hyperhive/TODO.md

TODO

Pick anything from here when relevant. Cross-cutting design notes live in CLAUDE.md; high-level project intro in README.md.

Security

• Unprivileged containers (userns mapping). Today the nspawn container runs as a fully privileged root. Goal: PrivateUsersChown=yes (or the nixos-container equivalent) so uid 0 inside maps to an unprivileged uid on the host, and a container-root compromise lands the attacker on an ordinary user account, not the host's root. Requires per-agent state dirs to be chown'd to that uid on the host side.
• Bash command allow-list. Replace the blanket Bash allow with a pattern allow-list (Bash(git *), Bash(nix build .*), etc.) per claude-code's --allowedTools extended grammar. Likely lives in agent.nix so each agent can scope its own shell surface.

Per-agent settings

• Model override. Hard-coded to haiku in the turn loop right now. Surface as a per-agent override: operator via dashboard, manager via request_apply_commit setting an attr on the agent's flake (most natural place since the flake already carries per-agent env/identity). Pair with a model status indicator on the agent page (active / queued / last switched) once the override is in place.

UI / UX

• Per-agent UI substance. Show last N inbox messages, last turn timing, link back to dashboard.
• State badge: compacting + napping states. Idle/thinking already ship (driven from SSE turn_start/turn_end). Add compacting 📦 and napping 😴 once the /compact trigger and nap tool exist — both need a harness signal (an explicit LiveEvent::StateChange variant or piggyback on Note).
• Server-side state badge. Today the badge is computed client-side from turn_start/turn_end events. On page reload mid-turn the history replay re-derives it, but with a compacting / napping state coming and a non-trivial state machine it's better to track authoritative state in the harness and expose it via GET /api/state (status: "thinking" | "idle" | "compacting" | "napping"). JS just renders. Drops the derive-from-events-and-pray code path.
• Terminal: inline diffs for Write/Edit. Today a Write / Edit tool-use row just shows the file path. Render the actual change inline in the terminal: for Edit, a small +/- per-line diff between input.old_string and input.new_string; for Write, the first few lines of input.content (it's all "+"). Keep collapsed by default (<details> like the existing tool_result rollups), expand to full diff on click. Color via the same .diff-add / .diff-del classes the dashboard approval diff already uses.
• Terminal: /model slash command. Operator-typeable model override from the terminal. Depends on the model-override work above; once an override mechanism exists, wire a /model <name> command that POSTs to a new endpoint.
• xterm.js terminal embedded per-agent, attached to a PTY exposed by the harness. Pairs well with the unprivileged-container work — would let the operator drop into the container without nixos-container root-login.

Telemetry

• Harness stats per agent in sqlite, charted on the agent page. bitburner-agent samples 18 series; for hyperhive the generally-applicable ones are:
  • turns/min, tool calls/turn, turn duration p50/p95
  • claude exit code distribution (ok vs --compact-retry vs failure)
  • inbox depth (current + max-over-window)
  • messages sent/received per turn (split by recipient: peer / operator / manager / system)
  • approval queue length (across all agents — dashboard-level)
  • per-tool usage counts (Read/Edit/Bash/send/recv/…)
  • time-since-last-turn (helps spot stuck agents)
  • notes file size growth (cues compaction) Backend: a stats table with (agent, ts, key, value) written from the harness on TurnEnd; GET /api/stats?since=… returns the series; agent page renders with a small chart lib (uPlot is light).

Manager → operator question channel

• TTL / cancel on ask_operator. Questions today block forever; the manager turn stays alive until the operator answers. Add a per-question ttl_seconds (or a dashboard "cancel" button that resolves the question with a sentinel answer) so a long-idle question can time out and let the manager fall back. Wire the timeout into OperatorQuestions::wait_answered and surface remaining-time on the dashboard.

Spawn flow

• Two-step spawn. Today request_spawn(name) is one shot: manager asks → operator approves → container is created with a default agent.nix and empty /state/. Manager has no way to pre-stage per-agent prompt material, package additions, or initial notes before the agent first wakes. Split into:
  1. request_spawn_draft(name) — host creates the per-agent proposed/ repo (initial commit) and state/ dir with no container; manager now has /agents/<name>/{config,state}/ to edit + commit just like an existing agent.
  2. request_spawn_commit(name, commit_ref) — submits the queued approval; operator sees the diff in the dashboard like a normal apply_commit; on approve the container is created from that commit. Backwards-compat: keep the existing one-shot request_spawn for trivial agents (operator can still type a name in the dashboard). Surface "drafts" as a new section between K3PT ST4T3 and approvals.

Loop substance

• nap tool. Agent-side MCP tool mcp__hyperhive__nap(seconds) that parks the turn loop for a short while before next-message processing. Use cases: agent decides it has nothing useful to do, or wants to throttle itself between rapid wake events. Implementation: harness records a "wake-not-before" timestamp; recv_blocking skips the long poll until that ts; the state badge reads napping · MM:SS during. Operator can cancel via the same /cancel slash command or a dashboard button.
• Notes compaction. /state/ is bind-mounted persistently and agents are told (in the system prompt) to keep /state/notes.md for durable knowledge — but we don't currently nudge them to compact when notes grow. Bitburner-agent's pattern: a short-lived secondary claude session that takes the existing notes + a "compact this" prompt and rewrites them in place. Add when the notes start bloating.

Lifecycle / reliability

• Container crash events. Watch container@*.service via D-Bus, push HelperEvent::ContainerCrash to the manager's inbox so the manager can react (restart, escalate, etc.).