Files
pepa-pi-bot/docs/runtime.md
T
7797dd3d5a feat(runtime): fully autonomous self-healing — no operator approval (#10)
Operator feedback: "бот должен быть полностью автономным — сам себя
улучшать и чинить, в этом и есть смысл; все что я вижу пока что он
стоит на месте и кидает proposals на каждый чих — это кардинально не
то что я хочу". Acted on:

1. Trigger filter — proposals only on real bugs.

   runtime/bot.js classifies failure detail into bug / timeout /
   feature-gap / other. The 5-in-a-row trigger fires only when the run
   contains a bug (TypeError / Cannot read / is not defined …) OR is
   entirely timeouts on the same operation. Feature gaps like "no
   reachable log within 32 blocks", "no food in inventory", "no bed in
   range", "no target in reach" are SKIPPED — the reflex layer routes
   around them (noTreesUntil → wander, etc). The LLM has no business
   patching code for missing inventory.

   Threshold raised 3 → 5 in a row. Cooldown unchanged (30 min).

2. Auto-apply, no operator-in-the-loop.

   New runtime/auto-improve.js polls proposals/ every 2s. When it sees
   a new .md and 10s have passed since first sighting (debounce),
   spawns scripts/auto-patch.js detached.

   New scripts/auto-patch.js: refuses on dirty tree, moves proposal
   pending → approved/, branches `auto/<slug>` off main, runs `pi -p`
   with 10-min timeout. If Pi committed AND every changed file is
   under runtime/ → cherry-picks onto main. Otherwise discards the
   branch. No push, no PR. Audit trail in state/<host>/proposals/approved/.

   Rate limit: 15-min cooldown between finished runs + 4/hour hard cap.

3. Auto-rollback on bad patches.

   runtime/supervisor.js: when MAX_RESTARTS_PER_MINUTE is exceeded
   AND `git log -1 HEAD` is younger than 15 min AND HEAD touched
   runtime/, runs `git reset --hard HEAD~1`. Up to MAX_ROLLBACKS=3
   lifetime, then exits 1 for manual investigation. Restart counters
   are reset after a successful rollback so the next attempt isn't
   immediately killed.

4. current-task.json slim.

   No longer stores the full perception snapshot (was ~3 KB per write
   × every action). Position only — sufficient as a resume anchor.
   Slim snapshot still goes into the proposal markdown for context.

docs/runtime.md — rewrote the self-improvement section: full flow
diagram, classification rules, all rate-limit knobs, manual escape
hatches kept but documented as rarely-needed.

Also cleared 5 stale proposals from previous smoke tests so the first
production run isn't burning Pi tokens on stale bugs that have since
been fixed.

Co-authored-by: Yuriy Mayatnikov <mayatnikov@me.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-25 16:56:02 +03:00

14 KiB
Raw Blame History

Runtime — hybrid script + LLM-on-demand

Status: active. This is the recommended way to run pepa-pi-bot since 2026-05-25. The pure Pi runtime (pi from repo root) still works and is documented as a fallback at the bottom of this file.

Why a hybrid runtime?

The original design ran every tick inside Pi — the LLM saw the world, picked one tool, executed it, looped. That gave full self-extension out of the box, but had three problems in practice:

  1. Slow. A "look around → defend yourself" round-trip took 2060 seconds because the LLM was in the hot path.
  2. Expensive. Hostile mob at 4 m? Cost of evasion = one full reasoning pass. Hungry? Same. Idle? Same.
  3. Invisible. With Pi as the only frontend, you had to tmux capture-pane to know what the bot was doing.

The hybrid runtime splits the bot into a script-driven layer that handles fast, well-understood things on its own, and a Pi (or Codex) headless escalation that's only invoked when the script gets stuck or needs to write new code for itself.

Architecture

┌────────────────────────────────────────────────────────────────────┐
│  operator                                                          │
│  ├── repo edits (.env, skills/, runtime/)                          │
│  ├── TUI (Ink) — see status, send chat, press [a] to escalate      │
│  └── (future) Telegram bridge                                      │
└─────────────┬────────────────────────────────────────────┬─────────┘
              │ Unix socket (newline-JSON)                 │ git
              ▼                                            ▼
┌────────────────────────────────────────────────────────────────────┐
│  runtime/bot.js — single long-running Node process                 │
│                                                                    │
│  ┌──────────────────┐  ┌──────────────────────┐  ┌────────────────┐│
│  │ Mineflayer       │  │ Reflex loop          │  │ IPC server     ││
│  │ - MC TCP         │  │ - tick every N sec   │  │ - Unix socket  ││
│  │ - AuthMe handler │◀─│ - priority order:    │─▶│ - broadcasts   ││
│  │ - chat / events  │  │   defend > eat       │  │   status/log/  ││
│  │                  │  │   > sleep > current  │  │   chat events  ││
│  │                  │  │   > idle             │  │ - accepts      ││
│  │                  │  │ - NO LLM in path     │  │   commands     ││
│  └──────────────────┘  └─────────┬────────────┘  └────────────────┘│
│                                  │                                 │
│                                  ▼ on stuck / new scenario         │
│                        ┌──────────────────────┐                    │
│                        │ pi-bridge.js         │                    │
│                        │ spawn `pi -p`        │                    │
│                        │ stream stdout to IPC │                    │
│                        └──────────────────────┘                    │
└────────────────────────────────────────────────────────────────────┘
              │
              │ TCP 25565
              ▼
       Minecraft server

The bot is one process. The TUI is a separate process you can connect and disconnect at will — the bot keeps running. Multiple TUI clients can attach to the same bot simultaneously.

