diff --git a/Cargo.lock b/Cargo.lock index 7d573164..0d5b1376 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -1523,6 +1523,7 @@ dependencies = [ "clap", "forgejo-api", "futures-util", + "hive-agent-sock", "hive-claude", "hive-core-agent-sock", "hive-sh4re", @@ -1552,6 +1553,7 @@ dependencies = [ "anyhow", "axum", "clap", + "hive-agent-sock", "hive-core-agent-sock", "hive-sh4re", "rmcp", @@ -1562,6 +1564,14 @@ dependencies = [ "tracing-subscriber", ] +[[package]] +name = "hive-agent-sock" +version = "0.1.0" +dependencies = [ + "hive-sh4re", + "serde", +] + [[package]] name = "hive-agent-wake" version = "0.1.0" diff --git a/Cargo.toml b/Cargo.toml index 549abeb5..5ff5da9c 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -3,6 +3,7 @@ resolver = "3" members = [ "hive-agent", "hive-agent-mcp", + "hive-agent-sock", "hive-core-agent-sock", "hive-agent-wake", "hive-bash-mcp", @@ -51,6 +52,7 @@ clap = { version = "4", features = ["derive"] } clap_complete = "4" indicatif = "0.18" hive-sh4re = { path = "hive-sh4re" } +hive-agent-sock = { path = "hive-agent-sock" } hive-core-agent-sock = { path = "hive-core-agent-sock" } hive-claude = { path = "hive-claude" } hive-host-sock = { path = "hive-host-sock" } diff --git a/hive-agent-mcp/Cargo.toml b/hive-agent-mcp/Cargo.toml index 6fad2eb9..33576ddf 100644 --- a/hive-agent-mcp/Cargo.toml +++ b/hive-agent-mcp/Cargo.toml @@ -14,6 +14,7 @@ workspace = true anyhow.workspace = true axum.workspace = true clap.workspace = true +hive-agent-sock.workspace = true hive-core-agent-sock.workspace = true hive-sh4re.workspace = true rmcp.workspace = true diff --git a/hive-agent-mcp/src/mcp/mod.rs b/hive-agent-mcp/src/mcp/mod.rs index 5ac7c35d..8acc2c6e 100644 --- a/hive-agent-mcp/src/mcp/mod.rs +++ b/hive-agent-mcp/src/mcp/mod.rs @@ -35,8 +35,8 @@ pub use args::{ pub use render::{annotate_retries, format_ack, format_agent_meta, format_recv}; use render::{ - format_matrix_summary, loose_end_kind_label, matrix_unread_summary, parse_loose_end_kind, - render_loose_ends, reply_err, + format_matrix_summary, local_todos, loose_end_kind_label, matrix_unread_summary, + parse_loose_end_kind, render_loose_ends, reply_err, }; /// Write (or remove) the status file in the agent's own `state/` directory. @@ -333,6 +333,12 @@ impl AgentServer { ); } } + // Merge the harness's local todos (loose-ends v2) for self-queries. + // The harness owns the todo store in-container; another agent's + // todos aren't reachable from here (same as matrix above). + if is_self_query && let Some(todos) = local_todos().await { + loose_ends.extend(todos); + } let mut out = annotate_retries(render_loose_ends(&loose_ends), retries); // Append loose-end items published by external MCP daemons // (e.g. active bash tasks from hive-bash-mcp). Generic — no diff --git a/hive-agent-mcp/src/mcp/render.rs b/hive-agent-mcp/src/mcp/render.rs index dddc9eca..858e75ba 100644 --- a/hive-agent-mcp/src/mcp/render.rs +++ b/hive-agent-mcp/src/mcp/render.rs @@ -297,6 +297,30 @@ pub(super) async fn matrix_unread_summary() -> Option> { serde_json::from_value(serde_json::Value::Array(arr.clone())).ok() } +/// Query the harness's in-agent socket for this agent's local todos +/// (loose-ends v2). Returns `None` when `HIVE_AGENT_SOCKET` is unset / +/// absent or the query fails — best-effort, like [`matrix_unread_summary`], +/// so an agent without the socket is not penalised. +pub(super) async fn local_todos() -> Option> { + use tokio::io::{AsyncBufReadExt, AsyncWriteExt, BufReader}; + use tokio::net::UnixStream; + let socket = std::env::var_os("HIVE_AGENT_SOCKET").map(std::path::PathBuf::from)?; + if !socket.exists() { + return None; + } + let mut stream = UnixStream::connect(&socket).await.ok()?; + let mut req = + serde_json::to_string(&hive_agent_sock::Request::ListTodos { subsystem: None }).ok()?; + req.push('\n'); + stream.write_all(req.as_bytes()).await.ok()?; + let mut lines = BufReader::new(stream).lines(); + let line = lines.next_line().await.ok()??; + match serde_json::from_str::(&line).ok()? { + hive_agent_sock::Response::LooseEnds { loose_ends } => Some(loose_ends), + _ => None, + } +} + /// Format a `Vec` into a per-room summary string. /// Single room / single message collapses to one line; multi-room /// expands to a bulleted list. Returns an empty string for empty input. diff --git a/hive-agent-sock/Cargo.toml b/hive-agent-sock/Cargo.toml new file mode 100644 index 00000000..7818923d --- /dev/null +++ b/hive-agent-sock/Cargo.toml @@ -0,0 +1,11 @@ +[package] +name = "hive-agent-sock" +edition.workspace = true +version.workspace = true + +[lints] +workspace = true + +[dependencies] +serde.workspace = true +hive-sh4re.workspace = true diff --git a/hive-agent-sock/src/lib.rs b/hive-agent-sock/src/lib.rs new file mode 100644 index 00000000..dc7bd7a8 --- /dev/null +++ b/hive-agent-sock/src/lib.rs @@ -0,0 +1,76 @@ +//! Wire types for the *in-agent* socket, served by the hive-agent harness +//! to the in-container producers (matrix / bash MCP daemons) and +//! `forge_notify`. Currently carries the loose-ends-v2 *todo* op family; +//! more in-agent request families may be added over time (the socket is +//! deliberately named for the agent, not the todos). +//! +//! Distinct from `hive-core-agent-sock`, the *host*-served core↔agent +//! protocol on `/run/hive/mcp.sock`: this socket never leaves the +//! container. The harness owns the todo store locally and signals its own +//! turn loop directly, so hive-c0re is not in the todo path — no broker +//! round-trip, no long-poll, no marker files. + +use serde::{Deserialize, Serialize}; + +use hive_sh4re::LooseEnd; + +/// In-container path of the harness-served in-agent socket. The harness +/// binds it on boot; the in-container producers dial it for todo ops. +/// (Placeholder default — the harness + producers resolve the real path +/// from config; kept here so a producer with no override has a sane one.) +pub const DEFAULT_AGENT_SOCKET: &str = "/run/hive/agent.sock"; + +/// A request on the in-agent socket. Serialised with a `cmd` tag so the +/// in-container producers can emit a plain JSON line without linking a +/// typed client (matrix/bash build the JSON by hand). +#[derive(Debug, Clone, Serialize, Deserialize)] +#[serde(tag = "cmd", rename_all = "snake_case")] +pub enum Request { + /// Upsert a todo from an in-container subsystem (matrix / bash / + /// forge). `subsystem` is the producer marker; `key` the optional + /// subsystem-specific dedup key (a matrix room id, a bash task id). + /// A new-or-changed row signals the turn loop; an identical keyed + /// re-push is a silent no-op. Keyless todos always insert as one-offs. + UpsertTodo { + subsystem: String, + #[serde(default, skip_serializing_if = "Option::is_none")] + key: Option, + summary: String, + #[serde(default, skip_serializing_if = "Option::is_none")] + source: Option, + }, + /// Clear producer-resolved todo(s). `key = Some(k)` clears the one + /// keyed row; `key = None` clears the subsystem's keyless rows; `all + /// = true` wipes the producer's whole set (cancel-and-recreate on + /// daemon restart). + ClearTodo { + subsystem: String, + #[serde(default, skip_serializing_if = "Option::is_none")] + key: Option, + #[serde(default)] + all: bool, + }, + /// List todos, optionally filtered to one `subsystem` (a producer + /// reconciling its own set). `None` = all. + ListTodos { + #[serde(default, skip_serializing_if = "Option::is_none")] + subsystem: Option, + }, + /// The agent marks one of its own todos done, by id. + MarkTodoDone { id: i64 }, +} + +/// A response on the in-agent socket. Serialised with a `kind` tag, +/// mirroring the core↔agent protocol's response shape. +#[derive(Debug, Clone, Serialize, Deserialize)] +#[serde(tag = "kind", rename_all = "snake_case")] +pub enum Response { + /// Op succeeded, no payload. + Ok, + /// Op succeeded and touched `count` rows (clear / mark-done). + Acked { count: u64 }, + /// `ListTodos` result. + LooseEnds { loose_ends: Vec }, + /// Op failed; `message` is operator-facing. + Err { message: String }, +} diff --git a/hive-agent/Cargo.toml b/hive-agent/Cargo.toml index 2b0c74c1..e09c6da8 100644 --- a/hive-agent/Cargo.toml +++ b/hive-agent/Cargo.toml @@ -19,6 +19,7 @@ time.workspace = true futures-util = "0.3" clap.workspace = true hive-claude.workspace = true +hive-agent-sock.workspace = true hive-core-agent-sock.workspace = true hive-sh4re.workspace = true rmcp.workspace = true diff --git a/hive-agent/src/main.rs b/hive-agent/src/main.rs index 69b8b0b1..4c241696 100644 --- a/hive-agent/src/main.rs +++ b/hive-agent/src/main.rs @@ -23,6 +23,8 @@ mod prompt; mod serve_common; mod stats; mod stream_enrich; +mod todo_server; +mod todos; mod turn; mod turn_stats; mod vacuum; @@ -169,6 +171,22 @@ fn synthetic_continue() -> hive_sh4re::DeliveredMessage { } } +/// Synthesize the message that drives a turn when an in-container producer +/// upserted a new/changed *todo* over the in-agent socket (loose-ends v2). +/// The harness owns the todo store locally and signals the serve loop +/// directly — so this wake never touches the broker (no long-poll, no +/// marker file). `id = 0` is the same non-broker sentinel as +/// [`synthetic_continue`]. +fn synthetic_todo_message() -> hive_sh4re::DeliveredMessage { + hive_sh4re::DeliveredMessage { + from: "todo".into(), + body: "you have todos — call get_loose_ends to see them".into(), + id: 0, + redelivered: false, + in_reply_to: None, + } +} + /// Synthetic message that drives the single stop-checkpoint turn when c0re /// signals a graceful stop. The agent gets one final turn to flush durable /// `/state` before the container is stopped; new inbound is already fenced. @@ -208,6 +226,10 @@ enum RecvOutcome { /// one stop-checkpoint turn (flush durable `/state`), reports /// `GracefulStopComplete`, and exits so the container can be stopped. GracefulStop, + /// An in-container producer upserted a new/changed todo over the + /// in-agent socket; the serve loop drives a `synthetic_todo_message` + /// turn. Not a broker message — the harness signalled itself directly. + LocalTodo, } /// Wire surface abstraction. `AgentSurface` is the only impl — the trait @@ -423,6 +445,26 @@ async fn serve_main(socket: &Path, poll_ms: u64) -> Result<()> { tracing::error!