vinzenz/hyperhive
vinzenz/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Activity
hyperhive/hive-bash-mcp
History
Repository files (latest commit first)
Filename Latest commit message Latest commit date
atlas b31db6c7d9 subagent, bash: run each subagent run and bash task in its own cgroup 
hive-subagent-daemon and hive-bash-daemon are now delegated their cgroup
(Delegate=memory pids, DelegateSubgroup=daemon). At start each daemon
enables memory+pids below its unit and creates runs/ (subagent) or
tasks/ (bash). Each subagent run gets runs/<name>-<n>/ and each bash
task tasks/<id>/, both with memory.oom.group=1, and the daemon removes
the directory when the run ends. An OOM kill now takes one whole run
instead of one process out of the daemon's shared cgroup, and never the
daemon itself.

Subagent runs get oom_score_adj +1000; bash tasks and both daemons stay
at 0; the harness stays at -500. The unit-level OOMScoreAdjust=500 on
the subagent daemon goes: it was both the daemon's own score and a floor
no run below it could be lowered past.

Bash tasks enter their cgroup from the existing pre_exec hook, writing
to a cgroup.procs fd opened by the parent. hive-claude builds its
Command itself and exposes neither a pre_exec hook nor the child's pid,
so a subagent's claude (and an ACP agent's process) is started as the
daemon's own binary with the run's cgroup in its environment; main()
enters the cgroup, raises oom_score_adj and execs the real program, the
trampoline swarm-review-classifier already uses. Both enter before the
program executes, so nothing it forks starts outside the run's cgroup.

Without a delegated `daemon` leaf (tests, a dev shell) each daemon logs
one warning and runs everything in its own cgroup, as before.

The shared helper is hive-agent-sock's run_cgroup module, the crate
both daemons already use for what in-container plugins share.

Closes #5007
2026-10-13 12:25:27 +02:00
..
src subagent, bash: run each subagent run and bash task in its own cgroup 2026-10-13 12:25:27 +02:00
Cargo.toml hive-bash-mcp: sweep harness/bash-tasks/ of files older than 30d at startup 2026-09-11 19:25:33 +02:00
README.md docs(#2627): crate READMEs for the harness column (hive-agent, hive-agent-mcp, hive-agent-wake, hive-bash-mcp) 2026-07-23 16:41:05 +02:00

README.md

hive-bash-mcp

Per-agent background bash-task runner: a long-running daemon (hive-bash-daemon) plus the thin stdio MCP bridge (hive-bash-mcp) claude spawns each turn to talk to it. This is what backs the bash_run / bash_status tools agents use to kick off long-lived shell commands (builds, test suites, anything that shouldn't block a turn) and check on them later.

When to use it

Look here when changing how background bash tasks are spawned, tracked, or surfaced. The daemon owns all subprocess lifecycle (sh -c spawn, completion monitoring, task state files under harness/bash-tasks/) and pushes task-completion todos onto the harness's in-agent socket so they show up in get_loose_ends. The bridge binary is deliberately dumb: no subprocess management, just a unix-socket round-trip per tool call, so it cold-starts in milliseconds every turn.

Shape

Two bins from one shared lib (src/lib.rs):

  • hive-bash-daemon (src/main.rs) — the long-running daemon. runner.rs is the spawn/monitor loop and pushes todos on task transitions; socket.rs serves the daemon's own unix socket for tool-call requests from the bridge.
  • hive-bash-mcp (src/bin/mcp.rs) — the stdio MCP server claude spawns per turn. Forwards every tool call to the daemon over the socket via protocol.rs's DaemonRequest/DaemonResponse and returns the result.

Supporting modules: paths.rs (daemon socket + agent-socket resolution), stats.rs (the bash_commands favorite-tool stat recorded into turn-stats.sqlite).