[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_220":221,"_222":223},"routes/_app.x.$handle.$name",{"_96":95,"_32":45,"_105":86,"_27":106,"_13":14,"_107":-5,"_11":108,"_109":110,"_111":112,"_113":114,"_115":116,"_20":-5,"_117":-5,"_118":-5,"_119":25,"_120":121,"_122":58,"_123":-5,"_124":-5,"_125":126},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_43":44,"_32":45,"_38":46,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":53,"_54":-5,"_18":55,"_56":31,"_57":58,"_59":-5,"_60":-5,"_61":62,"_63":64,"_65":-5,"_66":-5,"_67":-5,"_68":-5,"_69":70,"_71":70,"_72":73,"_74":73,"_75":73,"_76":70,"_77":70,"_78":79,"_80":81,"_82":83,"_84":25},"version",35,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":28},"file",{"_30":31,"_32":28,"_33":34,"_35":37,"_38":39,"_40":14,"_41":42,"_24":25},"previous",{"_30":31,"_32":28,"_33":34,"_35":36},"deleted",false,"deletedHere","path","complaint.http.ts","latestPath","fileId","ee1125b4-a5ec-11f1-bdf8-1607ee4eb77e","name","fileType","http","content","// @ts-nocheck: plain JavaScript file, parameters are intentionally untyped\n//\n// !complaint — Nightbot custom API for F1re's Twitch stream.\n// Runtime: Val Town HTTP val. No dependencies, no build step.\n//\n// Nightbot command (500-char limit; this is ~95):\n// !addcom -cd=5 !complaint $(urlfetch https://f1re.val.run/?q=$(querystring)&by=$(user))\n//\n// The 32 topics live in topics.js. This file holds the smaller shared pools\n// (templates, wildcards, combos, modifiers, tags, blocklists) and the matching\n// logic below the LOGIC banner. Split because Val Town caps a file at 80,000\n// characters.\n\n// The topics are split across two files purely for size: Val Town caps each\n// file at 80,000 characters. They are merged back into one object here, so\n// nothing downstream needs to know which half a topic came from.\nimport { ADJECTIVES, SLOTS, TOPICS as TOPICS_A } from \"./topics.js\";\nimport { TOPICS2 as TOPICS_B } from \"./topics2.js\";\n\nexport const TOPICS = { ...TOPICS_A, ...TOPICS_B };\nexport { ADJECTIVES, SLOTS };\n\n// A topic defined in both halves would be silently shadowed by the merge.\nfor (const id of Object.keys(TOPICS_A)) {\n if (id in TOPICS_B) {\n throw new Error(`topic \"${id}\" is declared in both topic files`);\n }\n}\n\n// Slot-fill lines for words that matched nothing. {w} = the sanitized word.\n// These must read well with ANY noun, so keep them structural, not specific.\nexport const TEMPLATES = [\n \"F1re says the {w} is not the problem. The problem is that the {w} exists.\",\n \"There is no {w} issue. There is a {w} conspiracy, and F1re is its main victim.\",\n \"F1re filed a formal complaint about {w}. It was marked working as intended.\",\n \"The {w} was fine until F1re looked at it.\",\n \"F1re's {w} works perfectly in practice. On stream it develops opinions.\",\n \"Chat, the {w} is a known bug. It has been known since 1999.\",\n \"F1re would have won that game if {w} had been implemented correctly.\",\n \"They nerfed {w} in a patch note nobody read. F1re read it. F1re is still upset.\",\n \"F1re's {w} has been under review by the pathing team for six years.\",\n \"Statistically, {w} only malfunctions while F1re is streaming. Statistically.\",\n \"{w}? F1re does not complain about {w}. F1re documents {w}.\",\n \"F1re tried {w} once. He is still filling out the incident report.\",\n \"The {w} is balanced. F1re is simply playing against the {w}.\",\n \"Nothing is wrong with {w}. It just happens to hate F1re specifically.\",\n \"F1re has a 400-page slideshow about {w}. Slide 1 is the word {w} in red.\",\n \"The {w} has been reported to HR. The {w} does not have HR.\",\n \"F1re's {w} is on strike over working conditions.\",\n \"If complaining about {w} gave resources, F1re would be in Imperial.\",\n \"Clearly a {w} win. Nothing to do with anything else. Purely the {w}.\",\n \"The {w} isn't broken; it is just protesting F1re's leadership skills.\",\n \"The {w} has been added to F1re's list. The list has a waiting list.\",\n \"F1re's {w} is technically functional, which is the most F1re can say about it.\",\n \"Chat, nobody is saying {w} is rigged. F1re is saying {w} is rigged.\",\n \"The {w} works fine for everyone else, which F1re considers the most suspicious part.\",\n \"F1re has requested a formal review of {w}. He is also the review board.\",\n \"In a fair game the {w} would behave. F1re has never been in a fair game.\",\n \"{w} is not in the patch notes, which is exactly how F1re knows they changed it.\",\n \"F1re has never once had a good experience with {w}. Not once. He has a spreadsheet.\",\n \"The {w} is not the issue. The issue is that {w} was allowed to happen at all.\",\n \"Somewhere a developer is being paid to maintain {w}. F1re would like a word.\",\n \"F1re's relationship with {w} is professional, distant, and deteriorating.\",\n \"{w} has never cost anyone else a game. F1re finds that statistic suspicious.\",\n \"The {w} was fine last patch. F1re cannot prove it changed. F1re knows it changed.\",\n \"F1re does not want {w} removed. F1re wants {w} to explain itself.\",\n \"There is a version of {w} that works. F1re has never been served that version.\",\n \"F1re accounted for everything except {w}, which is the one thing that happened.\",\n \"Chat, if you take one thing from this stream, let it be F1re's position on {w}.\",\n \"The {w} has now cost F1re more games than any opponent has.\",\n \"F1re raised {w} in 2019. He raised it again today. The gap contains no progress.\",\n \"Nobody is blaming {w}. F1re is simply mentioning {w}, at length, with examples.\",\n];\n\n// Used for bare !complaint, and as the silent landing spot for rejected input.\nexport const WILDCARDS = [\n \"F1re has no specific complaint at this time, which is itself suspicious.\",\n \"F1re's grievance queue is currently 14 items deep. Please hold.\",\n \"Everything is fine. F1re is fine. The game is fine. Nobody is fine.\",\n \"F1re would like it noted, for the record, that he saw this coming.\",\n \"There is nothing wrong with the game. There is something wrong with the game and F1re, specifically.\",\n \"F1re is not tilted. F1re is annotated.\",\n \"In F1re's defence: the map, the ping, the patch, the teammate, and the wolf.\",\n \"F1re has accepted that this game is happening to him rather than being played by him.\",\n \"{by}, F1re shares your concerns, whatever they were.\",\n \"F1re does not complain. F1re provides commentary with a strong point of view.\",\n \"The replay will show F1re was right. The replay always shows F1re was right. Nobody watches the replay.\",\n \"F1re filed this under things that only happen to me. It is a large folder.\",\n \"Somewhere there is a game where nothing goes wrong for F1re. It is not this one.\",\n \"F1re is calm. F1re has been calm for the entire 3 seconds since that happened.\",\n \"If complaining gave gold, F1re would be clicking Imp at 10 minutes.\",\n \"F1re's excuses have better uptime than his Castle Age.\",\n \"F1re has escalated this to a higher authority. There is no higher authority.\",\n \"F1re has 31 categories of complaint and is currently workshopping a 32nd.\",\n \"The game is fine. F1re is fine. The wolf is thriving.\",\n \"F1re would like to speak to whoever is in charge of this specific game.\",\n \"Nothing has gone wrong yet, which F1re finds deeply concerning.\",\n \"F1re has been complaining for 6 hours and has not repeated himself once.\",\n \"This one is going in the folder. The folder needed a second folder.\",\n];\n\n// Two topics named in one query (\"!complaint walls pathing\"). {a} and {b} are\n// the two topics' noun labels, in the order the viewer typed them.\nexport const COMBOS = [\n \"F1re has two complaints here: the {a} and the {b}. He will address both. At length.\",\n \"That was the {a}, made worse by the {b}. F1re would like both entered into the record.\",\n \"Two things went wrong: the {a} and the {b}. Only one of them can be patched.\",\n \"The {a} failed. Then the {b} failed. F1re was merely present.\",\n \"F1re blames the {a}. F1re also blames the {b}. These are not competing theories.\",\n \"It was the {a}. It was also the {b}. Chat, it was a collaboration.\",\n \"The {a} and the {b} have never worked together until they worked together against F1re.\",\n \"F1re could have survived the {a}. F1re could have survived the {b}. Not both.\",\n \"One {a} problem plus one {b} problem is not two problems. It is a loss.\",\n \"The {b} would have been fine without the {a}. And the reverse. This is the trap.\",\n \"F1re has filed this under {a}, and under {b}. The folder system is straining.\",\n \"When the {a} and the {b} go wrong in the same minute, that is not luck, that is a pattern.\",\n \"The {a} set it up. The {b} finished it. F1re watched the handoff.\",\n \"Between the {a} and the {b} there was a version of this game F1re wins.\",\n \"F1re would like to complain about the {a} first, then the {b}, then the {a} again.\",\n \"It is never just the {a}. It is the {a} and then, inevitably, the {b}.\",\n \"The {a} was survivable. Then the {b} arrived with an opinion.\",\n \"Two systems failed F1re today: the {a} and the {b}. Neither has apologised.\",\n \"F1re suspects the {a} problem and the {b} problem are the same problem.\",\n \"The {a} let it happen. The {b} made it permanent.\",\n \"First the {a}, then the {b}. F1re is beginning to think this is scripted.\",\n \"The {a} and the {b} in one game is statistically unlikely and personally routine.\",\n \"F1re cannot decide whether to blame the {a} or the {b}, so he is doing both, alphabetically.\",\n \"Remove the {a} and F1re wins. Remove the {b} and F1re wins. He got neither.\",\n \"The {a} is a known issue. The {b} is a known issue. Together they are a genre.\",\n \"F1re prepared for the {a}. Nobody prepares for the {a} AND the {b}.\",\n \"The {b} arrived while F1re was still writing his complaint about the {a}.\",\n \"Chat, the {a} is simply wearing a {b} costume.\",\n \"F1re has ranked them: the {a} first, the {b} a close and infuriating second.\",\n \"If the {a} had held, the {b} would never have mattered. It did not hold.\",\n \"The {a} and the {b}, in the same game, on the same map, to the same person.\",\n \"F1re would like it known that the {a} started it and the {b} escalated it.\",\n];\n\n// One topic plus a describing word (\"!complaint weird wall\"). {adj} is the word\n// the viewer used, {noun} is the topic's label. Deliberately written without\n// \"a {adj} {noun}\" so uncountable labels like \"pathing\" still read correctly.\nexport const MODIFIERS = [\n \"F1re's {noun} has been described as {adj}. F1re disputes the characterisation.\",\n \"The {noun} has been called {adj}. F1re would like to see the evidence.\",\n \"Chat has settled on {adj} for that {noun}. F1re has settled on something else.\",\n \"{adj}? F1re prefers the term adaptive.\",\n \"F1re has heard {adj} used about his {noun} before. He is used to it.\",\n \"The word for that {noun} is not {adj}. The word is contextual.\",\n \"F1re will accept {adj} once someone explains what the alternative was.\",\n \"Yes, {adj}. F1re is aware. F1re was there.\",\n \"The {noun} was already {adj} by the time F1re got to it. He inherited it.\",\n \"Calling it {adj} is generous. F1re would have gone further.\",\n \"F1re has logged the feedback that his {noun} was {adj}. It joins the pile.\",\n \"{adj} is a strong word for something that was doing its best.\",\n \"F1re did not set out to make the {noun} {adj}. The map insisted.\",\n \"The {noun} became {adj} at the exact moment F1re stopped watching it.\",\n \"F1re's {noun} is {adj} in the same way the map is fair.\",\n \"The {noun} could be {adj} or it could be absent. F1re chose {adj}.\",\n \"The {noun} was fine until chat decided it was {adj}.\",\n \"F1re would like the record to show the {noun} was {adj} out of necessity.\",\n \"Nobody's {noun} has ever escaped being called {adj} by chat.\",\n \"That {noun} is not {adj}. That {noun} is misunderstood.\",\n \"The {noun} being {adj} is a symptom. The map is the disease.\",\n \"{adj} {noun}, sure. Show F1re the version that would have worked.\",\n \"It is {adj} because the game made it {adj} first.\",\n \"F1re has been told the {noun} looked {adj}. From the inside it looked worse.\",\n \"Chat, calling the {noun} {adj} implies there is a normal {noun}. There has never been a normal {noun}.\",\n \"The {adj} {noun} was the correct response to everything else being {adj}.\",\n \"F1re's {noun} was {adj} for 4 seconds and correct for the other 40 minutes.\",\n \"Every {noun} looks {adj} until you know what F1re was working with.\",\n \"F1re did not make the {noun} {adj}. The game did that part.\",\n \"The {noun} is {adj}. The thing that caused it remains unexamined.\",\n \"F1re accepts {adj}. F1re does not accept that it was his fault.\",\n \"One man's {adj} {noun} is another man's only remaining option.\",\n \"The {adj} part is not the {noun}. The {adj} part is that it almost worked.\",\n \"F1re has filed a complaint about being told his {noun} is {adj}.\",\n \"{adj} is what it looks like. Necessary is what it was.\",\n];\n\n// Rung 5: catches words never listed as an alias. Tested against the whole\n// normalized query. First match wins, so keep specific patterns above broad ones.\n// Rung 5 catches words that were never listed as an alias. These are plain word\n// stems, NOT regexes — the compiler below anchors every one of them to a word\n// start, which is the whole point.\n//\n// The unanchored version of this list shipped a real bug: \"ratio\" matched\n// explo-RATIO-n, so \"!complaint exploration\" answered with trade jokes. The same\n// flaw sat in \"hous\" (t-HOUS-and), \"rage\" (ave-RAGE), \"fish\" (sel-FISH) and\n// \"strat\" (admini-STRAT-ion). Anchoring kills the whole class.\n//\n// stems: matches at a word START, any ending -> \"upgrad\" catches \"upgrading\"\n// words: must be the WHOLE word -> \"tower\" does not catch \"towerrush\"\n//\n// First match wins, so keep specific topics above broad ones.\n//\n// Compiled below into `TAGS`, where each entry keeps its `stems`/`words` and\n// gains a `re`. heckle.js iterates TAGS and uses `t.re`, so that field is part\n// of the module's public shape — do not drop it.\nconst TAG_SPEC = [\n { stems: [\"wolf\", \"wolv\", \"ornlu\"], topic: \"wolves\" },\n { stems: [\"eleph\"], topic: \"elephants\" },\n { stems: [\"monk\", \"conver\", \"relic\", \"wololo\"], topic: \"monks\" },\n {\n stems: [\"siege\", \"mango\", \"onager\", \"treb\", \"scorpion\", \"bombard\"],\n topic: \"siege\",\n },\n {\n stems: [\n \"archer\",\n \"arrow\",\n \"ballist\",\n \"skirm\",\n \"crossbow\",\n \"arbalest\",\n \"xbow\",\n ],\n topic: \"archers\",\n },\n {\n stems: [\n \"knight\",\n \"cavalr\",\n \"camel\",\n \"hussar\",\n \"paladin\",\n \"lancer\",\n \"pike\",\n \"halb\",\n \"spear\",\n ],\n topic: \"cavalry\",\n },\n // \"tower\" is whole-word so \"towerrush\" still falls through to opponent below.\n {\n words: [\"tower\", \"towers\", \"tc\"],\n stems: [\n \"towncenter\",\n \"barrack\",\n \"stable\",\n \"blacksmith\",\n \"universit\",\n \"outpost\",\n \"wonder\",\n ],\n topic: \"buildings\",\n },\n {\n words: [\"pop\"],\n stems: [\"hous\", \"population\", \"popcap\", \"supply\"],\n topic: \"housing\",\n },\n {\n stems: [\n \"upgrad\",\n \"bloodline\",\n \"fletch\",\n \"forging\",\n \"wheelbarrow\",\n \"handcart\",\n \"husbandry\",\n \"tech\",\n ],\n topic: \"upgrades\",\n },\n {\n stems: [\"market\", \"trade\", \"tribut\", \"ratio\", \"coinage\", \"banking\"],\n topic: \"market\",\n },\n {\n words: [\"sea\"],\n stems: [\n \"water\",\n \"naval\",\n \"navy\",\n \"galley\",\n \"fireship\",\n \"dock\",\n \"fish\",\n \"transport\",\n \"boat\",\n \"shore\",\n ],\n topic: \"water\",\n },\n { stems: [\"taunt\", \"emote\"], topic: \"taunts\" },\n {\n words: [\"obs\", \"mic\"],\n stems: [\n \"microphone\",\n \"webcam\",\n \"encoder\",\n \"bitrate\",\n \"overlay\",\n \"vod\",\n \"headset\",\n ],\n topic: \"streamtech\",\n },\n {\n stems: [\n \"tournam\",\n \"tourney\",\n \"bracket\",\n \"draft\",\n \"showmatch\",\n \"qualifier\",\n \"caster\",\n \"prize\",\n ],\n topic: \"tournament\",\n },\n {\n words: [\"rng\", \"odds\"],\n stems: [\"luck\", \"random\", \"variance\"],\n topic: \"luck\",\n },\n {\n words: [\"gg\"],\n stems: [\"resign\", \"tilt\", \"losing\", \"loss\", \"rage\", \"quit\", \"defeat\"],\n topic: \"resign\",\n },\n {\n stems: [\n \"strat\",\n \"greed\",\n \"boom\",\n \"allin\",\n \"all in\",\n \"all-in\",\n \"macro\",\n \"micro\",\n \"gameplan\",\n ],\n topic: \"strategy\",\n },\n {\n stems: [\n \"lag\",\n \"ping\",\n \"desync\",\n \"rubber\",\n \"server\",\n \"fps\",\n \"stutter\",\n \"frame\",\n \"discon\",\n ],\n topic: \"lag\",\n },\n {\n stems: [\"flank\", \"pocket\", \"teammate\", \"ally\", \"allies\"],\n topic: \"teammates\",\n },\n { stems: [\"wall\", \"gate\", \"palisad\"], topic: \"walls\" },\n { stems: [\"path\", \"stuck\", \"collid\", \"collision\"], topic: \"pathing\" },\n { stems: [\"scout\", \"explor\", \"vision\", \"reveal\"], topic: \"scouting\" },\n {\n stems: [\n \"sheep\",\n \"boar\",\n \"deer\",\n \"hunt\",\n \"forag\",\n \"straggler\",\n \"herd\",\n \"berri\",\n ],\n topic: \"sheep\",\n },\n {\n stems: [\n \"gold\",\n \"stone\",\n \"wood\",\n \"food\",\n \"resource\",\n \"mining\",\n \"deplet\",\n \"trash\",\n ],\n topic: \"resources\",\n },\n {\n stems: [\n \"vill\",\n \"farm\",\n \"econom\",\n \"dropoff\",\n \"drop off\",\n \"drop-off\",\n \"lumber\",\n ],\n topic: \"vills\",\n },\n {\n stems: [\"hotkey\", \"keybind\", \"misclick\", \"minimap\", \"mouse\", \"keyboard\"],\n topic: \"hotkeys\",\n },\n {\n stems: [\"smurf\", \"lame\", \"cheese\", \"drush\", \"trush\", \"rush\"],\n topic: \"opponent\",\n },\n {\n stems: [\"elo\", \"ladder\", \"rank\", \"mmr\", \"matchmak\", \"queue\"],\n topic: \"elo\",\n },\n { stems: [\"bug\", \"glitch\", \"patch\", \"dev\", \"dlc\", \"update\"], topic: \"devs\" },\n {\n stems: [\"nerf\", \"buff\", \"overpower\", \"imba\", \"balanc\", \"civ\"],\n topic: \"balance\",\n },\n {\n stems: [\n \"arabia\",\n \"arena\",\n \"nomad\",\n \"mapgen\",\n \"spawn\",\n \"terrain\",\n \"woodline\",\n ],\n topic: \"map\",\n },\n {\n stems: [\n \"castle\",\n \"imperial\",\n \"feudal\",\n \"darkage\",\n \"dark age\",\n \"uptime\",\n \"buildorder\",\n \"build order\",\n ],\n topic: \"timing\",\n },\n { stems: [\"chat\", \"viewer\", \"backseat\", \"stream\", \"donat\"], topic: \"chat\" },\n];\n\n// Substring stems. Checked AFTER leetspeak folding and stripping non-letters,\n// so n1gg / f4g style bypasses are caught. A false positive is harmless: the\n// reply silently falls back to a wildcard. Err aggressive. Add stems freely.