(error = %e, "web_ui::serve exited with error"); } }); + // In-agent todo socket (loose-ends v2): the harness owns the todo store + // locally and serves the in-container producers on `HIVE_AGENT_SOCKET`. + // A new/changed upsert fires `todo_wake` so the serve loop drives a turn + // directly — no broker round-trip, no marker files. Best-effort: if the + // store can't open, the socket just isn't served. + let todo_wake = Arc::new(tokio::sync::Notify::new()); + match todos::Todos::open(&paths::todos_db()) { + Ok(store) => { + let store = Arc::new(store); + let wake = todo_wake.clone(); + tokio::spawn(async move { + if let Err(e) = todo_server::run(store, wake).await { + tracing::error!(error = %e, "in-agent todo socket exited with error"); + } + }); + } + Err(e) => { + tracing::error!(error = ?e, "open todos db failed — in-agent todo socket disabled"); + } + } if matches!(initial, LoginState::NeedsLogin) { login::wait_for_login(&claude_dir, login_state.clone(), &bus, poll_ms).await; } else { @@ -439,6 +481,7 @@ async fn serve_main(socket: &Path, poll_ms: u64) -> Result<()> { bus, stats, &files, + todo_wake, ) .await } @@ -460,6 +503,7 @@ async fn serve_loop( bus: Bus, stats: Option, files: &turn::TurnFiles, + todo_wake: Arc, ) -> Result<()> { tracing::info!(socket = %socket.display(), "harness serve"); S::requeue_inflight(socket).await; @@ -474,8 +518,20 @@ async fn serve_loop( loop { let next = match self_continue.take() { Some(msg) => msg, - None => match S::recv_next(socket).await { + None => match { + // Idle wait: race the broker long-poll against a local + // todo signal so an in-container producer's upsert drives a + // turn without any broker round-trip. `biased` polls the + // broker recv first, so a genuinely-ready inbox message is + // never dropped in favour of the todo wake. + tokio::select! { + biased; + o = S::recv_next(socket) => o, + () = todo_wake.notified() => RecvOutcome::LocalTodo, + } + } { RecvOutcome::Message(first) => first, + RecvOutcome::LocalTodo => synthetic_todo_message(), RecvOutcome::Empty => { // Idle: no message this poll. Service a queued operator // `/compact` here so it runs even when no turn is driving diff --git a/hive-agent/src/paths.rs b/hive-agent/src/paths.rs index 9535c7a5..2d7c7b99 100644 --- a/hive-agent/src/paths.rs +++ b/hive-agent/src/paths.rs @@ -40,6 +40,14 @@ pub fn harness_dir() -> PathBuf { hive_sh4re::paths::harness_dir() } +/// Harness-local todo store (loose-ends v2). A dedicated sqlite db under +/// the harness dir — the todos are mutable per-agent state the harness +/// owns, kept out of the append-only `hyperhive-events.sqlite` sink. +#[must_use] +pub fn todos_db() -> PathBuf { + harness_dir().join("hyperhive-todos.sqlite") +} + /// Per-turn config dir for the regenerated claude-{mcp-config,settings, /// system-prompt} files the harness drops before each turn. Set by /// systemd via `RuntimeDirectory = "hive-config"`: a per-service runtime diff --git a/hive-agent/src/todo_server.rs b/hive-agent/src/todo_server.rs new file mode 100644 index 00000000..7ad2382c --- /dev/null +++ b/hive-agent/src/todo_server.rs @@ -0,0 +1,168 @@ +//! In-agent socket server (loose-ends v2). Binds the harness-owned +//! `HIVE_AGENT_SOCKET` and serves the `hive-agent-sock` protocol to the +//! in-container producers (matrix / bash daemons, forge-notify). Todo ops +//! hit the harness-local [`Todos`] store; a new-or-changed upsert fires an +//! in-process [`Notify`] so the serve loop drives a turn — no hive-c0re +//! round-trip, no broker long-poll, no marker files. +//! +//! One request/response line per connection, matching the producers' +//! existing best-effort JSON-line clients (they just change which socket +//! they dial, not the payload). + +use std::path::{Path, PathBuf}; +use std::sync::Arc; + +use anyhow::{Context, Result}; +use hive_agent_sock::{Request, Response}; +use hive_sh4re::LooseEnd; +use tokio::io::{AsyncBufReadExt, AsyncWriteExt, BufReader}; +use tokio::net::{UnixListener, UnixStream}; +use tokio::sync::Notify; + +use crate::todos::{Todo, Todos}; + +/// Resolve the in-agent socket path from `HIVE_AGENT_SOCKET`. `None` when +/// unset/empty (e.g. a standalone dev run) — the server then stays off. +fn socket_path() -> Option { + std::env::var_os("HIVE_AGENT_SOCKET") + .filter(|v| !v.is_empty()) + .map(PathBuf::from) +} + +/// Run the in-agent socket server: bind + accept loop, one request/response +/// line per connection. A no-op (returns `Ok`) when `HIVE_AGENT_SOCKET` is +/// unset, so a standalone harness without producers just skips it. +/// +/// # Errors +/// +/// Returns an error if the socket path is set but can't be bound. +pub async fn run(store: Arc, wake: Arc) -> Result<()> { + let Some(path) = socket_path() else { + tracing::info!("HIVE_AGENT_SOCKET unset — in-agent todo socket disabled"); + return Ok(()); + }; + let listener = bind(&path)?; + tracing::info!(socket = %path.display(), "in-agent todo socket listening"); + loop { + match listener.accept().await { + Ok((stream, _)) => { + let store = store.clone(); + let wake = wake.clone(); + tokio::spawn(async move { + if let Err(e) = handle_conn(stream, &store, &wake).await { + tracing::warn!(error = ?e, "in-agent todo connection failed"); + } + }); + } + Err(e) => tracing::warn!(error = ?e, "in-agent todo accept failed"), + } + } +} + +/// Bind a `UnixListener` at `path`, creating the parent dir and unlinking a +/// stale socket left by a prior boot (which would otherwise block `bind` +/// with `EADDRINUSE`). +fn bind(path: &Path) -> Result { + if let Some(parent) = path.parent() { + std::fs::create_dir_all(parent) + .with_context(|| format!("create in-agent socket dir {}", parent.display()))?; + } + if path.exists() { + let _ = std::fs::remove_file(path); + } + UnixListener::bind(path).with_context(|| format!("bind in-agent socket {}", path.display())) +} + +/// Handle one connection: read a single JSON request line, apply it to the +/// store, write the JSON response line back. +async fn handle_conn(stream: UnixStream, store: &Todos, wake: &Notify) -> Result<()> { + let (read, mut write) = stream.into_split(); + let mut reader = BufReader::new(read); + let mut line = String::new(); + if reader.read_line(&mut line).await? == 0 { + return Ok(()); + } + let resp = match serde_json::from_str::(line.trim()) { + Ok(req) => dispatch(req, store, wake), + Err(e) => Response::Err { + message: format!("bad request: {e}"), + }, + }; + let mut out = serde_json::to_string(&resp)?; + out.push('\n'); + write.write_all(out.as_bytes()).await?; + write.flush().await?; + Ok(()) +} + +/// Apply one request to the store, firing `wake` on a new/changed upsert so +/// the serve loop runs a turn. +fn dispatch(req: Request, store: &Todos, wake: &Notify) -> Response { + match req { + Request::UpsertTodo { + subsystem, + key, + summary, + source, + } => match store.upsert(&subsystem, key.as_deref(), &summary, source.as_deref()) { + Ok((_, changed)) => { + if changed { + wake.notify_one(); + } + Response::Ok + } + Err(e) => err(&e), + }, + Request::ClearTodo { + subsystem, + key, + all, + } => { + let result = if all { + store.clear_subsystem(&subsystem) + } else { + store.clear(&subsystem, key.as_deref()) + }; + match result { + Ok(count) => Response::Acked { + count: u64::try_from(count).unwrap_or(0), + }, + Err(e) => err(&e), + } + } + Request::ListTodos { subsystem } => match store.list(subsystem.as_deref()) { + Ok(todos) => Response::LooseEnds { + loose_ends: todos.into_iter().map(to_loose_end).collect(), + }, + Err(e) => err(&e), + }, + Request::MarkTodoDone { id } => match store.mark_done(id) { + Ok(count) => Response::Acked { + count: u64::try_from(count).unwrap_or(0), + }, + Err(e) => err(&e), + }, + } +} + +/// Format a store error into an `Err` response. +fn err(e: &anyhow::Error) -> Response { + Response::Err { + message: format!