hyperhive/hive-c0re/src/job_queue/submit.rs

//! Request-level submit API — the surface the dashboard POST handlers,
//! the MCP socket handlers, and `hivectl` paths call.
//!
//! The **power ops** (`stop` / `start` / `restart`) are built here, not in
//! `templates.rs`: each agent's subgraph shape depends on its *live* running
//! state, which needs an async `lifecycle::is_running` read that a pure/sync
//! template can't do. So these fns are async — they read each agent's state,
//! assemble a per-agent subgraph out of the shared pure primitives
//! (`JobBuilder::node` + `templates::rebuild_nodes`), all declaring into ONE job
//! (independent per-agent roots, concurrent on their own leases).
//!
//! Dynamic shape rule: `stop`/`start` carry a head `SetWanted(w)` (durable
//! intent write) — `restart` does NOT (it bounces the container but leaves
//! `wanted` untouched, so a deliberately-stopped agent isn't forced up). The
//! tail `Reconcile` (the convergence guarantee — cheap, noops when already
//! converged) is ALWAYS present; only the *mechanical* nodes
//! (`Signal`/`Drain`/`StopForUpdate`) are state-conditional (skipped for a
//! down agent — nothing to quiesce/stop). Keeping `Reconcile` in every shape
//! closes the TOCTOU window: if an agent flips state between the `is_running`
//! read and node execution, the tail `Reconcile` still converges it in-DAG,
//! with `StopForUpdate`-noop as the backstop — no reliance on an external
//! reconcile sweep. Every helper emits a fresh queue snapshot so the
//! dashboard shows the new DAG immediately.

use std::sync::Arc;

use super::model::NodeKind;
use super::resource::Resource;
use super::templates::rebuild_nodes;
use super::{JobBuilder, Source, templates};
use crate::coordinator::Coordinator;
use crate::lifecycle;

fn submit_and_emit(
    coord: &Arc<Coordinator>,
    source: Source,
    reason: String,
    declare: impl FnOnce(&JobBuilder),
) -> u64 {
    let id = coord
        .job_queue
        .submit(source, reason, declare)
        .expect("template-declared shapes are acyclic");
    coord.emit_rebuild_queue_snapshot();
    id
}

/// Manual/approval-independent rebuild (always relocks the agent's
/// meta input — the meta-update cascade grows its own rebuild subgraphs
/// in-DAG instead of going through this surface).
pub fn rebuild(coord: &Arc<Coordinator>, agent: &str, source: Source, reason: String) -> u64 {
    submit_and_emit(coord, source, reason, |builder| {
        templates::rebuild(builder, agent, true);
    })
}

// ---- dynamic power-op DAG assembly ----------------------------------------
//
// The pure per-agent chain builders below take `running` (and `stale`)
// explicitly so they stay pure + unit-testable without a live container;
// the async `*_many` fns read the real state via `lifecycle::is_running`
// then hand it in. Each chain declares into the shared job it is handed, and
// names the nodes it depends on — so there is nothing to rebase.

/// One agent's **stop** subgraph. `SetWanted(Off)` head + `Reconcile` tail
/// always; the graceful `Signal → Drain` quiesce only when the agent is
/// actually running (nothing to drain on a down container). The `Reconcile`
/// stays even for a down agent so a race-up between the state read and exec
/// is still stopped in-DAG.
fn stop_chain(builder: &JobBuilder, agent: &str, graceful: bool, running: bool) {
    // `SetWanted` is the group root and owns the agent lease; the mechanical
    // steps are its children (borrow the lease, run once it reaches `Finishing`,
    // dep-ordered among themselves).
    let a = || agent.to_owned();
    let wanted = builder
        .node(NodeKind::SetWanted {
            agent: a(),
            up: false,
        })
        .needs(Resource::Agent(a()));
    // Declaration order is dependency order: the quiesce steps come first so
    // the `Reconcile` that waits on them can name them.
    if graceful && running {
        let signal = builder
            .node(NodeKind::Signal { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(wanted);
        let drain = builder
            .node(NodeKind::Drain { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(wanted)
            .after_ok(signal);
        let _ = builder
            .node(NodeKind::Reconcile { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(wanted)
            .after_ok(drain);
    } else {
        let _ = builder
            .node(NodeKind::Reconcile { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(wanted);
    }
}

