hive-c0re copies the hive's store read token to /run/hyperhive/forge-hive-token
for nix's git credential helper, and no longer mints, reads or falls back to
the core admin token. drops the core user sweep, avatar upload, webhook
cleanup and hive-priv's RunForgeAdmin.
SelectField and the refresh-interval picker are built on a new Listbox, a
select-only combobox: a button that keeps focus and a listbox in the top
layer, so it works the same in every current browser, is not clipped by a
scrolling ancestor and shows above a modal dialog. Keys, type-ahead and
placement are pure functions with tests. required becomes aria-required, as a
button has no native form validation.
Nothing reads core/meta: builds come from internal/meta. Remove push_meta, ensure_meta_remote, meta_read_access, ensure_repo, the per-agent forge sync that only served them, and the git commit path in meta.rs; remove_permission_files now just deletes the seed files.
Replaces the broadcast button in the messages card's header: a Disclosure
titled broadcast, folded by default, over the messages card. The composer is
built on the first open and kept, so it polls the agent list only once used.
The broadcast page wording in broadcast-types.ts and its test now says
composer.
The messages page gets a broadcast button that opens the agent picker and
message box above the stream, so a sent message shows up in the list below
it. The broadcast route and nav tab go, with no redirect: the page is now the
composer, BroadcastComposer, in the messages page.
A placing state is accepted only for an agent declared or queued on the hive; anything else gets 404 telling the caller to create it at swarm level. swarm-ui's wanted badge no longer offers a first declaration for an undeclared row.
Comments, docs and the hive-c0re README no longer say the hive renders, locks or pushes a meta flake. core/meta is the legacy permission-seed repo that meta::remove_permission_files still commits to; builds come from internal/meta (meta_rev.rs). Docs note that a start of a stale stopped agent can fail until the swarm's update pins the config.
queue_meta_update adds a TriggerDeploy behind each retry UpdateMeta (after_ok) for every agent whose latest deploy waited on a failed update. A created or merged agent whose deploy was skipped is deployed by the run that succeeds; a deploy skipped for another reason, or an agent already deployed, gets nothing.
A hive builds only from internal/meta, so a TriggerDeploy after a failed UpdateMeta could only fail its build. Both DAGs that queue an update (merged config PR, agent create) now gate the deploy with after_ok: the update shows Failed and the deploy Skipped.
The hive no longer renders, locks or commits a meta flake. A rebuild's
MetaSync node picks the internal/meta commit and records it with the
hyperhive and config pins it holds; Prebuild, Swap and Create build from it,
and a head that does not pin the agent's current config fails the build
naming the agent. hive-priv's meta_rev is required.
Removed: meta.rs sync/lock/render, the manual-rebuild relock, the boot
sweep's hyperhive bump (MetaLock), the update-meta-inputs form and its
panel, MetaWindow, the hive's hyperhive_rev setting.
Kept: agent_set::refresh holds what sync_agents also did (topology.json, the
gateway's agent-socket map and nginx include), called from spawn, destroy,
rebuild and the migration.
The boot reconcile and the stale-start upgrade compare the agent's marker
with the hyperhive pin of internal/meta; the dashboard's deployed chip reads
the config marker a successful build writes.
The swarm UI's user menu moves to @hive/shared and takes the whoami URL and
authelia's URL as props; the swarm UI keeps a thin wrapper, and the hive
dashboard and the per-agent page mount the same component. The dashboard's
nginx vhost proxies /api/whoami to authelia (the agent page is served from
the same vhost), and hive-c0re's state carries authelia_public_url, set from
services.hyperhive.swarm.authelia.url. The agent page shows the menu only
behind the gateway.
The small preview in the detail panel keeps only the connection badge and
expand, and no longer follows the agent's state stream; the full terminal
tab keeps all its badges. The table gets a state row with the turn state, a
paused badge, and the context and billed tokens (not the model, which the
inference row shows). The preview's top padding drops to one badge row, and
the showHeaderBadges prop goes since the full-tab mount is now the only one
with badges.
The schema's event fields gain values, a values_from marker (agents, repos)
and a closed flag; a node field gains the marker and the flag. kind is now
listed as issue and PR, what the router sends. The flow editor shows a
dropdown where the list is complete, suggestions where it is only examples,
and a plain box where there is none.
The editor's preview button posts the draft to POST /api/notify/dry-run and
puts each box's hit count on the canvas; any edit hides the numbers. Boxes of
kind request_reviewer label themselves "reviewer", so the long kind name no
longer runs into the hit count shown beside it.
Removes the HiveCopy hand-off: hive-c0re no longer writes each agent's
entry from swarm-agent-permissions.json into its runtime dir as
swarm-permissions.json, and the harness and subagent daemon no longer
read it. hive-c0re keeps the file as its own last read of the swarm for
the manage_root_agent mounts. Also removes permissions_import::pending(),
its only caller being the hand-off. No nix provisioned the copy, so no
nix option is removed and no tombstone is needed.
Rule after this change: until an agent's first successful read of its
key it holds no permissions (no execution, so no bash), and a subagent it
starts holds the same. After a read a set key wins, [] holds none and an
unset key is the default. A failed read keeps the last value, so an
unreachable queue at start keeps the agent closed until a read succeeds.
The harness logs once at warn that it starts closed.
Fixes#5315
/notifications/new could be a saved flow's own page, so the draft now opens at
/notifications/draft/new, which has a segment more than any flow's route. The
placeholder id no longer needs reserving. The .notify-node-hits rule styled the
removed tree rows; nothing renders it.
The flows list draws each flow as the editor's graph, read-only and with the
dry run's hit counts in the boxes, in place of the indented tree. A box's line
of text is cut by the browser at the box's edge, so it no longer runs out of it.
create flow (a primary button beside the refresh interval) and duplicate open
the editor on an unsaved draft at /notifications/new, with ?from= naming the
flow to copy; the flow is made by save, under an id taken from its name. Edit
and duplicate are buttons in the flow's header row. The editor's head and
inspector are laid out as sections.
The refusal comment still described collapsed sections the dialog no longer
has. The form's submit now returns while a create is in flight. The docs say
create another keeps only the hive.
The create dialog was one tall form that scrolled as a whole, so its header
and button moved out of view. It is now steps you go through in order:
identity, instructions (AGENTS.md in a box of its own), permissions,
resources and inference, then a summary of what will be sent and what the
agent creation queues, where create is. The header, step tabs and footer stay
put and only the step's own fields scroll. A refused AGENTS.md text sends the
dialog back to its step. The request is unchanged.
The detail panel had editors for inference, instructions, resource caps and
accounts, but nothing for the permissions the swarm grants, which could only
be set at create time or with swarmctl. A permissions row shows the grant
(the default set, none, or the list) and an edit dialog ticks the catalogue
from GET /api/permissions and saves with PUT /api/agents/{agent}/permissions.
Ticking exactly the default set sends the DELETE, as the create form sends no
list for it.
The controller counts review requests, verdicts and per-check results and times each run; the classifier service counts retried asks. A provisioned Grafana dashboard charts them.
on deploy, deploy.hive-controller.queue.agentNatsUrl must be one
scheme://host:port; an IPv6 literal or a comma list now fails at evaluation
instead of failing every container start.