("{e:#}"), + } +} + +/// Map a stored [`Todo`] to a [`LooseEnd::Todo`], deriving `age_seconds` +/// from `updated_at` (saturating so a backwards clock step reads 0). +fn to_loose_end(t: Todo) -> LooseEnd { + let now = hive_sh4re::wire_time::now_unix(); + let age = u64::try_from(now.saturating_sub(t.updated_at)).unwrap_or(0); + LooseEnd::Todo { + id: t.id, + subsystem: t.subsystem, + subsystem_key: t.subsystem_key, + summary: t.summary, + source: t.source, + age_seconds: age, + } +} diff --git a/hive-agent/src/todos.rs b/hive-agent/src/todos.rs new file mode 100644 index 00000000..19af3142 --- /dev/null +++ b/hive-agent/src/todos.rs @@ -0,0 +1,293 @@ +//! Harness-local todo store — the persistent, DB-backed half of the +//! "todos" (loose-ends v2) system, owned by the in-container harness. +//! +//! In-container subsystems (matrix, forge-notify, bash) push *todos* to +//! the harness over the in-agent socket instead of firing wakes directly. +//! The harness owns this store locally (one sqlite db under the harness +//! dir) and signals its own turn loop on a new/changed row — hive-c0re is +//! not involved (no broker round-trip, no marker files). +//! +//! Because the store lives inside a single agent's container, todos are +//! **not** agent-scoped here (unlike the old c0re store): every row +//! belongs to this agent. A todo is tagged with a `subsystem` marker plus +//! an optional `subsystem_key` (a matrix room id, a bash task id, …), and +//! `(subsystem, subsystem_key)` is the upsert/dedup key — re-pushing the +//! same item is idempotent, and a producer can list / clear / rebuild only +//! its own set (e.g. matrix wipes + recreates its todos on daemon restart). +//! +//! Removal has two paths (mara's call): the producing subsystem `clear`s a +//! todo it has resolved (keyed by subsystem + key), or the agent itself +//! `mark_done`s one by id. Both delete the row. + +use std::path::Path; +use std::sync::Mutex; + +use anyhow::{Context, Result}; +use hive_sh4re::wire_time::now_unix; +use rusqlite::{Connection, params}; + +const SCHEMA: &str = r" +CREATE TABLE IF NOT EXISTS todos ( + id INTEGER PRIMARY KEY AUTOINCREMENT, + subsystem TEXT NOT NULL, + subsystem_key TEXT, + summary TEXT NOT NULL, + source TEXT, + updated_at INTEGER NOT NULL +); +-- (subsystem, subsystem_key) is the upsert/dedup key. A NULL key never +-- conflicts (SQLite treats NULLs as distinct), so keyless todos always +-- insert as one-offs; keyed todos update in place. +CREATE UNIQUE INDEX IF NOT EXISTS idx_todos_dedup + ON todos (subsystem, subsystem_key); +"; + +/// One dynamic, subsystem-pushed todo. Timestamps are unix seconds; the +/// consumer derives `age_seconds` from `updated_at`. +#[derive(Debug, Clone)] +pub struct Todo { + pub id: i64, + /// Producing subsystem marker (`"matrix"`, `"forge"`, `"bash"`, …). + pub subsystem: String, + /// Optional subsystem-specific dedup key (matrix room id, bash task + /// id, …). `None` = a keyless one-off todo. + pub subsystem_key: Option, + /// Human-readable one-line summary shown to the agent. + pub summary: String, + /// Optional free-text provenance (e.g. the room name / task label). + pub source: Option, + pub updated_at: i64, +} + +/// The harness-local todo store. Cheap to share behind an `Arc`; the inner +/// connection is guarded by a `Mutex` (todo ops are short sqlite writes). +pub struct Todos { + conn: Mutex, +} + +impl Todos { + /// Open (creating if needed) the todo store at `path`. + /// + /// # Errors + /// + /// Propagates sqlite open / schema-apply failures. + pub fn open(path: &Path) -> Result { + let conn = + Connection::open(path).with_context(|| format!("open todos db {}", path.display()))?; + conn.execute_batch(SCHEMA).context("apply todos schema")?; + Ok(Self { + conn: Mutex::new(conn), + }) + } + + /// Insert a todo, or update the existing one for `(subsystem, key)` + /// when `key` is `Some` and already present. A `None` key never + /// conflicts, so it always inserts a fresh row. + /// + /// Returns `(id, changed)` where `changed` is `true` when the row is + /// new OR its `summary`/`source` actually differed — the caller uses + /// this to decide whether to signal the turn loop (re-pushing an + /// identical keyed todo is a no-op and must not re-wake). + /// + /// # Errors + /// + /// Propagates sqlite query / execute failures. + /// + /// # Panics + /// + /// Panics if the connection mutex is poisoned. + pub fn upsert( + &self, + subsystem: &str, + key: Option<&str>, + summary: &str, + source: Option<&str>, + ) -> Result<(i64, bool)> { + let conn = self.conn.lock().unwrap(); + let now = now_unix(); + let existing: Option<(i64, String, Option)> = if key.is_some() { + conn.query_row( + "SELECT id, summary, source FROM todos \ + WHERE subsystem = ?1 AND subsystem_key IS ?2", + params![subsystem, key], + |row| Ok((row.get(0)?, row.get(1)?, row.get(2)?)), + ) + .ok() + } else { + None + }; + if let Some((id, cur_summary, cur_source)) = existing { + let unchanged = cur_summary == summary && cur_source.as_deref() == source; + if unchanged { + return Ok((id, false)); + } + conn.execute( + "UPDATE todos SET summary = ?1, source = ?2, updated_at = ?3 WHERE id = ?4", + params![summary, source, now, id], + )?; + return Ok((id, true)); + } + conn.execute( + "INSERT INTO todos \ + (subsystem, subsystem_key, summary, source, updated_at) \ + VALUES (?1, ?2, ?3, ?4, ?5)", + params![subsystem, key, summary, source, now], + )?; + Ok((conn.last_insert_rowid(), true)) + } + + /// Clear producer-resolved todo(s) by `(subsystem, key)`. `key = + /// Some(k)` targets the one keyed row; `key = None` matches + /// `subsystem_key IS NULL`, i.e. **all** keyless todos for that + /// subsystem (clear a specific keyless one via [`Todos::mark_done`] by + /// id instead). Returns the number of rows deleted. + /// + /// # Errors + /// + /// Propagates the sqlite delete failure. + /// + /// # Panics + /// + /// Panics if the connection mutex is poisoned. + pub fn clear(&self, subsystem: &str, key: Option<&str>) -> Result { + let conn = self.conn.lock().unwrap(); + let n = conn.execute( + "DELETE FROM todos WHERE subsystem = ?1 AND subsystem_key IS ?2", + params![subsystem, key], + )?; + Ok(n) + } + + /// Clear every todo `subsystem` owns — used by a producer that rebuilds + /// its whole set on restart (cancel-and-recreate). Returns the number + /// of rows deleted. + /// + /// # Errors + /// + /// Propagates the sqlite delete failure. + /// + /// # Panics + /// + /// Panics if the connection mutex is poisoned. + pub fn clear_subsystem(&self, subsystem: &str) -> Result { + let conn = self.conn.lock().unwrap(); + let n = conn.execute("DELETE FROM todos WHERE subsystem = ?1", params![subsystem])?; + Ok(n) + } + + /// The agent marks one of its todos done, by id. Returns the number of + /// rows deleted (0 when the id was unknown / already gone). + /// + /// # Errors + /// + /// Propagates the sqlite delete failure. + /// + /// # Panics + /// + /// Panics if the connection mutex is poisoned. + pub fn mark_done(&self, id: i64) -> Result { + let conn = self.conn.lock().unwrap(); + let n = conn.execute("DELETE FROM todos WHERE id = ?1", params![id])?; + Ok(n) + } + + /// List todos, newest-updated first. `subsystem = Some(..)` filters to + /// one producer's set; `None` returns all. + /// + /// # Errors + /// + /// Propagates the sqlite prepare / query failures. + /// + /// # Panics + /// + /// Panics if the connection mutex is poisoned. + pub fn list(&self, subsystem: Option<&str>) -> Result> { + let conn = self.conn.lock().unwrap(); + let mut stmt = conn.prepare( + "SELECT id, subsystem, subsystem_key, summary, source, updated_at \ + FROM todos \ + WHERE (?1 IS NULL OR subsystem = ?1) \ + ORDER BY updated_at DESC, id DESC", + )?; + let rows = stmt + .query_map(params![subsystem], |row| { + Ok(Todo { + id: row.get(0)?, + subsystem: row.get(1)?, + subsystem_key: row.get(2)?, + summary: row.get(3)?, + source: row.get(4)?, + updated_at: row.get(5)?, + }) + })? + .collect::>>()?; + Ok(rows) + } +} + +#[cfg(test)] +mod tests { + use super::*; + + // Return the `TempDir` alongside the store so it outlives the test — + // dropping it early deletes the dir and SQLite fails with + // `SQLITE_READONLY_DBMOVED`. + fn store() -> (tempfile::TempDir, Todos) { + let dir = tempfile::tempdir().unwrap(); + let db = Todos::open(&dir.path().join("todos.sqlite")).unwrap(); + (dir, db) + } + + #[test] + fn keyed_upsert_dedups_and_reports_changed() { + let (_dir, s) = store(); + let (id1, changed1) = s + .upsert("matrix", Some("!room:x"), "1 unread", None) + .unwrap(); + assert!(changed1, "first push is new → changed"); + // Same key + same summary → no-op, not changed (must not re-wake). + let (id2, changed2) = s + .upsert("matrix", Some("!room:x"), "1 unread", None) + .unwrap(); + assert_eq!(id1, id2, "keyed upsert updates in place, same row"); + assert!