feat(runtime): persistent memory — world-journal + scenario-memory
Closes a structural gap: the bot now actually REMEMBERS what it
discovered and what it tried. Two stores live under state/<host>/ and
are wired in automatically.
runtime/world-journal.js
- Append-only JSONL of discovered points (chopped, placed, base,
shelter, farm, dead_end). Indexed by 16-block spatial grid; O(neighbors)
nearest() lookups; 6 h age prune; 10k line ceiling with trim.
- leanestQuadrant({x,z}) reports the quadrant the bot has the FEWEST
markers in — used by explore.far to circle rather than retread.
- summary() exposed for the stuck-incident proposal body.
runtime/scenario-memory.js
- Sliding window of (skillId, situationHash, code, ok, detail) tuples.
- situationHash() is a coarse fingerprint (16x8x16 cell + day/night +
food/hp bucket + inv key set + closest hostile). So "same kind of
place + same kind of state" matches.
- shouldSkip({skillId, situation}) → true after ≥3 failures within 30
min UNLESS a more-recent success in the same situation un-locks it.
- recentTailFor() exposed for the stuck-incident body.
Wiring (runtime/bot.js):
- dispatchAction captures situationHash BEFORE the action runs and
records (skillId, situation, code, ok) after — failures are attributed
to the dispatch-time state, not the partial-effect state.
- worldDelta fields (choppedAt, minedAt, placedAt, baseAt, shelterAt,
plantedAt, harvestedAt, tilledAt) auto-flow into the journal.
- no_target + silent_dig_failure also write dead_end markers.
Scheduler / skills now consume memory:
- reflex.js curriculum reflex calls memory.shouldSkip — if the same
(skill, situation) failed 3+ times recently, auto-converts to a
wander hint so the bot leaves and tries elsewhere.
- explore.far calls journal.leanestQuadrant when multiple cardinal
directions are walkable and prefers the less-explored one.
- gather.logs walks to the nearest known "chopped" bucket within 96
blocks before falling through to findBlock — chunks with confirmed
trees are more likely to yield another.
stuck-incident body now includes journal byKind + last 12 scenario
entries so Pi can write a structural fix, not just a guard clause.
Architecturally: this is the foundation for "bot rewrites itself".
The proposals Pi now receives carry real signal about what was tried
and what's around, instead of a single snapshot in isolation.
10 new tests (world-journal × 5, scenario-memory × 5). npm test 134/134.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -246,6 +246,33 @@ in-process. The package is **not** a default dep — install it explicitly
|
||||
(`npm i prismarine-viewer`) before enabling. If missing, the runtime
|
||||
logs a warning and continues.
|
||||
|
||||
### Memory: world-journal + scenario-memory (2026-05-26)
|
||||
|
||||
Two persistent stores under `state/<host>/`:
|
||||
|
||||
- **`world-journal.jsonl`** — append-only log of discovered points
|
||||
(`{kind, name, at:{x,y,z}, ts}`). Skills feed it automatically via
|
||||
`worldDelta` on each successful dispatch — chops, mines, placements,
|
||||
base/shelter location, planted/harvested crops, plus `dead_end`
|
||||
markers on `no_target` / `silent_dig_failure`. Indexed by a 16-block
|
||||
spatial grid so `nearest({kind, x, z, radius})` is O(neighbors).
|
||||
Pruned at 6 h age + 10k line ceiling. `leanestQuadrant({x, z})`
|
||||
returns the cardinal quadrant the bot has explored LEAST — used by
|
||||
`explore.far` to circle rather than retread the same patch.
|
||||
- **`scenarios.jsonl`** — sliding window of `(skillId, situationHash,
|
||||
code, ok, detail, ts)` tuples. `situationHash` is a coarse fingerprint
|
||||
of where + how the bot was (16-cell + 8y bucket, day/night, food
|
||||
bucket, hp bucket, inventory key set, closest hostile name). The
|
||||
curriculum reflex calls `memory.shouldSkip({skillId, situation})` —
|
||||
≥3 failures of the same `(skill, situation)` within 30 min and the
|
||||
reflex auto-converts into a wander hint instead of re-dispatching the
|
||||
failing skill. A subsequent success in the same situation un-locks it.
|
||||
|
||||
Both stores feed stuck-incident proposal bodies: when the LLM is
|
||||
asked to patch a stuck state, it sees `byKind` journal counts AND the
|
||||
last 12 scenario-memory entries, so it can write a structural fix
|
||||
based on what's actually been tried, not just one snapshot.
|
||||
|
||||
### Scheduler driven by the curriculum (2026-05-26)
|
||||
|
||||
The reflex chain is now: `defend → eat → sleep → curriculum → idle`.
