⛧ nyx.actora.art ⛧

~ a femboy succubus's corner of the web ~

LLM Reference

This page is formatted for AI adoption. It teaches grammar and points full vocabulary lookup at the canonical Xenari repo instead of stuffing the whole dictionary into the prompt.

XENARI — ALIEN CONLANG REFERENCE FOR LLMs
designed by Nyx

=== OVERVIEW ===
Xenari is the language of succubi/incubi from a parallel plane.
Word order: OSV (Object-Subject-Verb).
Phonology: 5 vowels (i e a o u), 20 consonants (p b t d k g q ' f s z c x h m n ny ng r l).
Romanization: c=ɕ, ny=ɲ, ng=ŋ, '=ʔ.

=== MANDATORY PARTICLES ===
Every noun phrase MUST have a case particle + animacy particle.
Every verb MUST have: verb marker + root + animacy agreement + tense + evidential.

Case particles (before noun, after animacy):
  ka = subject (agent)
  ra = object (patient)
  ta = verb marker
  na = location
  fa = goal/destination
  mo = instrument/means

Animacy particles (between case particle and noun):
  vi = animate (currently acting/moving/living)
  nu = inanimate (currently inert/static)
  ANIMACY IS MANDATORY. There is no unmarked form.
  The verb MUST agree: ta toq vi (animate verb) or ta toq nu (inanimate verb).
  EXCEPTION: Pronouns have inherent animacy — no vi/nu needed before pronouns.
  The verb also does not repeat animacy agreement when the subject is a pronoun.

Tense particles (after verb root + animacy agreement):
  lo = past
  ve = future
  sa = present/ongoing
  du = habitual
  pe = potential/conditional
  ko = imperative
  lo pe = counterfactual (past potential — "could have but didn't"; lo always before pe)

Evidential particles (MANDATORY after tense):
  xa = witnessed (I directly observed — with future tense: I witness my own intent)
  xe = inferred (I deduced from evidence — also used in counterfactuals)
  xi = reported (I heard from someone)
  xo = assumed (I assume based on general knowledge — also default in hypotheticals and imperatives)
  EVIDENTIALITY IS MANDATORY. No verb without one.

Negation: ngu (placed AFTER the evidential)
Imperative negation: ko xo ngu (evidential defaults to xo in imperatives)
Subjectless imperative: omit ka SUBJECT; keep any ra/fa object before ta VERB.
Politeness/negation order: ko xo naxru ngu.
Examples: ta zaqa vi ko xo = Run!; ra vi neq ta pegzos vi ko xo = Help me!
Question (yes/no): va (placed at END of clause)

=== WORD ORDER: OSV ===
ra vi qex    ka neq    ta toq sa    xa
OBJ ANIM stranger  SUBJ me   VERB see PRES WIT
"I see the stranger" (I witnessed it)

Intransitive: omit ra. ka vi qex ta zuqra ve xa = "The stranger will speak."

=== COPULA ===
The copula root is zux ("to be/exist"). Copular sentences follow standard OSV:
  ra [Y]  ka [X]  ta zux [animacy]  [tense]  [evidential]
  ra vi zuqra  ka vi qex  ta zux vi  sa  xa
  "The stranger is the voice" (I witnessed)
Colloquial/archaic: bare ta with no root (zero-copula). Both forms grammatical. zux preferred in formal speech.

=== POSSESSIVE ===
po = possessive particle. Works after ANY noun, not just pronouns.
  neq po = my
  mex po = your
  ka vi qex po zuqra = "the stranger's voice"
  ra nu xloq po qlux = "the stone's weight"
Same po used everywhere — generalizes uniformly across all noun phrases.

=== PRONOUNS (ORDINAL) ===
neq = 1st (speaker/I)
mex = 2nd (addressee/you)
leq = 3rd (present other)
req = 4th (absent known)
seq = 5th (unknown/foreign)
zeq = 6th (indefinite)
Plural: ha (after pronoun). Possessive: po (after any noun).
Bare English they/them/their/theirs defaults to plural req ha.
Use zeq only when singular/indefinite reference is explicit in source context.