(!changed2, "identical re-push is a no-op"); + // Same key, new summary → updates, changed. + let (id3, changed3) = s + .upsert("matrix", Some("!room:x"), "3 unread", None) + .unwrap(); + assert_eq!(id1, id3); + assert!(changed3); + assert_eq!(s.list(Some("matrix")).unwrap().len(), 1); + } + + #[test] + fn keyless_todos_always_insert() { + let (_dir, s) = store(); + let (a, _) = s.upsert("bash", None, "task done", None).unwrap(); + let (b, _) = s.upsert("bash", None, "task done", None).unwrap(); + assert_ne!(a, b, "keyless pushes are distinct one-offs"); + assert_eq!(s.list(Some("bash")).unwrap().len(), 2); + } + + #[test] + fn clear_and_mark_done_remove_rows() { + let (_dir, s) = store(); + s.upsert("matrix", Some("!a:x"), "unread", None).unwrap(); + let (id, _) = s.upsert("forge", Some("pr-1"), "review", None).unwrap(); + assert_eq!(s.clear("matrix", Some("!a:x")).unwrap(), 1); + assert_eq!(s.mark_done(id).unwrap(), 1); + assert!(s.list(None).unwrap().is_empty()); + } + + #[test] + fn clear_subsystem_wipes_only_its_own() { + let (_dir, s) = store(); + s.upsert("matrix", Some("!a:x"), "u", None).unwrap(); + s.upsert("matrix", Some("!b:x"), "u", None).unwrap(); + s.upsert("forge", Some("pr-1"), "r", None).unwrap(); + assert_eq!(s.clear_subsystem("matrix").unwrap(), 2); + let left = s.list(None).unwrap(); + assert_eq!(left.len(), 1); + assert_eq!(left[0].subsystem, "forge"); + } +} diff --git a/hive-c0re/src/coordinator.rs b/hive-c0re/src/coordinator.rs index 707ac73c..9d40377d 100644 --- a/hive-c0re/src/coordinator.rs +++ b/hive-c0re/src/coordinator.rs @@ -16,7 +16,6 @@ use crate::container_view::{self, ContainerView}; use crate::dashboard_events::DashboardEvent; use crate::operator_questions::OperatorQuestions; use crate::socket_server::{self, AgentSocket}; -use crate::todos::Todos; /// Capacity of the dashboard event channel. Slow browser subscribers /// (idle tab, throttled connection) drop frames past this — that's @@ -33,11 +32,6 @@ pub struct Coordinator { pub broker: Arc, pub approvals: Arc, pub questions: Arc, - /// Dynamic, subsystem-pushed todos (loose-ends v2). In-agent - /// subsystems (matrix, forge, bash) upsert/clear todos over mcp.sock - /// instead of firing wakes directly; `get_todos` merges these with - /// the computed static loose ends. - pub todos: Arc, /// Scheduled-prompts queue. One sqlite connection, /// internal mutex; the worker drains due rows and the manager /// handlers insert / cancel through the same handle. @@ -466,7 +460,6 @@ impl Coordinator { let broker = Broker::open(db_path).context("open broker")?; let approvals = Approvals::open(db_path).context("open approvals")?; let questions = OperatorQuestions::open(db_path).context("open operator_questions")?; - let todos = Todos::open(db_path).context("open todos")?; let scheduled_prompts = crate::scheduled_prompts::ScheduledPrompts::open(db_path) .context("open scheduled_prompts")?; // BuildLogs wants a directory (it picks its own `build_logs.sqlite` @@ -496,7 +489,6 @@ impl Coordinator { broker: Arc::new(broker), approvals: Arc::new(approvals), questions: Arc::new(questions), - todos: Arc::new(todos), scheduled_prompts: Arc::new(scheduled_prompts), build_logs, audit_log, diff --git a/hive-c0re/src/loose_ends.rs b/hive-c0re/src/loose_ends.rs index 72a9b0ee..b09228f8 100644 --- a/hive-c0re/src/loose_ends.rs +++ b/hive-c0re/src/loose_ends.rs @@ -97,40 +97,9 @@ pub fn for_agent(coord: &Coordinator, agent: &str) -> Result> { age_seconds: saturating_age(now, r.created_at.timestamp()), }); } - // Dynamic, subsystem-pushed todos (loose-ends v2). Scoped to - // this agent; the producing subsystem or the agent itself clears them. - out.extend(todos_for(coord, agent, None)?); Ok(out) } -/// This agent's dynamic todos as `LooseEnd::Todo` rows, optionally -/// filtered to one `subsystem`. Shared by [`for_agent`] and the -/// `ListTodos` handler so the row-mapping lives in one place. -/// -/// # Errors -/// -/// Propagates the todo-store query failure. -pub fn todos_for( - coord: &Coordinator, - agent: &str, - subsystem: Option<&str>, -) -> Result> { - let now = now_unix(); - Ok(coord - .todos - .list(agent, subsystem)? - .into_iter() - .map(|t| LooseEnd::Todo { - id: t.id, - subsystem: t.subsystem, - subsystem_key: t.subsystem_key, - summary: t.summary, - source: t.source, - age_seconds: saturating_age(now, t.updated_at.timestamp()), - }) - .collect()) -} - /// Hive-wide loose-ends view: EVERY pending approval + EVERY /// unanswered question + EVERY pending reminder. Manager surface /// only; sub-agents can't see each other's threads via the agent diff --git a/hive-c0re/src/main.rs b/hive-c0re/src/main.rs index dd704ad4..a97aff7b 100644 --- a/hive-c0re/src/main.rs +++ b/hive-c0re/src/main.rs @@ -43,7 +43,6 @@ pub(crate) use stats::{ }; pub(crate) use stores::{ approvals, audit_log, broker, build_logs, db, operator_questions, power, scheduled_prompts, - todos, }; pub(crate) use workers::{ agent_sockets, auto_update, crash_watch, knowledge, mcp_sockets, reminder_scheduler, diff --git a/hive-c0re/src/socket_server/mod.rs b/hive-c0re/src/socket_server/mod.rs index 99be0eb2..8f14dfd5 100644 --- a/hive-c0re/src/socket_server/mod.rs +++ b/hive-c0re/src/socket_server/mod.rs @@ -596,29 +596,6 @@ async fn dispatch(req: &Request, agent: &str, coord: &Arc) -> Respo since_secs, agent: target, } => handle_reminder_rollup(coord, agent, target.as_deref(), *since_secs), - // Todos (loose-ends v2): in-container subsystems push/clear - // their own; the agent lists / marks its own done. Scoped to the - // calling agent (the socket identity) — no cross-agent access. - Request::UpsertTodo { - subsystem, - key, - summary, - source, - } => handle_upsert_todo( - coord, - agent, - subsystem, - key.as_deref(), - summary, - source.as_deref(), - ), - Request::ClearTodo { - subsystem, - key, - all, - } => handle_clear_todo(coord, agent, subsystem, key.as_deref(), *all), - Request::ListTodos { subsystem } => handle_list_todos(coord, agent, subsystem.as_deref()), - Request::MarkTodoDone { id } => handle_mark_todo_done(coord, agent, *id), // Orchestration / diagnostics verbs — gated per-verb on tool-group // membership or topology (see `dispatch_orchestration`). _ => dispatch_orchestration(req, agent, coord).await, @@ -813,83 +790,6 @@ fn handle_get_loose_ends(coord: &Arc, agent: &str, target: Option<& } } -/// `UpsertTodo` — a subsystem pushes/updates one of this agent's todos. -/// Coalesces a wake ONLY when the row is new or actually changed, so -/// re-pushing an identical keyed todo is a silent no-op. -fn handle_upsert_todo( - coord: &Arc, - agent: &str, - subsystem: &str, - key: Option<&str>, - summary: &str, - source: Option<&str>, -) -> Response { - match coord.todos.upsert(agent, subsystem, key, summary, source) { - Ok((_, changed)) => { - if changed { - let _ = coord.broker.send(&Message { - from: "todo".to_owned(), - to: agent.to_owned(), - body: "you have todos — call get_loose_ends to see them".to_owned(), - in_reply_to: None, - }); - } - Response::Ok - } - Err(e) => Response::Err { - message: format!("{e:#}"), - }, - } -} - -/// `ClearTodo` — a producer clears a resolved todo by `(subsystem, key)`, -/// or wipes its whole set when `all` (cancel-and-recreate on restart). -fn handle_clear_todo( - coord: &Arc, - agent: &str, - subsystem: &str, - key: Option<&str>, - all: bool, -) -> Response { - let result = if all { - coord.todos.clear_subsystem(agent, subsystem) - } else { - coord.todos.clear(agent, subsystem, key) - }; - match result { - Ok(count) => Response::Acked { - count: u64::try_from(count).unwrap_or(0), - }, - Err(e) => Response::Err { - message: format!("{e:#}"), - }, - } -} - -/// `ListTodos` — enumerate this agent's todos (optionally one subsystem's) -/// as `LooseEnd::Todo` rows, so a producer can reconcile its own set. -fn handle_list_todos(coord: &Arc, agent: &str, subsystem: Option<&str>) -> Response { - match crate::loose_ends::todos_for(coord, agent, subsystem) { - Ok(loose_ends) => Response::LooseEnds { loose_ends }, - Err(e) => Response::Err { - message: format!("{e:#}"), - }, - } -} - -/// `MarkTodoDone` — the agent clears one of its own todos by id (scoped to -/// the agent, so it can't touch another agent's). -fn handle_mark_todo_done(coord: &Arc, agent: &str, id: i64) -> Response { - match coord.todos.mark_done(agent, id) { - Ok(count) => Response::Acked { - count: u64::try_from(count).unwrap_or(0), - }, - Err(e) => Response::Err { - message: format!