Per mara's direction: both nixpkgs and nixpkgs-unstable are now
top-level meta flake inputs with explicit store-path URLs. Hyperhive
follows them rather than the other way around:
inputs.nixpkgs.url = "path:${pkgs.path}";
inputs.nixpkgs-unstable.url = "path:${nixpkgs-unstable}";
inputs.hyperhive.url = "...";
inputs.hyperhive.inputs.nixpkgs.follows = "nixpkgs";
inputs.hyperhive.inputs.nixpkgs-unstable.follows = "nixpkgs-unstable";
New NixOS host options (auto-set at build time, overridable):
services.hyperhive.c0re.nixpkgsFlake
default: "path:${pkgs.path}" — host's evaluated nixpkgs.
services.hyperhive.c0re.nixpkgsUnstableFlake
default: "path:${nixpkgs-unstable}" from hyperhive's flake.nix —
the channel that carries claude-code. Operators can override to
track a different unstable snapshot.
Legacy fallback (both args empty) preserved for backward compat.
Two new Rust tests cover the full-URL and fallback paths.
meta flake was using `nixpkgs.follows = "hyperhive/nixpkgs"` but
`hyperhive` is a store-path input, so nix resolves hyperhive's own
pinned lock rather than the host's follows-substituted version.
When an operator sets `inputs.hyperhive.inputs.nixpkgs.follows =
"nixpkgs"` in their host flake, the meta flake was silently ignoring
it and using hyperhive's pinned nixpkgs instead.
Fix: hive-c0re.nix injects `--nixpkgs-flake path:${pkgs.path}` into
the daemon's ExecStart. `pkgs` IS the host's nixpkgs when follows is
set; otherwise it's hyperhive's own pin — so the meta flake gets the
right nixpkgs in both cases. render_flake emits `nixpkgs.url = "..."`
(explicit) when nixpkgs_flake is non-empty, falling back to the old
`follows` form when empty for backward compat.
Add `hivectl gateway {create-user,delete-user,list-users}` subcommands for
managing htpasswd files used by gateway Basic auth. Pure Rust bcrypt
(cost 12, $2y$ prefix nginx accepts). No external htpasswd binary required.
Also fix the NixOS module assertion: `cfg.auth ? htpasswdFile` is always
true in the module system (declared options always exist as keys); switch
to `nullOr path; default = null` + `!= null` check so the assertion
actually fires with a useful error when enable=true but no file is set.
Guard bind-mount and nginx config against null to prevent eval errors.
Update docs/gateway.md to show hivectl commands instead of raw htpasswd.
Every other Coordinator opener takes the broker.sqlite path and
derives its parent internally; BuildLogs alone wants a directory.
Passing the file path tripped create_dir_all on an existing file
(File exists os error 17) and blocked c0re boot.
Backend half of #726. The 32-line stderr ring buffer in
`lifecycle::run` and `lifecycle::prebuild_toplevel` routinely
truncated the actual eval error — a single 'tried alternatives'
block out of a failing module ref is often 30+ lines on its own,
which pushed the real cause out of the bailout message. With this
patch the full stream lands in sqlite where the dashboard can
surface it; bail-outs now point at the build log id instead of an
arbitrary tail.
### New module: `hive-c0re::build_logs`
`BuildLogs::open(db_path)` creates a sqlite db at
`<db_path>/build_logs.sqlite`. Schema: id, agent, kind, cmdline,
started_at, finished_at, status, stdout, stderr — indexed for
both per-agent latest-N queries and the status-driven retention
sweep. API: `start / append_stdout / append_stderr / finish` for
the streaming writer side (best-effort — every append handles
sqlite errors via tracing::warn so a transient blip never tears
down a rebuild), plus `list_recent_for_agent / get_full` for the
read side (50-row cap clamped server-side).
### Process-singleton handle
`build_logs::install / global()` install the `Arc<BuildLogs>` at
`Coordinator::open` so `lifecycle::run` and
`lifecycle::prebuild_toplevel` can write without us threading
the handle through every `pub async fn` entry point in the
lifecycle surface — there are 10+ call sites and the handle is
the same Arc everywhere anyway. Reads via `global()` return None
in early-startup / standalone-test paths so callers no-op cleanly.
### Lifecycle integration
`run` derives the kind from `args[0]` (the nixos-container verb)
and the agent name from `args[1]` (stripped of the `h-` agent
prefix so dashboard grouping matches the bare agent name). It opens
a row before spawning, pipes stdout/stderr into both tracing AND
the row, then `finish`es with the terminal status. `prebuild_toplevel`
does the same with kind = "prebuild" and the agent name already
in scope from its caller. On failure both bail with "see build log
#<id>" instead of the ring-buffer tail.