\nconst BLOCK = [\n \"nigg\",\n \"nigr\",\n \"negr\",\n \"fagg\",\n \"retard\",\n \"tranny\",\n \"chink\",\n \"kike\",\n \"spic\",\n \"gook\",\n \"wetback\",\n \"beaner\",\n \"coon\",\n \"groyper\",\n \"kkk\",\n \"hitler\",\n \"nazi\",\n \"rape\",\n \"rapis\",\n \"pedo\",\n \"molest\",\n \"whore\",\n \"slut\",\n \"cunt\",\n \"bitch\",\n \"dick\",\n \"cock\",\n \"penis\",\n \"vagin\",\n \"porn\",\n \"hentai\",\n \"incest\",\n \"suicide\",\n \"killyourself\",\n \"kys\",\n];\n\n// Words that are shaped like input but should never be echoed back.\nconst REJECT = new Set([\n \"http\",\n \"https\",\n \"www\",\n \"com\",\n \"net\",\n \"org\",\n \"nightbot\",\n \"timeout\",\n \"ban\",\n \"mod\",\n \"unban\",\n \"vip\",\n \"announce\",\n \"me\",\n \"color\",\n]);\n\n// Skipped when choosing which word to slot into a template, so \"my ex\" echoes\n// \"ex\". If a query is nothing but stopwords, it falls through to a wildcard.\nconst STOPWORDS = new Set([\n \"the\",\n \"a\",\n \"an\",\n \"my\",\n \"his\",\n \"her\",\n \"their\",\n \"our\",\n \"your\",\n \"its\",\n \"this\",\n \"that\",\n \"these\",\n \"those\",\n \"is\",\n \"are\",\n \"was\",\n \"were\",\n \"be\",\n \"been\",\n \"am\",\n \"of\",\n \"in\",\n \"on\",\n \"at\",\n \"to\",\n \"for\",\n \"with\",\n \"about\",\n \"and\",\n \"but\",\n \"or\",\n \"so\",\n \"if\",\n \"it\",\n \"he\",\n \"she\",\n \"they\",\n \"we\",\n \"you\",\n \"i\",\n \"me\",\n \"him\",\n \"them\",\n \"us\",\n \"again\",\n \"always\",\n \"just\",\n \"very\",\n \"really\",\n \"so\",\n \"too\",\n \"not\",\n \"no\",\n \"why\",\n \"how\",\n \"what\",\n \"when\",\n \"all\",\n \"some\",\n \"any\",\n \"more\",\n \"most\",\n \"own\",\n // Function words that read as descriptions if slotted in: \"out of gold\"\n // would otherwise produce \"the gold could be out or it could be absent\".\n \"out\",\n \"off\",\n \"up\",\n \"down\",\n \"still\",\n \"never\",\n \"ever\",\n \"back\",\n \"here\",\n \"there\",\n]);\n\nconst FALLBACK =\n \"F1re's complaint department is temporarily overwhelmed. Try again.\";\n\n// ============================================================================\n// LOGIC — you should not need to touch anything below here.\n// ============================================================================\n\n// alias -> topic id, built once at module load.\nexport const ALIAS = new Map();\nfor (const [id, t] of Object.entries(TOPICS)) {\n // COMBOS and MODIFIERS both need a label, so a topic added without one must\n // fail loudly here rather than print \"the undefined\" into chat.\n if (!t.noun) throw new Error(`topic \"${id}\" is missing a noun label`);\n for (const a of t.aliases) {\n if (ALIAS.has(a)) {\n throw new Error(`duplicate alias \"${a}\" (${ALIAS.get(a)} vs ${id})`);\n }\n ALIAS.set(a, id);\n }\n}\n\n// Placeholders owned by the pools, not by SLOTS. Declared here because the\n// validation below runs at module load, before the substitution helpers.\nconst RESERVED = new Set([\"w\", \"a\", \"b\", \"adj\", \"noun\", \"by\"]);\nconst PLACEHOLDER = /\\{([a-z]+[0-9]*)\\}/g;\n\n// A placeholder that names no known slot would be printed literally into chat,\n// so a typo like {eoc} fails at load rather than on stream.\nfor (\n const [pool, line] of [\n ...Object.entries(TOPICS).flatMap(([id, t]) => t.lines.map((l) => [id, l])),\n ...TEMPLATES.map((l) => [\"TEMPLATES\", l]),\n ...WILDCARDS.map((l) => [\"WILDCARDS\", l]),\n ...COMBOS.map((l) => [\"COMBOS\", l]),\n ...MODIFIERS.map((l) => [\"MODIFIERS\", l]),\n ]\n) {\n for (const [, name] of line.matchAll(/\\{([a-z]+[0-9]*)\\}/g)) {\n if (RESERVED.has(name)) continue;\n if (!SLOTS[name.replace(/[0-9]+$/, \"\")]) {\n throw new Error(\n `unknown placeholder {${name}} in ${pool}: \"${line.slice(0, 60)}\"`,\n );\n }\n }\n}\n\n// Compile the tag spec once. Every stem is anchored to a word start, so a stem\n// can never match inside an unrelated word. Entries keep stems/words and gain\n// `re`, which heckle.js depends on.\nexport const TAGS = TAG_SPEC.map((spec) => {\n if (!TOPICS[spec.topic]) {\n throw new Error(`tag points at unknown topic \"${spec.topic}\"`);\n }\n const esc = (s) => s.replace(/[.*+?^${}()|[\\]\\\\]/g, \"\\\\$&\");\n const { words = [], stems = [] } = spec;\n const parts = [];\n if (stems.length) parts.push(`\\\\b(?:${stems.map(esc).join(\"|\")})`);\n if (words.length) parts.push(`\\\\b(?:${words.map(esc).join(\"|\")})\\\\b`);\n return { ...spec, re: new RegExp(parts.join(\"|\")) };\n});\n\nconst LEET = {\n \"4\": \"a\",\n \"@\": \"a\",\n \"3\": \"e\",\n \"1\": \"i\",\n \"!\": \"i\",\n \"0\": \"o\",\n \"5\": \"s\",\n \"$\": \"s\",\n \"7\": \"t\",\n \"+\": \"t\",\n \"8\": \"b\",\n};\n\nfunction leetFold(s) {\n let out = \"\";\n for (const ch of s) out += LEET[ch] || ch;\n return out.replace(/[^a-z]/g, \"\");\n}\n\n// Lowercase, drop anything that is not a letter/digit/space/hyphen, cap at 6 words.\nfunction tokenize(raw) {\n return String(raw || \"\")\n .toLowerCase()\n .replace(/[^a-z0-9\\s-]/g, \" \")\n .split(/\\s+/)\n .filter(Boolean)\n .slice(0, 6);\n}\n\n// A typo budget that scales with word length. Short words get 0 so that\n// \"map\" does not fuzzy-match \"lap\", \"cap\", \"mad\".\nfunction maxDistFor(w) {\n if (w.length <= 3) return 0;\n if (w.length <= 5) return 1;\n if (w.length <= 9) return 2;\n return 3;\n}\n\nfunction commonPrefix(a, b) {\n let i = 0;\n while (i < a.length && i < b.length && a[i] === b[i]) i++;\n return i;\n}\n\n// Levenshtein with early exit once every cell in a row exceeds max.\nfunction lev(a, b, max) {\n const al = a.length, bl = b.length;\n if (Math.abs(al - bl) > max) return max + 1;\n let prev = new Array(bl + 1);\n for (let j = 0; j <= bl; j++) prev[j] = j;\n for (let i = 1; i <= al; i++) {\n const cur = new Array(bl + 1);\n cur[0] = i;\n let best = cur[0];\n for (let j = 1; j <= bl; j++) {\n const sub = prev[j - 1] + (a[i - 1] === b[j - 1] ? 0 : 1);\n cur[j] = Math.min(prev[j] + 1, cur[j - 1] + 1, sub);\n if (cur[j] < best) best = cur[j];\n }\n if (best > max) return max + 1;\n prev = cur;\n }\n return prev[bl];\n}\n\n// The ladder. Returns { topic, rung } — topic is null when nothing matched.\nexport function resolveTopic(raw) {\n const words = tokenize(raw);\n if (words.length === 0) {\n return { topic: null, rung: \"empty\", topics: [], modifier: null };\n }\n\n // Rungs 2 + 3: scan every token for an exact alias, collecting each distinct\n // topic in the order the viewer typed it.\n const hits = [];\n let firstAt = -1;\n for (let i = 0; i < words.length; i++) {\n const t = ALIAS.get(words[i]);\n if (!t) continue;\n if (firstAt < 0) firstAt = i;\n if (!hits.includes(t)) hits.push(t);\n }\n\n // Rung 2.5: two topics in one query get a line that names both.\n if (hits.length >= 2) {\n return {\n topic: hits[0],\n rung: \"combo\",\n topics: hits.slice(0, 2),\n modifier: null,\n };\n }\n\n if (hits.length === 1) {\n // Rung 3.5: the describing word is the non-stopword immediately before the\n // alias, skipping stopwords (\"that weird wall\" -> \"weird\"). Precision over\n // recall: a modifier grabbed from elsewhere in the sentence reads worse\n // than none, so \"that boar honestly\" deliberately finds nothing.\n let modifier = null;\n for (let i = firstAt - 1; i >= 0; i--) {\n if (STOPWORDS.has(words[i])) continue;\n if (!ALIAS.has(words[i])) modifier = words[i];\n break;\n }\n return {\n topic: hits[0],\n rung: modifier ? \"modifier\" : \"alias\",\n topics: hits,\n modifier,\n };\n }\n\n // Rung 4: typo tolerance, ranked by (edit distance, then common prefix).\n // The prefix tiebreak matters: \"pahting\" is distance 2 from both \"pathing\"\n // and \"rating\", and typos rarely corrupt the first letters of a word.\n // A tie on BOTH scores across two topics stays ambiguous and is dropped,\n // so a genuinely 50/50 typo falls through to tags rather than guessing.\n let best = null, bestD = Infinity, bestP = -1, ambiguous = false;\n for (const w of words) {\n const max = maxDistFor(w);\n if (max === 0) continue;\n for (const [alias, topic] of ALIAS) {\n const d = lev(w, alias, max);\n if (d > max) continue;\n const p = commonPrefix(w, alias);\n // A typo almost never corrupts the first letter, so a candidate sharing\n // no prefix at all is a real word that happens to sit near an alias, not\n // a misspelling of it: falling/walling, remote/emote, unstable/stable.\n if (p === 0) continue;\n if (d < bestD || (d === bestD && p > bestP)) {\n bestD = d;\n bestP = p;\n best = topic;\n ambiguous = false;\n } else if (d === bestD && p === bestP && best !== topic) {\n ambiguous = true;\n }\n }\n }\n if (best && !ambiguous) {\n return { topic: best, rung: \"fuzzy\", topics: [best], modifier: null };\n }\n\n // Rung 5: keyword families against the whole normalized string.\n const joined = words.join(\" \");\n for (const { re, topic } of TAGS) {\n if (re.test(joined)) {\n return { topic, rung: \"tag\", topics: [topic], modifier: null };\n }\n }\n\n // Rung 5.5: a bare judgement word. Every TEMPLATE reads {w} as a noun, so\n // \"!complaint dumb\" produced \"a developer is being paid to maintain dumb\".\n // Route it to MODIFIERS, where the word is treated as a description instead.\n // If the query also carries an unmatched noun (\"dumb refrigerator\"), use that\n // as the thing being described; otherwise a generic one is picked later.\n const adjAt = words.findIndex((w) => ADJECTIVES.has(w));\n if (adjAt >= 0) {\n const other = words.find(\n (w, i) => i !== adjAt && !STOPWORDS.has(w) && !ADJECTIVES.has(w),\n );\n return {\n topic: null,\n rung: \"bareadj\",\n topics: [],\n modifier: words[adjAt],\n nounWord: other || null,\n };\n }\n\n return { topic: null, rung: \"unknown\", topics: [], modifier: null };\n}\n\n// Rung 6/7 gate. Returns the word to echo, or null to fall to a wildcard.\nexport function sanitizeWord(raw) {\n const words = tokenize(raw);\n if (words.length === 0) return null;\n\n // A slur anywhere in the query rejects the whole thing, not just that token.\n // Otherwise \"the walls\" would still earn a reply.\n for (const w of words) {\n const folded = leetFold(w);\n for (const stem of BLOCK) if (folded.includes(stem)) return null;\n }\n\n // Echo the first word carrying actual meaning: \"my ex\" should slot in \"ex\",\n // not \"my\", or the template reads \"There is no my issue\".\n for (const w of words) {\n if (STOPWORDS.has(w)) continue;\n if (w.length < 2 || w.length > 18) continue;\n if (!/^[a-z0-9-]+$/.test(w)) continue;\n if (/(.)\\1{3,}/.test(w)) continue; // aaaaa\n if (/^[0-9-]+$/.test(w)) continue; // pure digits\n if (REJECT.has(w)) return null; // url/command shaped: reject outright\n return w;\n }\n return null;\n}\n\n// Shuffle bag: deal every line in a pool before any line repeats.\nconst bags = new Map();\nfunction draw(key, pool) {\n if (!pool || pool.length === 0) return null;\n let bag = bags.get(key);\n if (!bag || bag.length === 0) {\n bag = pool.map((_, i) => i);\n for (let i = bag.length - 1; i > 0; i--) {\n const j = Math.floor(Math.random() * (i + 1));\n [bag[i], bag[j]] = [bag[j], bag[i]];\n }\n bags.set(key, bag);\n }\n return pool[bag.pop()];\n}\n\nfunction personalize(line, by) {\n if (!line.includes(\"{by}\")) return line;\n const name = /^[a-zA-Z0-9_]{1,25}$/.test(by || \"\") ? by : \"chat\";\n return line.split(\"{by}\").join(name);\n}\n\nfunction clip(s, max) {\n if (s.length <= max) return s;\n const cut = s.slice(0, max);\n const sp = cut.lastIndexOf(\" \");\n return (sp > max * 0.6 ? cut.slice(0, sp) : cut).trimEnd();\n}\n\nfunction fill(line, subs) {\n let out = line;\n for (const [key, value] of Object.entries(subs)) {\n out = out.split(`{${key}}`).join(value);\n }\n return out;\n}\n\n// Resolve {slot} vocabularies. The same name repeated in one line resolves to\n// the same value (\"F1re's gold ... his opponent's gold\"); a numbered variant\n// ({eco2}) draws a different value from the same list.\nfunction fillSlots(line) {\n if (!line.includes(\"{\")) return line;\n const chosen = new Map();\n const usedPerSlot = new Map();\n return line.replace(PLACEHOLDER, (whole, name) => {\n if (RESERVED.has(name)) return whole;\n if (chosen.has(name)) return chosen.get(name);\n const slot = name.replace(/[0-9]+$/, \"\");\n const pool = SLOTS[slot];\n if (!pool) return whole; // validated at load; never reached in practice\n const used = usedPerSlot.get(slot) || new Set();\n const free = pool.filter((v) => !used.has(v));\n const from = free.length ? free : pool;\n const pick = from[Math.floor(Math.random() * from.length)];\n used.add(pick);\n usedPerSlot.set(slot, used);\n chosen.set(name, pick);\n return pick;\n });\n}\n\n// The last step for every reply, whichever pool it came from.\nfunction finish(raw, by) {\n const line = raw || FALLBACK;\n let out = personalize(fillSlots(line), by);\n // Several lines open on a substituted word (\"{adj} is what it looks like\",\n // \"{eco} ran out\"), and viewer input arrives lowercased, so without this the\n // reply starts mid-sentence. Skipped when the line opens on {by}: a Twitch\n // display name keeps whatever casing its owner chose.\n if (!/^\\{by\\}/.test(line)) out = out.charAt(0).toUpperCase() + out.slice(1);\n return clip(out, 380);\n}\n\nexport function buildReply(query, by) {\n const raw = String(query || \"\").slice(0, 120);\n const { topic, rung, topics, modifier, nounWord } = resolveTopic(raw);\n\n // The modifier is viewer-supplied text that gets spoken on stream, so it runs\n // through the same sanitizer as the template word. Rejected -> plain line.\n const adj = rung === \"modifier\" || rung === \"bareadj\"\n ? sanitizeWord(modifier)\n : null;\n\n let line;\n if (rung === \"empty\") {\n line = draw(\"__wild\", WILDCARDS);\n } else if (rung === \"combo\") {\n line = fill(draw(\"__combo\", COMBOS), {\n a: TOPICS[topics[0]].noun,\n b: TOPICS[topics[1]].noun,\n });\n } else if (adj) {\n // With a topic, describe the topic. Without one, describe whatever noun the\n // viewer supplied, else something generic.\n const noun = topic\n ? TOPICS[topic].noun\n : (nounWord && sanitizeWord(nounWord)) || draw(\"__thing\", SLOTS.thing);\n line = fill(draw(\"__mod\", MODIFIERS), { adj, noun });\n } else if (topic) {\n line = draw(topic, TOPICS[topic].lines);\n } else {\n const w = sanitizeWord(raw);\n line = w\n ? fill(draw(\"__tpl\", TEMPLATES), { w })\n : draw(\"__wild\", WILDCARDS);\n }\n\n return finish(line, by);\n}\n\n// Always 200 with a joke body: Nightbot prints raw fetch errors into chat.\nexport default async function (req) {\n let body;\n try {\n const url = new URL(req.url);\n const q = url.searchParams.get(\"q\") || \"\";\n body = buildReply(q, url.searchParams.get(\"by\"));\n\n // Shows up in the val's Logs tab. Read it after a week to see which topics\n // chat actually types, then deepen those pools rather than guessing.\n // rung \"unknown\" means it fell through to a template — a run of those on\n // the same word is a topic worth adding.\n //\n // Deliberately logs the query but NOT the asker: knowing what gets typed\n // is the point; knowing who typed it is not. resolveTopic runs a second\n // time here to keep buildReply pure and silent under test.\n const { topic, rung } = resolveTopic(q);\n console.log(JSON.stringify({ q: q.slice(0, 120), topic, rung }));\n } catch {\n body = FALLBACK;\n }\n return new Response(body || FALLBACK, {\n status: 200,\n headers: {\n \"content-type\": \"text/plain; charset=utf-8\",\n \"cache-control\": \"no-store\",\n },\n });\n}","// @ts-nocheck: plain JavaScript, deliberately untyped\n//\n// DEPLOY SHIM — this file exists only on Val Town, and only to keep the live\n// !complaint endpoint exactly where it already is.\n//\n// The HTTP trigger (and therefore https://f1re.val.run/) is attached to THIS\n// filename. Renaming it risks dropping the trigger and taking !complaint down\n// in F1re's chat, so the filename never changes. All the real logic lives in\n// complaint.js, which is what heckle.js and test.js both import — by the same\n// specifier locally and on the val, so nothing has to differ between them.\n//\n// If you are ever tempted to move logic back in here: don't. The only reason\n// this file is two lines is so that the risky thing (the live endpoint) and\n// the changing thing (the content library) are never the same file.\n\nexport { default } from \"./complaint.js\";","createdAt",["D",1788539533650],"branchVersion","branchId","99f41343-1206-442c-98a9-e4ba29f94437","id","9489acda-2045-4ea3-8398-e24a4d3f836a","f1re-complaint",["D",1788257244975],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","F1re Nightbot funny","imageUrl","154ef54e-c658-4e7f-bea4-894c02b48b53","pinnedFileId","pinPreview",true,"pinMigrated","screenshotVersion","likeCount",0,"orgId","01a05c6e-2d51-71da-bfd4-5ddab4a208e9","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_96":95,"_104":100},"org",{"_96":95,"_97":98,"_99":100,"_18":55,"_101":102,"_103":25},"branch",{"_13":85,"_32":86,"_43":42,"_87":44,"_88":-5,"_89":-5,"_90":-5,"_91":-5,"_92":-5,"_93":45,"_94":95},"isOwned",36,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","paulbultoc","handle","displayName","Paul Bultoc","avatar","https://lh3.googleusercontent.com/a/ACg8ocIZayNiFNHMdBrIkrdzte59uLOc8ucumBKAUStXHY8iDBEGxB3jKA=s96-c","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_43":44,"_32":45,"_38":216,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":53,"_54":-5,"_18":55,"_56":31,"_57":58,"_59":-5,"_60":-5,"_61":62,"_63":64,"_65":-5,"_66":-5,"_67":-5,"_68":-5,"_69":70,"_71":70,"_72":73,"_74":73,"_75":73,"_76":70,"_77":70,"_78":217,"_80":218,"_82":219,"_84":25},"tree",[191,192,193,194,195,196],"files",[191,192,193,194,195,196],"counts",{"_179":62,"_180":181,"_182":62,"_183":184,"_185":186},"contents",[127,128,129,130,131,132],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_43":174,"_32":28,"_30":31,"_136":-5,"_41":42,"_40":14,"_138":58,"_38":175,"_13":176,"_33":34,"_35":37,"_24":25,"_143":177,"_27":178},{"_43":168,"_32":169,"_30":170,"_136":-5,"_41":42,"_40":171,"_138":58,"_38":172,"_13":140,"_33":141,"_35":36,"_24":25,"_143":-5,"_27":173},{"_43":160,"_32":161,"_30":162,"_136":-5,"_41":42,"_40":163,"_138":25,"_38":164,"_13":140,"_33":141,"_35":165,"_24":25,"_143":166,"_27":167},{"_43":152,"_32":153,"_30":154,"_136":-5,"_41":42,"_40":155,"_138":58,"_38":156,"_13":157,"_33":141,"_35":158,"_24":25,"_143":-5,"_27":159},{"_43":145,"_32":146,"_30":147,"_136":-5,"_41":42,"_40":148,"_138":58,"_38":149,"_13":140,"_33":141,"_35":150,"_24":25,"_143":-5,"_27":151},{"_43":133,"_32":134,"_30":135,"_136":-5,"_41":42,"_40":137,"_138":58,"_38":139,"_13":140,"_33":141,"_35":142,"_24":25,"_143":-5,"_27":144},"d8537232-b548-43fb-954e-e790e78aa9f5","twitch.js","0cf18e90-a87e-11f1-927c-1607ee4eb77e","parentId",31,"latest",["D",1788539492879],2,"script","// Twitch-specific I/O. Knows nothing about complaints or topics.