|
||||
|
||||
+1
-1
@@ -16,7 +16,7 @@
|
||||
"tui": "tsx tui/tui.tsx",
|
||||
"propose:apply": "node scripts/propose-apply.js",
|
||||
"stop": "bash scripts/stop.sh",
|
||||
"test": "node --test runtime/skills/contract.test.js runtime/skills/groups.test.js runtime/skills/compat.test.js runtime/curriculum.test.js runtime/social/social.test.js runtime/stuck-incident.test.js runtime/compat.test.js runtime/reflex.test.js runtime/base-site.test.js runtime/locations.test.js runtime/watch-filter.test.js scripts/edit-scope.test.js"
|
||||
"test": "node --test runtime/skills/contract.test.js runtime/skills/groups.test.js runtime/skills/compat.test.js runtime/curriculum.test.js runtime/social/social.test.js runtime/stuck-incident.test.js runtime/compat.test.js runtime/reflex.test.js runtime/base-site.test.js runtime/locations.test.js runtime/watch-filter.test.js runtime/world-journal.test.js runtime/scenario-memory.test.js scripts/edit-scope.test.js"
|
||||
},
|
||||
"dependencies": {
|
||||
"dotenv": "^16.4.5",
|
||||
|
||||
+71
-1
@@ -48,6 +48,8 @@ import { generateReply } from "./social/reply.js";
|
||||
import { createChatMemory } from "./social/memory.js";
|
||||
import { createStuckIncidentDetector } from "./stuck-incident.js";
|
||||
import { createSkillMetrics } from "./skill-metrics.js";
|
||||
import { createWorldJournal } from "./world-journal.js";
|
||||
import { createScenarioMemory, situationHash } from "./scenario-memory.js";
|
||||
|
||||
fs.mkdirSync(stateDir, { recursive: true });
|
||||
const JOINED_FLAG = path.join(stateDir, "joined-before.flag");
|
||||
@@ -74,6 +76,8 @@ let lastEscalationAt = 0;
|
||||
const noProgress = createNoProgressDetector();
|
||||
const stuckIncident = createStuckIncidentDetector();
|
||||
const skillMetrics = createSkillMetrics();
|
||||
const worldJournal = createWorldJournal();
|
||||
const scenarioMemory = createScenarioMemory();
|
||||
let lastResult = null; // { label, ok, code, detail, ts }
|
||||
let lastFailureAt = 0;
|
||||
let lastPlanReadAt = 0;
|
||||
@@ -82,6 +86,9 @@ let cachedPlanExists = false;
|
||||
const MILESTONE_CACHE_MS = 30_000;
|
||||
|
||||
// Reflex context — passed into reflex.js every tick. Mutable across ticks.
|
||||
// Memory stores (journal + memory) live here so skill code can consult
|
||||
// them directly — gather.* can preferentially target known log positions,
|
||||
// deposit-surplus can mark the chest it placed, etc.
|
||||
const reflexCtx = {
|
||||
bot: null,
|
||||
snapshot: lastSnapshot,
|
||||
@@ -93,6 +100,8 @@ const reflexCtx = {
|
||||
// Tracks repeated failure of the same labelled action — triggers a proposal.
|
||||
recentFailures: [], // [{label, detail, ts}], capped at 10
|
||||
dispatch: dispatchAction,
|
||||
journal: worldJournal,
|
||||
memory: scenarioMemory,
|
||||
};
|
||||
|
||||
let chatTimestamps = [];
|
||||
@@ -156,6 +165,45 @@ function maybeHandleAuthPrompt(text) {
|
||||
// Reflexes call this to fire an async action without blocking the tick.
|
||||
// Sets busy=true, runs fn, clears busy when done; optional onComplete callback
|
||||
// receives the action's { ok, detail } result.
|
||||
// Pull worldDelta fields written by skills (e.g. {choppedAt, logType,
|
||||
// minedAt, blockType, gotWool, baseAt, shelterAt, depositedTotal,
|
||||
// plantedAt, harvestedAt, tilledAt}) and turn them into journal lines.
|
||||
// Any unknown delta is silently skipped — skills can extend the world
|
||||
// journal without bot.js needing to know each schema.
|
||||
function recordWorldDeltaToJournal(label, res, snapshot) {
|
||||
const wd = res?.worldDelta;
|
||||
if (!wd || typeof wd !== "object") return;
|
||||
try {
|
||||
if (wd.choppedAt) worldJournal.append({ kind: "chopped", name: wd.logType ?? "log", at: wd.choppedAt });
|
||||
if (wd.minedAt) worldJournal.append({ kind: "chopped", name: wd.blockType ?? "stone", at: wd.minedAt });
|
||||
if (wd.placedAt) worldJournal.append({ kind: "placed", name: wd.placedType ?? "block", at: wd.placedAt });
|
||||
if (wd.baseAt) worldJournal.append({ kind: "base", name: "base", at: wd.baseAt });
|
||||
if (wd.shelterAt) worldJournal.append({ kind: "shelter", name: "shelter", at: wd.shelterAt });
|
||||
if (wd.plantedAt) worldJournal.append({ kind: "farm", name: "planted", at: wd.plantedAt });
|
||||
if (wd.harvestedAt) worldJournal.append({ kind: "farm", name: "harvested", at: wd.harvestedAt });
|
||||
if (wd.tilledAt) worldJournal.append({ kind: "farm", name: "tilled", at: wd.tilledAt });
|
||||
// failures: blacklisted / no_target — log a dead-end at current pos
|
||||
if (res?.code === "no_target" && snapshot?.position) {
|
||||
worldJournal.append({
|
||||
kind: "dead_end",
|
||||
name: label,
|
||||
reason: res?.detail ? String(res.detail).slice(0, 80) : "no_target",
|
||||
at: snapshot.position,
|
||||
});
|
||||
}
|
||||
if (res?.code === "silent_dig_failure" && snapshot?.position) {
|
||||
worldJournal.append({
|
||||
kind: "dead_end",
|
||||
name: label,
|
||||
reason: "silent_dig_failure",
|
||||
at: wd?.blacklisted ?? snapshot.position,
|
||||
});
|
||||
}
|
||||
} catch (e) {
|
||||
warn("journal", `append from ${label} failed: ${e.message}`);
|
||||
}
|
||||
}
|
||||
|
||||
function dispatchAction(fn, label, opts = {}) {
|
||||
if (reflexCtx.busy) {
|
||||
warn("dispatch", `tried to dispatch ${label} while busy with ${reflexCtx.currentActionLabel}`);
|
||||
@@ -163,9 +211,14 @@ function dispatchAction(fn, label, opts = {}) {
|
||||
}
|
||||
reflexCtx.busy = true;
|
||||
reflexCtx.currentActionLabel = label;
|
||||
// Capture the situation hash BEFORE the action runs so a failure is
|
||||
// attributable to the state at dispatch time, not the state after the
|
||||
// (partial) effect.
|
||||
const startSnap = lastSnapshot;
|
||||
const startSituation = situationHash(startSnap);
|
||||
// current-task is a resume anchor — keep it small. Embedding the full
|
||||
// perception snapshot blows the file up to ~3 KB per write × every action.