### Retention
`spawn_vacuum` mirrors `stats_vacuum`/`events_vacuum` in shape —
hourly tick that calls `BuildLogs::vacuum()`. Rule: failures kept
30d (operators dig into them), successes 24h (mostly noise after
a day), in-flight rows never reaped regardless of age (running
builds shouldn't disappear from their own log viewer mid-stream).
### Out of scope (follow-ups)
- Dashboard endpoints (`GET /api/build-logs/{agent}`,
`GET /api/build-logs/{id}`) — wire layer
- ContainerView.build_logs field — agent-card chip data source
- Side-panel viewer + SSE `build_log_appended` event — UX
- Download-as-text link — operator workflow polish
These all stack cleanly on top of the data layer + writer this
PR ships. Filing as phase 2 PRs.
### Validation
- 5 new unit tests pass (start/append/finish flow, list ordering
+ clamp, get_full miss, vacuum per-status rule, post-finish
append fault tolerance)
- 157 hive-c0re lib tests pass overall
- cargo check workspace clean
Refs #726.
post_meta_update returns 200 immediately and runs the nix flake
update + agent-rebuild ripple in a background task, so the META
INPUTS panel looked idle for the whole multi-minute window (#259).
Track in-flight runs with a Coordinator atomic counter, exposed via
an RAII MetaUpdateGuard held across run_meta_update. Surface it as
the meta_update_running snapshot field plus a MetaUpdateRunning SSE
event (flipped only when the count crosses 0, so concurrent runs
flip the flag once). The panel shows a pulsing in-progress banner
and disables the update button while a run is active.
DashboardEvent::QuestionAdded gains question_refs and
QuestionResolved gains answer_refs — both populated via
scan_validated_paths at emit time, same helper the broker
forwarder uses for Sent/Delivered. cold-load snapshot wraps
each OpQuestion in QuestionView with the same fields computed
once per /api/state.
client threads refs through questionsState rows (pending +
history) and passes them to appendLinkified at every render
site (live pane, history details). path tokens in question and
answer bodies now linkify with the same server-vouched
guarantee broker messages already enjoyed.
questions pane now shows both operator-targeted threads
(target IS NULL) and agent-to-agent threads (target = some
agent). filter chips above the list: all / @operator / @peer /
per-participant. peer rows get a mauve left rule + a 0V3RR1D3
button that POSTs the same /answer-question endpoint
(OperatorQuestions::answer already permits the operator as
answerer on any target).
wire changes: OperatorQuestions gains pending_all +
recent_answered_all; QuestionAdded + QuestionResolved events
carry target: Option<String>; emit sites drop their
target.is_none() guard. answered-history rows show the
answerer prefix so override answers are auditable at a glance.
new DashboardEvent::ContainerStateChanged + ContainerRemoved
close the last refetch loop on the dashboard. Coordinator's
rescan_containers_and_emit diffs a fresh container_view::build_all
against a cached last_containers map and fires per-row events.
called from actions::approve (post-spawn), actions::destroy,
the lifecycle_action wrapper, auto_update::rebuild_agent, and
the existing 10s crash_watch poll.
ContainerView extracted to its own module so coordinator and
dashboard can both build it. dashboard endpoints flip to 200;
container-lifecycle forms carry data-no-refresh. client drops
the periodic poll entirely — initial cold load + SSE for
everything afterwards. pending overlay reads from the existing
transientsState since the new event payload doesn't carry it.
PURG3 + meta-update keep the post-submit refetch since
tombstones + meta_inputs aren't event-derived yet; tracked in
TODO.md.
manager has /agents bind-mounted too, so /agents/hm1nd/state
resolves there alongside the legacy /state. one canonical path in
the wake message instead of branching on MANAGER_NAME.
manager keeps /state (legacy mount); sub-agents see their state at
/agents/<name>/state. wake message hardcoded /state/ for everyone,
which is wrong for sub-agents post-refactor — they get a path they
can't ls. switch on MANAGER_NAME and format the right path.
bare set_transient/clear_transient pairs leak the in-memory transient
on task cancellation, panics, or any early return between the two
calls — dashboard then shows the agent stuck in 'rebuilding…'
forever (coder hit this today). add Coordinator::transient_guard
returning a TransientGuard whose Drop clears, and convert every
caller (dashboard lifecycle_action, auto_update::rebuild_agent,
manager_server Update, actions::destroy, actions Spawn task,
migrate phase 4). destroy() now takes &Arc<Coordinator> so it can
hold a guard. existing stuck transients clear on next hive-c0re
restart since transient state is in-memory only.