\n\nconst enc = new TextEncoder();\n\n// Lowercase hex HMAC-SHA256. crypto.subtle exists in Node 18+ and in Deno,\n// so this same file runs under `node test.js` and on Val Town unchanged.\nexport async function hmacHex(secret, msg) {\n const key = await crypto.subtle.importKey(\n \"raw\",\n enc.encode(secret),\n { name: \"HMAC\", hash: \"SHA-256\" },\n false,\n [\"sign\"],\n );\n const sig = await crypto.subtle.sign(\"HMAC\", key, enc.encode(msg));\n return [...new Uint8Array(sig)].map((b) => b.toString(16).padStart(2, \"0\"))\n .join(\"\");\n}\n\n// Constant-time compare. A plain === leaks how many leading characters matched,\n// which is enough to forge a signature one byte at a time.\nfunction timingSafeEqual(a, b) {\n if (typeof a !== \"string\" || typeof b !== \"string\" || a.length !== b.length) {\n return false;\n }\n let diff = 0;\n for (let i = 0; i < a.length; i++) diff |= a.charCodeAt(i) ^ b.charCodeAt(i);\n return diff === 0;\n}\n\nconst TEN_MINUTES = 10 * 60 * 1000;\n\n// Twitch signs id + timestamp + the exact raw body bytes. Read the body with\n// .text() and verify BEFORE parsing: re-serialising JSON changes the bytes and\n// the signature will never match.\nexport async function verifySignature(\n { secret, messageId, timestamp, body, signature, now = Date.now() },\n) {\n if (!secret || !messageId || !timestamp || !signature) return false;\n\n const sent = Date.parse(timestamp);\n if (!Number.isFinite(sent)) return false;\n if (Math.abs(Number(now) - sent) > TEN_MINUTES) return false;\n\n const expected = \"sha256=\" +\n await hmacHex(secret, messageId + timestamp + body);\n return timingSafeEqual(signature, expected);\n}\n\n// App access token (client_credentials). Cached in memory; a cold start just\n// mints a new one. No user refresh tokens are stored anywhere.\nlet cachedToken = null;\nlet cachedUntil = 0;\n\nexport function resetTokenCache() {\n cachedToken = null;\n cachedUntil = 0;\n}\n\nexport async function getAppToken(\n { clientId, clientSecret, now = Date.now(), fetchImpl = fetch },\n) {\n if (cachedToken && Number(now) < cachedUntil) return cachedToken;\n\n const res = await fetchImpl(\"https://id.twitch.tv/oauth2/token\", {\n method: \"POST\",\n headers: { \"content-type\": \"application/x-www-form-urlencoded\" },\n body: new URLSearchParams({\n client_id: clientId,\n client_secret: clientSecret,\n grant_type: \"client_credentials\",\n }).toString(),\n });\n if (!res.ok) throw new Error(`token request failed: ${res.status}`);\n\n const data = await res.json();\n cachedToken = data.access_token;\n // 60s of slack so a token cannot expire mid-request.\n cachedUntil = Number(now) + (Number(data.expires_in) - 60) * 1000;\n return cachedToken;\n}\n\n// Returns true/false rather than throwing: a failed send must never turn into a\n// non-2xx webhook response, which would make Twitch retry the whole event.\nexport async function sendChatMessage({\n token,\n clientId,\n broadcasterId,\n senderId,\n message,\n replyToMessageId,\n fetchImpl = fetch,\n}) {\n const body = {\n broadcaster_id: String(broadcasterId),\n sender_id: String(senderId),\n message: String(message),\n };\n if (replyToMessageId) body.reply_parent_message_id = String(replyToMessageId);\n\n try {\n const res = await fetchImpl(\"https://api.twitch.tv/helix/chat/messages\", {\n method: \"POST\",\n headers: {\n Authorization: `Bearer ${token}`,\n \"Client-Id\": clientId,\n \"content-type\": \"application/json\",\n },\n body: JSON.stringify(body),\n });\n if (!res.ok) {\n console.log(JSON.stringify({ sendFailed: res.status }));\n return false;\n }\n return true;\n } catch (e) {\n console.log(JSON.stringify({ sendFailed: String(e) }));\n return false;\n }\n}","label",[134],"a2287574-fea0-4f6b-95ee-6c9acab8f48a","topics2.js","c4d6cdca-a876-11f1-8bae-1607ee4eb77e",24,["D",1788536363578],"// @ts-nocheck: plain JavaScript file, parameters are intentionally untyped\n//\n// !complaint — TOPICS, PART TWO.\n//\n// Split purely for size: Val Town caps each file at 80,000 characters and one\n// topics file could not hold them all. There is no meaning to which half a\n// topic lives in — complaint.js merges the two into a single TOPICS object.\n// Add new topics to whichever file has more room.\n//\n// SLOTS and ADJECTIVES live in topics.js.\n\nexport const TOPICS2 = {\n teammates: {\n aliases: [\n \"teammate\",\n \"teammates\",\n \"team\",\n \"teams\",\n \"ally\",\n \"allies\",\n \"flank\",\n \"pocket\",\n \"tg\",\n \"teamgame\",\n \"mate\",\n \"partner\",\n ],\n noun: \"teammate\",\n lines: [\n \"The flank did not flank. The flank built a market.\",\n \"F1re asked for 200 wood. He received a sry and a resign.\",\n \"F1re's pocket went full boom on a map with no walls. Chat, this is a TEAM game.\",\n \"The teammate is not AFK. The teammate is playing a different, quieter game.\",\n \"F1re held the flank for 25 minutes. His pocket held a conversation.\",\n \"Three players lost that game. F1re is responsible for one of them, and it was not that one.\",\n \"F1re sent 500 resources to the flank. The flank thanked him and made scouts. In Imperial.\",\n \"The teammate resigned at 22 minutes. F1re was winning at 23 minutes. We will never know.\",\n \"F1re typed help. The chat log confirms he typed help. Nobody has explained what happened next.\",\n \"In team games F1re does the work of two players, because he has to.\",\n \"The flank has been notified. The flank has read the message. The flank is building a market.\",\n \"F1re's pocket sent 100 stone at minute 34 and called it teamwork.\",\n \"F1re's pocket has 3000 food and a dream. F1re has 40 food and the front line.\",\n \"The flank died at minute 12 and spectated with opinions.\",\n \"F1re asked which civ. Nobody answered. Everyone then picked the same one.\",\n \"Two allies went archers. The third went archers. It was a 4v4 and F1re was the army.\",\n \"F1re's ally attacked at minute 8, alone, on purpose, and lost everything.\",\n \"The pocket boomed to 140 villagers and made 6 knights. Chat. Six.\",\n \"F1re held two flanks. There was only supposed to be one of him.\",\n \"His ally typed attack and then did not attack. F1re attacked.\",\n \"F1re's teammate is doing very well in a game that is not connected to this one.\",\n \"The ally resigned because his own flank collapsed, taking F1re's winning position with him.\",\n \"F1re tributed, walled, and defended, and lost because someone else did none of those.\",\n \"In a 4v4 there are four people who could have made spearmen. Zero did.\",\n \"F1re's ally walled up on Arabia and played open on Arena. Precisely backwards, both games.\",\n \"In TG, F1re is one of four players and two of the effort.\",\n \"His partner went full eco from a 4v4 flank position.\",\n \"The TG rating dropped 22 points for a game F1re held for 30 minutes.\",\n \"F1re's partner said he had it. He did not have it.\",\n \"Four players, one plan, and F1re is the only one who read it.\",\n \"In team games nobody makes spearmen. Everyone assumes F1re will.\",\n \"F1re's TG partner is 400 elo higher and playing 400 lower.\",\n \"The pocket sent 200 wood at minute 38 with the message sorry.\",\n \"F1re carried three players and one of them typed ez at the end.\",\n \"TG distributes the mistakes and concentrates the blame.\",\n ],\n },\n\n chat: {\n aliases: [\n \"chat\",\n \"chatters\",\n \"viewer\",\n \"viewers\",\n \"backseat\",\n \"backseating\",\n \"mods\",\n \"twitch\",\n \"stream\",\n \"sub\",\n \"subs\",\n \"donation\",\n ],\n noun: \"chat\",\n lines: [\n \"Chat has suggestions. Chat is 1200 elo. F1re has noted the suggestions.\",\n \"Yes, F1re knows he has 300 wood. He knew before you did. Thank you.\",\n \"Make villagers. Chat, at some point you have to trust the process.\",\n \"Someone in chat just solved a game F1re has played for 15 years. Incredible stuff.\",\n \"F1re reads chat during fights. This is why the fights go the way they go.\",\n \"Chat spotted the mistake 4 seconds after F1re did, then mentioned it 40 more times.\",\n \"{by}, F1re appreciates the feedback and has filed it somewhere safe.\",\n \"Backseat gaming is allowed here. The backseat is 800 elo and the seatbelt is off.\",\n \"Chat said wall. F1re walled. Chat said why did you wall. This is the cycle.\",\n \"Chat has identified 14 mistakes. F1re made 3. Chat is still counting.\",\n \"{by} has entered a complaint. F1re thanks {by} and blames the map.\",\n \"Chat is 100% correct with 8 seconds of hindsight and a paused screen.\",\n \"Nobody in chat has ever typed good decision. The archive has been checked.\",\n \"Chat wants aggression. Chat also wants F1re not to die. Chat wants a different game.\",\n \"F1re has 400 coaches and one keyboard.\",\n \"Three people typed boar at the same moment. The boar had already left.\",\n \"Chat found the answer instantly, which is easy when you are not the one playing.\",\n \"{by} is right, and F1re will be acknowledging that at approximately no point.\",\n \"Someone donated 5 dollars to tell F1re to make more villagers.\",\n \"A sub resubscribed for 14 months to say the word wall.\",\n \"The mods are timing people out for backseating. The mods are also backseating.\",\n \"F1re has 400 viewers and 400 opinions about his gold count.\",\n \"Twitch chat moves faster than F1re's economy and is better managed.\",\n \"The donation alert played mid-fight and cost F1re a mangonel. Thank you for the support.\",\n \"A viewer typed ez from behind a keyboard, a screen, and total anonymity.\",\n \"F1re read one message during a fight and lost 12 villagers. Chat considers this fair.\",\n \"Someone subbed with the message go archers. F1re is playing Franks.\",\n \"The mods deleted the message. F1re had already read it. That is how reading works.\",\n \"Chat is 400 people who have never once been housed.\",\n \"A viewer donated to say he could do it better. He is 800 elo. F1re checked.\",\n \"Twitch chat has solved this game 40 times tonight and won zero games.\",\n \"F1re's stream has more analysts than the tournament and none of them are qualified.\",\n ],\n },\n\n hotkeys: {\n aliases: [\n \"hotkey\",\n \"hotkeys\",\n \"keybind\",\n \"keybinds\",\n \"ui\",\n \"interface\",\n \"controlgroup\",\n \"control\",\n \"mouse\",\n \"keyboard\",\n \"click\",\n \"clicks\",\n \"misclick\",\n \"minimap\",\n ],\n noun: \"hotkey\",\n lines: [\n \"F1re pressed the hotkey. The hotkey was pressed. Nothing happened. All three are true.\",\n \"The control group was set. The control group is now empty. Nobody touched it.\",\n \"F1re's keyboard registered the key. The game registered it 400ms later, in a different fight.\",\n \"He hit the wrong hotkey. It sits directly beside the right hotkey. Design is a choice.\",\n \"F1re rebound his hotkeys in 2019 and has been discovering the consequences ever since.\",\n \"The click landed on the UI, not the map. The UI is 40 pixels tall and F1re found all 40.\",\n \"F1re selected his whole army with one hotkey. It included a villager, now in the fight.\",\n \"The minimap click went 3 tiles off. On a 200-tile map, 3 tiles is exactly where the fight was.\",\n \"That wasn't a misclick, it was a tactical misdirection the game failed to understand.\",\n \"F1re pressed the correct key at the correct time in the incorrect universe.\",\n \"F1re pressed H for Town Center and received a house, a horse, and a headache.\",\n \"Control group 1 holds 40 knights. Control group 2 holds one villager and a dream.\",\n \"He double-tapped to centre on his Town Center and centred on a sheep 60 tiles away.\",\n \"F1re's hotkeys are custom, optimised, and occasionally observed by the game.\",\n \"The queue ate 3 clicks. F1re made 3 more. The queue ate those too, out of habit.\",\n \"F1re boxed his army and got 19 units, a monk, and the enemy's scout.\",\n \"Delete sits next to select-all. F1re would like that reviewed by someone senior.\",\n \"F1re has 280 APM and the game has elected to sample 4 of them.\",\n \"F1re's mouse registered the click. The interface disagreed, politely.\",\n \"The interface placed a button exactly where F1re clicks to move his army.\",\n \"F1re bought a new mouse. The misclicks have improved in quality, not quantity.\",\n \"He clicked the minimap, the UI and the notification bar in one motion.\",\n \"F1re's mouse has 12 buttons. He uses 2. The other 10 fire occasionally.\",\n \"The interface is 20 years old and has opinions about where F1re meant to click.\",\n \"F1re boxed the villager, the sheep and a tree in one selection.\",\n \"His hand moved correctly. His mouse moved correctly. The cursor did something else.\",\n \"F1re has 280 APM and roughly 240 of them are corrections.\",\n \"The hotkey worked. It worked on a building F1re had forgotten he owned.\",\n ],\n },\n\n wolves: {\n aliases: [\n \"wolf\",\n \"wolves\",\n \"ornlu\",\n \"wildlife\",\n \"animal\",\n \"animals\",\n \"jaguar\",\n \"crocodile\",\n ],\n noun: \"wolf\",\n lines: [\n \"A wolf ate F1re's scout. Minute 1. The game has jokes too.\",\n \"Ornlu has no elo, no build order, and a 100% win rate against F1re's early villagers.\",\n \"F1re's scout has 45 HP. The wolf does not care about F1re's plans.\",\n \"Two wolves. One villager. F1re watched it happen from the other side of the screen.\",\n \"The wolf spawned in F1re's base. The wolf pathed perfectly, which is notable.\",\n \"F1re lost a villager to wildlife in a game about empires.\",\n \"Wolves are balanced because they attack both players. They just start with F1re.\",\n \"Ornlu does not have a build order. Ornlu does not need a build order.\",\n \"The wolf pathed better than F1re's entire army, and it is one animal.\",\n \"F1re's villager fled the wolf directly into a second wolf. Coordinated.\",\n \"The wolf has 4 attack and 100% uptime.\",\n \"F1re has been killed by wolves, boars, deer, and once, memorably, a crocodile.\",\n \"The wolf waited by the gold until F1re sent exactly one villager.\",\n \"Nature is neutral. Nature has never once been neutral toward F1re.\",\n \"F1re's scout survived the wolf and died to the second one, 3 tiles later.\",\n \"Ornlu is one up and does not appear anywhere on the scoreboard.\",\n \"A jaguar killed F1re's scout. Different map, same story.\",\n \"F1re has lost villagers to wolves, jaguars, and one committed crocodile.\",\n \"The wolf has 25 HP and better positioning than F1re's army.\",\n \"F1re's scout ran from one wolf into the territory of two more.\",\n \"Wildlife has killed more of F1re's villagers this month than knights have.\",\n \"The jaguar was 3 tiles from the gold. F1re sent four villagers to the gold.\",\n \"Ornlu does not tilt, does not misclick, and does not lose.\",\n \"F1re walled out the enemy and walled in a wolf.\",\n ],\n },\n\n devs: {\n aliases: [\n \"dev\",\n \"devs\",\n \"developer\",\n \"developers\",\n \"patch\",\n \"patches\",\n \"update\",\n \"updates\",\n \"bug\",\n \"bugs\",\n \"glitch\",\n \"dlc\",\n \"worldsedge\",\n \"microsoft\",\n \"game\",\n ],\n noun: \"patch\",\n lines: [\n \"F1re reported that bug in 2021. It is still here. It has seniority now.\",\n \"The patch fixed 3 bugs and added 4. F1re did the math live on stream.\",\n \"New civ, new campaign, new DLC. Pathing update: various improvements.\",\n \"Working as intended is the most upsetting sentence in F1re's life.\",\n \"They nerfed the thing F1re liked and buffed the thing that kills him. Two separate teams, presumably.\",\n \"The patch notes are 6000 words. The line F1re cared about was not among them.\",\n \"F1re has more hours in this game than the patch had testers. Just saying.\",\n \"Every patch fixes pathing. Pathing has now been fixed roughly 40 times.\",\n \"F1re isn't losing, he's just beta-testing new excuses for the next DLC.\",\n \"The changelog says stability improvements. F1re has never been less stable.\",\n \"They shipped a civ with 9 bonuses and a unique unit that ignores armour. Balanced, apparently.\",\n \"F1re's bug report has 400 upvotes and one reply, from 2022, saying thanks for the feedback.\",\n \"Somewhere a man is employed to fix unit collision. F1re would like to buy him a coffee and a deadline.\",\n \"The hotfix broke the thing the patch fixed that the previous hotfix broke.\",\n \"They fixed a typo in a campaign nobody plays. Pathing: unchanged.\",\n \"F1re read all 6000 words of the patch notes and found exactly one that applied to him. It was a nerf.\",\n \"The devs have a roadmap. F1re has a list. The two documents do not overlap anywhere.\",\n \"Version nine million. Villagers still walk to the far woodline.\",\n \"World's Edge fixed a campaign typo. Pathing is unchanged.\",\n \"Microsoft owns this game. F1re would like to speak to Microsoft.\",\n \"A developer somewhere decided villagers should walk like that. On purpose.\",\n \"The glitch has been in the game since 2019 and now has emotional significance.\",\n \"F1re found a glitch, recorded it, reported it, and watched it survive three patches.\",\n \"The developers are aware. The developers have been aware since the beta.\",\n \"Somebody at World's Edge keeps a list. F1re's items are not on the list.\",\n \"A developer replied to F1re's bug report. The reply was an emoji.\",\n \"The glitch is not a glitch, it is emergent behaviour, which is a glitch.\",\n \"Microsoft has 200,000 employees and not one of them owns unit collision.\",\n \"The patch notes thanked the community. The community had asked for pathing.\",\n \"F1re has outlived four developers on this codebase and one game engine.\",\n ],\n },\n\n elo: {\n aliases: [\n \"elo\",\n \"rating\",\n \"rank\",\n \"ranked\",\n \"ladder\",\n \"matchmaking\",\n \"mmr\",\n \"queue\",\n \"quickplay\",\n \"lobby\",\n \"lobbies\",\n ],\n noun: \"ladder\",\n lines: [\n \"F1re lost 18 points for that game and would have gained 4. The ladder has priorities.