/// One agent's **start** subgraph. `SetWanted(Up)` head; a down + stale-rev
/// agent gets the rebuild subgraph (its tail `Reconcile` starts it on
/// current derivations), otherwise a plain `Reconcile` (which starts a down
/// agent and noops an already-running one).
fn start_chain(builder: &JobBuilder, agent: &str, running: bool, stale: bool) {
    let wanted = builder
        .node(NodeKind::SetWanted {
            agent: agent.to_owned(),
            up: true,
        })
        .needs(Resource::Agent(agent.to_owned()));
    if !running && stale {
        // Rebuild subtree chained behind the `SetWanted` head. `MetaSync`,
        // `Prebuild` + `Reconcile` are their own group roots (top-level, per
        // `rebuild_nodes`).
        rebuild_nodes(builder, agent, true, Some(wanted));
    } else {
        let _ = builder
            .node(NodeKind::Reconcile {
                agent: agent.to_owned(),
            })
            .needs(Resource::Agent(agent.to_owned()))
            .part_of(wanted);
    }
}

/// One agent's **restart** subgraph. Restart NEVER rewrites `wanted`
/// intent (no `SetWanted` head, unlike stop/start): it bounces the
/// container and lets the tail `Reconcile` converge to the agent's
/// EXISTING intent, so a deliberately-stopped (`wanted = Off`) agent is
/// not forced back up by a hive-wide restart. A running agent gets the
/// mechanical stop (`Signal → Drain` when graceful, then `StopForUpdate`)
/// before `Reconcile`; a down agent gets just `Reconcile`, which
/// converges to intent — a stopped (`wanted = Off`) agent stays stopped,
/// a crashed (`wanted = Up`) agent comes back up.
fn restart_chain(builder: &JobBuilder, agent: &str, graceful: bool, running: bool) {
    let a = || agent.to_owned();
    if !running {
        // Nothing to bounce — a lone Reconcile converges to intent.
        let _ = builder
            .node(NodeKind::Reconcile { agent: a() })
            .needs(Resource::Agent(a()));
        return;
    }
    // Running: mechanical stop then Reconcile. The first stop node is the group
    // ROOT (no SetWanted head) and owns the agent lease; the rest are its
    // children (borrow the lease, dep-ordered), so the bounce holds one
    // continuous lease and `Reconcile` cancel-cascades if a stop step fails.
    //
    // `Reconcile` gates on the last mechanical step. For a non-graceful bounce
    // that step *is* the root, and the parent gate already orders it — a child
    // must NOT dep on its own parent (dep-scope), so it takes no sibling edge.
    if graceful {
        let signal = builder
            .node(NodeKind::Signal { agent: a() })
            .needs(Resource::Agent(a()));
        let drain = builder
            .node(NodeKind::Drain { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(signal);
        let stop = builder
            .node(NodeKind::StopForUpdate { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(signal)
            .after_ok(drain);
        let _ = builder
            .node(NodeKind::Reconcile { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(signal)
            .after_ok(stop);
    } else {
        let stop = builder
            .node(NodeKind::StopForUpdate { agent: a() })
            .needs(Resource::Agent(a()));
        let _ = builder
            .node(NodeKind::Reconcile { agent: a() })
            .needs(Resource::Agent(a()))
            .part_of(stop);
    }
}

// The `*_nodes` declarers below are the PURE core the async `*_many` fns call
// after reading live state — they take the per-agent running (and stale)
// flags explicitly, so unit tests exercise the online/offline shapes without
// a live container. `*_many` = gather state + declare + submit.
//
// A power op has no tail node: its effect is its nodes (`SetWanted` +
// `Reconcile`), with nothing left to do once they settle.
//
// There is no concatenation step either: every chain declares into the same
// builder and each keeps its own root, so the per-agent subgraphs are
// independent and run concurrently, each on its own lease. Rebasing one
// subgraph's indices onto another's used to be a function.