new NixOS module option services.hive-c0re.operatorPronouns
(free text, default 'she/her', example 'they/them'). hive-c0re
takes it as a CLI flag (--operator-pronouns, lib.escapeShellArg'd
in the systemd unit), stores it on Coordinator, threads it into
the meta flake's mkAgent so each agent's systemd service gets
HIVE_OPERATOR_PRONOUNS set. the harness reads the env at boot
and substitutes {operator_pronouns} into the agent / manager
system prompt alongside {label}. nix string is escaped against
backslash + double-quote so non-ascii / quoted values
round-trip safely. prompt addendum: both agent.md and
manager.md mention the operator's pronouns up front so claude
uses them naturally in third-person reference. propagates on
next ↻ R3BU1LD (meta lock bump, no per-agent approval).
new AgentRequest::AskOperator + AgentResponse::QuestionQueued on
the per-agent socket — same shape as the manager flavor, agent
gets the same wire surface (still uses the same operator_questions
table). agent_server::dispatch wires AskOperator through coord
.questions.submit(agent, ...) so the row's asker is the sub-agent
name; the ttl watchdog already in manager_server gets shared and
spawn_question_watchdog goes pub.
answer routing: operator_questions::answer now returns (question,
asker). post_answer_question + post_cancel_question + the watchdog
fire OperatorAnswered through new coord.notify_agent(asker, event)
instead of always notify_manager — the event lands in whichever
agent originally asked. notify_manager is now a thin wrapper.
agent socket plumbing: agent_server::start takes Arc<Coordinator>
instead of Arc<Broker> so dispatch has access to questions +
notify path; coordinator::{register_agent,ensure_runtime} take
self: &Arc<Self>. mcp::AgentServer grows the ask_operator tool;
allowed_mcp_tools(Agent) adds it; prompts/agent.md replaces the
'message the manager to ask the operator' guidance with the
direct tool description.
new Coordinator::kick_agent(name, reason) drops a system message
into the agent's inbox so the next turn picks it up with a 'you
were just (re)started, check /state/ for notes, --continue session
is intact' hint. wakes the turn loop without any harness-side
handling needed — it's just another inbox message with sender =
'system'.
wired from:
- dashboard /start /restart /rebuild handlers (via lifecycle_action's
on-success tail)
- manager mcp_hyperhive_start / restart
dashboard: pending approvals + tombstones + questions now refresh on
a 5s heartbeat when nothing else is happening. previously refresh
only fired on async-form submit or on broker traffic addressed to
operator — manager-queued approvals went through neither, so the
operator had to reload to see them. 5s is the slow-path; 2s
remains for in-flight transients.
new section between containers and questions: lists every name with a
state dir under /var/lib/hyperhive/agents/ that doesn't correspond to
a live container. shows state size + last-modified age + whether
claude creds are kept. two actions per row:
- R3V1V3 — queues a spawn approval with the same name (operator
approves to recreate; spawn flow reuses prior config + claude
creds, no re-login needed)
- PURG3 — wipes the agent's state + applied dirs (post /purge-tombstone/
endpoint; refuses if a live container with that name still exists)
dashboard also opens agent links in new tabs now (target=_blank +
rel=noopener) so the operator's overview tab stays put when they
dive into an agent.
backend:
- TransientKind grows Starting / Stopping / Restarting / Rebuilding /
Destroying alongside the existing Spawning. each dashboard handler
(start/restart/kill/rebuild/destroy) wraps the lifecycle call with
set_transient + clear_transient so the dashboard knows what's in
flight. transient kind is surfaced inline on ContainerView.pending
(existing-container actions) — only Spawning (pre-creation) lands
in the separate transients list.
frontend:
- container row is now two lines: identity + meta on top, action
buttons below. less cluttered, leaves room for the pending state
pill. pending rows dim their actions and surface a pulsing
'◐ spawning… / starting… / stopping… / restarting… / rebuilding…
/ destroying…' indicator next to the name.
- 'needs login' / 'needs update' chips moved into a unified .badge
styling for consistency.
- auto-refresh kicks in not only on transient spawn but on any
container with a pending action.
new mcp tool on the manager surface that queues a question on the
dashboard and returns the question id immediately. operator submits an
answer via /answer-question/<id>; the dashboard fires
HelperEvent::OperatorAnswered { id, question, answer } into the manager
inbox so the next turn picks it up.
also: fix async-form button stuck on spinner after successful submit
(refreshState skipped re-rendering, so the button was never re-enabled).