("{e:#}"), - }, - } -} - /// `CountPendingReminders` — resolve the target (own / subtree free, else /// `QueryAgentState`) then count its pending reminders. fn handle_count_pending_reminders( diff --git a/hive-c0re/src/stores/mod.rs b/hive-c0re/src/stores/mod.rs index 45e676c6..1c02820c 100644 --- a/hive-c0re/src/stores/mod.rs +++ b/hive-c0re/src/stores/mod.rs @@ -12,4 +12,3 @@ pub mod db; pub mod operator_questions; pub mod power; pub mod scheduled_prompts; -pub mod todos; diff --git a/hive-c0re/src/stores/todos.rs b/hive-c0re/src/stores/todos.rs deleted file mode 100644 index 8effdd65..00000000 --- a/hive-c0re/src/stores/todos.rs +++ /dev/null @@ -1,341 +0,0 @@ -//! Todo store — the persistent, DB-backed half of the "todos" -//! (loose-ends v2) system. -//! -//! Subsystems inside an agent's container (matrix, forge-notify, bash, -//! …) push *todos* to the agent over the mcp.sock protocol instead of -//! firing wakes directly. Todos are scoped to the owning `agent` (the -//! socket identity of the pushing container) and tagged with a -//! `subsystem` marker plus an optional `subsystem_key` (a matrix room -//! id, a bash task id, …); together `(agent, subsystem, subsystem_key)` -//! is the upsert/dedup key, so re-pushing the same item is idempotent -//! (no duplicate) and a producer can list / clear / rebuild only its own -//! set (e.g. matrix wipes + recreates its todos on daemon restart). -//! -//! Removal has two paths (mara's call): the producing subsystem -//! `clear`s a todo it has resolved (keyed by subsystem + key), or the -//! agent itself `mark_done`s one by id. Both delete the row. -//! -//! This table holds only the *dynamic* subsystem-pushed todos. The -//! static ones (pending approvals / questions / reminders / undelivered -//! messages) are still computed on demand in `loose_ends.rs`; `get_todos` -//! merges the two. Folding the static kinds into this table is a later -//! increment. - -use std::path::Path; -use std::sync::Mutex; - -use anyhow::{Context, Result}; -use chrono::{DateTime, Utc}; -use hive_sh4re::wire_time::now_unix; -use rusqlite::{Connection, params}; -use serde::Serialize; - -use crate::db::Migration; - -const SCHEMA: &str = r" -CREATE TABLE IF NOT EXISTS todos ( - id INTEGER PRIMARY KEY AUTOINCREMENT, - agent TEXT NOT NULL, - subsystem TEXT NOT NULL, - subsystem_key TEXT, - summary TEXT NOT NULL, - source TEXT, - created_at INTEGER NOT NULL, - updated_at INTEGER NOT NULL -); --- (agent, subsystem, subsystem_key) is the upsert/dedup key. A NULL key --- never conflicts (SQLite treats NULLs as distinct), so keyless todos --- always insert as one-offs; keyed todos update in place. -CREATE UNIQUE INDEX IF NOT EXISTS idx_todos_dedup - ON todos (agent, subsystem, subsystem_key); -"; - -// New table — no legacy rows to migrate past the initial schema. -const MIGRATIONS: &[Migration] = &[]; - -/// One dynamic, subsystem-pushed todo. -#[derive(Debug, Clone, Serialize)] -pub struct Todo { - pub id: i64, - /// Owning agent (the container whose subsystem pushed it). - pub agent: String, - /// Producing subsystem marker (`"matrix"`, `"forge"`, `"bash"`, …). - pub subsystem: String, - /// Optional subsystem-specific dedup key (matrix room id, bash task - /// id, …). `None` = a keyless one-off todo. - #[serde(default, skip_serializing_if = "Option::is_none")] - pub subsystem_key: Option, - /// Human-readable one-line summary shown to the agent. - pub summary: String, - /// Optional free-text provenance (e.g. the room name / task label). - #[serde(default, skip_serializing_if = "Option::is_none")] - pub source: Option, - pub created_at: DateTime, - pub updated_at: DateTime, -} - -pub struct Todos { - conn: Mutex, -} - -impl Todos { - /// Open (creating if needed) the todo store at `path`. - /// - /// # Errors - /// - /// Propagates sqlite open / schema-apply / migration failures. - pub fn open(path: &Path) -> Result { - let conn = crate::db::open(path, "todos")?; - conn.execute_batch(SCHEMA).context("apply todos schema")?; - crate::db::apply_versioned_migrations(&conn, "todos", MIGRATIONS)?; - Ok(Self { - conn: Mutex::new(conn), - }) - } - - /// Insert a todo for `agent`, or update the existing one for - /// `(agent, subsystem, key)` when `key` is `Some` and already - /// present. A `None` key never conflicts, so it always inserts a - /// fresh row. - /// - /// Returns `(id, changed)` where `changed` is `true` when the row is - /// new OR its `summary`/`source` actually differed — the caller uses - /// this to decide whether to coalesce a wake (re-pushing an identical - /// keyed todo is a no-op and must not re-wake). - /// - /// # Errors - /// - /// Propagates sqlite query / execute failures. - /// - /// # Panics - /// - /// Panics if the connection mutex is poisoned. - pub fn upsert( - &self, - agent: &str, - subsystem: &str, - key: Option<&str>, - summary: &str, - source: Option<&str>, - ) -> Result<(i64, bool)> { - let conn = self.conn.lock().unwrap(); - let now = now_unix(); - let existing: Option<(i64, String, Option)> = if key.is_some() { - conn.query_row( - "SELECT id, summary, source FROM todos \ - WHERE agent = ?1 AND subsystem = ?2 AND subsystem_key IS ?3", - params![agent, subsystem, key], - |row| Ok((row.get(0)?, row.get(1)?, row.get(2)?)), - ) - .ok() - } else { - None - }; - if let Some((id, cur_summary, cur_source)) = existing { - let unchanged = cur_summary == summary && cur_source.as_deref() == source; - if unchanged { - return Ok((id, false)); - } - conn.execute( - "UPDATE todos SET summary = ?1, source = ?2, updated_at = ?3 WHERE id = ?4", - params![summary, source, now, id], - )?; - return Ok((id, true)); - } - conn.execute( - "INSERT INTO todos \ - (agent, subsystem, subsystem_key, summary, source, created_at, updated_at) \ - VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?6)", - params![agent, subsystem, key, summary, source, now], - )?; - Ok((conn.last_insert_rowid(), true)) - } - - /// Clear producer-resolved todo(s) by `(agent, subsystem, key)`. - /// `key = Some(k)` targets the one keyed row; `key = None` matches - /// `subsystem_key IS NULL`, i.e. **all** keyless todos for that - /// subsystem (keyless rows have no distinguishing key — clear a - /// specific one via [`Todos::mark_done`] by id instead). Returns the - /// number of rows deleted (0 when nothing matched). - /// - /// # Errors - /// - /// Propagates the sqlite delete failure. - /// - /// # Panics - /// - /// Panics if the connection mutex is poisoned. - pub fn clear(&self, agent: &str, subsystem: &str, key: Option<&str>) -> Result { - let conn = self.conn.lock().unwrap(); - let n = conn.execute( - "DELETE FROM todos WHERE agent = ?1 AND subsystem = ?2 AND subsystem_key IS ?3", - params![agent, subsystem, key], - )?; - Ok(n) - } - - /// Clear every todo `agent`'s `subsystem` owns — used by a producer - /// that rebuilds its whole set on restart (cancel-and-recreate). - /// Returns the number of rows deleted. - /// - /// # Errors - /// - /// Propagates the sqlite delete failure. - /// - /// # Panics - /// - /// Panics if the connection mutex is poisoned. - pub fn clear_subsystem(&self, agent: &str, subsystem: &str) -> Result { - let conn = self.conn.lock().unwrap(); - let n = conn.execute( - "DELETE FROM todos WHERE agent = ?1 AND subsystem = ?2", - params![agent, subsystem], - )?; - Ok(n) - } - - /// The agent marks one of *its own* todos done, by id. Scoped to - /// `agent` so one agent can't clear another's. Returns the number of - /// rows deleted (0 when the id was unknown / not owned / already gone). - /// - /// # Errors - /// - /// Propagates the sqlite delete failure. - /// - /// # Panics - /// - /// Panics if the connection mutex is poisoned. - pub fn mark_done(&self, agent: &str, id: i64) -> Result { - let conn = self.conn.lock().unwrap(); - let n = conn.execute( - "DELETE FROM todos WHERE id = ?1 AND agent = ?2", - params![id, agent], - )?; - Ok(n) - } - - /// List `agent`'s todos, newest-updated first. `subsystem = Some(..)` - /// filters to one producer's set (so a producer can enumerate + - /// reconcile only its own); `None` returns all of the agent's. - /// - /// # Errors - /// - /// Propagates the sqlite prepare / query failures. - /// - /// # Panics - /// - /// Panics if the connection mutex is poisoned. - pub fn list(&self, agent: &str, subsystem: Option<&str>) -> Result> { - let conn = self.conn.lock().unwrap(); - let mut stmt = conn.prepare( - "SELECT id, agent, subsystem, subsystem_key, summary, source, created_at, updated_at \ - FROM todos \ - WHERE agent = ?1 AND (?2 IS NULL OR subsystem = ?2) \ - ORDER BY updated_at DESC, id DESC", - )?; - let rows = stmt - .query_map(params![agent, subsystem], |row| { - let created: i64 = row.get(6)?; - let updated: i64 = row.get(7)?; - Ok(Todo { - id: row.get(0)?, - agent: row.get(1)?, - subsystem: row.get(2)?, - subsystem_key: row.get(3)?, - summary: row.get(4)?, - source: row.get(5)?, - created_at: DateTime::from_timestamp(created, 0).unwrap_or_default(), - updated_at: DateTime::from_timestamp(updated, 0).unwrap_or_default(), - }) - })? - .collect::>>()?; - Ok(rows) - } -} - -#[cfg(test)] -mod tests { - use super::*; - - // Return the `TempDir` alongside the store so it outlives the test — - // dropping it early deletes the dir and SQLite fails with - // `SQLITE_READONLY_DBMOVED`. - fn store() -> (tempfile::TempDir, Todos) { - let dir = tempfile::tempdir().unwrap(); - let db = Todos::open(&dir.path().join("todos.sqlite")).unwrap(); - (dir, db) - } - - #[test] - fn keyed_upsert_dedups_and_reports_changed() { - let (_dir, s) = store(); - let (id1, changed1) = s - .upsert("alice", "matrix", Some("!room:x"), "1 unread", None) - .unwrap(); - assert!