Surface reminder activity statistics (scheduled, delivered, pending counts)
for each agent over configurable time windows. Needed by the per-agent
stats page to display reminder metrics.
Adds:
- ReminderStats struct and ReminderRollup request/response variants
- Broker::reminder_rollup_for(agent, since_secs) method
- Agent and manager socket handlers for the new RPC
- SocketReply mapping for response conversion
each turn_stats row already records the model; roll it up per bucket
so the /stats page can show which model ran when. model choice
greatly affects token cost, so the new stacked-bar chart sits right
under the cost chart for eyeball correlation across the window.
Snapshot gains a sorted `models` series list; each Bucket carries a
`model_counts` map.
new hive-ag3nt::stats module reads turn_stats.sqlite read-only and
aggregates over 24h/7d/30d windows (hourly/daily buckets) — turn
rate, p50/p95/avg duration, ctx tokens (avg/max), cost token
components, top tools, wake mix, result mix. served by the agent
itself so per-MCP extensions can register more providers without
the host knowing their schemas.
/stats route + /api/stats?window=... on the per-agent web ui.
chart.js v4.4.4 pulled from jsdelivr (SRI hash deferred). nav
links: 📊 chip on the dashboard container row + 📊 stats → on
the per-agent header.
todo housekeeping: softened damocles-area note at the top,
new reverse-proxy + deferred reminder-rollup items, removed
the two telemetry-ui items absorbed by this page.
the existing ctx badge was misnamed: it summed `result.usage`, which is
the cumulative tokens billed across every inference in the turn. for
tool-heavy turns that easily exceeds the model's context window (a 600k
cached prefix × 15 sub-calls = 9M cache_read), making it useless as a
"should i compact?" signal.
now two separate badges:
ctx · N last inference's prompt size = actual context window in
use right now. parsed from each `assistant` event's
`.message.usage`; the harness tracks the most recent one
across the stream and snapshots it when the `result`
event lands.
cost · M cumulative tokens billed across the whole turn (the
previous behaviour, now correctly labelled).
both update via a single `TokenUsageChanged { ctx, cost }` SSE event at
turn-end. turn_stats grows four columns (`last_input_tokens`,
`last_output_tokens`, `last_cache_read_input_tokens`,
`last_cache_creation_input_tokens`) so the cold-load seed can paint both
badges on page load. migrations run try-and-ignore ALTERs so existing
agent dbs catch up; pre-migration rows have last-inference zeros and
yield no `ctx` seed (badge stays empty until next turn) rather than a
misleading 0.
layout
- unified prefix-column for every row kind: padding-left + negative
text-indent so the glyph (→ ← · ◆ ✓ ✗ ⌁ !) sits in the same column
whether the row is flat or a <details>. wraps hang under the body,
not under the glyph.
- expandable rows drop the directional glyph from their summary text;
the ▸/▾ disclosure marker from CSS sits in the prefix column instead,
and the row's colour still carries cyan = outbound, muted = inbound.
- turn-start / turn-end de-weighted: bold/margin/tint dropped, the
coloured left rule alone marks the boundary.
note classification
- stderr lines render orange with a `!` glyph (was muted `·`)
- operator-initiated notes (cancel/compact/model/new-session) render
mauve italic (was muted `·` indistinguishable from harness chatter)
- catch-all .sys row escalates to orange `!` so unrecognised stream-json
shapes surface for follow-up instead of hiding in muted noise
message-bearing rows
- send / ask / answer tool_use rich renderers default-open with the
body inline; new ask + answer renderers (previously fell through to
the generic JSON dump). recv tool_result also default-open, keyed by
tracking tool_use_id → name across the stream so we know which
result came from which tool.
- assistant text rows render markdown.
- bodies use vendored marked v4.0.2 (hive-fr0nt::MARKED_JS); falls
back to plain text when the asset doesn't load.
extra-mcp tool pretty-print
- generic args formatter replaces the raw JSON dump for unknown tools
(single-string field → `name k: "v"`; single dict / multi-field →
trimmed `k: v · k: v …` summary)
dashboard .live .msgrow gets a text-indent: 0 reset so the new
hanging-indent metrics from TERMINAL_CSS don't leak into the flex-grid
broker rows.
new /api/open-threads endpoint on hive-ag3nt proxies the agent's
own GetOpenThreads RPC (manager flavour proxies the hive-wide
ManagerRequest::GetOpenThreads). same data the
mcp__hyperhive__get_open_threads tool sees from inside claude.
