mara on #563: 'we fixed the agent to not start turns in that
state (it fell back to online before), but this does not show on
dashboard properly'.
Root cause: the per-agent harness flips LoginState::NeedsLogin in
memory on three entry paths (cold-boot without a session, 401
mid-turn, /api/logout) and parks in wait_for_login. But
wait_for_login itself never called bus.emit_status('needs_login_idle')
at entry — only the /api/logout handler does that today. So:
- Cold boot: agent has no session, harness shows 'needs login'
on its own web UI (via LoginState mutex), but the dashboard's
needs_login field stays false because the
{state_dir}/hyperhive-needs-login sentinel was never written.
- 401 mid-turn: same — the 'after a turn failed' path in
hive-ag3nt.rs / hive-m1nd.rs flips LoginState directly without
emitting status, then calls wait_for_login, which now waits
silently with no sentinel write.
Fix: hoist the emit_status('needs_login_idle') call into
wait_for_login itself. All three entry paths get the sentinel
write for free; the /api/logout handler's explicit call (web_ui.rs
line 966) becomes redundant but idempotent — no behaviour change
there. The 'online' clear at session refresh stays exactly where
it was at the loop's exit.
Both hive-ag3nt and hive-m1nd binaries share wait_for_login, so the
manager harness benefits without a separate change.
Cuts every `include_bytes!`/`include_str!` of a non-rust path in
the workspace over to runtime file loads from `$HIVE_ASSETS_DIR`
(the `hyperhive-assets` derivation introduced in the previous
commit). After this commit the rust derivation has no compile-time
dependency on `branding/*` or `hive-ag3nt/prompts/*` anymore.
Call-site flips:
- `hive-c0re/src/forge.rs::CORE_AVATAR_PNG` /
`CONFIG_ORG_AVATAR_PNG`: were `include_bytes!` of
`branding/hyperhive.png` and `$OUT_DIR/agent-configs.png`. Now
`ensure_core_avatar` / `ensure_config_org_avatar` `tokio::fs::read`
via `hive_sh4re::assets::{core_avatar_png, config_org_avatar_png}`
at startup. The `agent-configs.png` is now rendered by the
`hyperhive-assets` derivation's rsvg-convert step (was
`hive-c0re/build.rs` + librsvg on the rust derivation's
nativeBuildInputs — both gone in the next commit).
- `hive-ag3nt/src/prompt.rs::TEMPLATE`: `render` now takes the
template as an argument; `write_system_prompt` reads it once from
`$HIVE_ASSETS_DIR/prompts/system.md` before calling render. The
test module still `include_str!`s the production template so
`cargo test --workspace` doesn't need `HIVE_ASSETS_DIR` set —
this is the only remaining compile-time reference to the file
from the rust workspace, gated to `#[cfg(test)]`.
- `hive-ag3nt/src/turn.rs::CLAUDE_SETTINGS`: was `include_str!`'d
and written via `tokio::fs::write`; now `tokio::fs::copy` from
`$HIVE_ASSETS_DIR/prompts/claude-settings.json` into the
per-agent socket dir.
- `hive-ag3nt/src/web_ui.rs::DEFAULT_ICON`: was `include_str!`'d;
now read on-demand from `$HIVE_ASSETS_DIR/branding/hyperhive.svg`
inside `serve_icon`. Falls back to an empty body if missing so
the endpoint never panics on a misconfigured container (matches
the existing "per-agent icon.svg override" fallthrough).
`HIVE_ASSETS_DIR` wiring:
- Inside containers: `nix/templates/harness-base.nix`
`environment.variables` sets it to
`${pkgs.hyperhive-assets}/share/hyperhive` (resolved through
the default overlay applied in `mkContainer`). Verified by
building `agent-base-toplevel` and grepping the resulting
`/etc/set-environment`.
- Host-side: `nix/modules/hive-c0re.nix` adds an `assets` option
defaulting to `hyperhive.packages.${system}.assets`, threaded
in from the flake's nixosModules wiring, and sets the same env
var on the `hive-c0re` systemd unit so the daemon's
`forge::ensure_*_avatar` startup hooks find the PNGs.