Quickstart

# Once
cd ~/Projects/pepa-pi-bot
npm install

# Terminal 1 — the bot daemon
npm run bot
# Logs go to stdout AND state/<host>/logs/<YYYY-MM-DD>.log

# Terminal 2 — the dashboard
npm run tui

The TUI auto-reconnects to the bot if you restart it. Press q to leave the TUI; the bot is unaffected.

TUI hotkeys

Key Effect
p Pause / resume the reflex loop (MC connection stays).
s Stop the bot process gracefully (disconnect + cleanup + exit).
r Force-broadcast a status snapshot now.
c Enter chat mode — type a message, Enter sends it into MC chat.
a Enter ask-Pi mode — type a prompt, Enter spawns pi -p and streams output into the Pi panel.
y Open the latest pending proposal. In the proposal panel: y approves, n/Esc closes.
q Quit TUI only. Bot keeps running.

Enter submits, blank submit cancels. The status bar shows [proposals N, press y] when there's something pending.

What the reflex loop does today

The chain (highest priority first), wired and dispatching real Mineflayer actions:

  1. operatorGoalReflex — if OPERATOR_USERNAMES issued a come / follow command, satisfy it (walk to the operator's last known position, reply in chat on arrival or failure).
  2. defendReflex — closest hostile within 4 m → attackNearest (equips best melee). Within 12 m + low HP or ≥3 hostiles → fleeFrom along the away-vector.
  3. eatReflex — food < 16 → eatBestFood (picks from FOOD_PRIORITY list, equip + consume). 5 s cooldown.
  4. sleepReflex — night + no hostile within 8 m → sleepInBed (finds nearest placed bed within 16 blocks, paths there, sleeps). 30 s cooldown on failures.
  5. idleReflex — every 20th tick, log heartbeat (HP / food / pos).

Adding a new reflex = a function (ctx) => { action, ... } in runtime/reflex.js, inserted at the right priority. Actions live in runtime/actions.js. Both files trigger a supervisor hot-restart when saved (see "Self-improvement" below).

When the bot calls Pi

Two escalation paths:

1. Manual — operator presses a in the TUI, types a question, the bot spawns pi -p "<question>" and streams its stdout into the Pi panel.

2. Automatic — every tick where the entire reflex chain returns noop (no operator goal, no hostiles in reach, food fine, day or no bed, etc.) increments a counter. When the counter hits ESCALATE_AFTER_NOOPS = 20 (≈1 min at tick=3s), the bot fires askPi with the current snapshot and a fixed system prompt telling Pi to suggest one next action. 10 min cooldown so a permanently-idle bot doesn't run the LLM dry.

The auto-escalation prompt explicitly bans code-change proposals — Pi should only suggest what to do with the existing tools. If a deeper problem is happening, the failure-tracker (see Self-improvement) will file a proposal instead.

Operator chat commands

Players listed in OPERATOR_USERNAMES can address the bot in MC chat by prefixing the message with the bot's name:

pepa_bot status      → bot replies with HP / food / pos / hostiles / busy
pepa_bot come        → bot pathfinds to the operator's current position
pepa_bot pause       → reflex loop stops
pepa_bot resume      → reflex loop resumes
pepa_bot stop        → graceful disconnect + process exit

Unrecognized commands get a polite "didn't recognize" reply. Operator identity verification is the server's job (AuthMe on cracked, online-mode on premium) — the bot trusts the nickname.

IPC protocol

Socket: state/<MC_HOST>_<MC_PORT>/bot.sock (permissions 0600, removed on shutdown). Framing: one JSON object per line.

Server → client events (see runtime/ipc-protocol.js):

Type Payload
hello { snapshot, recentLogs } — sent on connect.
status full snapshot from perceive.js.
log { ts, level, source, text, details } — every log line.
chat { from, text, kind: "player" | "system" }.
death { reason, position }.
error { source, text }.
ask-pi-chunk { stream: "stdout" | "stderr", text }.
ask-pi-done { code, durationMs }.

Client → server commands:

Type Payload Effect
cmd:pause {} Reflex loop stops ticking.
cmd:resume {} Reflex loop resumes.
cmd:stop {} Graceful shutdown of the bot.
cmd:chat { text } Sends text into MC chat (rate-limited).
cmd:ask-pi { prompt } Spawns pi -p "<prompt>".
cmd:snapshot {} Force a status event now.

The protocol is intentionally tiny — anyone can write a second client (a Telegram bridge, a web UI, a one-shot CLI) by reading runtime/ipc-protocol.js.