frontend renders a collapsible <details> section above the
terminal, listing each pending row (approval / question) with
asker → target, age, and free-form body. auto-expands on the
first appearance of any open thread; sticky after that.
refreshed on cold load + after every turn_end (turns are when
threads land or resolve).
new sqlite table at /state/hyperhive-turn-stats.sqlite on each
agent's state dir. one row per claude turn captures identity
(model, wake_from, result_kind), timing (started/ended_at,
duration_ms), cost (input/output/cache_read/cache_creation token
counts), behaviour (tool_call_count + per-tool breakdown JSON),
and post-turn snapshot metrics (open_threads_count,
open_reminders_count).
wire additions:
- AgentRequest/ManagerRequest::CountPendingReminders +
Broker::count_pending_reminders_for(agent)
- Bus::observe_stream + take_tool_calls — pumps the existing
stdout stream-json, picks out tool_use blocks, accumulates per
turn. bin loops fold the breakdown into each row.
- TurnStats::open_default + TurnStatRow + record() — best-effort
inserts; failures log + don't block the harness.
both ag3nt and m1nd bins capture started_at + duration via
Instant::elapsed, fetch open-thread + reminder counts from
hive-c0re via the existing socket (post-turn, best-effort), and
record one row at turn_end. record_kind splits ok / failed /
prompt_too_long; failures carry the error message in note.
todo entries for host-side vacuum sweep + reading the table back
into agent/dashboard badges.
new LiveEvent variants on the per-agent bus —
status_changed / model_changed / token_usage_changed /
turn_state_changed — replace the per-agent web UI's
/api/state polling for the badge row.
emit sites:
- Bus::set_model → model_changed
- Bus::record_usage → token_usage_changed
- Bus::set_state → turn_state_changed
- turn::wait_for_login → status_changed("online") on creds detect
- post_login_start / post_login_cancel → status_changed("needs_login_*")
per-agent endpoints (post_set_model / post_compact / post_new_session
/ post_cancel_turn / post_login_*) now all return 200; client
drops the post-submit refetch except on login transitions, which
still need /api/state to render the OAuth form + session stream.
client adds dispatch on the four new event kinds, threads
`currentLabel` through so the composer re-enables on a live
status flip, and no longer fires refreshState() from turn_end or
postModel — the events carry the same signal faster.
closes the per-agent half of the dashboard event-channel
refactor; TODO entry dropped.
new `hyperhive.claudeMarketplaces` option (list of strings — URL,
path, or github:owner/repo). harness boot adds each via
`claude plugin marketplace add` before updating + installing the
configured plugins, so specs like `foo@some-marketplace` resolve
on a fresh container. idempotent: 'already exists' stderr is
treated as success.
install_configured now takes an optional notify recipient. on a
non-zero or spawn-failed 'claude plugin install', sub-agents send
the spec + stderr to manager via the hyperhive socket; manager
passes None so it doesn't message itself. boot still proceeds either
way — notification is best-effort.
run_claude now keeps a 20-line stderr ring buffer and bails with it
inline (was just 'exit <status>'). agent serve loop, on Failed (not
PromptTooLong — that's already absorbed by drive_turn's compaction
retry), sends the error body to manager via the normal hyperhive
send. swallows transport errors — failure is already in journald
and the events sqlite. manager-only harness (hive-m1nd) is unchanged
so it doesn't try to notify itself.
claude-code rejects --dangerously-skip-permissions / defaultMode=
bypassPermissions when running as root, which all hyperhive
containers do. revert to the previous explicit allow-list plumbing
(per-flavor list spliced into permissions.allow + --tools enable
list), keep TodoWrite out of the built-in allow set, and keep the
deny list (TodoWrite, WebFetch, WebSearch, Task) as belt-and-braces
in case anything sneaks past the allow gate.
bus-only note made post-mortems require the web UI / events sqlite;
now stderr lines also land in 'journalctl -M <container> -b' alongside
the existing LiveEvent::Note for the dashboard.
socket client now retries connect/IO failures with 2-4-8-16-30s
backoffs (60s total budget). transparent for non-tool callers via
request(); tool handlers go through request_retried() which also
returns the retry count, then annotate_retries() appends a one-line
note to the tool result so claude knows the slow round-trip was a
c0re flicker, not a content failure — avoids burning tokens on an
LLM-level retry.