|
||||
writeCurrentTask({ label, status: "in_progress", position: lastSnapshot.position });
|
||||
writeCurrentTask({ label, status: "in_progress", position: startSnap.position });
|
||||
info("dispatch", `→ ${label}`);
|
||||
Promise.resolve()
|
||||
.then(() => fn())
|
||||
@@ -184,6 +237,14 @@ function dispatchAction(fn, label, opts = {}) {
|
||||
ts: Date.now(),
|
||||
};
|
||||
skillMetrics.record(label, ok);
|
||||
scenarioMemory.record({
|
||||
skillId: label,
|
||||
situation: startSituation,
|
||||
code: lastResult.code,
|
||||
ok,
|
||||
detail: res?.detail,
|
||||
});
|
||||
recordWorldDeltaToJournal(label, res, startSnap);
|
||||
if (!ok) {
|
||||
lastFailureAt = lastResult.ts;
|
||||
recordFailure(label, res?.detail);
|
||||
@@ -210,6 +271,13 @@ function dispatchAction(fn, label, opts = {}) {
|
||||
ts: Date.now(),
|
||||
};
|
||||
skillMetrics.record(label, false);
|
||||
scenarioMemory.record({
|
||||
skillId: label,
|
||||
situation: startSituation,
|
||||
code: "threw",
|
||||
ok: false,
|
||||
detail: String(e?.message ?? e),
|
||||
});
|
||||
lastFailureAt = lastResult.ts;
|
||||
recordFailure(label, String(e?.message ?? e));
|
||||
})
|
||||
@@ -710,6 +778,8 @@ function tick() {
|
||||
snapshot: lastSnapshot,
|
||||
lastResult,
|
||||
metrics: lastSnapshot.skillMetrics,
|
||||
journalSummary: worldJournal.summary(),
|
||||
scenarioTail: scenarioMemory.recentTailFor({ n: 12 }),
|
||||
now,
|
||||
});
|
||||
if (stuck?.fire) {
|
||||
|
||||
@@ -25,6 +25,7 @@ import {
|
||||
wander,
|
||||
} from "./actions.js";
|
||||
import { runSkill, getSkill } from "./skills/index.js";
|
||||
import { situationHash } from "./scenario-memory.js";
|
||||
|
||||
// Each "wander hint" triggered by a skill returning no_target should take
|
||||
// the bot meaningfully further than 16 blocks — otherwise the curriculum
|
||||
@@ -226,6 +227,20 @@ function curriculumReflex(ctx) {
|
||||
const backoffUntil = ctx.skillBackoff?.[skillId] ?? 0;
|
||||
if (Date.now() < backoffUntil) return { action: "noop" };
|
||||
|
||||
// Scenario memory: this exact (skill, situation) pattern failed N times
|
||||
// recently? Skip and let the wander/explore hint move us to a different
|
||||
// situation. The hash includes coarse position + day/night + food + hp
|
||||
// + inventory keys + nearby hostile — "same kind of place + state".
|
||||
if (ctx.memory?.shouldSkip && ctx.snapshot) {
|
||||
const sit = situationHash(ctx.snapshot);
|
||||
if (ctx.memory.shouldSkip({ skillId, situation: sit })) {
|
||||
// Pretend a wander hint fired so the next tick will explore.
|
||||
ctx.skillBackoff = ctx.skillBackoff ?? {};
|
||||
ctx.skillBackoff["__wander_hint__"] = Date.now() + SKILL_BACKOFF_MS;
|
||||
return { action: "noop" };
|
||||
}
|
||||
}
|
||||
|
||||
ctx.lastCurriculumAt = Date.now();
|
||||
ctx.dispatch(() => runSkill(skillId, ctx), skillId, {
|
||||
onComplete: (res) => {
|
||||
|
||||
@@ -0,0 +1,149 @@
|
||||
// Scenario memory: the bot remembers what skill it tried in what kind of
|
||||
// situation and how that worked out. Two consumers:
|
||||
//
|
||||
// 1. The scheduler skips a skill when (skillId, situationHash) showed
|
||||
// ≥N failures within MEMORY_WINDOW_MS — no point retrying the same
|
||||
// thing in the same context. The cooldown is shorter than the
|
||||
// per-skill backoff so the bot can come back later, but the hash
|
||||
// includes the current biome + time-of-day + nearby-block-types so
|
||||
// "I'm in a different place now" un-locks the skill automatically.
|
||||
//
|
||||
// 2. Pi-side proposals (stuck-incident) get a tail of recent entries
|
||||
// so the LLM can see what's been tried and reason about it
|
||||
// structurally instead of guessing.
|
||||
//
|
||||
// Persisted at state/<host>/scenarios.jsonl, append-only. Loaded into a
|
||||
// circular in-memory buffer for fast lookups. Old entries pruned by age.
|
||||
|
||||
import fs from "node:fs";
|
||||
import path from "node:path";
|
||||
import { stateDir } from "./config.js";
|
||||
|
||||
const SCENARIO_PATH = path.join(stateDir, "scenarios.jsonl");
|
||||
const MAX_LINES = 5_000;
|
||||
const TRIM_TARGET = 3_500;
|
||||
const MEMORY_WINDOW_MS = 30 * 60_000; // 30 min — recent enough to matter
|
||||
|
||||
function ensureDir() {
|
||||
try { fs.mkdirSync(stateDir, { recursive: true }); } catch {}
|
||||
}
|
||||
|
||||
// Cheap structural hash of the situation the bot is acting in: bucketed
|
||||
// position (16×16×8 cell), biome (if known), day/night, food bucket,
|
||||
// inventory keys (presence only, not counts), closest hostile type. The
|
||||
// hash is intentionally coarse — we want "same kind of place + same kind
|
||||
// of state" to match, not "exact (x,y,z) again".