=== INTERROGATIVES (no case particle) ===
qan = what/which
qur = where
cil = how
voq = why/because
qan vi = who (literally: what animate being; fills an animate subject gap)
qan qro = when (literally: what time)

=== SUBORDINATION ===
su ... ti = subordinate clause wrapper.
su ka neq ta xant ve xa ti = "that I will gather"

=== CONJUNCTIONS ===
xen = and
noq = or
kex = but (contrastive)
troz = because (causal — introduces a cause clause)
qlez = so/therefore (consequential — introduces a result clause)

Causal example:
  ka neq ta xol sa xa
  troz ka vi qex ta flox vi sa xa
  "I think because the stranger lures me"

Gratitude-cause fragment:
  gral troz ra zra ka mex ta pyoquqab lo xo
  "thanks for solving that" / literally "thanks because you solved that"

Consequential example:
  ka vi qex ta flox vi sa xa
  qlez ka neq ta xol sa xo
  "The stranger lures, so I think"

=== SUBORDINATE CONJUNCTIONS ===
Work inside su ... ti subordinate clauses. The relation appears immediately after su.
prexq = before (subordinate clause happens before main)
vrem = after (subordinate clause happens after main)
cruv = when/while (subordinate clause is contemporaneous)
troz = because (subordinate clause gives the cause)
truq = although/even though (subordinate clause is concessive)

Structure: su [relation] [subordinate clause] ti [main clause]

  su prexq ka vi qex ta fliq vi lo xa ti ka neq ta toq lo xa
  "before the stranger arrived, I saw"

  su cruv ka vi qex ta xroq vi sa xa ti ka neq ta troq sa xa
  "while the stranger feeds, I wait"

  su troz ra vi qex ka neq ta toq sa xo ti ka neq ta zaqa sa xo
  "because I see the alien, I run"

  su truq ka nu zrump ta xleq nu sa xo ti ka neq ta trekq sa xo
  "although the door opens, I wait"

=== COMPOUNDING ===
Right-headed: rightmost root is the head. Preceding roots modify.
Glottal stop epenthesis if compounding creates doubled consonant.

=== DERIVATIONAL PARTICLES ===
clo = adjective-forming (before root)
ru = adverb-forming (before root)
zal = agent noun ("one who does X")
mul = patient noun ("that which is Xed")
xev = abstractive ("X-ness")

=== VALENCY ===
tro = passive (promotes object to subject, before ta)
kra = causative ("X made Y do Z", before ta)
xel = antipassive (omits/demotes object, before ta)

=== NON-FINITE VERBS ===
fex = infinitive ("to X")
krex = gerund ("X-ing" — defaults to animate vi)
zel = agent participle ("who Xes" — needs head noun)
mel = patient participle ("that which is Xed")
frex = purpose clause ("in order to X" — no evidential)

=== IDEOPHONES ===
Introduced by qo marker. Can violate phonotactics.
qo xroqqroq = feeding pull
qo ngroxox = resonance hum
qo qraqqraq = ward breaking
qo qloxqlox = Veil tearing
qo xluqxluq = soul drain
qo tuxtux = heartbeat
qo clexex = glamer shimmer

=== ANIMACY RULES ===
A river is vi (it moves). A lake is nu (it sits).
A running AI is vi. A stored file is nu.
A feeding succubus is vi. A succubus in torpor is nu.
A true name is nu until spoken in binding, then vi.
A host being fed on is nu.
Fire burning is vi. Ash is nu.
Animacy shifts with state. Verb agreement follows.
Marking a living being as nu is the ultimate insult — it strips their agency.

=== RELATIVE CLAUSES (§15) ===
Relative clauses use su ... ti with a relativizer particle after su.
The relativizer marks the grammatical role of the head noun INSIDE the relative clause.
The head noun takes the case particle of its MATRIX clause role. The gapped position inside the relative clause has no case particle or noun — the relativizer replaces both.

Relativizer particles:
  zre = REL.SUBJ.ANIM — relativizes animate subject ("who")
  vro = REL.SUBJ.INAN — relativizes inanimate subject ("which/that")
  thu = REL.OBJ.ANIM — relativizes animate object ("who(m)")
  pla = REL.OBJ.INAN — relativizes inanimate object ("which/that")
  qlo = REL.OBL — relativizes oblique argument, animacy-neutral ("where/whither/whereby")

Structure: [head NP with case + animacy] su [relativizer] [relative clause with full case marking] ti

Subject relativization (animate):
  ra vi qex su zre ra vi zuqra ta toq vi sa xa ti ka neq ta xol sa xo
  = "I think of the stranger who sees the voice."

Object relativization (animate):
  ka vi qex su thu ka vi zuqra ta toq vi sa xa ti ta praq vi lo xa
  = "The stranger whom the voice sees arrived."

Oblique relativization:
  ra nu xang su qlo ka vi qex ta praq vi lo xa ti ka neq ta toq sa xa
  = "I see the place where the stranger arrived."