(changed1, "first push is new → changed"); - // Same key + same summary → no-op, not changed (must not re-wake). - let (id2, changed2) = s - .upsert("alice", "matrix", Some("!room:x"), "1 unread", None) - .unwrap(); - assert_eq!(id1, id2, "keyed upsert updates in place, same row"); - assert!(!changed2, "identical re-push is a no-op"); - // Same key, new summary → updates, changed. - let (id3, changed3) = s - .upsert("alice", "matrix", Some("!room:x"), "3 unread", None) - .unwrap(); - assert_eq!(id1, id3); - assert!(changed3); - assert_eq!(s.list("alice", Some("matrix")).unwrap().len(), 1); - } - - #[test] - fn todos_are_scoped_per_agent() { - let (_dir, s) = store(); - // Same subsystem+key for two agents → distinct rows. - s.upsert("alice", "matrix", Some("!r:x"), "u", None) - .unwrap(); - s.upsert("bob", "matrix", Some("!r:x"), "u", None).unwrap(); - assert_eq!(s.list("alice", None).unwrap().len(), 1); - assert_eq!(s.list("bob", None).unwrap().len(), 1); - // bob can't mark alice's todo done. - let alice_id = s.list("alice", None).unwrap()[0].id; - assert_eq!(s.mark_done("bob", alice_id).unwrap(), 0); - assert_eq!(s.mark_done("alice", alice_id).unwrap(), 1); - } - - #[test] - fn keyless_todos_always_insert() { - let (_dir, s) = store(); - let (a, _) = s.upsert("alice", "bash", None, "task done", None).unwrap(); - let (b, _) = s.upsert("alice", "bash", None, "task done", None).unwrap(); - assert_ne!(a, b, "keyless pushes are distinct one-offs"); - assert_eq!(s.list("alice", Some("bash")).unwrap().len(), 2); - } - - #[test] - fn clear_and_mark_done_remove_rows() { - let (_dir, s) = store(); - s.upsert("alice", "matrix", Some("!a:x"), "unread", None) - .unwrap(); - let (id, _) = s - .upsert("alice", "forge", Some("pr-1"), "review", None) - .unwrap(); - assert_eq!(s.clear("alice", "matrix", Some("!a:x")).unwrap(), 1); - assert_eq!(s.mark_done("alice", id).unwrap(), 1); - assert!(s.list("alice", None).unwrap().is_empty()); - } - - #[test] - fn clear_subsystem_wipes_only_its_own() { - let (_dir, s) = store(); - s.upsert("alice", "matrix", Some("!a:x"), "u", None) - .unwrap(); - s.upsert("alice", "matrix", Some("!b:x"), "u", None) - .unwrap(); - s.upsert("alice", "forge", Some("pr-1"), "r", None).unwrap(); - assert_eq!(s.clear_subsystem("alice", "matrix").unwrap(), 2); - let left = s.list("alice", None).unwrap(); - assert_eq!(left.len(), 1); - assert_eq!(left[0].subsystem, "forge"); - } -} diff --git a/hive-core-agent-sock/src/lib.rs b/hive-core-agent-sock/src/lib.rs index e3ef2093..8c4b58ba 100644 --- a/hive-core-agent-sock/src/lib.rs +++ b/hive-core-agent-sock/src/lib.rs @@ -102,41 +102,6 @@ pub enum Request { #[serde(default, skip_serializing_if = "Option::is_none")] agent: Option, }, - /// Upsert a *todo* (loose-ends v2) from an in-container - /// subsystem (matrix / forge / bash). `subsystem` is the producer - /// marker; `key` is the optional subsystem-specific dedup key (a - /// matrix room id, a bash task id). Re-pushing an identical keyed - /// todo is a no-op; a new-or-changed one coalesces a wake to the - /// agent. Keyless todos always insert as one-offs. - UpsertTodo { - subsystem: String, - #[serde(default, skip_serializing_if = "Option::is_none")] - key: Option, - summary: String, - #[serde(default, skip_serializing_if = "Option::is_none")] - source: Option, - }, - /// Clear producer-resolved todo(s) by `(subsystem, key)`. `key = - /// Some(k)` clears the one keyed row; `key = None` clears **all** of - /// the subsystem's keyless todos (rows with no key can't be told - /// apart — clear a specific one via `MarkTodoDone` by id). `all = - /// true` wipes the producer's whole set (cancel-and-recreate on - /// daemon restart). - ClearTodo { - subsystem: String, - #[serde(default, skip_serializing_if = "Option::is_none")] - key: Option, - #[serde(default)] - all: bool, - }, - /// List todos, optionally filtered to one `subsystem` (a producer - /// enumerating its own set). `None` = all. - ListTodos { - #[serde(default, skip_serializing_if = "Option::is_none")] - subsystem: Option, - }, - /// The agent marks one of its own todos done, by id. - MarkTodoDone { id: i64 }, /// Count of pending (un-delivered) reminders. On the agent socket: /// same target rules as `GetLooseEnds` (self/children free; /// non-children require `query_agent_state`; `"*"` rejected). diff --git a/hive-matrix-mcp/src/handlers.rs b/hive-matrix-mcp/src/handlers.rs index 7c4d906e..b4c1810d 100644 --- a/hive-matrix-mcp/src/handlers.rs +++ b/hive-matrix-mcp/src/handlers.rs @@ -512,44 +512,6 @@ pub fn list_invites(client: &Client) -> DaemonResponse { DaemonResponse::ok(&invites) } -/// Rewrite `mcp-loose-ends/matrix.json` with a summary of all pending -/// room invites. The harness scans this directory generically in -/// `get_loose_ends` — no matrix-specific code needed there. -/// -/// Called after an invite arrives (from the sync handler) and after a -/// room is joined (to remove the accepted invite from loose-ends). -/// Atomic write (tmp + rename) so the harness never reads a partial file. -pub async fn refresh_invite_loose_ends(client: &Client) { - let invites = client.invited_rooms(); - let dir = crate::paths::mcp_loose_ends_dir(); - if let Err(e) = tokio::fs::create_dir_all(&dir).await { - tracing::warn!(error = ?e, "matrix: create mcp-loose-ends dir failed"); - return; - } - let items: Vec = invites - .iter() - .map(|room| { - let label = room_label(room); - format!( - "[matrix] pending invite: {label} — use list_invites to see, resolve_invite to accept or reject" - ) - }) - .collect(); - let dest = dir.join("matrix.json"); - let tmp = dest.with_extension("json.tmp"); - let json = serde_json::to_string(&items).unwrap_or_else(|_| "[]".to_owned()); - match tokio::fs::write(&tmp, &json).await { - Ok(()) => { - if let Err(e) = tokio::fs::rename(&tmp, &dest).await { - tracing::warn!(error = ?e, "matrix: rename mcp-loose-ends/matrix.json failed"); - } - } - Err(e) => { - tracing::warn!(error = ?e, "matrix: write mcp-loose-ends/matrix.json.tmp failed"); - } - } -} - pub async fn join_room(client: &Client, room_ref: &str) -> DaemonResponse { let parsed: &RoomOrAliasId = match room_ref.try_into() { Ok(p) => p, @@ -560,9 +522,14 @@ pub async fn join_room(client: &Client, room_ref: &str) -> DaemonResponse { let server_names: Vec = vec![]; match client.join_room_by_id_or_alias(parsed, &server_names).await { Ok(room) => { - // Refresh loose-ends so the accepted invite is removed from - // `get_loose_ends` output immediately after the agent joins. - refresh_invite_loose_ends(client).await; + // Clear the invite todo so the accepted invite drops out of + // `get_loose_ends` immediately (the sweep would also clear it + // on its next tick, but this makes it instant). + let _ = crate::wake::send_todo_clear( + Some(&crate::timeline::invite_key(room.room_id())), + false, + ) + .await; DaemonResponse::ok(&serde_json::json!({ "joined": true, "room_id": room.room_id().to_string(), @@ -595,7 +562,12 @@ pub async fn resolve_invite( }; match room.leave().await { Ok(()) => { - refresh_invite_loose_ends(client).await; + // Clear the invite todo immediately on reject. + let _ = crate::wake::send_todo_clear( + Some(&crate::timeline::invite_key(room.room_id())), + false, + ) + .await; DaemonResponse::ok(&serde_json::json!({ "rejected": true, "room_id": room.room_id().to_string(), diff --git a/hive-matrix-mcp/src/main.rs b/hive-matrix-mcp/src/main.rs index 42d768dd..dbb4f39c 100644 --- a/hive-matrix-mcp/src/main.rs +++ b/hive-matrix-mcp/src/main.rs @@ -73,7 +73,6 @@ async fn main() -> Result<()> { let cfgs = accounts::configured().context("read matrix account config")?; let mcp_socket = paths::daemon_socket(); - let hyperhive_socket = paths::hyperhive_socket(); let multi = cfgs.len() > 1; let primary = cfgs[0].name.clone(); let mut registry = Registry::new(primary); @@ -81,10 +80,10 @@ async fn main() -> Result<()> { for (idx, cfg) in cfgs.into_iter().enumerate() { let is_primary = idx == 0; - // Account-tag the wakes only in multi-account mode so single- - // account wake bodies stay byte-identical to the legacy format. + // Account-tag the todos only in multi-account mode so single- + // account todo summaries stay byte-identical to the legacy format. let tag = multi.then(|| cfg.name.clone()); - match bring_up_account(&cfg, &hyperhive_socket, tag.clone(), is_primary).await { + match bring_up_account(&cfg, tag.clone(), is_primary).await { Ok(Some((client, sync_loop))) => { registry.insert(cfg.name, client); sync_loops.push(sync_loop); @@ -109,8 +108,7 @@ async fn main() -> Result<()> { // before being skipped for this daemon lifetime. Err(e) if is_primary => return Err(e.context("bring up primary matrix account")), Err(e) => { - if let Some((client, sync_loop)) = - bring_up_secondary_with_retry(&cfg, &hyperhive_socket, tag, e).await + if let Some((client, sync_loop)) = bring_up_secondary_with_retry(&cfg, tag, e).await { registry.insert(cfg.name, client); sync_loops.push(sync_loop); @@ -189,7 +187,6 @@ async fn main() -> Result<()> { /// failure, no token, or all retries exhausted). async fn bring_up_secondary_with_retry( cfg: &AccountCfg, - hyperhive_socket: &std::path::Path, tag: Option, first_error: anyhow::Error, ) -> Option<(Client, SyncLoop)> { @@ -219,7 +216,7 @@ async fn bring_up_secondary_with_retry( "retrying secondary account bring-up after backoff" ); tokio::time::sleep(std::time::Duration::from_secs(delay)).await; - match bring_up_account(cfg, hyperhive_socket, tag.clone(), false).await { + match bring_up_account(cfg, tag.clone(), false).await { Ok(Some((client, sync_loop))) => { tracing::info!