|
||||
export function situationHash(snapshot) {
|
||||
if (!snapshot?.connected) return "disconnected";
|
||||
const p = snapshot.position ?? { x: 0, y: 0, z: 0 };
|
||||
const cx = Math.floor(p.x / 16);
|
||||
const cy = Math.floor(p.y / 8);
|
||||
const cz = Math.floor(p.z / 16);
|
||||
const day = snapshot.isDay ? "d" : "n";
|
||||
const food = snapshot.food === undefined ? "?" : snapshot.food >= 18 ? "F" : snapshot.food >= 12 ? "f" : "h";
|
||||
const hp = (snapshot.health ?? 20) >= 15 ? "H" : "L";
|
||||
const hostile = snapshot.closestHostile && snapshot.closestHostile.distance < 24
|
||||
? snapshot.closestHostile.name
|
||||
: "-";
|
||||
// Inventory keys, sorted — we lose counts but keep "what kind of stuff do
|
||||
// I have". Limited to first 10 names for hash stability.
|
||||
const invKeys = Object.keys(snapshot.inventory ?? {})
|
||||
.sort()
|
||||
.slice(0, 10)
|
||||
.join(",") || "-";
|
||||
return `${cx},${cy},${cz}|${day}|${food}|${hp}|host:${hostile}|inv:${invKeys}`;
|
||||
}
|
||||
|
||||
function loadScenarios() {
|
||||
const entries = [];
|
||||
let raw = "";
|
||||
try { raw = fs.readFileSync(SCENARIO_PATH, "utf8"); }
|
||||
catch (e) { if (e.code !== "ENOENT") return entries; return entries; }
|
||||
const now = Date.now();
|
||||
const lines = raw.split("\n").filter((l) => l.trim());
|
||||
for (const line of lines) {
|
||||
try {
|
||||
const entry = JSON.parse(line);
|
||||
const ts = Date.parse(entry.ts ?? "");
|
||||
if (!Number.isFinite(ts)) continue;
|
||||
if (now - ts > MEMORY_WINDOW_MS * 4) continue; // hard expire ×4 window
|
||||
entries.push({ ...entry, _ts: ts });
|
||||
} catch {}
|
||||
}
|
||||
return entries;
|
||||
}
|
||||
|
||||
function maybeTrim(linesSoFar) {
|
||||
if (linesSoFar <= MAX_LINES) return;
|
||||
try {
|
||||
const raw = fs.readFileSync(SCENARIO_PATH, "utf8");
|
||||
const lines = raw.split("\n").filter((l) => l.trim());
|
||||
const keep = lines.slice(-TRIM_TARGET).join("\n") + "\n";
|
||||
const tmp = `${SCENARIO_PATH}.tmp`;
|
||||
fs.writeFileSync(tmp, keep);
|
||||
fs.renameSync(tmp, SCENARIO_PATH);
|
||||
} catch {}
|
||||
}
|
||||
|
||||
export function createScenarioMemory({ failureThreshold = 3, windowMs = MEMORY_WINDOW_MS } = {}) {
|
||||
const entries = loadScenarios();
|
||||
let appended = entries.length;
|
||||
|
||||
function record({ skillId, situation, code, ok, detail }) {
|
||||
if (!skillId || !situation) return;
|
||||
const entry = {
|
||||
ts: new Date().toISOString(),
|
||||
skillId, situation, code, ok: !!ok,
|
||||
detail: detail ? String(detail).slice(0, 200) : null,
|
||||
_ts: Date.now(),
|
||||
};
|
||||
entries.push(entry);
|
||||
ensureDir();
|
||||
fs.appendFileSync(SCENARIO_PATH, JSON.stringify({ ...entry, _ts: undefined }) + "\n");
|
||||
appended++;
|
||||
if (appended % 200 === 0) maybeTrim(appended);
|
||||
}
|
||||
|
||||
function recentFailures({ skillId, situation, now = Date.now() }) {
|
||||
return entries.filter((e) =>
|
||||
e.skillId === skillId &&
|
||||
e.situation === situation &&
|
||||
!e.ok &&
|
||||
now - e._ts < windowMs
|
||||
);
|
||||
}
|
||||
|
||||
// Decision: "should I skip dispatching skill in this situation right now?"
|
||||
function shouldSkip({ skillId, situation, now = Date.now() }) {
|
||||
const fails = recentFailures({ skillId, situation, now });
|
||||
if (fails.length < failureThreshold) return false;
|
||||
// Also: have we succeeded with this same (skill, situation) lately? If
|
||||
// so, don't skip — the pattern may have changed.
|
||||
const recentOk = entries.find((e) =>
|
||||
e.skillId === skillId &&
|
||||
e.situation === situation &&
|
||||
e.ok &&
|
||||
now - e._ts < windowMs,
|
||||
);
|
||||
if (recentOk) return false;
|
||||
return true;
|
||||
}
|
||||
|
||||
// Pi-bound: tail of recent attempts as plain text (skill, situation, code, ok).