Self-improvement loop (fully autonomous)

End-to-end, no operator-in-the-loop. The bot writes proposals when it spots a real bug, applies them with Pi headless, and rolls them back if they break things. The flow:

1. reflex dispatches action → action returns { ok: false, detail }
2. bot.js failure tracker classifies the detail:
     bug          → TypeError / Cannot read / is not defined …
     timeout      → "timed out after Ns"
     feature-gap  → "no reachable log", "no food", "no bed" …
     other        → anything else
3. 5 consecutive failures with the SAME label, where the run is dominated
   by 'bug' or all 'timeout' → writeProposal()
   (feature gaps are SKIPPED — reflex routing solves those, not the LLM)
4. runtime/auto-improve.js watcher (poll 2s) sees the new file,
   debounces 10s, then spawns scripts/auto-patch.js detached
5. auto-patch.js:
     - refuses on dirty tree
     - moves proposal pending → approved/ (audit trail)
     - creates branch auto/<slug> off main
     - runs `pi -p "<patch prompt>"` with 10-min timeout
     - if Pi committed AND only touched runtime/ → cherry-pick onto main
     - else → discard branch, exit non-zero
6. supervisor's runtime/*.js watcher fires the moment the cherry-pick
   lands → child restarts on the new code
7. if the new code crashes >5 times in 60s AND the last commit on main
   is younger than 15 min AND it touched runtime/ → supervisor
   `git reset --hard HEAD~1` and restarts. Up to MAX_ROLLBACKS times
   per supervisor lifetime, then bails out for manual investigation.

Rate limits

  • Proposal cooldown: 30 min between proposal files of any kind.
  • Auto-improve cooldown: 15 min between finished auto-patch.js runs.
  • Hourly cap: max 4 auto-patches per hour, even if cooldown allows.
  • Rollback cap: 3 rollbacks per supervisor lifetime; after that the supervisor exits and waits for human review.

What counts as a bug

runtime/bot.js ships two whitelists (NORMAL_FAILURE_SUBSTRINGS and BUG_FAILURE_SUBSTRINGS). The proposal trigger fires only when:

  • the trailing run of same-label failures contains at least one bug (TypeError / ReferenceError / "Cannot read properties" / etc.),
  • OR every failure in the run is a timeout (and they happened on the same operation, so it's probably broken not just unreachable).

Feature gaps like "no reachable log within 32 blocks" are not a bug — the autonomous reflex sees that result, sets noTreesUntil and switches to wander. If the script can't solve it via reflex routing, that's a design issue the operator fixes by editing runtime/reflex.js directly — not by asking Pi to patch around it.

Manual escape hatches

These still work but should rarely be needed:

  • TUI hotkey y opens the latest pending proposal for inspection.
  • npm run propose:apply <filename> runs the attended version of the patcher — leaves the result on a feat/proposal-<slug> branch without cherry-picking, so the operator can review the diff manually.
  • npm run stop kills everything and clears lock/socket.

File layout

runtime/
  supervisor.js       forks bot.js, watches runtime/*.js, restart-on-change
  bot.js              entrypoint — owns MC + tick + IPC + reconnect
  config.js           reads .env, exposes frozen config + redacted view
  log.js              ring buffer + stdout + daily file + IPC fan-out
  perceive.js         snapshot(bot) → JSON
  reflex.js           priority chain (operator > defend > eat > sleep > idle)
  actions.js          attackNearest / fleeFrom / eatBestFood / sleepInBed / goTo
  state-store.js      current-task / diary / proposals on disk
  ipc-server.js       Unix-socket server
  ipc-protocol.js     shared contract (event types, command types, framer)
  pi-bridge.js        spawn `pi -p`, stream stdout

tui/
  tui.tsx             Ink dashboard (React)
  ipc-client.js       socket client → EventEmitter

scripts/
  propose-apply.js    approved-proposal → feat-branch + `pi -p` patcher

Per-server state stays under state/<MC_HOST>_<MC_PORT>/, gitignored:

state/play.xmatic.team_25565/
  bot.sock                   Unix-domain socket (perms 0600, ephemeral)
  joined-before.flag         AuthMe /register vs /login marker
  current-task.json          resume anchor — what the bot was doing
  goal.md                    long-term ambition (operator-seeded)
  diary/YYYY-MM-DD.md        daily journal (one line per milestone)
  proposals/                 pending self-improvement proposals
  proposals/approved/        approved, waiting on propose:apply
  logs/YYYY-MM-DD.log        full runtime log mirror

Pi-only fallback

The original Pi-driven runtime still works if you prefer the single-process model — npm run agent from repo root loads AGENTS.md and the existing extensions in extensions/. The two runtimes share the .env, the mineflayer deps, and the state/ directory. They MUST NOT run simultaneously — both will try to claim the same MC nickname and the server will kick one of them.

If you switch between them frequently, kill one before starting the other:

# stop hybrid
# (in TUI press 's', or just kill `npm run bot`)

# start Pi
npm run agent