/// Declare the stop DAG from explicit `(agent, running)` targets.
pub(crate) fn stop_nodes(builder: &JobBuilder, targets: &[(String, bool)], graceful: bool) {
    for (agent, running) in targets {
        stop_chain(builder, agent, graceful, *running);
    }
}

/// Assemble the start DAG from explicit `(agent, running, stale)` targets.
///
/// No DAG-level pill: each agent's dashboard label is derived from the node
/// running under its lease, so a down+stale agent that grew a rebuild subgraph
/// reports `rebuilding` during its swap and `starting` at its reconcile,
/// without the DAG having to guess one label covering every target.
pub(crate) fn start_nodes(builder: &JobBuilder, targets: &[(String, bool, bool)]) {
    for (agent, running, stale) in targets {
        start_chain(builder, agent, *running, *stale);
    }
}

/// Declare the restart DAG from explicit `(agent, running)` targets.
pub(crate) fn restart_nodes(builder: &JobBuilder, targets: &[(String, bool)], graceful: bool) {
    for (agent, running) in targets {
        restart_chain(builder, agent, graceful, *running);
    }
}

/// Restart a single agent. Thin wrapper over [`restart_many`].
pub async fn restart(coord: &Arc<Coordinator>, agent: &str, source: Source, reason: String) -> u64 {
    restart_many(coord, &[agent.to_owned()], false, source, reason).await
}

/// Graceful restart of a single agent (signal → drain → stop → reconcile,
/// when running). Thin wrapper over [`restart_many`] with `graceful = true`.
pub async fn graceful_restart(
    coord: &Arc<Coordinator>,
    agent: &str,
    source: Source,
    reason: String,
) -> u64 {
    restart_many(coord, &[agent.to_owned()], true, source, reason).await
}

/// Restart `agents` (one or many) in a **single** DAG — one per-agent
/// subgraph each, built dynamically from live running state and run
/// concurrently on their own leases. A running agent gets the stop→reconcile
/// chain (`graceful` prepends signal→drain); a down agent gets just a lone
/// `Reconcile` (nothing to stop). Restart never writes `wanted`, so the
/// tail `Reconcile` converges each agent to its EXISTING intent — a
/// deliberately-stopped agent stays down. The whole hive-wide
/// `hivectl restart` is one DAG.
pub async fn restart_many(
    coord: &Arc<Coordinator>,
    agents: &[String],
    graceful: bool,
    source: Source,
    reason: String,
) -> u64 {
    let mut targets = Vec::with_capacity(agents.len());
    for agent in agents {
        targets.push((agent.clone(), lifecycle::is_running(agent).await));
    }
    submit_and_emit(coord, source, reason, |builder| {
        restart_nodes(builder, &targets, graceful);
    })
}

/// Start a single agent. Thin wrapper over [`start_many`].
pub async fn start(coord: &Arc<Coordinator>, agent: &str, source: Source, reason: String) -> u64 {
    start_many(coord, &[agent.to_owned()], source, reason).await
}