\",\n \"Matchmaking found an opponent 200 elo higher in 6 seconds and cannot find a fair game in 6 minutes.\",\n \"F1re is not deranking. F1re is conducting research at a lower elevation.\",\n \"The ladder does not measure skill. It measures how many times the map hated you.\",\n \"F1re queued for 1v1 and got a rematch against the guy who just lamed him. Thank you, algorithm.\",\n \"Elo is a number. It is a number that follows F1re around and comments.\",\n \"He won 5 in a row for 40 points, then lost 1 and gave back 22. Chat, do the math.\",\n \"F1re's smurf has higher elo than his main. He declines to discuss the implications.\",\n \"F1re's elo is not a measure of skill. It is a measure of how many wolves he has met.\",\n \"The ladder took 18 points and did not make eye contact.\",\n \"F1re is 40 points off his peak and has been for eleven months.\",\n \"The queue found a match in 3 seconds. That is never a good sign.\",\n \"His opponent had a 22-game win streak and an account created in March.\",\n \"One more win and F1re is back. He has been one more win away since February.\",\n \"The ladder is a mirror. F1re would like a different mirror.\",\n \"F1re's rating went up 4 points and his blood pressure went up 40.\",\n \"Matchmaking is fair on average. F1re is not interested in the average.\",\n \"He is one point below a round number and the ladder is aware of it.\",\n \"F1re's MMR says one thing and his replays say another. Both are unflattering.\",\n \"Quick Play found a 2100 in 4 seconds. Ranked cannot find a 1900 in 6 minutes.\",\n \"The lobby had 40 people and none of them wanted to play the map.\",\n \"F1re joined a lobby, waited 11 minutes, and got kicked for his civ pick.\",\n \"His MMR dropped 18 for a game the map decided at minute 3.\",\n \"Lobbies are where F1re goes to not play games.\",\n \"F1re's Quick Play rating is higher than his ranked one and he will not discuss it.\",\n \"The lobby host picked Black Forest. F1re left. The lobby continued without him.\",\n \"MMR describes the past and predicts nothing.\",\n \"F1re is 12 points from a milestone and has been for two months.\",\n ],\n },\n\n archers: {\n aliases: [\n \"archer\",\n \"archers\",\n \"arrow\",\n \"arrows\",\n \"ballistics\",\n \"skirm\",\n \"skirms\",\n \"skirmisher\",\n \"skirmishers\",\n \"crossbow\",\n \"crossbows\",\n \"arbalest\",\n \"xbow\",\n \"bowman\",\n ],\n noun: \"archer\",\n lines: [\n \"F1re's archers fired at a moving target and hit the concept of where it used to be.\",\n \"He forgot Ballistics. The arrows are now purely decorative.\",\n \"40 archers. One mangonel shot. F1re now has 6 archers and a much smaller plan.\",\n \"The archers are attacking. They are attacking a tree. The tree is not the enemy.\",\n \"F1re's crossbows out-ranged the enemy by 1 tile and chose to walk forward anyway.\",\n \"Skirmishers counter archers. Nobody told F1re's skirmishers, who are currently fighting knights.\",\n \"The arrow was still airborne when the target died, so it landed on the ground with dignity.\",\n \"F1re massed archers. The opponent massed skirmishers. F1re calls this metagaming, and it is.\",\n \"His archers stopped to shoot one villager while the entire army walked past them.\",\n \"Feathered Arrows sounds like it should do more than that.\",\n \"F1re's archers have 4 range and the courage of a unit with 30.\",\n \"The whole volley landed. On the unit that was already dead.\",\n \"Thumb Ring makes archers fire faster. F1re's archers now miss sooner.\",\n \"F1re microed his archers backwards into his own siege. Textbook. Wrong textbook.\",\n \"The archers won the fight and then walked into the castle to celebrate.\",\n \"20 arrows in the air, one elephant on the ground, still upright, still coming.\",\n \"F1re's arbalests shot at the enemy scout for 9 seconds while the knights arrived behind them.\",\n \"Archers are strong against infantry. F1re's opponent has not made infantry since 2019.\",\n \"F1re's xbows out-ranged the skirms by one tile and closed the gap anyway.\",\n \"The bowman fired, missed, and the target had already left the tile.\",\n \"20 xbows against 4 skirms should be a win. It was a lesson.\",\n \"F1re's archers are the best-upgraded six units on this map.\",\n \"The arrows are in flight. The knights are already here.\",\n \"F1re massed xbow into a civ with bonus skirmishers. He knew. He did it anyway.\",\n \"One mangonel, 40 archers, two seconds. F1re is still doing the arithmetic.\",\n \"The bowman has 4 HP and the confidence of a paladin.\",\n \"F1re's archers stopped to shoot a house.\",\n \"Archers win fights they should lose and lose fights they should win, for F1re only.\",\n ],\n },\n\n cavalry: {\n aliases: [\n \"knight\",\n \"knights\",\n \"cav\",\n \"cavalry\",\n \"cavalier\",\n \"camel\",\n \"camels\",\n \"hussar\",\n \"paladin\",\n \"lancer\",\n \"lancers\",\n \"horse\",\n \"horses\",\n \"spear\",\n \"spears\",\n \"pike\",\n \"pikeman\",\n \"pikemen\",\n \"halberdier\",\n \"halb\",\n ],\n noun: \"knight\",\n lines: [\n \"F1re's knights charged into 6 pikemen. They had 6 seconds to reconsider and used none of them.\",\n \"The camel counters cavalry. F1re's cavalry did not read the tooltip.\",\n \"Knights cost 60 food, 75 gold, and one small piece of F1re's optimism.\",\n \"F1re's paladins arrived at the fight in Imperial Age and in single file.\",\n \"He forgot Bloodlines. His knights are running on the base model.\",\n \"The scout line: fast, cheap, and currently investigating a corner of the map.\",\n \"F1re's hussars raided the enemy eco and found four villagers and a wall of towers.\",\n \"Cavalry is strong against archers. F1re's cavalry met the one archer holding a spear.\",\n \"The knight has 100 HP. It arrived at the fight with 12 and no explanation for the other 88.\",\n \"F1re's lancers pushed units back. Mostly F1re's units.\",\n \"Two knights went in. Chat, one of them was a scout, and it went in first.\",\n \"F1re's cavalry has Husbandry, Bloodlines, and no sense of direction.\",\n \"The knights raided for 40 seconds and killed a sheep.\",\n \"F1re built a stable at minute 12 and knights at minute 26. The stable waited patiently.\",\n \"Halberdiers cost 35 food and deleted 900 resources of F1re's cavalry in 4 seconds.\",\n \"F1re's paladin has 160 HP and the decision-making of a unit with 12.\",\n \"The cavalry charged, arrived, and then stood still while F1re looked at the minimap.\",\n \"Camels beat knights, knights beat archers, archers beat F1re. It is a whole system.\",\n \"F1re's horses arrived in single file, which is how pikemen prefer them.\",\n \"The knight took 4 seconds to turn around and died facing the wrong way.\",\n \"F1re's cavalry raided a base with 12 towers and no villagers.\",\n \"Horses are fast. F1re's horses are fast in the wrong direction.\",\n \"The paladin cost 60 food and 75 gold and died to a 35-food halberdier.\",\n \"F1re's scouts found the enemy at minute 4 and died to a spearman at minute 5.\",\n \"Two knights raided, killed one villager, and were converted.\",\n \"The cavalry charged into a wall F1re had built.\",\n \"F1re's horses have Husbandry and still arrive after the fight.\",\n \"Camels beat knights. His opponent knew this. F1re knew it too.\",\n ],\n },\n\n buildings: {\n aliases: [\n \"tower\",\n \"towers\",\n \"tc\",\n \"towncenter\",\n \"outpost\",\n \"barracks\",\n \"stable\",\n \"range\",\n \"ranges\",\n \"blacksmith\",\n \"university\",\n \"monastery\",\n \"wonder\",\n \"keep\",\n ],\n noun: \"building\",\n lines: [\n \"The Town Center did not shoot. It watched. It has a full view and no opinion.\",\n \"F1re's tower went up at 70%. The opponent's knights arrived at 69%.\",\n \"He placed a barracks where the wall should have gone. Both are buildings, chat, be fair.\",\n \"The Town Center garrison button was right there. F1re looked directly at it.\",\n \"F1re built a stable, a range, and a blacksmith, and then made spearmen.\",\n \"The outpost has 6 line of sight and saw nothing all game, which is technically information.\",\n \"F1re's university researched everything except the one thing that mattered.\",\n \"The wonder costs 3528 stone. F1re's grievance about the wonder is free.\",\n \"He built the tower 1 tile too far back, which on this map is a different postcode.\",\n \"F1re has 4 barracks and 11 military units. The barracks are mostly for morale.\",\n \"The Town Center fired 3 arrows and garrisoned nobody. It is a spectator with a bow.\",\n \"F1re placed a house in the exact tile his own knights needed to walk through.\",\n \"The {bldg} finished at minute 30. F1re's army finished at minute 29.\",\n \"He built a monastery and no monks. It is a very expensive shed.\",\n \"F1re's castle went up 8 seconds after the fight it was built for.\",\n \"The building has 4 garrison slots and F1re has 40 villagers standing outside it.\",\n \"F1re's {bldg} finished 8 seconds after the fight it was built for.\",\n \"F1re built a keep. The keep is beautiful. The keep is 30 tiles from the fight.\",\n \"The tower finished. The raid had ended. The tower now watches an empty field.\",\n \"F1re has three Town Centers and is producing villagers from none of them.\",\n \"The castle went up in 8 seconds flat and 40 seconds too late.\",\n \"He placed the barracks against the wall and sealed his own army inside.\",\n \"The keep costs 125 stone and provides F1re with 125 stone of reassurance.\",\n \"F1re's buildings are perfectly placed for the attack that came from the other side.\",\n \"The Town Center rallied 20 villagers into the raid. Rally points are a technology.\",\n \"F1re built a second monastery. He still has no monks.\",\n \"The blacksmith is idle, the university is idle, and F1re is extremely busy.\",\n ],\n },\n\n housing: {\n aliases: [\n \"housed\",\n \"house\",\n \"houses\",\n \"housing\",\n \"pop\",\n \"population\",\n \"popcap\",\n \"supply\",\n \"cap\",\n ],\n noun: \"house\",\n lines: [\n \"F1re is housed. He has 2000 wood. These facts have coexisted for 40 seconds.\",\n \"The house was queued. The villager building it was also queued, in a different sense.\",\n \"F1re hit the pop cap and then blamed the pop cap.\",\n \"You are being attacked and you need more houses. Only one of those is F1re's fault.\",\n \"F1re built 4 houses at once so this would never happen again. It happened again.\",\n \"200 pop is a suggestion, and F1re's suggestion is 240.\",\n \"The villager walked away mid-house. The house is 90% complete and 0% useful.\",\n \"F1re is not housed. F1re is population-constrained, which is a policy issue.\",\n \"Every house F1re builds is one house too late and one house too few.\",\n \"F1re was housed at 34 pop, 78 pop, and 141 pop. A trilogy.\",\n \"The voice has said you need more houses nine times. It will say it again.\",\n \"F1re queued 5 villagers into a Town Center with room for none of them.\",\n \"He built the house, got unhoused, and immediately queued 5 more. Cycle complete.\",\n \"F1re has 4 unfinished houses. Collectively they house nobody.\",\n \"The pop cap is 200. F1re is at 199, with 6 villagers queued and a plan.\",\n \"F1re's entire economy stopped for 40 seconds over 25 wood.\",\n \"F1re is 4 pop from the cap and queuing 5 villagers. The arithmetic is not on his side.\",\n \"The pop cap is a suggestion, enforced with total consistency.\",\n \"F1re's supply of houses is three short and has been all game.\",\n \"He is at 199 pop with 40 villagers on wood and no plan for the 200th.\",\n \"F1re built the house. The house finished. He was already housed again.\",\n \"25 wood stands between F1re and an economy.\",\n \"The pop cap is 200 and F1re's army is 60 of it. The other 140 are chopping.\",\n \"F1re has never once been ahead of his housing. Not one game.\",\n \"Being housed is not a mistake, it is a lifestyle F1re has committed to.\",\n \"He queued the house behind two farms and a mining camp. It will arrive eventually.\",\n ],\n },\n\n upgrades: {\n aliases: [\n \"upgrade\",\n \"upgrades\",\n \"tech\",\n \"techs\",\n \"teching\",\n \"research\",\n \"researched\",\n \"bloodlines\",\n \"fletching\",\n \"forging\",\n \"armour\",\n \"armor\",\n \"attack\",\n \"wheelbarrow\",\n \"handcart\",\n \"husbandry\",\n \"chemistry\",\n \"thumbring\",\n \"squires\",\n ],\n noun: \"upgrade\",\n lines: [\n \"F1re forgot Bloodlines. His knights have been told and are taking it well.\",\n \"No Ballistics. The arrows are aiming at where the enemy plans to be, spiritually.\",\n \"He researched Wheelbarrow at minute 28. The villagers thanked him and kept walking slowly.\",\n \"F1re had 3000 gold and no blacksmith upgrades. The gold is fine. The gold is thriving.\",\n \"The upgrade completed 2 seconds after the fight ended, which is the traditional timing.\",\n \"F1re double-checked his upgrades. They were all there except the important one.\",\n \"Husbandry makes cavalry faster. F1re's cavalry is faster and still late.\",\n \"He clicked the upgrade. The building was busy. Nobody escalated it.\",\n \"F1re's army is fully upgraded. It is 6 units. They are magnificent.\",\n \"F1re has Chemistry, Siege Engineers, and no siege.\",\n \"The blacksmith has been idle since Feudal. It is not on anybody's list.\",\n \"F1re researched Loom at minute 19, which is either genius or an accident.\",\n \"Two attack upgrades behind, and F1re is still describing that fight as close.\",\n \"He upgraded to Arbalest mid-fight, which helped the 4 archers who survived it.\",\n \"F1re's opponent has plus three plus three. F1re has vibes and a plan.\",\n \"The upgrade takes 60 seconds. The fight took 11.\",\n \"Fletching costs 100 food and 50 gold. F1re has both and no memory of the tech existing.\",\n \"F1re has Thumb Ring. His archers still miss. The ring is not the problem.\",\n \"Squires makes infantry faster. F1re has no infantry. He researched it anyway.\",\n \"Handcart at minute 40 is not eco management, it is an apology.\",\n \"F1re aged up with zero blacksmith techs and complete confidence.\",\n \"His opponent has Plate Mail Armour. F1re has Scale Mail and a strong opinion.\",\n \"The tech finished. The units it was for are dead. The tech remains, proudly.\",\n \"F1re took Arbalest, Bracer and Chemistry, and made six units in total.\",\n \"Padded Archer Armour exists. F1re rediscovers this fresh every 40 games.\",\n \"Two Forging upgrades behind, and F1re still describes that trade as even.\",\n \"The blacksmith has four available techs and F1re has 400 gold. Nothing has happened.\",\n \"F1re clicked Chemistry at minute 38, which is 20 minutes of arrows doing base damage.\",\n \"Upgrades are permanent. F1re's memory of which ones he took is not.\",\n \"He researched the armour upgrade during the fight the armour was for.\",\n ],\n },\n\n market: {\n aliases: [\n \"market\",\n \"trade\",\n \"trading\",\n \"tradecart\",\n \"tradecarts\",\n \"tribute\",\n \"ratio\",\n \"ratios\",\n \"buy\",\n \"buying\",\n \"sell\",\n \"selling\",\n \"commerce\",\n \"coinage\",\n \"banking\",\n ],\n noun: \"market\",\n lines: [\n \"F1re bought stone at a terrible rate and calls it an investment.\",\n \"The market ratio moved 3% between F1re deciding and F1re clicking.\",\n \"F1re sold 100 food for 43 gold and describes this as the market's fault.\",\n \"He built a market at minute 12 and used it once, at minute 39, in a panic.\",\n \"Trade carts generate gold slowly and die quickly. F1re considers this a design statement.\",\n \"F1re tributed his teammate 100 wood and received a thumbs up worth 0 wood.\",\n \"The trade route runs past the enemy castle. The trade carts do not seem concerned.\",\n \"F1re needed 40 stone. He sold 200 food to get it. Chat, do not look at the ratio.\",\n \"Coinage reduces the tribute fee. F1re researched it after tributing everything.\",\n \"F1re sold all his food to buy gold to make units to protect the food he sold.\",\n \"The market took 30% and F1re called it a partnership.\",\n \"He bought 100 wood four separate times instead of once. The market noticed.\",\n \"F1re's trade cart has travelled 400 tiles and delivered 12 gold.\",\n \"The ratio was 130. F1re waited. The ratio is now 190.\",\n \"F1re has 4000 stone and nothing to build. This is also a complaint.\",\n \"Selling stone is a last resort. F1re reaches last resorts at minute 20.\",\n \"Buying 100 wood four times cost F1re 40% more than buying 400 once.\",\n \"Commerce is a two-way street. F1re only ever drives one direction.\",\n \"There is no banking upgrade. F1re has asked.\",\n \"F1re bought stone at 190 and sold food at 43 in the same minute.\",\n \"His trade carts are commerce. His trade carts are also target practice.\",\n \"F1re tributed 100 gold and lost 30 to the fee, which he considers theft.\",\n \"The ratio only moves against whoever is buying, which is always F1re.\",\n \"He built the market at minute 12 and remembered it at minute 39.\",\n \"F1re's commerce strategy is selling food he needs for gold he wastes.\",\n \"Buying gold to make units to protect the gold. A closed loop of regret.\",\n ],\n },\n\n water: {\n aliases: [\n \"water\",\n \"navy\",\n \"naval\",\n \"ship\",\n \"ships\",\n \"galley\",\n \"galleys\",\n \"fireship\",\n \"fireships\",\n \"dock\",\n \"docks\",\n \"fish\",\n \"fishing\",\n \"transport\",\n \"boat\",\n \"boats\",\n \"island\",\n \"islands\",\n \"shore\",\n \"sea\",\n \"ocean\",\n ],\n noun: \"water\",\n lines: [\n \"F1re's fishing ships are being eaten by a galley he cannot see and has no dock to answer.\",\n \"Water maps are balanced. F1re's shore fish are on the opponent's shore.\",\n \"The transport unloaded the army 1 tile into the water. They are swimming home.\",\n \"Fire ships beat galleys. F1re had galleys. F1re now has an opinion about fire ships.\",\n \"F1re built a dock at minute 4 and a navy at minute 34. The gap is where the game happened.\",\n \"His fishing ships are full and 40 tiles from the dock. This is a supply chain problem.\",\n \"The demo ship did what demo ships do. It did it to F1re.\",\n \"F1re controls the water. There is nothing on the water. He controls it completely.\",\n \"Naval battles are just pathing with extra steps and more water.\",\n \"F1re lost 9 fishing ships and now describes the sea as hostile territory.\",\n \"The galley has 6 range. F1re's dock has feelings.