /// explicit `@mention` on every follow-up.
    pub bot_sent_event_ids: Arc<TokioMutex<HashSet<OwnedEventId>>>,
    /// Per-room htop monitoring sessions.  Keyed by room ID; each entry holds
    /// a stop-signal sender that the background task watches.
    pub htop_sessions: super::super::htop::HtopSessions,
    /// Chat transport used for sending and editing messages.
    ///
    /// All message I/O goes through this abstraction so the bot logic works
    /// with any platform, not just Matrix.
    pub transport: Arc<dyn ChatTransport>,
    /// Persistent store for pending deferred-start timers.
    pub timer_store: Arc<TimerStore>,
    /// In gateway mode: the currently active project (shared with the gateway HTTP handler).
    /// `None` in standalone single-project mode.
    pub gateway_active_project: Option<Arc<RwLock<String>>>,
    /// In gateway mode: shared live projects map from [`GatewayState`].
    ///
    /// The `new project` command writes here so HTTP handlers see the new entry
    /// immediately without requiring a gateway restart.  `None` in standalone mode.
    pub gateway_projects_store: Option<Arc<RwLock<BTreeMap<String, ProjectEntry>>>>,
    /// In gateway mode: shared configured release channels (story 1169).
    ///
    /// Read by the `pull <channel>` command to resolve a channel's
    /// `base_url`, pinned `pubkey`, and optional `bearer_token`. `None` in
    /// standalone mode.
    pub gateway_channels_store: Option<
        Arc<RwLock<BTreeMap<String, crate::service::gateway::config::ReleaseChannelConfig>>>,
    >,
    /// Bounded FIFO set of already-handled incoming event IDs.
    ///
    /// The Matrix sync loop can replay events on reconnect. This set ensures
    /// each event is processed at most once. Insert the event ID before any
    /// side-effecting work; return early if the insert returns `false`.
    pub handled_incoming_event_ids: Arc<TokioMutex<SeenEventIds>>,
    /// In gateway mode: the port the gateway is listening on.
    ///
    /// Used by the "rebuild gateway" command to construct the health-check URL
    /// passed to the trampoline.  `None` in standalone single-project mode.
    pub gateway_port: Option<u16>,
    /// Timestamp (ms since Unix epoch) of the last Matrix event received in any
    /// configured room.  Updated atomically on every `on_room_message` call so
    /// the `health` command can detect a stale or dead sync loop.
    pub last_matrix_event_ms: Arc<AtomicI64>,
    /// Optional model override from bot.toml. Passed as `--model` to the
    /// `claude` CLI when set.
    pub model: Option<String>,
    /// Maximum size in bytes of the digest the `compact` command writes as a
    /// seed file. From `bot.toml`'s `compact_seed_max_bytes`.
    pub compact_seed_max_bytes: usize,
    /// `cache_read_input_tokens` threshold above which the bot suggests
    /// running `compact` after a turn. From `bot.toml`'s
    /// `cache_read_suggest_threshold`.
    pub cache_read_suggest_threshold: u64,
    /// Minimum seconds between repeated `compact` suggestions for the same
    /// room. From `bot.toml`'s `compact_suggest_cooldown_secs`.
    pub compact_suggest_cooldown_secs: i64,
}

impl BotContext {
    /// Resolve the effective project root for command dispatch.
    ///
    /// In gateway mode the bot's `project_root` is the gateway config directory.
    /// Each project lives in a subdirectory named after the project, so the
    /// effective root for commands is `project_root / active_project_name`.
    /// In standalone (single-project) mode this returns `project_root` unchanged.
    pub async fn effective_project_root(&self) -> std::path::PathBuf {
        if let Some(ref ap) = self.gateway_active_project {
            let name = ap.read().await.clone();
            self.services.project_root.join(&name)
        } else {
            self.services.project_root.clone()
        }
    }

    /// Returns `true` if the bot is running in gateway mode.
    pub fn is_gateway(&self) -> bool {
        self.gateway_active_project.is_some()
    }