/// Start `agents` (one or many) in a **single** DAG — one per-agent subgraph
/// each, built dynamically from live state and run concurrently on their own
/// leases. A down agent gets `SetWanted(Up) → Reconcile` (or, rev stale, a
/// rebuild-then-start so it comes up on current derivations); an already-
/// running agent gets `SetWanted(Up) → Reconcile` (the reconcile noops). The
/// whole hive-wide `hivectl start` is one DAG.
pub async fn start_many(
    coord: &Arc<Coordinator>,
    agents: &[String],
    source: Source,
    reason: String,
) -> u64 {
    let current = crate::auto_update::current_flake_rev(&coord.hyperhive_flake);
    let mut targets = Vec::with_capacity(agents.len());
    for agent in agents {
        let running = lifecycle::is_running(agent).await;
        let stored = std::fs::read_to_string(crate::paths::applied_rev_marker(agent)).ok();
        let stale = current
            .as_ref()
            .is_some_and(|rev| stored.as_deref() != Some(rev.as_str()));
        if !running && stale {
            tracing::info!(%agent, "start: rev stale + agent down — rebuild-then-start");
        }
        targets.push((agent.clone(), running, stale));
    }
    submit_and_emit(coord, source, reason, |builder| {
        start_nodes(builder, &targets);
    })
}

/// Hard stop a single agent. Thin wrapper over [`stop_many`].
pub async fn stop(coord: &Arc<Coordinator>, agent: &str, source: Source, reason: String) -> u64 {
    stop_many(coord, &[agent.to_owned()], false, source, reason).await
}

/// Graceful stop of a single agent (signal → drain → reconcile, when
/// running). Thin wrapper over [`stop_many`] with `graceful = true`.
pub async fn graceful_stop(
    coord: &Arc<Coordinator>,
    agent: &str,
    source: Source,
    reason: String,
) -> u64 {
    stop_many(coord, &[agent.to_owned()], true, source, reason).await
}

/// Stop `agents` (one or many) in a **single** DAG — one per-agent subgraph
/// each, built dynamically from live state and run concurrently on their own
/// leases. A running agent gets `SetWanted(Off) → [Signal → Drain →](graceful)
/// Reconcile`; a down agent gets just `SetWanted(Off) → Reconcile` (skips the
/// pointless quiesce, keeps the Reconcile as the race-up backstop). The whole
/// hive-wide `hivectl stop` is one DAG.
pub async fn stop_many(
    coord: &Arc<Coordinator>,
    agents: &[String],
    graceful: bool,
    source: Source,
    reason: String,
) -> u64 {
    let mut targets = Vec::with_capacity(agents.len());
    for agent in agents {
        targets.push((agent.clone(), lifecycle::is_running(agent).await));
    }
    submit_and_emit(coord, source, reason, |builder| {
        stop_nodes(builder, &targets, graceful);
    })
}

/// Perm change: commit the JSON file(s) then rebuild.
pub fn perm_change(
    coord: &Arc<Coordinator>,
    agent: &str,
    source: Source,
    reason: String,
    payload: super::PermPayload,
) -> u64 {
    submit_and_emit(coord, source, reason, |builder| {
        templates::perm_change(builder, agent, payload);
    })
}

/// Meta-input lock bump; cascade rebuilds fan out on completion.
pub fn meta_update(
    coord: &Arc<Coordinator>,
    inputs: Vec<String>,
    source: Source,
    reason: String,
) -> u64 {
    submit_and_emit(coord, source, reason, |builder| {
        templates::meta_update(builder, inputs, None);
    })
}

/// Topology move(s) as a queue DAG. `moves` is `(child, new_parent)` pairs —
/// one entry for `set-parent`, N for `set-parent-bulk`. Fire-and-forget like
/// everything else in this module: submits and returns a DAG id
/// immediately, the caller learns the outcome async (dashboard job view /
/// `hivectl`'s `QueueDag` poll). Wired from `server.rs`'s `HostRequest::
/// SetParent` (hivectl) and `dashboard/topology.rs`'s `set-parent`/
/// `set-parent-bulk` handlers.
pub fn reparent(
    coord: &Arc<Coordinator>,
    moves: Vec<(hive_types::Ident, Option<hive_types::Ident>)>,
    source: Source,
    reason: String,
) -> u64 {
    submit_and_emit(coord, source, reason, |builder| {
        templates::reparent(builder, moves);
    })
}