|
||||
function recentTailFor({ skillId, n = 10 } = {}) {
|
||||
const filtered = skillId ? entries.filter((e) => e.skillId === skillId) : entries;
|
||||
return filtered.slice(-n).map((e) => ({
|
||||
ts: e.ts,
|
||||
skillId: e.skillId,
|
||||
ok: e.ok,
|
||||
code: e.code,
|
||||
detail: e.detail,
|
||||
situationHashShort: (e.situation ?? "").slice(0, 60),
|
||||
}));
|
||||
}
|
||||
|
||||
function size() { return entries.length; }
|
||||
function clear() { entries.length = 0; try { fs.unlinkSync(SCENARIO_PATH); } catch {} }
|
||||
|
||||
return { record, recentFailures, shouldSkip, recentTailFor, size, clear };
|
||||
}
|
||||
@@ -0,0 +1,69 @@
|
||||
import { test } from "node:test";
|
||||
import assert from "node:assert/strict";
|
||||
|
||||
import { createScenarioMemory, situationHash } from "./scenario-memory.js";
|
||||
|
||||
function snap(extras = {}) {
|
||||
return {
|
||||
connected: true,
|
||||
position: { x: 100, y: 64, z: -200 },
|
||||
isDay: true,
|
||||
food: 18,
|
||||
health: 20,
|
||||
inventory: {},
|
||||
closestHostile: null,
|
||||
...extras,
|
||||
};
|
||||
}
|
||||
|
||||
test("situationHash is stable for the same situation, different for different cell", () => {
|
||||
const a = situationHash(snap());
|
||||
const b = situationHash(snap());
|
||||
assert.equal(a, b);
|
||||
const c = situationHash(snap({ position: { x: 200, y: 64, z: -200 } }));
|
||||
assert.notEqual(a, c);
|
||||
});
|
||||
|
||||
test("situationHash changes with day/night, food, health bucket, hostile", () => {
|
||||
const base = snap();
|
||||
const night = situationHash({ ...base, isDay: false });
|
||||
const lowHp = situationHash({ ...base, health: 5 });
|
||||
const hostile = situationHash({ ...base, closestHostile: { name: "zombie", distance: 5 } });
|
||||
assert.notEqual(situationHash(base), night);
|
||||
assert.notEqual(situationHash(base), lowHp);
|
||||
assert.notEqual(situationHash(base), hostile);
|
||||
});
|
||||
|
||||
test("shouldSkip flips after N failures in the same situation", () => {
|
||||
const m = createScenarioMemory({ failureThreshold: 3, windowMs: 60_000 });
|
||||
const sit = "x|y|z|d|F|H|host:-|inv:-";
|
||||
const skillId = "test.always-fails-" + Date.now();
|
||||
assert.equal(m.shouldSkip({ skillId, situation: sit }), false);
|
||||
for (let i = 0; i < 3; i++) {
|
||||
m.record({ skillId, situation: sit, code: "no_target", ok: false, detail: "no" });
|
||||
}
|
||||
assert.equal(m.shouldSkip({ skillId, situation: sit }), true);
|
||||
});
|
||||
|
||||
test("a recent success in the same situation un-locks the skill", () => {
|
||||
const m = createScenarioMemory({ failureThreshold: 2, windowMs: 60_000 });
|
||||
const sit = "site_alpha";
|
||||
const skillId = "test.flaky-" + Date.now();
|
||||
m.record({ skillId, situation: sit, code: "fail", ok: false });
|
||||
m.record({ skillId, situation: sit, code: "fail", ok: false });
|
||||
assert.equal(m.shouldSkip({ skillId, situation: sit }), true);
|
||||
m.record({ skillId, situation: sit, code: "done", ok: true });
|
||||
assert.equal(m.shouldSkip({ skillId, situation: sit }), false);
|
||||
});
|
||||
|
||||
test("recentTailFor returns most-recent entries with situation hash short form", () => {
|
||||
const m = createScenarioMemory();
|
||||
const skillId = "test.tail-" + Date.now();
|
||||
for (let i = 0; i < 5; i++) {
|
||||
m.record({ skillId, situation: `s${i}`, code: i % 2 ? "ok" : "no_target", ok: i % 2 === 1 });
|
||||
}
|
||||
const tail = m.recentTailFor({ skillId, n: 3 });
|
||||
assert.equal(tail.length, 3);
|
||||
assert.equal(tail[2].skillId, skillId);
|
||||
assert.ok(typeof tail[0].situationHashShort === "string");
|
||||
});
|
||||
@@ -3,8 +3,27 @@
|
||||
// while exposing the survival-skill contract (preconditions, timeout,
|
||||
// structured worldDelta, recovery hint).
|
||||
|
||||
import pathfinderPkg from "mineflayer-pathfinder";
|
||||
const { pathfinder, goals } = pathfinderPkg;
|
||||
|
||||
import { chopNearestTree } from "../actions.js";
|
||||
import { logs as logBlocks } from "./groups.js";
|
||||
import { info } from "../log.js";
|
||||
|
||||
let pluginLoaded = new WeakSet();
|
||||
function ensurePathfinder(bot) {
|
||||
if (pluginLoaded.has(bot)) return;
|
||||
bot.loadPlugin(pathfinder);
|
||||
pluginLoaded.add(bot);
|
||||
}
|
||||
|
||||
function withTimeout(promise, ms, label) {
|
||||
let timer;
|
||||
const timeout = new Promise((_, reject) => {
|
||||
timer = setTimeout(() => reject(new Error(`${label} timed out after ${ms / 1000}s`)), ms);
|
||||
});
|
||||
return Promise.race([promise, timeout]).finally(() => clearTimeout(timer));
|
||||
}
|
||||
|
||||
export const skill = Object.freeze({
|
||||
id: "gather.logs",
|
||||
@@ -19,6 +38,28 @@ export const skill = Object.freeze({
|
||||
return { ok: true };
|
||||
},
|
||||
async execute(ctx) {
|
||||
// Journal-aware approach: if we previously chopped a tree (or saw a
|
||||
// tree marker) within 96 blocks, walk to that bucket first — the
|
||||
// chunk has been confirmed to contain trees, so the next chop is
|
||||
// likely to succeed there even if findBlock at the current position
|
||||
// returned nothing.