    /// Return the base URL for the currently active project, if in gateway mode.
    pub async fn active_project_url(&self) -> Option<String> {
        let ap = self.gateway_active_project.as_ref()?;
        let name = ap.read().await.clone();
        self.project_url(&name).await
    }

    /// Return the base URL for a named project from the live gateway store.
    pub async fn project_url(&self, name: &str) -> Option<String> {
        let store = self.gateway_projects_store.as_ref()?;
        store
            .read()
            .await
            .get(name)
            .and_then(|entry| entry.url.clone())
    }

    /// Proxy a bot command to the active project over a WebSocket RPC call.
    ///
    /// Connects to `{base_url}/ws`, sends an `rpc_request` frame for the
    /// `bot.command` method, and returns the Markdown response from the
    /// `rpc_response` frame.  Returns an error message string if the
    /// connection or command fails.
    pub async fn proxy_bot_command(&self, command: &str, args: &str) -> Option<String> {
        let base_url = self.active_project_url().await?;
        Some(Self::run_proxy_bot_command(&base_url, command, args).await)
    }

    /// Run the `bot.command` WebSocket RPC call against `base_url` and return
    /// the Markdown response, or an error message string on failure.
    ///
    /// `pub(crate)` so callers that need to target a project by name (rather
    /// than always the active one, as [`Self::proxy_bot_command`] does — e.g.
    /// the `status <project>` command) can resolve their own URL and reuse
    /// this transport logic.
    pub(crate) async fn run_proxy_bot_command(base_url: &str, command: &str, args: &str) -> String {
        use futures::{SinkExt, StreamExt};
        use tokio_tungstenite::tungstenite::Message as WsMsg;

        // Convert http(s):// → ws(s)://
        let ws_base = if let Some(rest) = base_url.strip_prefix("https://") {
            format!("wss://{rest}")
        } else if let Some(rest) = base_url.strip_prefix("http://") {
            format!("ws://{rest}")
        } else {
            base_url.to_string()
        };
        let ws_url = format!("{ws_base}/ws");

        let correlation_id = uuid::Uuid::new_v4().to_string();
        let request = serde_json::json!({
            "kind": "rpc_request",
            "version": 1,
            "correlation_id": correlation_id,
            "ttl_ms": 30_000u64,
            "method": "bot.command",
            "params": { "command": command, "args": args },
        });
        let request_text = match serde_json::to_string(&request) {
            Ok(t) => t,
            Err(e) => return format!("Failed to serialize RPC request: {e}"),
        };

        let connect_timeout = std::time::Duration::from_secs(5);
        let ws_stream = match tokio::time::timeout(
            connect_timeout,
            tokio_tungstenite::connect_async(&ws_url),
        )
        .await
        {
            Ok(Ok((stream, _))) => stream,
            Ok(Err(e)) => {
                return format!("Failed to connect to project server at {ws_url}: {e}");
            }
            Err(_) => {
                return format!(
                    "Project server at {ws_url} is unreachable (connect timed out after {connect_timeout:?})"
                );
            }
        };

        let (mut sink, mut stream) = ws_stream.split();

        if let Err(e) = sink.send(WsMsg::Text(request_text.into())).await {
            return format!("Failed to send RPC request: {e}");
        }

        let response_timeout = std::time::Duration::from_secs(30);
        let deadline = tokio::time::Instant::now() + response_timeout;
        while let Ok(Some(msg)) = tokio::time::timeout_at(deadline, stream.next()).await {
            match msg {
                Ok(WsMsg::Text(text)) => {
                    let Ok(frame) = serde_json::from_str::<serde_json::Value>(&text) else {
                        continue;
                    };
                    if frame.get("kind").and_then(|v| v.as_str()) != Some("rpc_response") {
                        continue;
                    }
                    if frame
                        .get("correlation_id")
                        .and_then(|v| v.as_str())
                        .map(|id| id != correlation_id)
                        .unwrap_or(true)
                    {
                        continue;
                    }
                    let ok = frame.get("ok").and_then(|v| v.as_bool()).unwrap_or(false);
                    if ok {
                        return frame
                            .get("result")
                            .and_then(|r| r.get("response"))
                            .and_then(|v| v.as_str())
                            .map(String::from)
                            .unwrap_or_else(|| {
                                "Command succeeded with no response text".to_string()
                            });
                    } else {
                        let err = frame
                            .get("error")
                            .and_then(|v| v.as_str())
                            .unwrap_or("unknown error");
                        return format!("Project server command failed: {err}");
                    }
                }
                Ok(WsMsg::Close(_)) => break,
                Err(e) => return format!("WebSocket error: {e}"),
                _ => continue,
            }
        }