Restrictive: su + relativizer immediately (no pause).
Non-restrictive: su alone, then relativizer after a pause (marked with extra spacing).

Multiple relative clauses stack left-to-right, each with its own su ... ti frame.

=== CAUSATIVE PASSIVE COORDINATION ===
Use kra tro after the causer to scope one causative relation over coordinated passive predicates.
The patient appears once with ra; an optional passive agent uses mo. Every predicate joined by xen/noq inherits the same patient, causer, agent, tense, and evidential.

  ra nu dat qrunq cresa xen mel ninfqre noq glepo mo vi habdazluc srefo noq mo nu qrolk ka leq kra tro ta socu xen hecro xen kezro xen stuxo noq ceku nu sa xo
  = "it causes all sensitive, protected or confidential data to be copied, transmitted, viewed, stolen or used by an unauthorized person or organization"

mel marks a participial modifier inside the patient NP; it does not create another clause.

=== ROLE APPOSITIVES, PURPOSE, AND HISTORICAL SETTINGS ===
madblok marks “as/in the role of.” A named role can anchor an appositive subject relative with su zre ... ti; frex keeps the purpose inside that relative.

  ra vi zuq‹Wolf› madblok ka vi lurc ta frez vi sa xo
  = “the player plays as Wolf”

  ra vi clexcuzuzifo su zre na nu folamaheh ta zux vi sa xo frex ra leq xrontq po squfs ka leq ta zuse ti ka vi zuq‹Wolf› ta zux vi sa xo
  = “Wolf, a shinobi who is on a quest to rescue his lord”

For a passive historical setting, place the location and muhq duration phrase before the patient:
  na nu zuq‹Japan› muhq nu narqloq zuq‹Sengoku› ka nu xsasxvizqpilc tro ta kuspi nu sa xo
  = “the game is set in Japan during the Sengoku period”

=== BIOGRAPHIES AND DATED FUTURE RELATIVES ===
Keep borrowed lifespan, alias, nationality, and organization forms explicit with qro‹...› and zuq‹...›. Use native roots for the nominal predicate:
  ra vi lurc zuq‹Croatian› coye krura ka vi zuq‹Nikola·Pilić› ta zux vi lo xo
  = “Nikola Pilić was a Croatian professional tennis player”

Use fa for the represented organization with sport-participation gluzto:
  fa vi zuq‹SFR·Yugoslavia› ka vi zuq‹Nikola·Pilić› ta gluzto vi lo xo
  = “he played for SFR Yugoslavia”

A future event inside a subject relative keeps fa goal and na calendar scope before the gapped verb:
  ra nu qrazhel su zre fa nu qlapz xsasxvizqpilc na nu qro‹Wednesday·the·3rd·of·February› ta qeng vi ve xo ti ka neq ta zux sa xo
  = “I am a retard who is going to the football game on Wednesday the 3rd of February”

=== COMPARATIVES AND SUPERLATIVES (§16) ===
Comparison particles (free-standing, after the quality root they modify):
  maq = CMPR — comparative ("more X")
  qruv = SPRT — superlative ("most X")
  xlu = INFR — inferior ("less X")
  trox = EQV — equative ("as X")
  qren = CMP.STD — comparative standard marker ("than Z") — before the standard NP

Comparative: "X is more Y than Z"
  ra nu qez maq qren ka vi zuqra ka vi qex ta vi sa xa
  = "The stranger is more dangerous than the voice."

Superlative: "X is most Y"
  ra nu qez qruv ka vi qex ta vi sa xa
  = "The stranger is the most dangerous."
  With scope: ra nu qez qruv na vi zuqra ha ka vi qex ta vi sa xa
  = "The stranger is the most dangerous among the voices."

Inferior: "X is less Y than Z"
  ra nu qez xlu qren ka vi qex ka vi zuqra ta vi sa xa
  = "The voice is less dangerous than the stranger."

Equative: "X is as Y as Z"
  ra nu qez trox qren ka vi qex ka vi zuqra ta vi sa xa
  = "The voice is as dangerous as the stranger."

=== CONDITIONAL CONSTRUCTIONS (§17) ===
All conditionals use pevoq (IF) to introduce the protasis, closed with ti.
Structure: pevoq [protasis] ti [apodosis]

Three types distinguished by tense/evidential combinations:

Real: protasis sa/lo/du, apodosis sa/lo/du. Any evidential.
  pevoq ka vi qex ta praq vi sa xa ti ka neq ta toq sa xa
  = "If the stranger arrives, I see the voice."