`hive-c0re/build.rs` deleted entirely; `[package].build` removed
from `hive-c0re/Cargo.toml`; rsvg-convert dependency lives in the
assets derivation only.
Validated: `nix build .#default .#checks.x86_64-linux.clippy
.#agent-base-toplevel .#manager-toplevel --fallback` all succeed.
`/etc/set-environment` in the toplevel shows
`HIVE_ASSETS_DIR="/nix/store/.../hyperhive-assets-0.1.0/share/hyperhive"`.
The naersk → crane swap in the parent commit flips clippy from
silently passing to actually failing on `-D warnings` (naersk's
`mode = "clippy"` mangled the `--` separator so the deny never took
effect). This commit clears the surfaced lints so the workspace
builds clean under the new enforcement — every fix is mechanical and
preserves behaviour. Tests still pass (160 across the workspace).
Auto-fixes via `cargo clippy --fix`:
- `doc_markdown` (19 sites): bare identifiers in doc comments
wrapped in backticks
- `format_in_format_args`, `explicit_into_iter_loop`,
`redundant_closure_for_method_calls`, `useless_conversion`, and
a few more — mechanical rewrites of the kind cargo can apply
safely.
Hand-fixed:
- `match_same_arms` (forge_notify::is_atx_heading): two arms returning
`true` collapsed into a single `matches!` pattern.
- `cast_sign_loss` + `format_push_string` (mcp.rs status formatter):
guarded `i64 → u64` through `u64::try_from(…).unwrap_or(0)` (status
timestamps are always positive in practice; clamp the skew edge to
0) and swapped `out.push_str(&format!(…))` for `write!` into the
buffer with an infallible-writer `let _ =`.
- `doc_lazy_continuation` in turn.rs + manager_server.rs + sh4re/lib.rs:
doc paragraphs that the markdown parser was treating as list-item
continuations got either a separating blank line or a `/`-for-`+`
word swap so the parser stops seeing a list.
- `unused_async` (manager_server::handle_request_schedule_prompt):
function has no `.await`; dropped the `async` and its `.await` call
site.
- `needless_pass_by_value` (scheduled_prompts::submit): take
`&NewSchedule` instead of moving the struct in; updated two prod
callers and eight test sites to pass references.
- `type_complexity` (approvals::mark_cancelled): hoisted the
7-tuple SELECT row shape into a `type CancelLookupRow = (…);` alias.
Allow-with-reason for intentional patterns:
- `option_option` (6 sites across dashboard / scheduled_prompts /
manager_server): `Option<Option<T>>` carries three-state PATCH
semantics (missing key = leave alone, `Some(None)` = clear,
`Some(Some(v))` = set). Collapsing to `Option<T>` loses the
"clear" state.
- `dead_code` (rebuild_queue::QueueKind::Destroy /
QueueSource::CrashRecover; topology::parent_of / default_seed):
wire-shape variants + API surfaces kept for the upcoming features
(#361 follow-ups, future `Destroy` queue routing, crash-recovery
path). Allowed at the variant / function level with the rationale
in `reason = "…"`.
- `too_many_lines` on three specific call-sites: a 117-line
exhaustive-variant test (dashboard_events::kind_tag_matches_…),
the meta-flake string template renderer
(meta::render_flake_with_lookup), and the notification poll loop
(forge_notify::poll_once) — splitting any of them would just hide
the contiguous shape they exist to keep visible.
`nix flake check` formatting target is still broken on main itself
(pre-existing nixfmt drift across ~28 files unrelated to this PR);
left alone here so the scope stays "crane port + lints the port
exposed" and the operator's review doesn't have to triage drive-by
nixfmt churn.
the stream-json result event carries modelUsage.<model>.contextWindow
which is the actual per-inference active window the model enforces.
for claude-sonnet-4-6 this is 200k even though the full prompt cache
can hold millions of tokens via accumulated cache reads.
with the nix-configured sonnet = 1000000 the proactive compact watermark
sat at 750k and was never reached. agents grew context until prompt_too_long
at ~170k — reactive compact, no checkpoint turn.
changes:
- bus gains api_context_window field seeded from modelUsage.*.contextWindow
in each turn's result event. authoritative; falls back to env var, then 200k.