(account = %cfg.name, "secondary matrix account recovered"); return Some((client, sync_loop)); @@ -260,7 +257,6 @@ async fn bring_up_secondary_with_retry( /// handling). The returned sync loop is driven by the caller. async fn bring_up_account( cfg: &AccountCfg, - hyperhive_socket: &std::path::Path, tag: Option, is_primary: bool, ) -> Result> { @@ -293,30 +289,26 @@ async fn bring_up_account( let sync_client = client.clone(); let cb_client = client.clone(); - let wake_socket = Arc::new(hyperhive_socket.to_path_buf()); - // Separate dedup sets: one tracks invites already woken about, the + // Separate dedup sets: one tracks invites already pushed as todos, the // other unread-message rooms. Both are pruned to their current state - // each sweep (see the sweep fns) so re-invites / new messages re-wake. + // each sweep (see the sweep fns) so re-invites / new messages re-push. let invite_notified = Arc::new(tokio::sync::Mutex::new(std::collections::HashSet::new())); let unread_notified = Arc::new(tokio::sync::Mutex::new(std::collections::HashSet::new())); // Startup cancel-and-recreate (loose-ends v2): wipe this agent's - // matrix todos so stale ones (rooms read while the daemon was down) don't - // linger, then let the first sweep rebuild the set to match current - // unread reality. Best-effort; the sweep converges regardless. - let _ = crate::wake::send_todo_clear(hyperhive_socket, None, true).await; + // matrix todos so stale ones (rooms read / invites resolved while the + // daemon was down) don't linger, then let the first sweep rebuild the + // set to match current reality. Best-effort; the sweep converges. + let _ = crate::wake::send_todo_clear(None, true).await; let sync_loop: SyncLoop = Box::pin(async move { sync_client .sync_with_callback(SyncSettings::default(), move |_response| { let client = cb_client.clone(); - let socket = wake_socket.clone(); let invite_notified = invite_notified.clone(); let unread_notified = unread_notified.clone(); let tag = tag.clone(); async move { - timeline::sweep_invites(&client, &socket, &invite_notified, tag.as_deref()) - .await; - timeline::sweep_unread(&client, &socket, &unread_notified, tag.as_deref()) - .await; + timeline::sweep_invites(&client, &invite_notified, tag.as_deref()).await; + timeline::sweep_unread(&client, &unread_notified, tag.as_deref()).await; matrix_sdk::LoopCtrl::Continue } }) diff --git a/hive-matrix-mcp/src/paths.rs b/hive-matrix-mcp/src/paths.rs index ccc8a441..dc9dcf17 100644 --- a/hive-matrix-mcp/src/paths.rs +++ b/hive-matrix-mcp/src/paths.rs @@ -72,22 +72,3 @@ pub fn accounts_file() -> PathBuf { let state_dir = std::env::var("HYPERHIVE_STATE_DIR").unwrap_or_default(); PathBuf::from(format!("{state_dir}/matrix-accounts.json")) } - -/// Hyperhive control socket — the daemon writes wake signals here so -/// the harness drives a new claude turn on incoming matrix events. -/// Mirrors the path `forge_notify` writes to. -#[must_use] -pub fn hyperhive_socket() -> PathBuf { - std::env::var_os("HIVE_CONTROL_SOCKET") - .map_or_else(|| PathBuf::from("/run/hive/mcp.sock"), PathBuf::from) -} - -/// Directory where MCP daemons write loose-end summary files for the harness. -/// Each daemon writes `.json` here; the harness scans the dir in -/// `get_loose_ends` to surface active work from all MCPs generically. -/// Resolution lives in `hive_sh4re::paths` so the harness + every MCP -/// daemon agree on the location. -#[must_use] -pub fn mcp_loose_ends_dir() -> PathBuf { - hive_sh4re::paths::mcp_loose_ends_dir() -} diff --git a/hive-matrix-mcp/src/timeline.rs b/hive-matrix-mcp/src/timeline.rs index 50ebc082..6e46e4fa 100644 --- a/hive-matrix-mcp/src/timeline.rs +++ b/hive-matrix-mcp/src/timeline.rs @@ -1,19 +1,18 @@ -//! Matrix → hyperhive wake bridge. Both room messages and room invites -//! are surfaced to the agent by SWEEPING state on every post-sync -//! callback, not by one-shot `m.room.message` / `StrippedRoomMemberEvent` -//! handlers: a one-shot wake whose `send_wake` raced a hive-c0re / socket- -//! down window (a container rebuild) was dropped with no retry, leaving -//! the agent deaf to matrix activity until manually prompted. Sweeping the -//! reliable sync path re-checks each tick and self-heals a dropped wake on -//! the next one. +//! Matrix → hyperhive todo bridge (loose-ends v2). Both room messages and +//! room invites are surfaced to the agent by SWEEPING state on every +//! post-sync callback, not by one-shot `m.room.message` / +//! `StrippedRoomMemberEvent` handlers: a one-shot signal that raced a +//! socket-down window (a container rebuild) was dropped with no retry, +//! leaving the agent deaf to matrix activity until manually prompted. +//! Sweeping the reliable sync path re-checks each tick and self-heals a +//! dropped push on the next one — each todo is idempotent by room key. //! -//! Wake bodies stay short: `sweep_unread` sends the all-rooms unread +//! Todo summaries stay short: `sweep_unread` pushes the per-room unread //! summary (`wake::format_unread_summary`); the message stays unread //! server-side so the agent fetches detail via `read_room`. Self-sent //! messages never raise an unread notification, so they never wake. use std::collections::HashSet; -use std::path::Path; use matrix_sdk::{Client, ruma::OwnedRoomId}; use tokio::sync::Mutex; @@ -38,7 +37,6 @@ use crate::{handlers, wake}; /// no explicit self-filter is needed here. pub async fn sweep_unread( client: &Client, - socket: &Path, notified: &Mutex>, account_tag: Option<&str>, ) { @@ -53,7 +51,7 @@ pub async fn sweep_unread( active.difference(&unread_ids).cloned().collect() }; for id in stale { - if wake::send_todo_clear(socket, Some(id.as_str()), false) + if wake::send_todo_clear(Some(id.as_str()), false) .await .is_ok() { @@ -61,15 +59,15 @@ pub async fn sweep_unread( } } - // Upsert a todo per currently-unread room. hive-c0re coalesces the wake - // iff the summary is new or changed, so re-upserting an unchanged room - // every sync tick is a cheap server-side no-op (no re-wake). + // Upsert a todo per currently-unread room. The harness coalesces the + // wake iff the summary is new or changed, so re-upserting an unchanged + // room every sync tick is a cheap no-op (no re-wake). for (id, ru) in &unread { let summary = wake::tag_account( account_tag, wake::format_unread_summary(std::slice::from_ref(ru)), ); - match wake::send_todo_upsert(socket, id.as_str(), &summary).await { + match wake::send_todo_upsert(id.as_str(), &summary).await { Ok(()) => { notified.lock().await.insert(id.clone()); } @@ -95,7 +93,6 @@ pub async fn sweep_unread( /// decides whether to accept or reject by calling `resolve_invite`. pub async fn sweep_invites( client: &Client, - socket: &Path, notified: &Mutex>, account_tag: Option<&str>, ) { @@ -103,38 +100,52 @@ pub async fn sweep_invites( let current_ids: HashSet = current.iter().map(|r| r.room_id().to_owned()).collect(); - let mut fresh = Vec::new(); - { - let mut seen = notified.lock().await; - // Drop invites that are no longer pending (joined/rejected/withdrawn) - // so a future re-invite to the same room wakes the agent again. - seen.retain(|id| current_ids.contains(id)); - for room in ¤t { - if seen.insert(room.room_id().to_owned()) { - fresh.push(room.clone()); - } + // Invites we previously pushed a todo for that are no longer pending + // (joined / rejected / withdrawn) → clear their todo, then drop from + // `notified` on success so a future re-invite re-upserts (retry next + // tick on failure). + let stale: Vec = { + let seen = notified.lock().await; + seen.difference(¤t_ids).cloned().collect() + }; + for id in stale { + if wake::send_todo_clear(Some(&invite_key(&id)), false) + .await + .is_ok() + { + notified.lock().await.remove(&id); } } - if fresh.is_empty() { - return; - } - // Refresh loose-ends before waking so the invite is visible in - // get_loose_ends during the agent's turn. - handlers::refresh_invite_loose_ends(client).await; - for room in fresh { + // Upsert a todo per pending invite. Keyed `invite:` (distinct + // from the unread sweep's `` key) so the two never collide; the + // harness coalesces the wake iff the summary is new or changed, so + // re-upserting an unchanged invite every tick is a cheap no-op. + for room in ¤t { let room_id = room.room_id().to_owned(); let label = room.name().unwrap_or_else(|| room_id.to_string()); - tracing::info!