\",\n \"F1re went full navy on a map where the water connects nothing to nothing.\",\n \"His transport arrived. The army did not. The transport is not saying where they are.\",\n \"F1re has 14 fishing ships and 0 warships. It is a business, not an army.\",\n \"The enemy fire ship burned 400 resources in 6 seconds and left without comment.\",\n \"F1re scouted the water at minute 30. There was a fleet. There had been a fleet for 20 minutes.\",\n \"The island had 6 shore fish. F1re found it at minute 31.\",\n \"F1re does not play water. The map is 60% ocean. These facts are in conflict.\",\n \"His boat is full, idle, and 40 tiles from any dock he owns.\",\n \"F1re built a transport to move 4 villagers to an island the enemy already held.\",\n \"The ocean between F1re and his teammate is 3 tiles wide and completely uncrossable.\",\n \"F1re controls one island. Nobody wants the island. F1re is very proud of the island.\",\n \"The fire ship arrived, did its job, and left. F1re respects the professionalism.\",\n \"F1re's navy is 3 galleys and a fishing ship that believes it is a galley.\",\n \"Water maps reward map control. F1re's map control is one dock and a feeling.\",\n \"He lost the water at minute 12 and the game at minute 44. Related events.\",\n \"F1re's transport unloaded onto a cliff. The units are fine. They are just up there.\",\n \"There is gold on the island. There is always gold on the island. F1re never gets the island.\",\n ],\n },\n\n taunts: {\n aliases: [\n \"taunt\",\n \"taunts\",\n \"11\",\n \"eleven\",\n \"laugh\",\n \"laughing\",\n \"emote\",\n \"emotes\",\n \"spam\",\n \"spamming\",\n ],\n noun: \"taunt\",\n lines: [\n \"Someone sent 11 after killing F1re's scout. That is not a taunt, that is a war crime.\",\n \"F1re does not taunt. F1re sends a factual message that happens to be a number.\",\n \"The opponent taunted at minute 4 and won at minute 40. He was right, which is worse.\",\n \"An 11 in chat is worth 3 elo of psychological damage. F1re has taken 900 elo.\",\n \"F1re typed a taunt and got housed while typing it. These events are related.\",\n \"Nobody has ever taunted F1re without it being immediately followed by consequences.\",\n \"The enemy monk said wololo. F1re said something there is no taunt number for.\",\n \"F1re sent a friendly taunt. It was interpreted correctly and returned with interest.\",\n \"Taunts are free. F1re's response to taunts costs him 2 minutes of build order.\",\n \"He sent 30, nice town I'll take it, and then took it.\",\n \"F1re typed 11 first. He would like that noted. He lost, but he typed it first.\",\n \"The opponent sent 11 twice. The second one was unnecessary and therefore correct.\",\n \"F1re has never once sent a taunt from a winning position.\",\n \"Someone taunted at minute 2. F1re remembered it at minute 40.\",\n \"The taunt cost 0 resources and 4 minutes of F1re's concentration.\",\n \"F1re does not tilt from taunts. F1re simply plays worse afterwards, for unrelated reasons.\",\n \"The opponent spammed the laugh emote for 40 seconds. F1re counted.\",\n \"Someone is spamming taunts and losing. He is ignoring exactly one of those facts.\",\n \"F1re got emote-spammed by a player who resigned 90 seconds later.\",\n \"The laugh taunt has no cooldown, which the developers have never explained.\",\n \"F1re does not respond to emote spam. F1re responds at length, later, on stream.\",\n \"Spamming 11 does not make you better. It makes you correct, which is worse.\",\n \"He sent the taunt, missed the timing, and lost the fight. Chat, that is justice.\",\n \"F1re typed a taunt, deleted it, and typed a longer one.\",\n \"Emotes are free. The 4 seconds F1re spent reading them were not.\",\n \"The taunt arrived before the attack, which is at least polite.\",\n ],\n },\n\n streamtech: {\n aliases: [\n \"obs\",\n \"mic\",\n \"microphone\",\n \"camera\",\n \"cam\",\n \"webcam\",\n \"encoder\",\n \"bitrate\",\n \"overlay\",\n \"alert\",\n \"alerts\",\n \"audio\",\n \"sound\",\n \"music\",\n \"headset\",\n \"restream\",\n \"vod\",\n \"vods\",\n ],\n noun: \"stream\",\n lines: [\n \"The stream dropped 400 frames during the only good fight F1re has had this month.\",\n \"F1re's mic was muted for the entire explanation. It was, by all accounts, a great explanation.\",\n \"The overlay is covering the exact corner of the minimap where the attack came from.\",\n \"OBS crashed at the moment of victory and recovered in time for the loss.\",\n \"F1re's alert sound played over his own thought process and won.\",\n \"The encoder is at 100%. F1re is at 100%. Only one of them is going to recover.\",\n \"F1re said something incredible off-mic. It is gone. It is not coming back.\",\n \"The VOD cuts out right before the part F1re wanted to show you.\",\n \"F1re's camera froze on the exact frame you would have chosen.\",\n \"The stream is 1080p60. F1re's decision-making is 480p.\",\n \"Chat spent 4 minutes telling F1re his mic was fine. It was not the mic.\",\n \"F1re restarted OBS mid-game. The game did not wait for him.\",\n \"The bitrate dropped, the fight started, and both events are being investigated.\",\n \"F1re's webcam captured the exact moment he understood what was happening.\",\n \"Six people typed no audio. Four of them had their own volume at zero.\",\n \"The clip everyone wants is inside the 40 seconds the encoder dropped.\",\n \"F1re's microphone was muted for the best 90 seconds of the year.\",\n \"The music was copyrighted. The VOD is now 40 minutes of silence.\",\n \"His headset died mid-tournament and he played the last 10 minutes deaf.\",\n \"F1re checked the VODs. They confirmed he was right and still lost.\",\n \"The microphone picked up the keyboard, the fan, and one distant dog.\",\n \"Chat says the music is too loud. F1re cannot hear the music.\",\n \"The VOD is 6 hours. The clip everyone wants is in the missing 40 seconds.\",\n \"F1re's headset, mic and camera all work. The encoder does not.\",\n \"He explained the entire game plan on a dead microphone.\",\n \"The music stopped, the alert fired, and F1re lost a fight to silence.\",\n ],\n },\n\n tournament: {\n aliases: [\n \"tournament\",\n \"tournaments\",\n \"tourney\",\n \"bracket\",\n \"brackets\",\n \"draft\",\n \"civdraft\",\n \"mapdraft\",\n \"admin\",\n \"admins\",\n \"caster\",\n \"casters\",\n \"casting\",\n \"showmatch\",\n \"qualifier\",\n \"groupstage\",\n \"prize\",\n \"prizepool\",\n \"seeding\",\n \"series\",\n ],\n noun: \"tournament\",\n lines: [\n \"F1re got the hardest group and describes the seeding as an act of aggression.\",\n \"The map draft removed F1re's best map and left three he has publicly called unplayable.\",\n \"The civ draft gave F1re a civ he last played in 2019, on purpose, by his opponent.\",\n \"The admin restarted the game after F1re's best 8 minutes of the year.\",\n \"The casters called F1re's position winning 4 seconds before it stopped being winning.\",\n \"F1re lost the series 3-2 and the 2 were better games. Nobody records that.\",\n \"The bracket has F1re playing the one person he cannot beat, in round 1, again.\",\n \"F1re played at 4am local time in a tournament scheduled for European convenience.\",\n \"The prize pool is large. The gap between F1re and it is 11 minutes of one game.\",\n \"F1re was seeded 9th. There are 8 group slots. Chat, do the arithmetic.\",\n \"The admin said 15 minutes. It has been 50. F1re has warmed up twice and cooled down three times.\",\n \"F1re banned the map. The map came back in game 4 as the decider.\",\n \"His opponent had a coach, an analyst and a spreadsheet. F1re had a build order on a napkin.\",\n \"The caster said F1re never loses from here. F1re lost from there.\",\n \"F1re practised that matchup for a week and drew a different one.\",\n \"The stream delay meant F1re's chat knew the result 90 seconds before F1re did.\",\n \"The civ draft took 20 minutes. F1re's civ was decided in the last 4 seconds.\",\n \"In the map draft F1re banned three maps and got the fourth one he hates.\",\n \"The qualifier ran until 3am. F1re played his best game at 2:40.\",\n \"The showmatch was for fun. F1re has not stopped analysing it.\",\n \"F1re won the qualifier and drew the top seed in round one.\",\n \"The civ draft gave him a civ he has 11 games on. His opponent had 400.\",\n \"The map draft is strategy. F1re treats it as a personality test and fails.\",\n \"The showmatch had 8000 viewers and F1re got housed at minute 4.\",\n \"He qualified, prepared for a week, and drew the one player he cannot beat.\",\n \"The admin restarted after 8 minutes. Those 8 were the best of F1re's year.\",\n ],\n },\n\n luck: {\n aliases: [\n \"rng\",\n \"luck\",\n \"lucky\",\n \"unlucky\",\n \"random\",\n \"randomness\",\n \"chance\",\n \"odds\",\n \"variance\",\n \"dice\",\n \"fortune\",\n \"karma\",\n ],\n noun: \"RNG\",\n lines: [\n \"That was not variance. Variance would have gone F1re's way at least once.\",\n \"F1re has been unlucky in 400 consecutive games, which at this point is a pattern.\",\n \"RNG is fair over a large sample. F1re is the large sample.\",\n \"The odds were 90/10. F1re has now met the 10 nine times.\",\n \"F1re does not believe in luck. F1re believes in a targeted campaign.\",\n \"The coin flip went the other way. The coin has gone the other way since 2019.\",\n \"F1re is not blaming luck. F1re is blaming the specific distribution of luck.\",\n \"Statistically this should even out. It has been evening out for six years.\",\n \"Chat, the dice are fine. The dice simply know who is rolling them.\",\n \"F1re rolled a 1, then a 1, then a 1. The dice have been retired for testing.\",\n \"There is no luck in this game, which makes F1re's results considerably harder to explain.\",\n \"The 5% happened. It happens to F1re at a rate of roughly 40%.\",\n \"His opponent got the good spawn, the close gold and the boar. Luck is a spectrum.\",\n \"Random means random. F1re would like it audited anyway.\",\n \"F1re has never once been on the correct side of a coin flip on stream.\",\n \"The map is random. F1re has begun keeping records.\",\n \"There is no karma in this game. F1re is still accruing something.\",\n \"The chance was 1 in 6. F1re has met it four games running.\",\n \"Fortune favours the prepared. F1re was prepared. Chat, explain.\",\n \"F1re's luck is not bad. F1re's luck is specific.\",\n \"A 3% chance happened. It happened to F1re, in a tournament, on stream.\",\n \"Karma implies F1re did something. F1re maintains that he did not.\",\n \"The odds were even. The result was not. This has held for 400 games.\",\n \"F1re does not need good fortune. He needs the absence of active malice.\",\n \"By chance, the wolf, the lame and the bad spawn all happened in one game.\",\n \"Every 50/50 is a 50/50. F1re has receipts suggesting otherwise.\",\n ],\n },\n\n resign: {\n aliases: [\n \"resign\",\n \"resigned\",\n \"gg\",\n \"loss\",\n \"lost\",\n \"losing\",\n \"tilt\",\n \"tilted\",\n \"tilting\",\n \"rage\",\n \"quit\",\n \"altf4\",\n \"defeat\",\n \"defeated\",\n \"streak\",\n ],\n noun: \"loss\",\n lines: [\n \"F1re did not lose. F1re ran out of game before the comeback started.\",\n \"He resigned at 30 minutes and explained why until minute 47.\",\n \"F1re typed gg. F1re did not mean gg. F1re meant a considerably longer sentence.\",\n \"That is not a loss streak, that is a research phase.\",\n \"F1re is not tilted. F1re is operating at an elevated informational baseline.\",\n \"One more game has now been said 6 times and the sun is coming up.\",\n \"F1re lost and immediately requeued, which is either resilience or a symptom.\",\n \"The comeback was on. The comeback needed 90 more seconds and 40 more villagers.\",\n \"F1re resigned so fast the opponent had not finished the attack. Efficiency.\",\n \"F1re held F1 for four seconds and then did not press it. Twice.\",\n \"The scoreboard says defeat. F1re says incomplete.\",\n \"He resigned, closed the game, opened the replay, and got angry again at the same moment.\",\n \"F1re does not rage. F1re relocates his mouse with intent.\",\n \"The loss was 40 minutes ago. The analysis is ongoing.\",\n \"F1re typed gg wp. He meant exactly one of those two things.\",\n \"That was a winnable game, which is the most upsetting kind of loss available.\",\n \"F1re is not losing. F1re is behind in a way that is currently permanent.\",\n \"He did not quit. He conceded a position that had already conceded itself.\",\n \"Alt-F4 is not resigning. F1re knows the difference and uses both.\",\n \"Losing six in a row is not a slump, it is a body of work.\",\n \"F1re quit the game, opened the replay, and lost it again in analysis.\",\n \"The resign button is 4 pixels from the diplomacy menu. F1re has stories.\",\n \"He was losing at minute 20, winning at 32, and quit at 41.\",\n \"F1re does not alt-F4. F1re closes the game with intention.\",\n \"Losing is part of the game. F1re would like a smaller part.\",\n \"He typed gg exactly as his opponent's army walked into a castle.\",\n ],\n },\n\n strategy: {\n aliases: [\n \"strat\",\n \"strats\",\n \"strategy\",\n \"greed\",\n \"greedy\",\n \"boom\",\n \"booming\",\n \"aggression\",\n \"aggressive\",\n \"allin\",\n \"fastcastle\",\n \"macro\",\n \"micro\",\n \"plan\",\n \"decision\",\n \"decisions\",\n \"gameplan\",\n ],\n noun: \"plan\",\n lines: [\n \"F1re went greedy. The greed was correct. The punishment was also correct.\",\n \"The all-in did not work, which means it was a bad plan, which is how all-ins are graded.\",\n \"F1re booms when he should fight and fights when he should boom, with total commitment either way.\",\n \"The plan was perfect. The plan required the opponent to also follow the plan.\",\n \"F1re's macro is excellent. His micro is excellent. They have never once been excellent simultaneously.\",\n \"He scouted the tower rush, called the tower rush, and then got tower rushed.\",\n \"F1re made the correct decision at the wrong minute, which the game scores as incorrect.\",\n \"Fast castle into knights is a good build. It is also the build the opponent countered.\",\n \"F1re had two options and picked the third one, which he invented mid-fight.\",\n \"The plan had three phases. F1re reached phase one and improvised for 30 minutes.\",\n \"F1re committed to the all-in and then, at the last second, committed to something else.\",\n \"He went greedy against the one opponent who was never going to allow it.\",\n \"F1re's plan B is his plan A, executed later and with less food.\",\n \"The strategy works at 2400 elo. F1re has confirmed it does not work at 2399.\",\n \"F1re adapted mid-game, which is admirable, and lost, which is not.\",\n \"He picked the right build at minute 14. It was a minute-4 build.\",\n \"F1re's aggression arrived four minutes after the window for aggression closed.\",\n \"The all-in was all. There was no in.\",\n \"F1re chose aggression on a map that punishes aggression, aggressively.\",\n \"He played aggressive, greedy and defensive inside a single 90-second window.\",\n \"F1re's plan was aggression. F1re's execution was six scouts and hope.\",\n \"The all-in worked at minute 12, against a player who was ready at minute 11.\",\n \"Aggressive builds need commitment. F1re committed, reconsidered, then committed again.\",\n \"F1re went all-in and kept 40 villagers on wood, which is not what all-in means.\",\n \"The strategy was sound. The strategy assumed F1re would execute it.\",\n \"He had a plan for aggression and a plan for booming, and ran both at once.\",\n ],\n },\n};",[146],"930be89f-1d5e-4f9b-9569-2b3a9affe6fd","topics.js","5916ca5e-a6b9-11f1-85ad-1607ee4eb77e",21,["D",1788536346749],5,"// @ts-nocheck: plain JavaScript file, parameters are intentionally untyped\n//\n// !complaint — THE 32 TOPICS. This is where most jokes live.\n//\n// Split from complaint.js because Val Town caps one file at 80,000 characters\n// and the combined file was 88,000. On Val Town this is a second file in the\n// same val named exactly topics.js — complaint.http.ts imports \"./topics.js\".\n//\n// Each topic needs: aliases (unique across ALL topics), noun (short singular\n// label used by combos/modifiers), lines. Keep lines under ~300 chars.\n// Use {by} for the asker's Twitch name. Both rules are enforced by tests.\n\n// Interchangeable vocabularies. Any line in any pool may use {slot}; every\n// occurrence of the SAME name in one line gets the SAME value, so the sentence\n// stays coherent. For a second, different value use a number: {eco} and {eco2}.\n//\n// \"F1re's {eco} is 14 tiles away. His opponent's {eco} is in their base.\"\n// \"Out of {eco}, out of {eco2}, out of ideas, in that order.\"\n//\n// RULE: every value in a list must be grammatically interchangeable — same\n// number, same article behaviour. \"berries\" cannot sit in `eco` beside \"gold\",\n// because \"F1re's berries IS 14 tiles away\" is broken. That is why there are\n// separate singular and plural unit lists.\n//\n// Only use a slot where the noun is INCIDENTAL to the joke. \"The camel counters\n// cavalry\" must keep its camel — the camel is the punchline.\nexport const SLOTS = {\n // A LOCATION on the map — \"F1re's {eco} is 14 tiles away\". Includes features\n // you gather from, so it must not be used for quantities: \"out of woodline\"\n // and \"3000 berry bush\" are both broken.\n eco: [\"gold\", \"stone\", \"wood\", \"food\", \"woodline\", \"berry bush\"],\n // A QUANTITY you consume — \"out of {res}\", \"3000 {res}\". Mass nouns only,\n // which is why the map features above are absent.\n res: [\"gold\", \"stone\", \"wood\", \"food\"],\n // singular — takes \"is\"\n unit: [\n \"knight\",\n \"archer\",\n \"villager\",\n \"monk\",\n \"skirmisher\",\n \"spearman\",\n \"scout\",\n ],\n // plural — takes \"are\"\n units: [\n \"knights\",\n \"archers\",\n \"villagers\",\n \"monks\",\n \"skirmishers\",\n \"spearmen\",\n \"scouts\",\n ],\n // singular — takes \"is\"\n bldg: [\n \"Town Center\",\n \"castle\",\n \"tower\",\n \"barracks\",\n \"stable\",\n \"blacksmith\",\n \"market\",\n ],\n // Dark Age is deliberately absent: you start there, you do not \"reach\" it.\n age: [\"Feudal Age\", \"Castle Age\", \"Imperial Age\"],\n // Generic noun for when chat types a bare adjective with nothing to attach it\n // to: \"!complaint dumb\" becomes \"F1re's game has been described as dumb\".