Build first cognitive response path
This commit is contained in:
Shay
parent
b9cd6684f5
commit
71e99c5c51
15 changed files with 423 additions and 0 deletions
3
chat/__init__.py
Normal file
3
chat/__init__.py
Normal file
|
|
@ -0,0 +1,3 @@
|
|||
from .runtime import ChatRuntime
|
||||
|
||||
__all__ = ["ChatRuntime"]
|
||||
53
chat/runtime.py
Normal file
53
chat/runtime.py
Normal file
|
|
@ -0,0 +1,53 @@
|
|||
from __future__ import annotations
|
||||
|
||||
from language_packs import OOVPolicy, load_pack, load_pack_entries
|
||||
from persona.motor import PersonaMotor
|
||||
from field.state import FieldState
|
||||
from session.context import SessionContext
|
||||
|
||||
|
||||
class ChatRuntime:
|
||||
def __init__(self, pack_id: str = "en_minimal_v1") -> None:
|
||||
manifest, manifold = load_pack(pack_id)
|
||||
self._manifest = manifest
|
||||
self._context = SessionContext(manifold, persona=PersonaMotor.identity())
|
||||
self._index_by_surface = {w: i for i, w in enumerate(self._context.vocab._words)}
|
||||
self._pos_by_surface = {
|
||||
e.surface: (e.pos or e.part_of_speech or "X") for e in load_pack_entries(pack_id)
|
||||
}
|
||||
|
||||
def _apply_oov_policy(self, tokens: list[str]) -> list[str]:
|
||||
kept: list[str] = []
|
||||
for token in tokens:
|
||||
try:
|
||||
self._context.vocab.get_versor(token)
|
||||
kept.append(token)
|
||||
except KeyError:
|
||||
if self._manifest.oov_policy is OOVPolicy.FAIL_CLOSED:
|
||||
raise
|
||||
if self._manifest.oov_policy is OOVPolicy.PROPOSE_VOCAB_EXPANSION:
|
||||
raise KeyError(f"OOV token requires vocab proposal: {token}")
|
||||
return kept
|
||||
|
||||
def _syntactic_guard(self, tokens: tuple[str, ...]) -> list[str]:
|
||||
out: list[str] = []
|
||||
prev_pos: str | None = None
|
||||
for token in tokens:
|
||||
pos = self._pos_by_surface.get(token, "X")
|
||||
if pos == prev_pos:
|
||||
continue
|
||||
out.append(token)
|
||||
prev_pos = pos
|
||||
return out
|
||||
|
||||
def respond(self, text: str, max_tokens: int = 32) -> str:
|
||||
tokens = [t.strip() for t in text.split() if t.strip()]
|
||||
filtered = self._apply_oov_policy(tokens)
|
||||
if not filtered:
|
||||
return ""
|
||||
self._context.ingest(filtered)
|
||||
node_idx = self._index_by_surface.get(filtered[0], 0)
|
||||
self._context.state = FieldState(F=self._context.state.F, node=node_idx, step=self._context.state.step)
|
||||
result = self._context.respond(max_tokens=max_tokens)
|
||||
guarded = self._syntactic_guard(result.tokens)
|
||||
return " ".join(guarded)
|
||||
|
|
@ -16,6 +16,12 @@ from .schema import (
|
|||
MorphologyEntry,
|
||||
OOVPolicy,
|
||||
)
|
||||
from .compiler import (
|
||||
compile_entries_to_manifold,
|
||||
compile_entries_to_modality_vocab,
|
||||
load_pack,
|
||||
load_pack_entries,
|
||||
)
|
||||
|
||||
__all__ = [
|
||||
"AlignmentEdge",
|
||||
|
|
@ -26,4 +32,8 @@ __all__ = [
|
|||
"LexicalEntry",
|
||||
"MorphologyEntry",
|
||||
"OOVPolicy",
|
||||
"compile_entries_to_manifold",
|
||||
"compile_entries_to_modality_vocab",
|
||||
"load_pack",
|
||||
"load_pack_entries",
|
||||
]
|
||||
|
|
|
|||
123
language_packs/compiler.py