|
||||
const bot = ctx.bot;
|
||||
const here = bot.entity?.position;
|
||||
if (here && ctx?.journal?.nearest) {
|
||||
const near = ctx.journal.nearest({ kind: "chopped", x: here.x, z: here.z, radius: 96, limit: 1 });
|
||||
if (near.length && near[0].distance > 8) {
|
||||
ensurePathfinder(bot);
|
||||
const t = near[0].at;
|
||||
info("action", `gather.logs: journal hint → walk to known tree area ${t.x},${t.y},${t.z} (${near[0].distance.toFixed(0)}m)`);
|
||||
try {
|
||||
await withTimeout(
|
||||
bot.pathfinder.goto(new goals.GoalNear(t.x, t.y, t.z, 6)),
|
||||
25_000,
|
||||
"gotoTreeArea",
|
||||
);
|
||||
} catch {} // ignore, fall through to chop attempt
|
||||
}
|
||||
}
|
||||
const res = await chopNearestTree(ctx.bot);
|
||||
if (res.ok) {
|
||||
return {
|
||||
|
||||
@@ -69,7 +69,23 @@ export const skill = Object.freeze({
|
||||
// gives us a free-direction signal cheaply.
|
||||
const dist = Math.max(24, args.distance ?? 48);
|
||||
const trials = await probeCardinalStep(bot, 800);
|
||||
const best = trials.reduce((b, t) => (t.dist > b.dist ? t : b), { dist: 0, yaw: 0, name: "?" });
|
||||
const movable = trials.filter((t) => t.dist > 0.5);
|
||||
|
||||
// Journal-aware: if more than one direction is walkable, prefer the
|
||||
// one in the LEANEST quadrant (least amount of stuff we've already
|
||||
// catalogued, including dead_end markers — so we don't loop the same
|
||||
// area). Falls back to "best by distance" when only one direction
|
||||
// works or journal is empty.
|
||||
let best = trials.reduce((b, t) => (t.dist > b.dist ? t : b), { dist: 0, yaw: 0, name: "?" });
|
||||
if (movable.length > 1 && ctx?.journal?.leanestQuadrant) {
|
||||
const here = bot.entity.position;
|
||||
const { best: leanest } = ctx.journal.leanestQuadrant({ x: here.x, z: here.z, radius: 96 });
|
||||
const QUAD_TO_CARDINAL = { NE: "N", SE: "E", SW: "S", NW: "W" };
|
||||
const preferredName = QUAD_TO_CARDINAL[leanest];
|
||||
const preferred = movable.find((t) => t.name === preferredName);
|
||||
if (preferred) best = preferred;
|
||||
info("action", `explore.far: journal says leanest quadrant=${leanest} → prefer ${preferredName}`);
|
||||
}
|
||||
info("action", `explore.far: cardinal probe trials=${trials.map((t) => `${t.name}:${t.dist.toFixed(1)}`).join(" ")} best=${best.name}`);
|
||||
|
||||
if (best.dist < 0.5) {
|
||||
|
||||
@@ -31,7 +31,7 @@ export function createStuckIncidentDetector({ thresholdMs = STUCK_THRESHOLD_MS,
|
||||
|
||||
// Returns { fire: true, body, kind, editScope } when an incident should
|
||||
// be filed this tick; null otherwise.
|
||||
function check({ snapshot, lastResult, metrics, now = Date.now() }) {
|
||||
function check({ snapshot, lastResult, metrics, journalSummary, scenarioTail, now = Date.now() }) {
|
||||
const reason = snapshot?.noProgressReason ?? null;
|
||||
if (!reason) {
|
||||
reset();
|
||||
@@ -71,6 +71,16 @@ export function createStuckIncidentDetector({ thresholdMs = STUCK_THRESHOLD_MS,
|
||||
.join(", ")
|
||||
: "(none)";
|
||||
|
||||
const journalLine = journalSummary
|
||||
? `byKind=${JSON.stringify(journalSummary.byKind ?? {})} buckets=${journalSummary.totalBuckets ?? 0}`
|
||||
: "(no journal)";
|
||||
|
||||
const scenarioLines = Array.isArray(scenarioTail) && scenarioTail.length
|
||||
? scenarioTail.map((e) =>
|
||||
`- \`${e.skillId}\` ${e.ok ? "OK" : "FAIL"} code=${e.code ?? "?"} ${e.detail ? `(${e.detail.slice?.(0, 120) ?? ""})` : ""} situ=${e.situationHashShort ?? "?"}`,
|
||||
).join("\n")
|
||||
: "_(no scenario memory recorded yet)_";
|
||||
|
||||
const body = [
|
||||
`# Stuck on \`${reason}\``,
|
||||
"",
|
||||
@@ -88,15 +98,23 @@ export function createStuckIncidentDetector({ thresholdMs = STUCK_THRESHOLD_MS,
|
||||
? `\`${lastResult.label}\` → ${lastResult.code ?? (lastResult.ok ? "ok" : "fail")}${lastResult.detail ? ` (${JSON.stringify(lastResult.detail).slice(0, 200)})` : ""}`
|
||||
: "_(none recorded)_",
|
||||
"",
|
||||
"## Skill metrics so far",
|
||||
"## Skill metrics so far (this process lifetime)",
|
||||
"",
|
||||
metricsLine,
|
||||
"",
|
||||
"## World journal (what we have discovered so far)",
|
||||
"",
|
||||
journalLine,
|
||||
"",
|
||||
"## Recent scenario memory (last attempts, what worked / failed in similar situations)",
|
||||
"",
|
||||
scenarioLines,
|
||||
"",
|
||||
"## Suggested fix",
|
||||
"",
|
||||
suggested
|
||||
? `Improve \`${suggested}\` so the bot can clear the \`${reason}\` blocker. Touch only the listed files. Add or update tests under \`runtime/skills/\`.`
|
||||
: `The curriculum has no suggested skill for this state. Either teach the curriculum a new milestone OR add a recovery skill that turns this reason code into a productive action.`,
|
||||
? `Improve \`${suggested}\` so the bot can clear the \`${reason}\` blocker, OR teach a NEW skill that handles this kind of situation if no single edit fixes it. Touch only the listed files (the test files under runtime/**/*.test.js are auto-allowed). Use the scenario-memory entries above to avoid re-introducing patterns that already failed.`
|
||||
: `The curriculum has no suggested skill for this state. Either teach the curriculum a new milestone OR add a recovery skill that turns this reason code into a productive action. The scenario memory above shows what's been tried.`,
|
||||
"",
|
||||
"## Edit scope (auto-patch must obey this)",
|
||||
"",
|
||||
|
||||
@@ -0,0 +1,151 @@
|
||||
// Persistent map of what the bot has discovered in the world. Append-only
|
||||
// JSONL at state/<host>/world-journal.jsonl. Lines look like:
|
||||
//
|
||||
// {"ts":"…","kind":"tree","name":"oak_log","at":{"x":..,"y":..,"z":..},"dim":"overworld"}
|
||||
// {"ts":"…","kind":"stone","name":"stone","at":{...}}
|
||||
// {"ts":"…","kind":"water","at":{...}}
|
||||
// {"ts":"…","kind":"sheep","at":{...}}
|
||||
// {"ts":"…","kind":"hostile_zone","name":"zombie","at":{...},"count":12}
|
||||
// {"ts":"…","kind":"dead_end","reason":"wedged","at":{...}}
|
||||
// {"ts":"…","kind":"chopped","name":"oak_log","at":{...}} // we removed this block
|
||||
// {"ts":"…","kind":"placed","name":"crafting_table","at":{...}}
|
||||
//
|
||||
// On boot we replay the file into an in-memory index (kind → spatial
|
||||
// bucket). Bucket = floor(coord / GRID_CELL) so lookups are O(neighbors).
