Closes the gap rc.2 left open. Live observation showed:
- Pi-coach extracted 5 high-quality lessons (do not explore.far at
night near zombies, etc.) but none of them fired (applied_count=0
across the board). Root cause: dispatcher consulted advice only on
the main curriculum path; the bot was falling into the wander/
explore.far FALLBACK after each gather attempt bailed, which
bypassed consult().
- Bot was wedged in a pit on (608, 90) with stone walls. recovery.
tunnel-out kept failing ("Digging aborted") because mining stone
with fists takes ~10s/block; pathfinder watchdog kills it.
This patch:
1. survive.pillar-up (runtime/skills/pillar-up.js) — new escape skill.
Places a placeable block under the bot and jumps onto it; repeats
up to 8 steps. No pickaxe required. Works in dirt/cobble/planks/
sand/gravel/wool/etc. The bot's vertical exit from any pit it can
stand in.
2. Wedged-emergency reflex (runtime/reflex.js). At the top of
curriculumReflex, if noProgressReason is wedged-like AND position
hasn't shifted ≥16 blocks in 60s AND no hostile in 6m AND pillar
block in inventory → dispatch survive.pillar-up. 2-min cooldown
between attempts.
3. consult() now also runs on the WANDER/explore.far fallback path
(runtime/reflex.js curriculumReflex). Pi-coach lessons can finally
take effect. If the fallback skill is overridden to a non-eligible
skill but the bot has a placeable block, falls back to pillar-up.
Outcomes feed reportAdviceOutcome so confidence stays grounded.
4. recordPOI("danger") on death (runtime/coach/postmortem.js). Spatial
memory now flags where the bot died, expires after 6h. POI table
was empty in rc.2.
5. SAFE_OVERRIDES extended (runtime/coach/advice.js): adds
survive.pillar-up and village.choose-base so coach lessons can
route there.
Tests: 255/255 green (+9 pillar-up).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
116 lines
4.2 KiB
JavaScript
116 lines
4.2 KiB
JavaScript
// coach/advice.js — turn knowledge.lessons into actionable dispatch overrides.
|
|
//
|
|
// The reflex chain calls consult() right before it would dispatch a
|
|
// planned skill. If a high-confidence lesson in the knowledge DB says
|
|
// "avoid that skill in this situation", we either swap in the lesson's
|
|
// preferred alternative or back off (which the curriculum reflex
|
|
// translates into wander / cooldown).
|
|
//
|
|
// This is the closing of the learning loop: post-mortem → lesson →
|
|
// recall → behavioural change. Without this, the DB is just a log.
|
|
|
|
import { isAvailable as knowledgeAvailable, topAdvice, markApplied } from "../knowledge/index.js";
|
|
import { info } from "../log.js";
|
|
|
|
// Skills we will not blindly swap into — they require their own
|
|
// preconditions (e.g. survive.flee needs a known threat direction).
|
|
// The dispatcher will still run runSkill on them, which performs the
|
|
// real precondition check.
|
|
const SAFE_OVERRIDES = new Set([
|
|
"survive.flee",
|
|
"survive.sleep",
|
|
"survive.eat",
|
|
"survive.pillar-up",
|
|
"recovery.tunnel-out",
|
|
"explore.far",
|
|
"explore.wander",
|
|
"village.build-shelter",
|
|
"village.choose-base",
|
|
]);
|
|
|
|
// Pi-coach occasionally suggests prefer_skill values that are mode names
|
|
// (from runtime/modes.js) rather than registered skill ids. We translate
|
|
// them to the closest equivalent skill before the SAFE_OVERRIDES check.
|
|
// Unknown values are returned as-is and will fall through to 'avoid'.