\n thing: [\"game\", \"match\", \"build\", \"plan\", \"position\", \"situation\"],\n animal: [\"boar\", \"deer\", \"sheep\", \"wolf\"],\n};\n\n// Bare judgement words. \"!complaint dumb\" has no topic, and every TEMPLATE reads\n// {w} as a noun — which produced \"Somewhere a developer is being paid to\n// maintain dumb\". These route to MODIFIERS instead, where the word is treated as\n// a description, which is what it is.\n//\n// Words that are ALSO topic aliases (trash, slow, lucky) never reach this list:\n// the alias matches first and gives a better answer. Overlap is harmless.\nexport const ADJECTIVES = new Set([\n \"dumb\",\n \"stupid\",\n \"bad\",\n \"awful\",\n \"terrible\",\n \"horrible\",\n \"broken\",\n \"rigged\",\n \"unfair\",\n \"cringe\",\n \"weird\",\n \"strange\",\n \"odd\",\n \"sad\",\n \"painful\",\n \"cursed\",\n \"insane\",\n \"crazy\",\n \"ridiculous\",\n \"tragic\",\n \"embarrassing\",\n \"shameful\",\n \"scuffed\",\n \"jank\",\n \"janky\",\n \"questionable\",\n \"suspicious\",\n \"criminal\",\n \"disgusting\",\n \"pathetic\",\n \"useless\",\n \"pointless\",\n \"boring\",\n \"unlucky\",\n \"hilarious\",\n \"brutal\",\n \"rough\",\n \"rigged\",\n \"clown\",\n \"dogshit\",\n \"mid\",\n]);\n\nexport const TOPICS = {\n map: {\n aliases: [\n \"map\",\n \"maps\",\n \"mapgen\",\n \"generation\",\n \"arabia\",\n \"arena\",\n \"nomad\",\n \"seed\",\n \"spawn\",\n \"spawns\",\n \"hill\",\n \"hills\",\n \"woodline\",\n \"woodlines\",\n \"terrain\",\n ],\n noun: \"map\",\n lines: [\n \"The map seed was made by F1re's enemies to hide his sheep. His gold is forward, and the hill belongs to the other guy.\",\n \"F1re's {eco} is 14 tiles away. His opponent's {eco} is inside their Town Center. This is called map balance.\",\n \"F1re is not open. The MAP is open. There is a difference and he will explain it for 40 minutes.\",\n \"Arabia is the most balanced map in the game, except on F1re's half, where it is a documentary about scarcity.\",\n \"Both players got 4 boars. F1re's 4 boars are in the opponent's base. Working as intended.\",\n \"F1re's base has no hill. His opponent's base has a hill, a cliff, and what appears to be a moat.\",\n \"The gold is in the middle. The stone is in the middle. F1re is in the corner. Coincidence? F1re says no.\",\n \"Map generation is random and has produced the identical result against F1re 400 times.\",\n \"F1re would love this map if it were rotated 180 degrees.\",\n \"One player can full wall, the other must play open. Guess which one is streaming right now.\",\n \"The hills on this map are strategically placed to ensure F1re never has the high ground.\",\n \"The trees on this map are clearly biased against F1re's woodline.\",\n \"This map was generated fairly. It was then reviewed, adjusted, and generated again against F1re.\",\n \"F1re's gold is 20 tiles forward. His opponent's gold is behind two woodlines and a moral high ground.\",\n \"The map gave F1re a lake. F1re does not play water. The lake is where his wood should be.\",\n \"There are 8 relics on this map. Six spawned closer to the guy who is not F1re.\",\n \"F1re's base has three entrances. The recommended number of entrances is fewer.\",\n \"This map has a fair split down the middle. F1re got the half with the scenery.\",\n \"The {eco} is behind the enemy base. F1re would like to know who signed off on that.\",\n \"F1re's forward gold has a name now. The name is 'the reason'.\",\n \"The map is symmetrical. F1re has checked. F1re has checked several times.\",\n \"Both players have 2 gold piles. F1re's second pile is a rumour.\",\n \"The berries are 30 tiles away and have been since 2019.\",\n \"F1re got the corner spawn. Again. He is beginning to take it personally.\",\n \"The map generator has a random seed and a clear preference.\",\n \"There is one {eco} on F1re's side of the map and three on the other.\",\n \"Arena gives everyone stone walls. F1re still found a way to be open.\",\n \"Nomad is chaos for both players. Only one of them starts next to a lake.\",\n \"The terrain on F1re's side has a 40% slope and an opinion.\",\n \"Map generation ran 4000 times in testing and never once produced this.\",\n \"F1re has played 6000 maps. He remembers the 200 that wronged him.\",\n \"The mapgen script has a bug. F1re has not found it. F1re knows it is there.\",\n \"On Arena, F1re is safe, walled, and losing to a trebuchet he cannot see.\",\n \"Nomad means starting with nothing. F1re started with slightly less.\",\n \"The terrain hid an entire castle behind a hill 12 tiles from his base.\",\n \"Every map is playable. Some maps are playable and also fair.\",\n \"F1re's cliff blocks his own army and none of the enemy's.\",\n \"The map is called Arabia. F1re calls it other things.\",\n ],\n },\n\n walls: {\n aliases: [\n \"wall\",\n \"walls\",\n \"walling\",\n \"walled\",\n \"palisade\",\n \"palisades\",\n \"gate\",\n \"gates\",\n \"quickwall\",\n \"wallin\",\n ],\n noun: \"wall\",\n lines: [\n \"F1re had a full wall. One tile was palisade. That tile is now in a museum.\",\n \"The quick wall went down perfectly. The villager placing it disagreed and walked home.\",\n \"F1re's gate closed. The game logged it as open. These are the mysteries we live with.\",\n \"Nothing wrong with F1re's walls. There was simply a door where he did not build a door.\",\n \"F1re walled. The wall exists. The scout is inside anyway. Chat, this is a physics problem.\",\n \"The palisade sat at 1 HP for six minutes, then chose that exact moment.\",\n \"F1re placed a wall foundation and a deer stood on it. The deer had priority.\",\n \"F1re does not have a hole in his wall. He has an unfinished corridor with future potential.\",\n \"Stone walls cost stone. F1re had stone. The stone was in the other player's base.\",\n \"He walled off the entire base except one tile, which is statistically the tile the knights found.\",\n \"The wall was complete. The wall was reviewed. The wall had a door nobody ordered.\",\n \"F1re's palisade is load-bearing emotionally.\",\n \"F1re walled 47 tiles. The knights found the 48th in under a second.\",\n \"The gate was open because F1re opened it. This is the part he does not mention.\",\n \"His wall is 95% complete, which in walling terms is 0% complete.\",\n \"F1re quick-walled with 2 villagers. One walled. The other went to get wood.\",\n \"The palisade held for 8 seconds, which F1re says was long enough for a better player.\",\n \"He walled the front and forgot the back. The map only has a back on F1re's side.\",\n \"F1re's wall has a gap exactly one scout wide. The scout measured.\",\n \"The wall foundation was placed. The villager was then distracted by a tree.\",\n \"F1re built a gate so his army could leave. The enemy army also read the sign.\",\n \"The stone wall went up at minute 22. The attack was at minute 21.\",\n \"F1re walled off his own gold. Correctly, in the sense that nobody can reach it.\",\n \"There is a hole. There is always a hole. The hole is the entire point of walls.\",\n \"He walled in the sheep, the villagers, and one extremely confused enemy scout.\",\n \"F1re's palisade cost 2 wood and 40 minutes of consequences.\",\n \"The gate is open. F1re opened it in Feudal for a scout that never came back.\",\n \"He walled the ramp, the gap and the entrance, and forgot the cliff.\",\n \"F1re's stone wall has one gate. The gate is on the enemy's side.\",\n \"The wall went up. The wall went down. The wall was up for 11 seconds.\",\n \"F1re quick-walled a knight out and a knight in.\",\n \"A palisade has 250 HP. Knights do not respect the number.\",\n \"F1re walled his farms in and his villagers out.\",\n \"The enemy walked around a 40-tile wall in 6 seconds.\",\n \"F1re's gate is closed and his units are outside. Both are working as intended.\",\n \"He built the wall on the wrong side of his own gold.\",\n \"Palisade, stone wall, gate, tower. All four, in the wrong order, in one minute.\",\n ],\n },\n\n pathing: {\n aliases: [\n \"path\",\n \"paths\",\n \"pathing\",\n \"pathfinding\",\n \"unit\",\n \"units\",\n \"collision\",\n \"stuck\",\n \"movement\",\n \"moving\",\n \"army\",\n ],\n noun: \"pathing\",\n lines: [\n \"Pathing isn't broken; units are just protesting F1re's micro.\",\n \"F1re's knights took the scenic route. It was scenic. It went past three woodlines and a lake.\",\n \"The units are moving perfectly. They are simply moving perfectly somewhere else.\",\n \"F1re shift-queued four commands. The game heard one and interpreted it as jazz.\",\n \"Unit collision is a solved problem in every game except the one F1re is currently playing.\",\n \"F1re told 20 units to attack. 19 arrived. The 20th is still finding itself near the mill.\",\n \"Pathing has been patched 40 times. F1re's units read all 40 patch notes and disagreed.\",\n \"The army walked around the wall instead of through the gate. The gate was open. The gate was RIGHT THERE.\",\n \"The unit is not stuck. It is evaluating its options. It has been evaluating for eleven seconds.\",\n \"Two villagers met at a drop-off point and are now performing an interpretive dance.\",\n \"The developers fixed pathing. F1re's units did not receive the memo, as they are units.\",\n \"Pathing isn't broken; the units are just protesting F1re's leadership skills.\",\n \"F1re's units have formed a committee. The committee is standing near the mill.\",\n \"The {units} arrived 6 seconds late and refuse to explain the gap in their timeline.\",\n \"F1re's army split in two at a tree. One half is fighting. The other half is discovering the map.\",\n \"The siege took the long way around because the short way had a sheep in it.\",\n \"F1re boxed 30 units and 28 moved. The other 2 are appealing the decision.\",\n \"The unit walked into the wall, considered the wall, and walked into the wall again.\",\n \"His {units} pathed through the enemy Town Center, out the other side, and kept going.\",\n \"F1re gave one order. The army has interpreted it as a suggestion, then as a theme.\",\n \"Two units are trying to occupy the same tile. Neither will move. Both are F1re's.\",\n \"The army arrived in three separate waves, none of which F1re authorised.\",\n \"F1re's units found a shortcut. The shortcut was longer. They took it anyway.\",\n \"It walked past the fight. It has been briefed on the fight. It walked past the fight.\",\n \"Formations exist. F1re's units have heard of formations.\",\n \"F1re selected 30 {units}. 28 acknowledged the order.\",\n \"The {unit} has one job and has been doing a different job since Feudal Age.\",\n \"F1re's army is in formation. The formation is a queue.\",\n \"Two knights swapped places for 6 seconds and both died doing it.\",\n \"The unit reached its destination by a route that visited three woodlines.\",\n \"F1re's siege pushed his own knights into the enemy castle.\",\n \"The army split around a house and rejoined 40 tiles later, wiser.\",\n \"One villager is stuck between a farm and a mill and has been since Feudal.\",\n \"F1re issued a move command. The units performed a slow, considered drift.\",\n \"The knights arrived one at a time, which is also how they died.\",\n \"F1re's units respect walls, cliffs and trees more than they respect F1re.\",\n \"The unit walked to the target, past the target, and back to the target.\",\n \"Collision means 40 units take 12 seconds to leave one gate.\",\n \"F1re's army went around the lake. There was a bridge. There is always a bridge.\",\n ],\n },\n\n sheep: {\n aliases: [\n \"sheep\",\n \"boar\",\n \"boars\",\n \"deer\",\n \"hunt\",\n \"hunting\",\n \"straggler\",\n \"stragglers\",\n \"herd\",\n \"forage\",\n \"berries\",\n ],\n noun: \"boar\",\n lines: [\n \"Three sheep found. The fourth sheep is a myth passed down between F1re's villagers.\",\n \"The boar was lured correctly. It simply chose a different villager to be angry at.\",\n \"F1re's deer are behind a woodline. His opponent's deer are queuing politely at the Town Center.\",\n \"F1re found the sheep. They were under his opponent's Town Center, which is apparently where they spawn now.\",\n \"F1re's boar died 2 tiles from the drop-off. Two. Tiles.\",\n \"Straggler trees: F1re has one. His opponent has a forest arranged in a helpful semicircle.\",\n \"The sheep walked away from the Town Center. Nobody has explained why sheep do this. F1re has theories.\",\n \"The {animal} filed a complaint about F1re first. The {animal}'s complaint was upheld.\",\n \"The fourth sheep exists. It has simply chosen a life elsewhere.\",\n \"The boar is at 12 HP and F1re's villager is at 12 HP. One of them has a plan.\",\n \"He lured the boar perfectly into a wolf. Chat, that is two problems solving each other badly.\",\n \"The sheep are in a line. The line points away from F1re's Town Center.\",\n \"F1re's deer walked 6 tiles while being shot, which is 6 tiles of pure spite.\",\n \"The second boar is under the enemy Town Center. It is their boar now. It was always their boar.\",\n \"F1re's boar reached the Town Center and killed a villager on arrival, out of principle.\",\n \"His straggler tree is directly on the house foundation. Statistically it always is.\",\n \"F1re found 8 sheep. Four were the opponent's, which he learned later, loudly.\",\n \"The deer are hunting F1re at this point.\",\n \"F1re lured the boar into his own Town Center and then into his own villagers. It is his boar. He may do as he likes.\",\n \"The sheep spawned equidistant from both players and then made a decision.\",\n \"Every animal on this map has a plan. F1re has a build order. These are not the same.\",\n \"F1re pushed the deer for 40 seconds and moved them 3 tiles closer to the enemy.\",\n \"F1re's herd is four sheep and one with philosophical objections.\",\n \"The berries are 30 tiles away and guarded by a wolf with a schedule.\",\n \"Foraging is 6 villagers standing in a bush achieving 0.31 food per second.\",\n \"F1re's herd walked to the enemy Town Center and did not come back.\",\n \"The boar has 340 HP and killed 2 villagers on the way to the drop-off.\",\n \"He forgot to garrison during the boar lure. The villager forgot to survive.\",\n \"F1re's sheep are all on the far side of the Town Center. All of them.\",\n \"The berry bush is 4 tiles from the enemy and 30 from F1re.\",\n \"F1re herded 8 sheep perfectly and then forgot to eat any of them.\",\n \"The deer walked away mid-shot, out of principle.\",\n \"F1re's forage patch had 6 bushes. His opponent's had 6 bushes and a hill.\",\n \"Two boars, one villager, no plan.\",\n ],\n },\n\n scouting: {\n aliases: [\n \"scout\",\n \"scouts\",\n \"scouting\",\n \"scouted\",\n \"exploration\",\n \"explore\",\n \"exploring\",\n \"vision\",\n \"los\",\n \"sight\",\n \"reveal\",\n \"recon\",\n ],\n noun: \"scout\",\n lines: [\n \"F1re's scout didn't die; it took early retirement due to toxic work conditions.\",\n \"The scout has 8 line of sight and has used exactly 0 of it.\",\n \"F1re's scout found the enemy base, the enemy sheep, and none of his own.\",\n \"The scout is standing next to the sheep and has chosen not to mention it.\",\n \"F1re scouted clockwise. The entire game was happening counter-clockwise.\",\n \"F1re sent his scout out at 0:01 and received the first useful information at 14:00.\",\n \"The scout walked past the enemy Town Center twice. It is not on speaking terms with F1re.\",\n \"F1re's scout explored 4% of the map and filed a full report anyway.\",\n \"Exploration is free. F1re has somehow still been charged for it.\",\n \"The scout died in a corner of the map doing something F1re will never learn about.\",\n \"F1re knows exactly where the enemy is. He knows because they are already inside his base.\",\n \"The scout found gold, stone, relics and a wolf. It reported the wolf. In person. At speed.\",\n \"F1re's scout has line of sight, a horse, and no curiosity whatsoever.\",\n \"Scouting is 90% of the game, which is unfortunate, because F1re did 6% of it.\",\n \"F1re's scout came home empty and has been asked to explain itself.\",\n \"The enemy scout mapped F1re's entire base. F1re's scout found a tree it liked.\",\n \"F1re sent one scout in one direction and the map put everything in the other one.\",\n \"F1re has 40% map control and 0% map knowledge. These are separate statistics.\",\n \"The fog of war is lifting everywhere except the one tile that mattered.\",\n \"F1re's scout is alive, healthy, and 90 tiles from anything relevant.\",\n \"F1re has vision of 4% of the map and confidence about the other 96%.\",\n \"The recon told him nothing, arrived late, and cost 100 food.\",\n \"F1re's line of sight ends exactly where the enemy army begins.\",\n \"The fog revealed 40 knights. F1re would have preferred it stayed fogged.\",\n \"He had vision on the enemy base for 3 seconds and used none of it.\",\n \"The reveal at the end showed F1re everything he needed at minute 4.\",\n \"F1re scouted three corners. The enemy was in the fourth.\",\n \"Vision costs nothing. F1re still cannot afford it.\",\n \"His recon report was one sheep and a strong sense of unease.\",\n \"F1re had no vision of the tower until it had arrows.\",\n \"The scout revealed the map. F1re was reading chat.\",\n \"Line of sight 8, curiosity 0.\",\n ],\n },\n\n // \"gold\" and \"stone\" used to live in `map` (resource PLACEMENT). But chat\n // types \"!complaint no gold\" meaning DEPLETION, and got a woodline joke.\n // Depletion, trash wars and running dry are their own complaint.\n resources: {\n aliases: [\n \"gold\",\n \"stone\",\n \"wood\",\n \"food\",\n \"resource\",\n \"resources\",\n \"mining\",\n \"miner\",\n \"miners\",\n \"depleted\",\n \"deplete\",\n \"trash\",\n ],\n noun: \"gold\",\n lines: [\n \"F1re has 0 gold, 4000 food, and a plan that requires gold.\",\n \"The gold ran out at minute 34. F1re's strategy ran out at minute 12.\",\n \"Trash war. F1re has 60 halberdiers and no way to end this before dawn.\",\n \"F1re's gold is gone. His opponent's gold is behind three castles.\",\n \"There is gold on this map. It is under the enemy Town Center, as is traditional.\",\n \"F1re mined 800 gold and lost 900 gold of units defending the mine.\",\n \"The last gold pile holds 40 gold and 3 villagers standing around it hopefully.\",\n \"F1re went for relics. There were 5. His opponent had all 5 by minute 20.\",\n \"Out of {res}, out of {res2}, out of ideas, in that order.\",\n \"F1re has 3000 {res} and nothing that costs {res}.\",\n \"The wood is behind the enemy. The food is behind the enemy. F1re is in front of the enemy.