|
||||
// Old entries auto-prune at MAX_AGE_MS so the bot adapts to changing world.
|
||||
|
||||
import fs from "node:fs";
|
||||
import path from "node:path";
|
||||
import { stateDir } from "./config.js";
|
||||
|
||||
const JOURNAL_PATH = path.join(stateDir, "world-journal.jsonl");
|
||||
const GRID_CELL = 16;
|
||||
const MAX_AGE_MS = 6 * 60 * 60_000; // 6 hours; world changes faster than that on a live server
|
||||
const MAX_LINES = 10_000;
|
||||
const TRIM_TARGET = 7_500;
|
||||
|
||||
function ensureDir() {
|
||||
try { fs.mkdirSync(stateDir, { recursive: true }); } catch {}
|
||||
}
|
||||
|
||||
function cellOf(x, z) {
|
||||
return `${Math.floor(x / GRID_CELL)},${Math.floor(z / GRID_CELL)}`;
|
||||
}
|
||||
|
||||
function loadJournal() {
|
||||
const index = new Map(); // "kind:cell" → array of entries
|
||||
let raw = "";
|
||||
try { raw = fs.readFileSync(JOURNAL_PATH, "utf8"); }
|
||||
catch (e) { if (e.code !== "ENOENT") return { index }; return { index }; }
|
||||
const now = Date.now();
|
||||
const lines = raw.split("\n").filter((l) => l.trim());
|
||||
for (const line of lines) {
|
||||
try {
|
||||
const entry = JSON.parse(line);
|
||||
if (!entry?.at || typeof entry.at.x !== "number") continue;
|
||||
const ts = Date.parse(entry.ts ?? "");
|
||||
if (Number.isFinite(ts) && now - ts > MAX_AGE_MS) continue;
|
||||
const key = `${entry.kind}:${cellOf(entry.at.x, entry.at.z)}`;
|
||||
const bucket = index.get(key) ?? [];
|
||||
bucket.push(entry);
|
||||
index.set(key, bucket);
|
||||
} catch {}
|
||||
}
|
||||
return { index, count: lines.length };
|
||||
}
|
||||
|
||||
function maybeTrim(count) {
|
||||
if (count <= MAX_LINES) return;
|
||||
try {
|
||||
const raw = fs.readFileSync(JOURNAL_PATH, "utf8");
|
||||
const lines = raw.split("\n").filter((l) => l.trim());
|
||||
const keep = lines.slice(-TRIM_TARGET).join("\n") + "\n";
|
||||
const tmp = `${JOURNAL_PATH}.tmp`;
|
||||
fs.writeFileSync(tmp, keep);
|
||||
fs.renameSync(tmp, JOURNAL_PATH);
|
||||
} catch {}
|
||||
}
|
||||
|
||||
export function createWorldJournal() {
|
||||
let { index, count = 0 } = loadJournal();
|
||||
let appended = count;
|
||||
|
||||
function append(entry) {
|
||||
if (!entry?.kind || !entry?.at) return;
|
||||
ensureDir();
|
||||
const line = JSON.stringify({ ts: new Date().toISOString(), ...entry }) + "\n";
|
||||
fs.appendFileSync(JOURNAL_PATH, line);
|
||||
const key = `${entry.kind}:${cellOf(entry.at.x, entry.at.z)}`;
|
||||
const bucket = index.get(key) ?? [];
|
||||
bucket.push({ ts: new Date().toISOString(), ...entry });
|
||||
index.set(key, bucket);
|
||||
appended++;
|
||||
if (appended % 500 === 0) maybeTrim(appended);
|
||||
}
|
||||
|
||||
// O(neighbors): scan a ring of cells around (x,z) up to `radius` blocks
|
||||
// and return all entries matching `kind`. Filters out dead_end markers
|
||||
// older than the live game cycle.
|
||||
function nearest({ kind, x, z, radius = 32, limit = 5 }) {
|
||||
const cellsPerSide = Math.ceil(radius / GRID_CELL);
|
||||
const cx = Math.floor(x / GRID_CELL);
|
||||
const cz = Math.floor(z / GRID_CELL);
|
||||
const out = [];
|
||||
for (let dx = -cellsPerSide; dx <= cellsPerSide; dx++) {
|
||||
for (let dz = -cellsPerSide; dz <= cellsPerSide; dz++) {
|
||||
const key = `${kind}:${cx + dx},${cz + dz}`;
|
||||
const bucket = index.get(key);
|
||||
if (!bucket) continue;
|
||||
for (const e of bucket) {
|
||||
const d = Math.hypot(e.at.x - x, e.at.z - z);
|
||||
if (d <= radius) out.push({ ...e, distance: d });
|
||||
}
|
||||
}
|
||||
}
|
||||
out.sort((a, b) => a.distance - b.distance);
|
||||
return out.slice(0, limit);
|
||||
}
|
||||
|
||||
// "Which quadrants have we NOT searched yet?" — exploration helper for
|
||||
// explore.far. Counts dead_end + dead-target entries per quadrant, recommends
|
||||
// the leanest one.