|
|
const MODE_TO_SKILL = Object.freeze({
|
|
self_preservation: "survive.flee",
|
|
night_shelter: "survive.sleep",
|
|
hunger: "survive.eat",
|
|
shelter: "village.build-shelter",
|
|
flee: "survive.flee",
|
|
sleep: "survive.sleep",
|
|
eat: "survive.eat",
|
|
tunnel_out: "recovery.tunnel-out",
|
|
"tunnel-out": "recovery.tunnel-out",
|
|
explore: "explore.far",
|
|
wander: "explore.far",
|
|
});
|
|
|
|
function normalisePreferSkill(raw) {
|
|
if (!raw || typeof raw !== "string") return raw;
|
|
if (SAFE_OVERRIDES.has(raw)) return raw;
|
|
const lower = raw.toLowerCase().trim();
|
|
if (MODE_TO_SKILL[lower]) return MODE_TO_SKILL[lower];
|
|
// Pi sometimes writes "survive_flee" or "survive flee"; normalise.
|
|
const dot = lower.replace(/[_\s]+/g, ".");
|
|
if (SAFE_OVERRIDES.has(dot)) return dot;
|
|
if (MODE_TO_SKILL[dot]) return MODE_TO_SKILL[dot];
|
|
return raw;
|
|
}
|
|
|
|
/**
|
|
* consult({ plannedSkillId, snapshot })
|
|
* → { action: 'override'|'avoid'|'proceed', overrideSkillId?, lessonId?, lesson? }
|
|
*
|
|
* 'override' — dispatch overrideSkillId instead of plannedSkillId
|
|
* 'avoid' — don't dispatch plannedSkillId; caller falls back to wander/idle
|
|
* 'proceed' — no high-confidence lesson applies; dispatch as planned
|
|
*/
|
|
export function consult({ plannedSkillId, snapshot } = {}) {
|
|
if (!knowledgeAvailable()) return PROCEED;
|
|
if (!plannedSkillId) return PROCEED;
|
|
const hostile = snapshot?.closestHostile?.name ?? snapshot?.threats?.[0]?.name ?? null;
|
|
const situation = snapshot?.situationHash ?? null;
|
|
const advice = topAdvice({
|
|
skill: plannedSkillId,
|
|
hostile,
|
|
situation,
|
|
});
|
|
if (!advice.lessonId) return PROCEED;
|
|
|
|
// avoid_skill matches?
|
|
if (advice.avoid && advice.avoid === plannedSkillId) {
|
|
const normalisedPrefer = normalisePreferSkill(advice.prefer);
|
|
if (normalisedPrefer && SAFE_OVERRIDES.has(normalisedPrefer)) {
|
|
if (normalisedPrefer !== advice.prefer) {
|
|
info("coach", `advice: normalised prefer "${advice.prefer}" → "${normalisedPrefer}"`);
|
|
}
|
|
info("coach", `advice: override ${plannedSkillId} → ${normalisedPrefer} (lesson #${advice.lessonId})`);
|
|
return {
|
|
action: "override",
|
|
overrideSkillId: normalisedPrefer,
|
|
lessonId: advice.lessonId,
|
|
lesson: advice.lesson,
|
|
};
|
|
}
|
|
info("coach", `advice: avoid ${plannedSkillId} (lesson #${advice.lessonId})`);
|
|
return { action: "avoid", lessonId: advice.lessonId, lesson: advice.lesson };
|
|
}
|
|
return PROCEED;
|
|
}
|
|
|
|
/**
|
|
* After the dispatcher runs the (possibly overridden) skill, call this
|
|
* with the lesson id and whether the outcome was good. Increments the
|
|
* lesson's applied/succeeded counters and nudges its confidence.
|
|
*/
|
|
export function reportOutcome({ lessonId, succeeded }) {
|
|
if (!lessonId) return;
|
|
markApplied(lessonId, { succeeded: !!succeeded });
|
|
}
|
|
|
|
const PROCEED = Object.freeze({ action: "proceed", lessonId: null, lesson: null });
|
|
|
|
// Test exports
|
|
export const __testing = { SAFE_OVERRIDES, MODE_TO_SKILL, normalisePreferSkill };
|