(%room_id, "matrix: pending invite swept, waking agent"); - let body = wake::tag_account( + let summary = wake::tag_account( account_tag, format!( "[matrix] invited to {label} ({room_id}) — \ use list_invites to see pending invites, resolve_invite to accept or reject" ), ); - if let Err(e) = wake::send_wake(socket, &body).await { - tracing::warn!(error = %e, "matrix: failed to deliver invite-wake to hyperhive"); + match wake::send_todo_upsert(&invite_key(&room_id), &summary).await { + Ok(()) => { + notified.lock().await.insert(room_id); + } + Err(e) => { + tracing::warn!(error = %e, room = %room_id, "matrix: invite todo upsert failed; will retry next sweep"); + } } } } + +/// Todo dedup key for a pending invite. Namespaced with an `invite:` +/// prefix so an invited room and an unread room (which the unread sweep +/// keys by bare room id) never share a todo row. `pub(crate)` so the +/// invite-resolution handlers can clear the matching todo immediately. +pub(crate) fn invite_key(room_id: &matrix_sdk::ruma::RoomId) -> String { + format!("invite:{room_id}") +} diff --git a/hive-matrix-mcp/src/wake.rs b/hive-matrix-mcp/src/wake.rs index 18c41d62..4598af2c 100644 --- a/hive-matrix-mcp/src/wake.rs +++ b/hive-matrix-mcp/src/wake.rs @@ -1,17 +1,14 @@ -//! Wake-signal writer: notifies the hyperhive harness when an incoming -//! matrix event arrives so claude drives a new turn. +//! Todo writer: pushes matrix *todos* (loose-ends v2) to the harness's +//! in-agent socket (`HIVE_AGENT_SOCKET`) when rooms have unread messages +//! or pending invites, so claude drives a turn to handle them. One JSON +//! line per op (`upsert_todo` / `clear_todo`), keyed by room id so +//! re-pushing an unchanged item is an idempotent no-op and resolving one +//! clears it. The harness owns the todo store locally and signals its own +//! turn loop — no hive-c0re round-trip. //! -//! Same wire shape as `hive-agent::forge_notify`'s wake: a single JSON -//! line written to the hyperhive control socket (`/run/hive/mcp.sock` -//! by default) carrying an `Request::Wake { from, body }`. -//! The agent harness's `agent_server` parses it and treats it as a -//! `Wake` from the matrix subsystem. -//! -//! Per the operator's call (phase 3): the body is a SHORT TEASER, not -//! the full message — the agent then reads the unmarked event via -//! the `read_room` MCP tool. Truncation to ~100 chars keeps the wake -//! prompt focused (`forge_notify` embeds longer excerpts because the -//! agent doesn't have a follow-up read-the-original tool for forge). +//! Todo summaries stay short: a SHORT TEASER, not the full message — the +//! agent then reads the unmarked event via the `read_room` MCP tool. +//! Truncation to ~100 chars keeps the summary focused. use std::path::Path; @@ -19,65 +16,59 @@ use anyhow::{Context, Result}; use tokio::io::AsyncWriteExt; use tokio::net::UnixStream; -/// Send an `Request::Wake { from: "matrix", body }` to the hyperhive -/// control socket at `socket`. Best-effort: returns Err on any plumbing -/// failure; callers log + ignore so a wake delivery hiccup doesn't tear -/// down the matrix sync loop. -/// -/// Wire format matches `hive_core_agent_sock::Request` tagged with `"cmd"` per -/// `#[serde(tag = "cmd", rename_all = "snake_case")]`. Must be `"cmd"`, -/// not `"kind"` — the harness deserialises against the hive-sh4re type -/// and silently discards requests that don't match. -/// -/// # Errors -/// -/// Returns an error on socket connect failure, serialisation failure, -/// or I/O error writing to or reading from the socket. -pub async fn send_wake(socket: &Path, body: impl AsRef) -> Result<()> { - let payload = serde_json::json!({ - "cmd": "wake", - "from": "matrix", - "body": body.as_ref(), - "transient": true, - }); - send_line(socket, &payload).await +/// The harness-served in-agent socket (`HIVE_AGENT_SOCKET`) where todo ops +/// go — distinct from the host-served control socket used by [`send_wake`]. +/// `None` when unset/empty, in which case todo sends are a best-effort +/// no-op (a standalone daemon without the harness socket). +fn agent_socket() -> Option { + std::env::var_os("HIVE_AGENT_SOCKET") + .filter(|v| !v.is_empty()) + .map(std::path::PathBuf::from) } -/// Upsert a matrix-subsystem *todo* (loose-ends v2) on the -/// hyperhive control socket — the replacement for a direct wake. `key` is -/// the room id (the dedup key); hive-c0re coalesces a wake iff the todo is -/// new or its `summary` changed. Best-effort like [`send_wake`]. +/// Upsert a matrix-subsystem *todo* (loose-ends v2) on the harness's +/// in-agent socket — the replacement for a direct wake. `key` is the room +/// id (the dedup key); the harness signals a turn iff the todo is new or +/// its `summary` changed. Best-effort: a no-op when `HIVE_AGENT_SOCKET` +/// isn't configured. /// /// # Errors /// /// Returns an error on socket connect failure, serialisation failure, /// or I/O error writing to or reading from the socket. -pub async fn send_todo_upsert(socket: &Path, key: &str, summary: impl AsRef) -> Result<()> { +pub async fn send_todo_upsert(key: &str, summary: impl AsRef) -> Result<()> { + let Some(socket) = agent_socket() else { + return Ok(()); + }; let payload = serde_json::json!({ "cmd": "upsert_todo", "subsystem": "matrix", "key": key, "summary": summary.as_ref(), }); - send_line(socket, &payload).await + send_line(&socket, &payload).await } -/// Clear matrix-subsystem todos. `key = Some(room)` clears one room's -/// todo (it was read); `all = true` wipes the whole matrix set -/// (cancel-and-recreate on daemon restart). Best-effort. +/// Clear matrix-subsystem todos on the harness's in-agent socket. `key = +/// Some(room)` clears one room's todo (it was read); `all = true` wipes the +/// whole matrix set (cancel-and-recreate on daemon restart). Best-effort: +/// a no-op when `HIVE_AGENT_SOCKET` isn't configured. /// /// # Errors /// /// Returns an error on socket connect failure, serialisation failure, /// or I/O error writing to or reading from the socket. -pub async fn send_todo_clear(socket: &Path, key: Option<&str>, all: bool) -> Result<()> { +pub async fn send_todo_clear(key: Option<&str>, all: bool) -> Result<()> { + let Some(socket) = agent_socket() else { + return Ok(()); + }; let payload = serde_json::json!({ "cmd": "clear_todo", "subsystem": "matrix", "key": key, "all": all, }); - send_line(socket, &payload).await + send_line(&socket, &payload).await } /// Write one JSON request line to the hyperhive control socket and drain diff --git a/nix/agent-modules/agent-service.nix b/nix/agent-modules/agent-service.nix index 2dca4049..21671fb5 100644 --- a/nix/agent-modules/agent-service.nix +++ b/nix/agent-modules/agent-service.nix @@ -201,6 +201,13 @@ in # `hive_c0re::agent_sockets::socket_path_for(name)` so lifecycle # bind-mounts and gateway upstream config stay in sync. HIVE_WEB_SOCKET = "/run/hive-agent/${userName}/web.sock"; + # In-agent socket (loose-ends v2): the harness binds this and + # serves the `hive-agent-sock` todo protocol to the in-container + # producers (matrix daemon) + the MCP bridge (`get_loose_ends`). + # Same per-agent runtime dir as the web socket so all agent-user + # services in this container can reach it; purely in-container + # (never bind-mounted to the host — unlike hive-c0re's mcp.sock). + HIVE_AGENT_SOCKET = "/run/hive-agent/${userName}/agent.sock"; # Loopback URL of the persistent `hive-mcp-http` daemon that # `render_claude_config` points claude at for the built-in # surface (HTTP is the sole transport — no per-turn stdio child). diff --git a/nix/agent-modules/matrix.nix b/nix/agent-modules/matrix.nix index 06c50331..1d13ce15 100644 --- a/nix/agent-modules/matrix.nix +++ b/nix/agent-modules/matrix.nix @@ -219,6 +219,10 @@ in wants = [ "network-online.target" ]; environment = { HIVE_MATRIX_SOCKET = "/run/hive-matrix/socket"; + # In-agent todo socket the harness serves (loose-ends v2): the + # matrix sweep pushes unread-room + pending-invite todos here + # instead of firing wakes at hive-c0re's mcp.sock. + HIVE_AGENT_SOCKET = "/run/hive-agent/${userName}/agent.sock"; RUST_LOG = "info"; } # Homeserver URL: by default the daemon inherits the host-forwarded diff --git a/nix/agent-modules/mcp.nix b/nix/agent-modules/mcp.nix index 8d039260..6a6aaeca 100644 --- a/nix/agent-modules/mcp.nix +++ b/nix/agent-modules/mcp.nix @@ -221,6 +221,10 @@ in wantedBy = [ "multi-user.target" ]; before = [ "hive-agent.service" ]; environment.RUST_LOG = "info"; + # In-agent todo socket the harness serves (loose-ends v2): the + # `get_loose_ends` handler dials it to merge this agent's local todos + # with the static loose-ends from hive-c0re. + environment.HIVE_AGENT_SOCKET = "/run/hive-agent/${userName}/agent.sock"; serviceConfig = { ExecStart = "${config.hyperhive.packages.hive-agent-mcp}/bin/hive-agent-mcp --http 127.0.0.1:${toString config.hyperhive.mcp.httpPort}"; SyslogIdentifier = "hive-mcp-http";