- new effective_context_window(bus) helper used by both watermark functions
- compact_watermark (75%) and auto_reset_watermark (50%) call effective_context_window
- context_tokens() docstring clarified: all three token fields (input +
cache_read + cache_creation) count against the per-inference contextWindow
limit. the large cache_read values seen in the result event are cumulative
across all inferences in a turn, not per-inference.
- /api/state context_window_tokens now reflects the calibrated window
closes#129
raw line.contains() on stdout false-positives when the agent's own
text mentions rate_limit_error or similar strings. stderr keeps the
raw match (it is always claude CLI output, not conversation content);
stdout now only fires on parsed {"type":"error"} events.
Closes#69
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.
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.
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.
claude-settings.json now sets permissions.defaultMode=bypassPermissions
with a small deny list (WebFetch, WebSearch, Task, TodoWrite). The
per-flavor allow list and --tools / --allowedTools CLI flags are gone
— anything not denied auto-approves. mcp.rs loses ALLOWED_BUILTIN_TOOLS,
builtin_tools_arg, allow_list, allowed_mcp_tools. The extraMcpServers
allowedTools field is parsed for back-compat but no longer wired
anywhere; restrict via permissions.deny instead.
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 TurnFiles bundle (mcp_config + settings + system_prompt +
flavor) materialised once per harness boot, passed to drive_turn
and compact_session alike. operator-initiated /compact now uses
the exact same session shape as a normal turn — same MCP
surface, same allowed tools, same role prompt — only the stdin
payload differs (/compact vs the wake-up body). web_ui's
AppState carries the TurnFiles instead of (label + socket +
flavor + ad-hoc file writes per click). bin/hive-ag3nt and
bin/hive-m1nd prepare TurnFiles before spawning the web UI and
pass them to both surfaces. web_ui::Flavor folds into a type
alias for mcp::Flavor — no two-stage enum mapping.
removes ClaudeMode + the run_claude variant fork (system prompt
was Option, mcp args were skipped on Compact). dead 'mode'
plumbing gone.
every claude invocation now runs with current_dir set to
/state — relative paths in tool calls (Read notes.md, Bash ls,
Write blob) land in the agent's durable bind-mounted dir
instead of the harness's systemd cwd. /state is RW + survives
destroy/recreate so this matches where the agent is told to
keep notes anyway. dev/test setups without the bind silently
fall back to the parent cwd.
bus carries a one-shot AtomicBool armed by POST
/api/new-session (or the /new-session slash command). next
turn drops --continue, starting a fresh claude session; the
flag clears automatically so subsequent turns resume normal
behavior. /compact still always uses --continue — compacting
a non-existent session is a no-op anyway.
per-agent page grows an ↻ new session button next to the
cancel-turn one (always visible, amber, confirms before
posting since dropping --continue context isn't reversible).
slash-command surface picks up /new-session for parity with
the button. note row emitted on the live feed both at arm-
time and again when the turn actually consumes the flag, so
the operator can confirm it landed.
- runtime model override: Bus::{model,set_model} + POST /api/model
(form-encoded {model: name}). turn.rs reads bus.model() per turn
so a flip lands on the next claude invocation. /api/state grows
a model field; agent page shows a 'model · <name>' chip in the
state row. '/model <name>' slash command POSTs to the endpoint
and refreshes state.
- port regression fix: agent_web_port no longer probes forward for
*existing* agents (the previous fix shifted ports for any agent
without a port file, including legacy ones whose container was
already bound to the bare hashed port — dashboard rendered the
new port, container was still on the old one, conn errors). new
rule: port file exists → use it; absent + applied flake present
→ legacy, persist port_hash without probing; absent + no applied
flake → fresh spawn, probe forward.
- SO_REUSEADDR on both the dashboard and per-agent web UI binds
via tokio::net::TcpSocket. operator hit 12 retries failing on
manager :8000 — REUSEADDR handles the TIME_WAIT case cleanly
without a new dep; retry still covers the genuine
process-still-alive overlap.
todo: drops the model-override entry (shipped); adds two new
items — model persistence (optional, future), and custom
per-agent MCP tools (groundwork for moving bitburner-agent into
hyperhive).