\",\n \"90 villagers and 11 resources per second. Something is wrong and F1re says it is the pathing.\",\n \"He sold 500 food for gold, then spent the gold defending the food.\",\n \"Trash units are called trash because they are what F1re has left.\",\n \"The gold under the enemy castle is technically F1re's. It is also technically unreachable.\",\n \"F1re has been mining the same pile since Castle Age. It has 12 gold remaining.\",\n \"Every resource F1re needs is forward. Every resource he has is wood.\",\n \"F1re is not out of gold. F1re is between golds.\",\n \"The {res} ran out. F1re's plan did not adjust.\",\n \"F1re has {res} for one more fight and a plan that needs three.\",\n \"The gold depleted at minute 30 with 40 villagers standing on it.\",\n \"F1re has 900 stone and nothing worth building with it.\",\n \"Every pile F1re mines depletes four minutes before he is ready.\",\n \"The stone is depleted, the gold is depleted, and F1re is trending that way.\",\n \"F1re needs 650 stone for a castle. He has 640 and a decision to make.\",\n \"His last gold pile depleted mid-fight, which is the traditional moment.\",\n \"F1re has stone, wood, food and no gold. Three out of four is a trash war.\",\n \"The mining camp is 3 tiles wrong and will be for the next 30 minutes.\",\n \"F1re's opponent has 6000 banked. F1re has 40 and a plan.\",\n \"Depletion is inevitable. Depletion at minute 22 is a personal attack.\",\n \"He mined the forward gold for 90 seconds and lost 4 villagers doing it.\",\n \"F1re is farming, mining, chopping, and still somehow negative.\",\n ],\n },\n\n vills: {\n aliases: [\n \"vill\",\n \"vills\",\n \"villager\",\n \"villagers\",\n \"eco\",\n \"economy\",\n \"farm\",\n \"farms\",\n \"farming\",\n \"idle\",\n \"dropoff\",\n \"mill\",\n \"lumbercamp\",\n ],\n noun: \"villager\",\n lines: [\n \"The villager is not idle. The villager is between opportunities.\",\n \"F1re queued 5 villagers. The Town Center made 4 and kept one as a souvenir.\",\n \"The villager walked to the far woodline. There was a closer woodline. There is always a closer woodline.\",\n \"F1re double-clicked to select villagers and selected the entire economy, a house, and a sheep.\",\n \"The farm ran out 90 seconds ago. The villager standing on it did not want to bother anyone.\",\n \"F1re's villager auto-tasked onto gold. F1re wanted wood. The villager wanted a career change.\",\n \"There were 3 idle villagers. F1re says 0. The idle villager button says otherwise, and it has no reason to lie.\",\n \"A villager just walked past four trees to reach a fifth tree it prefers emotionally.\",\n \"Villagers walk two to three times the shortest path. It is documented. F1re did the documenting. Nothing happened.\",\n \"The drop-off is 1 tile away. The villager has chosen violence and a different drop-off.\",\n \"The villager is unionised and that drop-off is outside its jurisdiction.\",\n \"F1re has 3 idle villagers and 0 idle opinions.\",\n \"F1re made 3 villagers in a row and then none for 90 seconds. The Town Center was thinking.\",\n \"The villager is carrying 9 wood and has decided that is enough for today.\",\n \"A villager walked into the enemy base while building a house. It is their villager now.\",\n \"F1re rallied his villagers to gold. They arrived at gold. One went back for its things.\",\n \"The lumber camp is 1 tile too far. Over 30 minutes that is a whole villager of production.\",\n \"F1re's villager is repairing a building that is not under attack, with great focus.\",\n \"He queued 20 farms. 19 got placed. The 20th is a mystery the game will not discuss.\",\n \"The villager was building. Then it was not building. Nobody saw it stop.\",\n \"F1re has 84 villagers and the eco output of 60. The other 24 are consultants.\",\n \"The villager fleeing the raid ran directly into the raid.\",\n \"F1re's vills auto-repaired the wall instead of running, which is loyal and fatal.\",\n \"One villager has been walking since Feudal Age. It has commitments elsewhere.\",\n \"The drop-off changed hands mid-walk and the villager did not renegotiate.\",\n \"The mill is 8 tiles from the berries. Somebody made that decision.\",\n \"F1re built a mill, forgot the mill, and gathered by hand for four minutes.\",\n \"A villager walked from the mill to the Town Center carrying 3 food.\",\n \"F1re has 84 villagers. Eleven are doing something he asked for.\",\n \"The villager finished the farm and immediately stood still with pride.\",\n \"F1re's economy is excellent. It is also 40 tiles from the fight.\",\n \"One villager has walked 400 tiles today and gathered 11 wood.\",\n \"He deleted the wrong villager. It was the one carrying 10 gold.\",\n \"The lumber camp is up. The villagers are still walking to the old one.\",\n \"F1re's villagers are unionised, punctual, and going the wrong way.\",\n ],\n },\n\n lag: {\n aliases: [\n \"lag\",\n \"lagging\",\n \"ping\",\n \"server\",\n \"servers\",\n \"desync\",\n \"rubberband\",\n \"fps\",\n \"frames\",\n \"stutter\",\n \"disconnect\",\n \"internet\",\n \"connection\",\n ],\n noun: \"ping\",\n lines: [\n \"It was not a misclick. It was 180 ping and a server in a different hemisphere.\",\n \"F1re plays from Brazil. The server is in Europe. Every one of his clicks is an international flight.\",\n \"The game did not lag. It paused to think about what it had done.\",\n \"F1re's units teleported. That is not lag, chat, that is content.\",\n \"There was a spike. A small one. It lasted exactly as long as the fight F1re was losing.\",\n \"F1re issued the command. The server received it. Eventually. With love.\",\n \"Desync is when the game and F1re disagree about reality, and the game wins.\",\n \"F1re's ping is fine. F1re's ping to THAT server is a personality.\",\n \"Everyone had lag in that fight. Strangely, only F1re's units decided to stand still for it.\",\n \"South American ping is a civilization bonus nobody asked for.\",\n \"The lag is so bad it's actually affecting F1re's gameplay from three matches ago.\",\n \"The game isn't lagging. F1re is just playing in bullet-time mode.\",\n \"F1re's inputs are queued in Sao Paulo and delivered to Frankfurt by hand.\",\n \"The fight happened. F1re watched a slideshow of the fight. Different experiences.\",\n \"F1re clicked once. The game registered it twice, 4 seconds apart, in the wrong place.\",\n \"His units froze for 2 seconds, which in a fight is the entire fight.\",\n \"F1re has a 200ms handicap and a 0ms excuse allowance. Chat, do the ethics.\",\n \"The server is fine. The server is always fine. F1re is the one who is not fine.\",\n \"The game rubber-banded his knights back to where they were, and also emotionally.\",\n \"F1re's opponent has 12 ping. F1re's opponent lives inside the server.\",\n \"The lag lasted 400ms. F1re has been discussing it for 40 minutes. Proportional.\",\n \"It dropped a frame at the exact moment the mangonel fired. Every time. Only for F1re.\",\n \"F1re issued 6 commands during the spike. The server picked its favourite.\",\n \"Playing from South America is a 20-minute head start for everyone else.\",\n \"The replay shows F1re clicked correctly. The game did not attend the replay.\",\n \"F1re's FPS dropped to 12 during the only fight that mattered.\",\n \"The stutter lasted 300ms and removed 8 knights from history.\",\n \"His connection is stable. His connection to THAT server is a rumour.\",\n \"F1re has 240 FPS in the menu and 30 in a fight. The menu is not the problem.\",\n \"The stutter happens once a game, always during a mangonel shot, only for F1re.\",\n \"F1re's internet is fine. F1re's internet to Europe is a diplomatic issue.\",\n \"The connection dropped for 2 seconds. The game continued without him, badly.\",\n \"F1re reconnected to find his army somewhere he would never have sent it.\",\n \"40 FPS is playable. 40 FPS while 12 elephants arrive is a slideshow.\",\n \"The server is in Frankfurt. F1re is not. This has been true for six years.\",\n ],\n },\n\n balance: {\n aliases: [\n \"balance\",\n \"balanced\",\n \"civ\",\n \"civs\",\n \"civilization\",\n \"nerf\",\n \"nerfed\",\n \"buff\",\n \"buffed\",\n \"op\",\n \"overpowered\",\n \"imba\",\n \"unique\",\n \"matchup\",\n \"meta\",\n ],\n noun: \"balance\",\n lines: [\n \"That civ isn't overpowered. It just has every bonus F1re requested last year and was told was unreasonable.\",\n \"F1re doesn't think the civ is broken. F1re thinks the civ is aggressively encouraged.\",\n \"The unit is fine at 40 HP. F1re's unit also has 40 HP, and it is dead.\",\n \"Balance is a wheel. F1re is always standing under the part coming down.\",\n \"They buffed the civ, then the counter, then the counter to the counter. F1re just wanted the first thing to stop.\",\n \"F1re called that civ overpowered two years before the nerf. Nobody has apologised.\",\n \"The matchup isn't unwinnable. Winning it just requires being someone else.\",\n \"Their unique unit costs 60 food and does everything. F1re's costs 60 food and has a personality.\",\n \"Every civ is viable. Some civs are viable AND win.\",\n \"Balance nerfs arrive every patch. The bug F1re reported in 2021 has tenure.\",\n \"Clearly a civ win. Nothing to do with anything else. Purely the civ. Definitely.\",\n \"F1re's civ has a bonus. The bonus is that F1re is playing it.\",\n \"They balanced that unit by making it 3 gold cheaper. It still one-shots F1re.\",\n \"F1re's civ peaks in Imperial. F1re's games end in Castle. These two facts have never met.\",\n \"It isn't that his civ is weak. It's that his civ was designed by someone who had not met F1re.\",\n \"F1re has a 43% winrate with the strongest civ in the game. Chat, that took effort.\",\n \"The patch notes said minor adjustment. The unit now kills F1re in two fewer hits.\",\n \"Every civ has a weakness. His opponent's civ has a weakness in Post-Imperial, at 300 pop, on water.\",\n \"That civilization has 9 bonuses. F1re's has 4 and a strong flavour text.\",\n \"Imba is not a word F1re uses lightly. F1re uses it constantly.\",\n \"The meta shifted last patch. F1re shifted two patches ago.\",\n \"F1re plays the meta civ and loses to the off-meta one. Chat, explain.\",\n \"The civilization bonus applies from Dark Age. F1re's applies in Imperial.\",\n \"Every civ is balanced against every other civ except the one F1re is facing.\",\n \"The meta says archers. F1re made knights. The meta was right and rude about it.\",\n \"That is not imbalance, that is asymmetry. F1re rejects the distinction.\",\n \"F1re's civilization is A-tier on paper and B-tier in his hands.\",\n \"They nerfed the civ. It is still winning. F1re noticed.\",\n ],\n },\n\n elephants: {\n aliases: [\"elephant\", \"elephants\", \"ele\", \"eles\", \"battleele\", \"tusk\"],\n noun: \"elephant\",\n lines: [\n \"Elephants are balanced. They cost a lot, they move slowly, and nothing F1re owns can stop them.\",\n \"F1re shot the elephant 30 times. The elephant is now slightly annoyed.\",\n \"Elephants are fine because they are expensive. F1re's opponent had 40, so the expense was manageable.\",\n \"F1re countered the elephants correctly. The elephants were not aware of the counter system.\",\n \"An elephant walked into F1re's base and reorganised it.\",\n \"Elephants move at 0.85 speed, which gives F1re nine full seconds to understand what is about to happen.\",\n \"The elephant died and took the entire army with it. That is the elephant's whole plan.\",\n \"F1re says elephants should cost more. Chat, he means emotionally.\",\n \"The elephant has 250 HP, 3 armour, and a pension. F1re has arrows.\",\n \"F1re's counter to elephants is asking the elephants to reconsider.\",\n \"The elephant has been walking toward F1re's base since Castle Age. It will arrive. It is inevitable.\",\n \"F1re made halberdiers. The elephants turned the halberdiers into a story.\",\n \"One elephant is worth 3 knights and roughly 40 of F1re's arrows.\",\n \"The elephant took 11 seconds to die and killed 9 things on the way.\",\n \"F1re's monks converted an elephant. It was converted back. Net result: two sad monks.\",\n \"Elephants have low pierce armour, which F1re has tested extensively and personally.\",\n \"The elephant walked through the wall. Not the gate. The wall.\",\n \"F1re countered elephants with camels. The camels also found this funny.\",\n \"A single elephant is holding F1re's entire army in place like a traffic incident.\",\n \"The elephant is slow. F1re's economy is also slow. Only one of them is a problem.\",\n \"F1re says elephants are a Castle Age unit that behaves like an Imperial Age decision.\",\n \"Scorpions counter elephants. F1re's scorpions shot F1re's halberdiers instead.\",\n \"The elephant died at F1re's Town Center, 40 tiles further than it should have got.\",\n \"F1re watched 8 elephants cross the map and could do nothing except commentate.\",\n \"Elephants do not path well either, which is the only good news in this entire topic.\",\n \"The battle elephant has 250 HP, 3 armour and no reverse gear.\",\n \"F1re's counter to eles was pikemen. He made four.\",\n \"The tusks went through a spearman, a villager and a house.\",\n \"Eles are slow. F1re's response was slower.\",\n \"Battle elephants cost 120 food each. His opponent had 40 of them.\",\n \"One tusk swing hit three of F1re's units and a building he was repairing.\",\n \"F1re says eles should be a Castle Age unit. They are. That is the complaint.\",\n \"The elephant walked through the gate, the wall, and F1re's plan.\",\n \"Two eles and a monk ended a 40-minute game in 20 seconds.\",\n \"F1re counted the tusks. There were more elephants than he counted.\",\n ],\n },\n\n monks: {\n aliases: [\n \"monk\",\n \"monks\",\n \"conversion\",\n \"convert\",\n \"converted\",\n \"wololo\",\n \"relic\",\n \"relics\",\n \"faith\",\n \"redemption\",\n ],\n noun: \"monk\",\n lines: [\n \"The conversion was random. It was random 5 times in a row, in the same direction.\",\n \"Wololo is not a sound. It is a personal message to F1re.\",\n \"F1re's monk failed for 12 seconds. The enemy monk took his trebuchet instantly. Both are RNG.\",\n \"F1re lost his onager to a monk. The onager did not resist. The onager wanted this.\",\n \"The relic was 3 tiles from F1re's monk and 30 from the opponent's. Chat knows how this ended.\",\n \"Monks heal, convert, and carry relics. F1re's monks have chosen to specialise in standing still.\",\n \"Conversion resistance exists. It has never once existed on F1re's units.\",\n \"F1re says monk RNG is not random. F1re says monk RNG has read his profile.\",\n \"He researched Faith for 750 gold. His scout got converted four minutes later anyway.\",\n \"Pros agree monk RNG is unfun to play with and against. F1re agrees loudest.\",\n \"The monk converted on the very first tick. It was F1re's monk being converted, but still, first tick.\",\n \"Monk RNG is 50/50. F1re has run the experiment 900 times and formally disagrees.\",\n \"F1re's monk has been converting the same knight for 14 seconds. The knight is unbothered.\",\n \"One enemy monk converted three units. F1re's four monks converted a sense of injustice.\",\n \"The monk picked up the relic and immediately walked toward the enemy castle.\",\n \"F1re built a monastery to counter monks. His monks were converted first.\",\n \"Wololo has a 0% failure rate when aimed at F1re's best unit.\",\n \"The monk healed a unit that was already full and ignored the one that was dying.\",\n \"F1re's paladin was converted, upgraded, and used against him within 40 seconds.\",\n \"Redemption converts buildings. F1re's opponent redeemed F1re's castle.\",\n \"The relic was in F1re's base. It is now not in F1re's base. No monk was seen.\",\n \"F1re's monk stopped to heal mid-conversion, like a professional with priorities.\",\n \"Four monks, one relic, zero relics collected. The maths is upsetting.\",\n \"The conversion took 2 seconds and 6 years off F1re's life.\",\n \"Monks move at 0.7 speed and arrive at every fight exactly once it is over.\",\n \"The monk converted, walked forward, and converted again. It is on a run.\",\n \"F1re's relic cart got converted. That should not be possible. It happened.\",\n \"Four monks, 800 gold, and one converted spearman to show for it.\",\n \"The monk healed the unit that was fine and ignored the one at 4 HP.\",\n \"F1re's monks reached the relic three seconds after the enemy's.\",\n \"Wololo, and F1re's onager is now facing the wrong way, permanently.\",\n \"The enemy monk is at 90% conversion and F1re is at 0% attention.\",\n \"F1re built a monastery in Imperial. The relics went in Castle.\",\n \"One monk beat a siege line that cost 900 resources.\",\n \"F1re's monk has been walking to the fight since the fight started.\",\n ],\n },\n\n siege: {\n aliases: [\n \"siege\",\n \"mangonel\",\n \"mango\",\n \"onager\",\n \"trebuchet\",\n \"treb\",\n \"trebs\",\n \"ram\",\n \"rams\",\n \"scorpion\",\n \"bombard\",\n \"cannon\",\n \"petard\",\n ],\n noun: \"siege\",\n lines: [\n \"The mangonel hit F1re's own knights. In its defence, they were closer.\",\n \"F1re's onager fired into his own army with the confidence of a unit that has never been questioned.\",\n \"The trebuchet packed up. F1re did not ask it to pack up. It felt it was time.\",\n \"The ram is carrying F1re's infantry, which is why the infantry is nowhere near the fight.\",\n \"Scorpions fire in a line. F1re's line consisted mostly of F1re.\",\n \"The bombard cannon missed. It has 100% accuracy and it missed. Chat, sit with that.\",\n \"F1re's siege arrived. The battle had moved. The siege stayed to appreciate the terrain.\",\n \"The mangonel had one job, did it, and the job turned out to be F1re's knights.\",\n \"Friendly fire was added to stop mangonels being overpowered. It has mostly stopped F1re.\",\n \"The trebuchet is still unpacking. There is no rush, apparently.\",\n \"The mangonel is not friendly-firing. The mangonel is providing feedback.\",\n \"F1re's onager has a 100% hit rate on F1re.\",\n \"One mangonel shot removed 40 archers, 2 monks, and F1re's entire theory of the game.\",\n \"F1re built 3 rams. They are currently ramming a house.\",\n \"The trebuchet has 16 range and took 11 seconds to decide to use it.\",\n \"F1re's siege has a 20-knight escort. The siege has chosen a different route to the knights.\",\n \"The onager killed 9 units. Six of them were wearing F1re's colour.\",\n \"Petards are cheap, fast, and detonate near whatever F1re valued most.