        "Project server did not respond in time (connection closed or timed out)".to_string()
    }
}

// ---------------------------------------------------------------------------
// Tests
// ---------------------------------------------------------------------------

#[cfg(test)]
mod tests {
    use super::*;
    use crate::service::permission_router::{PendingPermReplies, ResponderRegistry};
    use std::collections::HashMap;
    use std::path::PathBuf;

    fn make_user_id(s: &str) -> OwnedUserId {
        s.parse().unwrap()
    }

    /// Build a test `Services` bundle with the given project root.
    fn test_services(project_root: PathBuf) -> Arc<Services> {
        Arc::new(Services {
            project_root,
            agents: Arc::new(crate::agents::AgentPool::new_test(3000)),
            bot_name: "Assistant".to_string(),
            bot_user_id: "@bot:example.com".to_string(),
            ambient_rooms: Arc::new(std::sync::Mutex::new(std::collections::HashSet::new())),
            permission_registry: ResponderRegistry::new(),
            pending_perm_replies: PendingPermReplies::new(),
            permission_timeout_secs: 120,
            status: Arc::new(crate::service::status::StatusBroadcaster::new()),
            chat_dispatcher: Arc::new(crate::chat::dispatcher::ChatDispatcher::new(1_500)),
        })
    }

    /// Build a minimal `BotContext` for testing with the given Services and
    /// optional gateway active project.
    fn test_bot_context(
        services: Arc<Services>,
        gateway_active_project: Option<Arc<RwLock<String>>>,
        gateway_projects_store: Option<
            Arc<RwLock<BTreeMap<String, crate::service::gateway::config::ProjectEntry>>>,
        >,
    ) -> BotContext {
        BotContext {
            services,
            matrix_user_id: make_user_id("@bot:example.com"),
            target_room_ids: vec![],
            allowed_users: vec![],
            history: Arc::new(TokioMutex::new(HashMap::new())),
            history_size: 20,
            bot_sent_event_ids: Arc::new(TokioMutex::new(std::collections::HashSet::new())),
            htop_sessions: Arc::new(TokioMutex::new(HashMap::new())),
            transport: Arc::new(crate::chat::transport::whatsapp::WhatsAppTransport::new(
                "test-phone".to_string(),
                "test-token".to_string(),
                "pipeline_notification".to_string(),
            )),
            timer_store: Arc::new(crate::service::timer::TimerStore::load(
                std::path::PathBuf::from("/tmp/timers.json"),
            )),
            gateway_active_project,
            gateway_projects_store,
            gateway_channels_store: None,
            handled_incoming_event_ids: Arc::new(TokioMutex::new(SeenEventIds::new(
                SEEN_EVENT_IDS_CAP,
            ))),
            gateway_port: None,
            last_matrix_event_ms: Arc::new(AtomicI64::new(chrono::Utc::now().timestamp_millis())),
            model: None,
            compact_seed_max_bytes: 8_000,
            cache_read_suggest_threshold: 50_000,
            compact_suggest_cooldown_secs: 3_600,
        }
    }

    #[test]
    fn bot_context_is_clone() {
        // BotContext must be Clone for the Matrix event handler injection.
        fn assert_clone<T: Clone>() {}
        assert_clone::<BotContext>();
    }