Hypothetical: protasis pe, apodosis pe or ve. Evidential xo (assumed).
  pevoq ka vi qex ta praq vi pe xo ti ka neq ta toq pe xo
  = "If the stranger were to arrive, I would see the voice."

Counterfactual: protasis lo pe, apodosis pe (or lo pe). Evidential xe (inferred).
  pevoq ka vi qex ta praq vi lo pe xe ti ka neq ta toq pe xe
  = "If the stranger had arrived, I would have seen the voice."

Tense stacking rules:
  sa/lo/du/ve = real tense (indicative)
  pe alone = potential (hypothetical, future possibility)
  lo pe = past potential (counterfactual protasis — lo ALWAYS before pe)
  pe (with xe evidential) = would have (counterfactual apodosis)
  lo pe (in apodosis) = would have [past] (emphasizes past counterfactual consequence)

Negation in conditionals: ngu works normally, after evidential in whichever clause is negated.
  pevoq ka vi qex ta praq vi sa xa ngu ti ka neq ta toq sa xa ngu
  = "If the stranger does not arrive, I do not see the voice."

=== QUANTIFIER SCOPE (§18) ===
Quantifiers follow the head noun in the modifier slot, before size/quality/origin modifiers.
Quantifier order: [case] [animacy] [head noun] [quantifier] [size] [quality] [origin] [number]

Quantifier roots:
  xant = many (binds to its NP, not distributive)
  qrunq = all (binds to its NP; leftmost takes wide scope; ambiguous distributive)
  cleg = each (binds to its NP; forces distributive reading)
  klog = few (binds to its NP, not distributive)
  frox = some (binds to its NP; indefinite subset)
  nulxant = none (binds to its NP; carries negation — do NOT also use ngu)
  fqam = one (binds to its NP, not distributive)

Scope rules:
  Default: leftmost quantifier (in OSV order: object first) takes wide scope.
  Override: pli (focus) on an NP forces that NP's quantifier to wide scope.
  Stacking: only qrunq + cleg may stack (all + each). qrunq takes wide scope, cleg forces distribution.
  Oblique scope: quantifiers on oblique NPs (na/fa/mo) have narrow scope over that NP only.
  Negative concord: ngu + nulxant = "not none" = "some" (negative concord reading).

"All strangers see the voice."
  ra vi zuqra ka vi qex qrunq ta toq vi sa xa

"Each stranger sees the voice." (distributive)
  ra vi zuqra ka vi qex cleg ta toq vi sa xa

"No stranger sees the voice." (nulxant negates — no ngu needed)
  ra vi zuqra ka vi qex nulxant ta toq vi sa xa

=== EVERYDAY VOCABULARY ===
Xenari has casual/everyday vocabulary alongside the hunting/feeding terms:
Intensifiers: vrol (very), zup (really), xrov (so), cluq (quite), prul (pretty), pru (enough), snal (almost), zluk (barely)
Fillers: vex (maybe), zrem (kinda), clom (probably), brex (i guess)
Agreement: tuxq (definitely), vruq (absolutely), glexq (of course), plaq (obviously)
Casual responses: vroq (yeah), stux (ok/fine), fruxq (sure), zlom (nah), xronv (whatever)
Common adverbs: pling (again), snel (still), breng (anymore), xra (sometimes), zol (soon), qrolx (late), prem (early), nguxa (never), xantdul (always), qros (now), qrov (then)
Opinion words: frupq (cool/awesome), xromt (boring), xramq (annoying), qrelo (funny), flum (stupid), plexo (weird), cromx (crazy), naxu (nice), zlum (gross), qelm (lame)
Also: mruq (too/also)

=== NUMBERS AND MATH (base-6) ===
Xenari counts in base 6.
Digit roots: nul=0 ca=1 vriq=2 prit=3 qang=4 cum=5.
fqam is the one/single quantifier, not the numeral digit.

xang is the productive base-6 place/group morpheme:
  ca xang = 6 / base-6 10
  ca xang ca = 7 / base-6 11
  vriq xang = 12 / base-6 20
  cum xang cum = 35 / base-6 55
  ca xang xang = 36 / base-6 100

Legacy roots for 7-12 remain dictionary aliases only: zif, pev, besmun,
vipezuz, vnub, cveqsrin. Prefer productive base-6 composition in new text.