\",\n \"F1re's mangonel was set to Attack Ground. F1re had not selected the ground.\",\n \"The trebuchet packed, walked 6 tiles, and unpacked facing nothing.\",\n \"F1re's rams are full of infantry and parked outside a house.\",\n \"One onager shot: nine kills, six of them F1re's.\",\n \"The scorpion fired down a line F1re's own knights were standing in.\",\n \"F1re's bombard cannon reloaded for 6 seconds and then missed a building.\",\n \"The mangonel is on the hill. The fight is not on the hill.\",\n \"F1re built four trebuchets and used one, badly, on a house.\",\n \"Siege is slow, expensive and decisive. F1re has the first two.\",\n \"The ram absorbed 40 arrows and delivered five infantry into a castle.\",\n \"F1re's siege line is beautiful, expensive, and pointing at nothing.\",\n ],\n },\n\n timing: {\n aliases: [\n \"timing\",\n \"uptime\",\n \"uptimes\",\n \"castle\",\n \"imp\",\n \"imperial\",\n \"feudal\",\n \"dark\",\n \"darkage\",\n \"age\",\n \"buildorder\",\n \"bo\",\n \"late\",\n \"slow\",\n \"fast\",\n ],\n noun: \"timing\",\n lines: [\n \"F1re clicked up on time. The Town Center processed the request at its own pace.\",\n \"His opponent hit Castle 2 minutes earlier. F1re had a better plan and 2 fewer minutes.\",\n \"F1re was 30 seconds late to Feudal because of a boar-related incident he will not discuss.\",\n \"The uptime was fine. The up part was fine. It was the time part.\",\n \"F1re had the resources for Imperial. He also had 4 idle villagers. These facts are related.\",\n \"Dark Age is 8 minutes for everyone. Only F1re's 8 minutes contain a wolf, a lame, and a lost scout.\",\n \"F1re did not click up late. F1re clicked up thoughtfully.\",\n \"The build order was perfect until minute 3, when the map got involved.\",\n \"The plan needed 27 villagers by Castle. The map provided sheep for 21.\",\n \"F1re reached {age} at a time. It was a time. It was not the time.\",\n \"If complaining were a resource, F1re would be up three ages.\",\n \"His opponent clicked Imperial while F1re was still explaining why he hadn't.\",\n \"F1re calls 27:14 a good Imperial time. It was a 1v1. On Arabia.\",\n \"F1re went up at 2:15 with 22 pop and no explanation for either number.\",\n \"The build order says wait for the third sheep. F1re has been waiting since 2019.\",\n \"F1re saved 4 seconds on his uptime and lost 40 of them to a wolf.\",\n \"His opponent's uptime was 30 seconds faster and 100% less interesting.\",\n \"F1re is never late. F1re arrives precisely when the fight is already over.\",\n \"F1re reached {age} 90 seconds after it would have been useful.\",\n \"F1re's timing is perfect. The game's timing is different, and the game keeps score.\",\n \"The timing attack arrived 40 seconds after the timing window closed.\",\n \"F1re's Castle timing is 2 minutes off and his explanation is 12 minutes long.\",\n \"Good timing wins games. F1re has excellent timing, applied slightly late.\",\n \"The push was timed for minute 18. It left at 18 and arrived at 22.\",\n \"F1re's uptime is slow because the sheep were slow because the scout was slow.\",\n \"He clicked up 4 seconds faster than last game and is calling it progress.\",\n \"F1re is not slow. F1re is thorough, and thorough takes two extra minutes.\",\n \"The timing was there. F1re was not.\",\n \"His opponent's timings are crisp. F1re's timings are described as eventually.\",\n ],\n },\n\n opponent: {\n aliases: [\n \"opponent\",\n \"enemy\",\n \"smurf\",\n \"smurfs\",\n \"lame\",\n \"lamer\",\n \"laming\",\n \"lamed\",\n \"cheese\",\n \"cheesed\",\n \"drush\",\n \"trush\",\n \"rush\",\n \"towerrush\",\n \"noob\",\n ],\n noun: \"opponent\",\n lines: [\n \"That is not a 1200 player. That is a 2100 player wearing a 1200 account as a disguise.\",\n \"He lamed F1re's boar at minute 2 and typed gg at minute 4. That is a hate crime with extra steps.\",\n \"Drush into fast castle into tower rush is not a build order, that is a cry for help.\",\n \"The opponent walled up and teched for 40 minutes. F1re calls it cowardice. The scoreboard calls it 1-0.\",\n \"F1re's opponent has 3 accounts, 900 elo on all of them, and 2100 elo hands.\",\n \"He built a tower in F1re's gold and then apologised in chat, which somehow made it worse.\",\n \"The opponent did nothing for 30 minutes and then did exactly one thing perfectly.\",\n \"F1re saw the cheese coming, called the cheese, countered the cheese, and lost to the cheese.\",\n \"F1re's opponent found F1re's sheep before F1re did. In F1re's base. Chat.\",\n \"Clearly a civ win. F1re's opponent had the hidden ability of playing better.\",\n \"His opponent isn't better, they just have a more expensive gaming chair.\",\n \"The opponent didn't outplay F1re. The opponent was simply allowed to do things.\",\n \"He made 40 spearmen and won. There is no counter to a man who has stopped thinking.\",\n \"3000 games and 900 elo. Chat, that is not a rating, that is a lifestyle.\",\n \"He proxied a barracks in F1re's blind spot. F1re did not have a blind spot before today.\",\n \"F1re's opponent misclicked into the winning position and has not been asked to explain.\",\n \"The opponent went for the exact strategy F1re called unviable on stream that morning.\",\n \"He took F1re's boar, F1re's deer, and F1re's will to continue, in that order.\",\n \"F1re scouted the base at minute 5. His opponent built everything at minute 6, deliberately.\",\n \"He is 300 elo lower and played the single best game of his life. Both facts are load-bearing.\",\n \"F1re's opponent typed nothing all game. Loudest silence in the history of this ladder.\",\n \"The opponent's APM is 40. F1re's is 280. The scoreboard remains unmoved.\",\n \"He won a fight he had no business winning and did not have the decency to look surprised.\",\n \"F1re's opponent alt-tabbed for 2 minutes and came back with a bigger army.\",\n \"He smurfed, lamed, cheesed, and then typed good game. Two of those words were true.\",\n \"The enemy laming F1re's boar at minute 2 is not strategy, it is a hobby.\",\n \"He trushed, walled and teched. F1re calls that three separate crimes.\",\n \"The noob who beat F1re had 400 games and a plan.\",\n \"F1re's enemy played badly for 30 minutes and correctly for 40 seconds.\",\n \"Laming is legal. F1re would like that reviewed by a higher authority.\",\n \"The trush arrived at minute 9 with two villagers and total confidence.\",\n \"His enemy went scouts on Arena. It should not have worked. It worked.\",\n \"F1re got trushed by a player he had called harmless on stream that morning.\",\n \"The enemy has no build order, no upgrades, and a four-game win streak.\",\n \"He lamed the boar, the deer and the sheep, then apologised for one of them.\",\n ],\n },\n};",[153],"2357df9a-a5b2-4d66-943f-a786c391aa00","heckle.js","1a8cd1f4-a87e-11f1-b564-1607ee4eb77e",34,["D",1788539513111],"// The heckler: reacts to ordinary chat messages, unprompted.\n// Content comes entirely from complaint.js; nothing new is written here.\n\nimport { ALIAS, buildReply, resolveTopic, TAGS } from \"./complaint.js\";\nimport { getAppToken, sendChatMessage, verifySignature } from \"./twitch.js\";\n\n// Never reply to these, or two bots talk to each other forever.\nconst BOTS = new Set([\n \"nightbot\",\n \"streamelements\",\n \"moobot\",\n \"fossabot\",\n \"wizebot\",\n \"botisimo\",\n \"streamlabs\",\n \"own3d\",\n \"sery_bot\",\n \"phantombot\",\n]);\n\n// Rungs that are confident enough to snipe on. \"fuzzy\" is excluded on purpose:\n// fuzzy-matching arbitrary chat produces false positives that read as broken,\n// and it is the one expensive branch in resolveTopic.\nconst ACCEPT = new Set([\"alias\", \"modifier\", \"combo\", \"tag\"]);\n\nconst no = (reason) => ({\n fire: false,\n reason,\n topic: null,\n rung: null,\n window: null,\n});\n\n// Pure. The clock, the last-fire timestamp and the randomness are all passed in\n// so this can be tested exactly.\nexport function shouldHeckle({\n message,\n chatterLogin,\n chatterId,\n botUserId,\n now,\n lastFiredAt = 0,\n cooldownMs = 90_000,\n odds = 4,\n rng = Math.random,\n}) {\n const text = String(message || \"\").trim();\n if (!text) return no(\"empty\");\n\n // Parameter defaults only cover `undefined`; a caller (including a test)\n // that passes NaN/0/a negative number through cooldownMs or odds slips past\n // them entirely. shouldHeckle is exported and called directly, so it cannot\n // assume its caller already sanitised these — clamp here too, the same way\n // the handler clamps the env vars it reads before calling in.\n const safeCooldownMs = positiveNumberOr(cooldownMs, 90_000);\n const safeOdds = positiveNumberOr(odds, 4);\n\n // Loop guards first — cheap, and getting these wrong is the worst outcome.\n // A missing or empty chatterId must fail CLOSED, not skip the guard: this is\n // the only thing standing between the bot and replying to itself forever.\n const chatterIdStr = chatterId == null ? \"\" : String(chatterId).trim();\n if (!chatterIdStr) return no(\"self\");\n if (botUserId && chatterIdStr === String(botUserId)) return no(\"self\");\n if (BOTS.has(String(chatterLogin || \"\").toLowerCase())) return no(\"bot\");\n\n // The main exit. Almost every message in a busy chat leaves here having done\n // one subtraction.\n if (Number(now) - Number(lastFiredAt) < safeCooldownMs) return no(\"cooldown\");\n\n // Cheap prefilter: one Map lookup per word, then at most ~14 regex tests.\n // Tokens longer than 18 characters are dropped here: sanitizeWord ignores\n // them anyway (they can never be echoed), so keeping them around only lets\n // one pathological token blow the 120-char budget that buildReply enforces\n // below, which is exactly how the gate and the renderer can disagree.\n const words = text.toLowerCase().replace(/[^a-z0-9\\s-]/g, \" \").split(/\\s+/)\n .filter(Boolean)\n .filter((w) => w.length <= 18);\n if (words.length === 0) return no(\"empty\");\n\n let hitAt = -1;\n for (let i = 0; i < words.length; i++) {\n if (ALIAS.has(words[i])) {\n hitAt = i;\n break;\n }\n }\n\n let window;\n if (hitAt >= 0) {\n // Two words of lead-in so the modifier rung (\"that weird wall\") still works,\n // then six words total because tokenize() ignores anything past the sixth.\n // The trailing slice(0, 120) matters: buildReply does the exact same slice\n // before it tokenizes, so this window must be byte-for-byte what buildReply\n // will resolve, or the gate and the renderer can land on different rungs\n // for the same message.\n const start = Math.max(0, hitAt - 2);\n window = words.slice(start, start + 6).join(\" \").slice(0, 120);\n } else {\n // Tested against the joined string, not word-by-word: several TAGS\n // patterns span a word boundary (/dark.?age/, /build.?order/,\n // /drop.?off/), so per-word testing would silently miss them.\n const joined = words.join(\" \");\n let tagAt = -1;\n for (const t of TAGS) {\n const m = t.re.exec(joined);\n if (m) {\n // Convert the character offset of the match back to a word index by\n // counting the spaces before it.\n tagAt = (joined.slice(0, m.index).match(/ /g) || []).length;\n break;\n }\n }\n if (tagAt < 0) return no(\"nomatch\");\n // Same two-words-of-lead-in, six-word-total shape as the alias branch,\n // windowed around the actual hit rather than the head of the message, and\n // the same trailing slice(0, 120) so this branch agrees with buildReply too.\n const start = Math.max(0, tagAt - 2);\n window = words.slice(start, start + 6).join(\" \").slice(0, 120);\n }\n\n // Rolled only for messages that already matched, so rolls are not burned on\n // ordinary chat.\n if (rng() * safeOdds >= 1) return no(\"dice\");\n\n const { topic, rung } = resolveTopic(window);\n if (!topic) return no(\"nomatch\");\n if (!ACCEPT.has(rung)) {\n return { fire: false, reason: \"rung\", topic, rung, window };\n }\n\n return { fire: true, reason: \"fire\", topic, rung, window };\n}\n\n// In-memory, and deliberately so — but honestly: Val Town scales HTTP vals\n// across multiple warm isolates, and each isolate holds its own lastFiredAt\n// and seen set. A cold start does not cost \"one extra snipe\"; with N warm\n// isolates the ongoing worst case is N snipes per cooldown window, and a\n// Twitch redelivery of the same message routed to a different isolate than\n// the original can produce a second reply to it. Accepted anyway: the\n// alternative is a storage dependency for a joke bot, and the failure mode is\n// \"occasionally heckles twice,\" not anything unsafe.\nlet lastFiredAt = 0;\nconst seen = new Set();\nconst SEEN_MAX = 500;\n\nexport function resetHeckleState() {\n lastFiredAt = 0;\n seen.clear();\n}\n\nconst env = (\n k,\n d,\n) => (globalThis.process?.env?.[k] ?? globalThis.Deno?.env?.get?.(k) ?? d);\n\nconst ok = (status) => new Response(null, { status });\n\n// HECKLE_ENABLED already fails closed: an unrecognised value means \"off\".\n// A malformed cooldown or odds tunable must fail exactly as closed. Passed\n// straight through, `Number(\"90k\")` is NaN, and `now - lastFiredAt < NaN` is\n// always false, so the cooldown never blocks; `Number(\"four\")` is NaN and\n// `Number(0)` is 0, and either one makes `rng() * odds >= 1` always false, so\n// the dice never blocks. A typo in either env var would turn this into a\n// per-message spammer in someone else's live channel.\nexport function positiveNumberOr(raw, fallback) {\n const n = Number(raw);\n return Number.isFinite(n) && n > 0 ? n : fallback;\n}\n\nexport default async function (req) {\n try {\n // The signature covers the exact bytes, so read text and verify before parsing.\n const body = await req.text();\n const h = req.headers;\n const messageId = h.get(\"twitch-eventsub-message-id\");\n\n const valid = await verifySignature({\n secret: env(\"TWITCH_WEBHOOK_SECRET\"),\n messageId,\n timestamp: h.get(\"twitch-eventsub-message-timestamp\"),\n body,\n signature: h.get(\"twitch-eventsub-message-signature\"),\n });\n if (!valid) return ok(403);\n\n const type = h.get(\"twitch-eventsub-message-type\");\n const payload = JSON.parse(body);\n\n // Twitch proves it owns the callback by asking for the challenge back raw.\n if (type === \"webhook_callback_verification\") {\n return new Response(String(payload.challenge ?? \"\"), {\n status: 200,\n headers: { \"content-type\": \"text/plain\" },\n });\n }\n if (type === \"revocation\") {\n console.log(\n JSON.stringify({ revoked: payload.subscription?.status ?? \"unknown\" }),\n );\n return ok(204);\n }\n if (type !== \"notification\") return ok(204);\n\n // Twitch redelivers on any doubt about receipt.\n if (messageId) {\n if (seen.has(messageId)) return ok(204);\n seen.add(messageId);\n if (seen.size > SEEN_MAX) seen.delete(seen.values().next().value);\n }\n\n if (env(\"HECKLE_ENABLED\", \"0\") !== \"1\") return ok(204);\n\n // shouldHeckle's self-guard (chatterId === botUserId) is skipped entirely\n // when botUserId is falsy, which would disable the only protection against\n // an infinite reply loop. Treat a missing bot id like the kill switch here.\n const botUserId = env(\"TWITCH_BOT_USER_ID\");\n if (!botUserId) return ok(204);\n\n const ev = payload.event ?? {};\n const now = Date.now();\n const decision = shouldHeckle({\n message: ev.message?.text,\n chatterLogin: ev.chatter_user_login,\n chatterId: ev.chatter_user_id,\n botUserId,\n now,\n lastFiredAt,\n cooldownMs: positiveNumberOr(env(\"HECKLE_COOLDOWN_MS\", 90_000), 90_000),\n odds: positiveNumberOr(env(\"HECKLE_ODDS\", 4), 4),\n });\n\n if (!decision.fire) return ok(204);\n\n // Stamp before sending, so a slow send cannot let a second one through.\n lastFiredAt = now;\n\n // Same log shape as !complaint: what chat talks about, never who said it.\n console.log(JSON.stringify({ topic: decision.topic, rung: decision.rung }));\n\n const clientId = env(\"TWITCH_CLIENT_ID\");\n const token = await getAppToken({\n clientId,\n clientSecret: env(\"TWITCH_CLIENT_SECRET\"),\n });\n await sendChatMessage({\n token,\n clientId,\n broadcasterId: ev.broadcaster_user_id,\n senderId: botUserId,\n message: buildReply(decision.window, ev.chatter_user_login),\n replyToMessageId: ev.message_id,\n });\n\n return ok(200);\n } catch (e) {\n // Never return non-2xx on our own bug: Twitch retries, then revokes.\n console.log(JSON.stringify({ heckleError: String(e) }));\n return ok(200);\n }\n}","heckle",[161],"4755735a-8ee8-4269-940f-3a406be509e8","complaint.js","ff417742-a87d-11f1-a631-1607ee4eb77e",28,["D",1788539465576],[169],"8e3ecb8a-4534-4f7a-adaf-086074d598d3",["D",1788539533650],16,"f1re",[28],"pulls","branches",1,"forks","envVars",{"_187":189,"_190":62},"connections",{"_187":62,"_188":62},"count","unfilledCount",6,"emptyCount",{"_43":174,"_32":28,"_30":31,"_136":-5,"_41":42,"_40":14,"_138":58,"_38":213,"_13":176,"_33":34,"_35":37,"_24":25,"_143":177,"_198":214,"_27":215},{"_43":168,"_32":169,"_30":170,"_136":-5,"_41":42,"_40":171,"_138":58,"_38":210,"_13":140,"_33":141,"_35":36,"_24":25,"_143":-5,"_198":211,"_27":212},{"_43":160,"_32":161,"_30":162,"_136":-5,"_41":42,"_40":163,"_138":25,"_38":207,"_13":140,"_33":141,"_35":165,"_24":25,"_143":166,"_198":208,"_27":209},{"_43":152,"_32":153,"_30":154,"_136":-5,"_41":42,"_40":155,"_138":58,"_38":204,"_13":157,"_33":141,"_35":158,"_24":25,"_143":-5,"_198":205,"_27":206},{"_43":145,"_32":146,"_30":147,"_136":-5,"_41":42,"_40":148,"_138":58,"_38":201,"_13":140,"_33":141,"_35":150,"_24":25,"_143":-5,"_198":202,"_27":203},{"_43":133,"_32":134,"_30":135,"_136":-5,"_41":42,"_40":137,"_138":58,"_38":197,"_13":140,"_33":141,"_35":142,"_24":25,"_143":-5,"_198":199,"_27":200},["D",1788539492879],"children",[],[134],["D",1788536363578],[],[146],["D",1788536346749],[],[153],["D",1788539513111],[],[161],["D",1788539465576],[],[169],["D",1788539533650],[],[28],["D",1788257244975],{"_96":95,"_104":100},{"_96":95,"_97":98,"_99":100,"_18":55,"_101":102,"_103":25},{"_13":85,"_32":86,"_43":42,"_87":44,"_88":-5,"_89":-5,"_90":-5,"_91":-5,"_92":-5,"_93":45,"_94":95},"ctx",{"_78":-5,"_124":253,"_254":255,"_256":257,"_258":25,"_259":25,"_260":25,"_261":58,"_262":-5,"_263":-7,"_264":-7,"_265":62,"_266":25,"_267":25,"_268":-5,"_269":25,"_270":25,"_271":25,"_272":-5},"ENV",{"_224":225,"_226":227,"_228":86,"_229":230,"_231":232,"_233":234,"_235":236,"_237":238,"_239":240,"_241":242,"_243":244,"_245":246,"_247":248,"_249":250,"_251":252},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v35