    #[tokio::test]
    async fn effective_project_root_standalone_returns_project_root() {
        let services = test_services(PathBuf::from("/projects/myapp"));
        let ctx = test_bot_context(services, None, None);
        assert_eq!(
            ctx.effective_project_root().await,
            PathBuf::from("/projects/myapp")
        );
    }

    #[tokio::test]
    async fn effective_project_root_gateway_uses_active_project_subdir() {
        let services = test_services(PathBuf::from("/gateway"));
        let active = Arc::new(RwLock::new("huskies".to_string()));
        let ctx = test_bot_context(services, Some(Arc::clone(&active)), None);
        assert_eq!(
            ctx.effective_project_root().await,
            PathBuf::from("/gateway/huskies")
        );
    }

    #[tokio::test]
    async fn effective_project_root_gateway_reflects_project_switch() {
        let services = test_services(PathBuf::from("/gateway"));
        let active = Arc::new(RwLock::new("huskies".to_string()));
        let ctx = test_bot_context(services, Some(Arc::clone(&active)), None);

        assert_eq!(
            ctx.effective_project_root().await,
            PathBuf::from("/gateway/huskies")
        );

        *active.write().await = "robot-studio".to_string();

        assert_eq!(
            ctx.effective_project_root().await,
            PathBuf::from("/gateway/robot-studio")
        );
    }

    // -- SeenEventIds deduplication ----------------------------------------

    fn make_event_id(s: &str) -> OwnedEventId {
        s.parse().unwrap()
    }

    /// AC3: the same event_id presented twice is deduplicated — the second
    /// insert returns false, so any downstream handler would execute only once.
    #[test]
    fn insert_same_event_id_twice_returns_false_on_second() {
        let mut seen = SeenEventIds::new(10);
        let id = make_event_id("$event1:example.com");
        assert!(seen.insert(id.clone()), "first insert must be new");
        assert!(!seen.insert(id), "second insert must be a duplicate");
    }

    /// AC4: two different event_ids with the same body content both return
    /// true — dedupe is keyed strictly on event_id, never on content.
    #[test]
    fn insert_different_event_ids_same_body_both_new() {
        let mut seen = SeenEventIds::new(10);
        let id1 = make_event_id("$event1:example.com");
        let id2 = make_event_id("$event2:example.com");
        assert!(seen.insert(id1), "first event_id must be new");
        assert!(
            seen.insert(id2),
            "second event_id with identical body must also be new"
        );
    }

    /// The set evicts the oldest entry (FIFO) when the cap is reached so that
    /// subsequent inserts still work and memory stays bounded.
    #[test]
    fn seen_event_ids_evicts_oldest_at_cap() {
        let cap = 3;
        let mut seen = SeenEventIds::new(cap);
        let id0 = make_event_id("$ev0:example.com");
        let id1 = make_event_id("$ev1:example.com");
        let id2 = make_event_id("$ev2:example.com");
        let id3 = make_event_id("$ev3:example.com");

        assert!(seen.insert(id0.clone())); // deque: [id0]
        assert!(seen.insert(id1.clone())); // deque: [id0, id1]
        assert!(seen.insert(id2.clone())); // deque: [id0, id1, id2] — full
        // Cap reached — inserting id3 evicts id0 (oldest).
        assert!(seen.insert(id3.clone())); // deque: [id1, id2, id3]
        // id0 was evicted, so re-inserting it returns true (treated as new).
        assert!(seen.insert(id0), "evicted entry should be re-insertable");
        // Re-inserting id0 evicts id1 (new oldest). id2 and id3 are still present.
        assert!(!seen.insert(id2.clone()), "id2 still in set");
        assert!(!seen.insert(id3), "id3 still in set");
        // id1 was evicted — treated as new again.
        assert!(
            seen.insert(id1),
            "id1 was evicted and should be re-insertable"
        );
    }

    #[test]
    fn bot_context_has_no_require_verified_devices_field() {
        let services = test_services(PathBuf::from("/tmp"));
        let ctx = test_bot_context(services, None, None);
        let _cloned = ctx.clone();
    }