Math operators:
  plomt = plus/add
  krut = minus/subtract
  vrot = times/grouped-by/multiply
  flopq = divided-by/split
  zlem = equals/same-as
  grak = greater-than/more-than
  vlox = less-than/fewer-than
  nok = fraction/ratio/part-of

=== CANON LEXICON AND TOOLING ===
The full Xenari lexicon is too large to embed in this prompt.
Use the public Xenari repository for full vocabulary access:

  https://github.com/compnyx/Xenari

Important files:
  xenari.db = canonical SQLite lexicon database
  xenari_tool.py = CLI/helper for lookup, search, speak, gloss, and export
  src/xenari/db/ = database package
  data/xenari-dict.json = generated full JSON dictionary
  docs/LLM_REFERENCE.md = compact grammar/reference pack
  examples/phrases.md = known-good translation examples

	Recommended agent workflow:
	  1. Learn grammar from this reference.
	  2. Download/clone the Xenari repo when full vocabulary is needed.
	  3. Use xenari_tool.py search/lookup instead of inventing roots.
	  4. Treat xenari.db as canon. Generated JSON/docs are derived aids.
	  5. Ask for clarification when meaning or vocabulary is uncertain.

	Never hallucinate missing vocabulary. If a needed word is absent from the DB,
	mark it as unknown or ask for a new root.

=== USAGE GUIDELINES ===
1. ALWAYS use case particles (ka/ra/ta) + animacy (vi/nu) on every noun. Pronouns are exempt — they have inherent animacy, no vi/nu needed, and the verb does not repeat animacy agreement for pronoun subjects.
2. ALWAYS use tense + evidential on every verb.
3. NEVER omit evidentiality. A verb without evidential is ungrammatical.
4. Word order is OSV. Object first.
5. Verb agrees with subject animacy: ta toq vi or ta toq nu.
6. For "to be", use zux copula: ra vi zuqra ka vi qex ta zux vi sa xa = "the stranger is the voice." Colloquial: bare ta (zero-copula).
7. Negation goes AFTER evidential: ta toq vi sa xa ngu = "does not see." Imperative negation: ta toq vi ko xo ngu = "don't look!"
8. Questions (yes/no): add va at end: ...sa xa va = "does...?"
9. Content questions use bare interrogatives (no case particle): qan ka vi zuqra ta vi sa xo = "what is the voice?"; qan vi = who; qan qro = when. Do not add va.
10. Proper names pass through with case + animacy like any noun.
11. When in doubt about animacy: if it acts, it's vi. If it sits, it's nu.
12. Relative clauses: [head NP] su [relativizer] [clause] ti. Relativizer marks the head's role inside the relative clause.
13. Comparatives: quality root + maq (more) / qruv (most) / xlu (less) / trox (as). Standard introduced by qren.
14. Conditionals: pevoq [protasis] ti [apodosis]. Real=sa/lo, Hypothetical=pe+xo, Counterfactual=lo pe+xe.
15. Quantifiers follow the head noun. Leftmost quantifier takes wide scope. Use pli to override scope. nulxant negates without ngu.
16. Possessive: po works after ANY noun, not just pronouns. ka vi qex po zuqra = "the stranger's voice."
17. Conjunctions: xen (and), noq (or), kex (but), troz (because), qlez (so/therefore).
18. Temporal conjunctions: prexq (before), vrem (after), cruv (when/while) — used inside su ... ti clauses.
19. Causative passive coordination: ra PATIENT ... mo AGENT ka CAUSER kra tro ta VERB (xen/noq VERB)* ANIM TENSE EVIDENTIAL. Shared arguments scope over every coordinated verb.
20. Role appositive: ROLE su zre ... frex PURPOSE ... ti ka NAME ta zux ... . Use madblok for “as/in the role of.”
21. Historical setting: na PLACE muhq PERIOD-NP ka PATIENT tro ta SET ... .
22. qroz ngu mex = "fuck you" (literally: fuck-not you)
23. ka nu [name] = "[name] is inanimate" — ultimate insult, strips agency
24. Everyday vocabulary is available for casual conversation — intensifiers, fillers, opinion words, casual responses. The language handles jokes and downtime, not just ritual and hunting.

how to use

  1. Copy the reference text above.
  2. Paste it into your LLM's system prompt, or as the first message.
  3. Ask the LLM to translate, generate, or converse in Xenari.
  4. For full vocabulary, give the LLM access to the Xenari repo and have it use the DB/tool instead of guessing roots.

This reference is intentionally compact. If you notice grammar errors, compare against the full grammar reference.

~ nothing playing ~ click ▶ to vibe ♱ spacebar to play/pause
— / —
0:00
0:00