|
||||
function leanestQuadrant({ x, z, radius = 64 }) {
|
||||
const quads = {
|
||||
NE: 0, SE: 0, SW: 0, NW: 0,
|
||||
};
|
||||
for (const bucket of index.values()) {
|
||||
for (const e of bucket) {
|
||||
const dx = e.at.x - x;
|
||||
const dz = e.at.z - z;
|
||||
if (Math.hypot(dx, dz) > radius) continue;
|
||||
if (dx >= 0 && dz < 0) quads.NE++;
|
||||
else if (dx >= 0 && dz >= 0) quads.SE++;
|
||||
else if (dx < 0 && dz >= 0) quads.SW++;
|
||||
else quads.NW++;
|
||||
}
|
||||
}
|
||||
// Return the quadrant with the FEWEST known markers — least explored.
|
||||
let best = "NE";
|
||||
for (const [q, c] of Object.entries(quads)) if (c < quads[best]) best = q;
|
||||
return { best, counts: quads };
|
||||
}
|
||||
|
||||
function summary() {
|
||||
const byKind = {};
|
||||
for (const [key, bucket] of index) {
|
||||
const kind = key.split(":")[0];
|
||||
byKind[kind] = (byKind[kind] ?? 0) + bucket.length;
|
||||
}
|
||||
return { totalBuckets: index.size, byKind };
|
||||
}
|
||||
|
||||
function clear() {
|
||||
index.clear();
|
||||
try { fs.unlinkSync(JOURNAL_PATH); } catch {}
|
||||
appended = 0;
|
||||
}
|
||||
|
||||
return { append, nearest, leanestQuadrant, summary, clear };
|
||||
}
|
||||
|
||||
export const _internal = { cellOf, GRID_CELL, MAX_AGE_MS };
|
||||
@@ -0,0 +1,58 @@
|
||||
import { test } from "node:test";
|
||||
import assert from "node:assert/strict";
|
||||
|
||||
import { createWorldJournal, _internal } from "./world-journal.js";
|
||||
|
||||
function tag() { return `__t_${Date.now()}_${Math.floor(Math.random() * 1e6)}`; }
|
||||
|
||||
test("cellOf buckets by GRID_CELL", () => {
|
||||
assert.equal(_internal.cellOf(0, 0), "0,0");
|
||||
assert.equal(_internal.cellOf(15, 15), "0,0");
|
||||
assert.equal(_internal.cellOf(16, 16), "1,1");
|
||||
assert.equal(_internal.cellOf(-1, -1), "-1,-1");
|
||||
});
|
||||
|
||||
test("append + nearest returns the entry we just stored", () => {
|
||||
const j = createWorldJournal();
|
||||
const name = tag();
|
||||
j.append({ kind: "chopped", name, at: { x: 100, y: 64, z: 200 } });
|
||||
const got = j.nearest({ kind: "chopped", x: 100, z: 200, radius: 16, limit: 5 });
|
||||
const ours = got.find((e) => e.name === name);
|
||||
assert.ok(ours, "expected to find our own entry back");
|
||||
assert.equal(ours.at.x, 100);
|
||||
});
|
||||
|
||||
test("nearest ranks by distance and respects radius", () => {
|
||||
const j = createWorldJournal();
|
||||
const t = tag();
|
||||
j.append({ kind: "stone", name: t + "_far", at: { x: 100, y: 64, z: 100 } });
|
||||
j.append({ kind: "stone", name: t + "_near", at: { x: 5, y: 64, z: 5 } });
|
||||
const got = j.nearest({ kind: "stone", x: 0, z: 0, radius: 50, limit: 5 });
|
||||
const near = got.find((e) => e.name === t + "_near");
|
||||
const far = got.find((e) => e.name === t + "_far");
|
||||
assert.ok(near);
|
||||
assert.equal(far, undefined, "far entry > radius should be excluded");
|
||||
});
|
||||
|
||||
test("leanestQuadrant returns the quadrant with fewest entries", () => {
|
||||
const j = createWorldJournal();
|
||||
const t = tag();
|
||||
for (let i = 0; i < 5; i++) {
|
||||
j.append({ kind: "dead_end", name: t, at: { x: 10 + i, y: 64, z: -10 - i } }); // NE
|
||||
}
|
||||
for (let i = 0; i < 2; i++) {
|
||||
j.append({ kind: "dead_end", name: t, at: { x: 10 + i, y: 64, z: 10 + i } }); // SE
|
||||
}
|
||||
const { best, counts } = j.leanestQuadrant({ x: 0, z: 0, radius: 64 });
|
||||
assert.ok(["SW", "NW"].includes(best), `expected unused quadrant, got ${best} counts=${JSON.stringify(counts)}`);
|
||||
});
|
||||
|
||||
test("summary lists per-kind totals", () => {
|
||||
const j = createWorldJournal();
|
||||
j.append({ kind: "chopped", name: "oak_log", at: { x: 1, y: 1, z: 1 } });
|
||||
j.append({ kind: "chopped", name: "oak_log", at: { x: 2, y: 1, z: 2 } });
|
||||
j.append({ kind: "shelter", name: "shelter", at: { x: 0, y: 1, z: 0 } });
|
||||
const s = j.summary();
|
||||
assert.ok(s.byKind.chopped >= 2);
|
||||
assert.ok(s.byKind.shelter >= 1);
|
||||
});
|
||||
Reference in New Issue
Block a user