    /// A bot command issued in gateway mode must round-trip over WebSocket
    /// (using the `bot.command` RPC method) and must NOT use HTTP transport.
    #[tokio::test]
    async fn proxy_bot_command_uses_websocket_not_http() {
        use futures::{SinkExt, StreamExt};
        use tokio::net::TcpListener;
        use tokio_tungstenite::tungstenite::Message as WsMsg;

        // Bind an ephemeral port for our mock WebSocket server.
        let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
        let port = listener.local_addr().unwrap().port();

        // Spawn a minimal WS server: accept one connection, verify the
        // request uses the `bot.command` RPC method (not HTTP), and reply.
        let server = tokio::spawn(async move {
            let (tcp, _addr) = listener.accept().await.unwrap();
            let mut ws = tokio_tungstenite::accept_async(tcp).await.unwrap();
            while let Some(Ok(msg)) = ws.next().await {
                if let WsMsg::Text(text) = msg {
                    let req: serde_json::Value =
                        serde_json::from_str(&text).expect("valid JSON from proxy");
                    assert_eq!(
                        req["kind"], "rpc_request",
                        "transport must use rpc_request, not HTTP"
                    );
                    assert_eq!(req["method"], "bot.command");
                    assert_eq!(req["params"]["command"], "status");
                    let correlation_id = req["correlation_id"].clone();
                    let resp = serde_json::json!({
                        "kind": "rpc_response",
                        "correlation_id": correlation_id,
                        "ok": true,
                        "result": { "response": "all systems go" },
                    });
                    ws.send(WsMsg::Text(resp.to_string().into())).await.unwrap();
                    break;
                }
            }
        });

        let base_url = format!("http://127.0.0.1:{port}");
        let services = test_services(PathBuf::from("/gateway"));
        let active = Arc::new(RwLock::new("huskies".to_string()));
        let store = Arc::new(RwLock::new(BTreeMap::from([(
            "huskies".to_string(),
            crate::service::gateway::config::ProjectEntry {
                url: Some(base_url),
                auth_token: None,
                ssh_port: None,
                host_path: None,
                expected_node_id: None,
            },
        )])));
        let ctx = test_bot_context(services, Some(Arc::clone(&active)), Some(store));

        let result = ctx.proxy_bot_command("status", "").await;
        assert_eq!(
            result.as_deref(),
            Some("all systems go"),
            "proxy must return the response text from the rpc_response frame"
        );

        server.await.unwrap();
    }

    /// Regression test for story 1132: `active_project_url` must read from the
    /// live `gateway_projects_store`, not a stale snapshot frozen at bot startup.
    /// Adding a project to the store after `BotContext` is created must be
    /// visible immediately — no restart required.
    #[tokio::test]
    async fn active_project_url_reflects_runtime_added_project() {
        let store: Arc<RwLock<BTreeMap<String, crate::service::gateway::config::ProjectEntry>>> =
            Arc::new(RwLock::new(BTreeMap::new()));
        let active = Arc::new(RwLock::new("new-project".to_string()));
        let services = test_services(PathBuf::from("/gateway"));
        let ctx = test_bot_context(
            services,
            Some(Arc::clone(&active)),
            Some(Arc::clone(&store)),
        );

        // Store is empty — must return None.
        assert!(
            ctx.active_project_url().await.is_none(),
            "URL must be None when store is empty"
        );

        // Insert the entry at runtime (simulates `new project` command).
        store.write().await.insert(
            "new-project".to_string(),
            crate::service::gateway::config::ProjectEntry {
                url: Some("http://localhost:3099".to_string()),
                auth_token: None,
                ssh_port: None,
                host_path: None,
                expected_node_id: None,
            },
        );

        // Now the live store has the entry — active_project_url must see it.
        assert_eq!(
            ctx.active_project_url().await.as_deref(),
            Some("http://localhost:3099"),
            "URL must be visible after runtime insertion without bot restart"
        );
    }
}