Normal file
123
language_packs/compiler.py
Normal file
|
|
@ -0,0 +1,123 @@
|
|||
from __future__ import annotations
|
||||
|
||||
import hashlib
|
||||
import json
|
||||
from pathlib import Path
|
||||
|
||||
import numpy as np
|
||||
|
||||
from algebra.cl41 import geometric_product
|
||||
from algebra.versor import unitize_versor
|
||||
from language_packs.schema import LanguagePackManifest, LanguageRole, LexicalEntry, OOVPolicy
|
||||
from sensorium.protocol import CL41_DIM, ModalityVocabulary
|
||||
from vocab.manifold import VocabManifold
|
||||
|
||||
|
||||
def _hash_to_blade(name: str, salt: str) -> int:
|
||||
digest = hashlib.sha256(f"{salt}:{name}".encode("utf-8")).digest()
|
||||
return int.from_bytes(digest[:2], "big") % CL41_DIM
|
||||
|
||||
|
||||
def _hash_unit(name: str, salt: str) -> float:
|
||||
digest = hashlib.sha256(f"{salt}:{name}".encode("utf-8")).digest()
|
||||
return int.from_bytes(digest[:4], "big") / 2**32
|
||||
|
||||
|
||||
def _feature_rotor(name: str, salt: str, weight: float) -> np.ndarray:
|
||||
# Grade-2 block indices 6..15.
|
||||
idx = 6 + (_hash_to_blade(name, f"{salt}:biv") % 10)
|
||||
theta = (0.2 + 0.8 * _hash_unit(name, f"{salt}:angle")) * weight
|
||||
rotor = np.zeros(CL41_DIM, dtype=np.float32)
|
||||
rotor[0] = np.cos(theta)
|
||||
rotor[idx] = np.sin(theta)
|
||||
return rotor
|
||||
|
||||
|
||||
def _entry_to_coordinate(entry: LexicalEntry) -> np.ndarray:
|
||||
vec = np.zeros(CL41_DIM, dtype=np.float32)
|
||||
vec[0] = 1.0
|
||||
|
||||
pos = (entry.pos or entry.part_of_speech or "").lower()
|
||||
for domain in entry.semantic_domains:
|
||||
vec = geometric_product(vec, _feature_rotor(domain.lower(), "domain", 0.7))
|
||||
|
||||
if pos:
|
||||
vec = geometric_product(vec, _feature_rotor(pos, "pos", 0.35))
|
||||
|
||||
for tag in entry.morphology_tags:
|
||||
vec = geometric_product(vec, _feature_rotor(tag.lower(), "morph", 0.15))
|
||||
|
||||
vec = geometric_product(vec, _feature_rotor(entry.lemma.lower(), "lemma", 0.1))
|
||||
vec = geometric_product(vec, _feature_rotor(entry.surface.lower(), "surface", 0.05))
|
||||
return unitize_versor(vec)
|
||||
|
||||
|
||||
def compile_entries_to_manifold(entries: list[LexicalEntry]) -> VocabManifold:
|
||||
manifold = VocabManifold()
|
||||
for entry in entries:
|
||||
versor = _entry_to_coordinate(entry)
|
||||
manifold.add(entry.surface, versor)
|
||||
return manifold
|
||||
|
||||
|
||||
def compile_entries_to_modality_vocab(entries: list[LexicalEntry]) -> ModalityVocabulary[str]:
|
||||
vocab: ModalityVocabulary[str] = ModalityVocabulary()
|
||||
for entry in entries:
|
||||
point = _entry_to_coordinate(entry)
|
||||
vocab.register_point(entry.surface, point)
|
||||
return vocab
|
||||
|
||||
|
||||
def _parse_entry(payload: dict) -> LexicalEntry:
|
||||
return LexicalEntry(
|
||||
entry_id=payload["entry_id"],
|
||||
surface=payload["surface"],
|
||||
lemma=payload.get("lemma", payload["surface"]),
|
||||
language=payload["language"],
|
||||
part_of_speech=payload.get("part_of_speech"),
|
||||
pos=payload.get("pos"),
|
||||
morphology_id=payload.get("morphology_id"),
|
||||
morphology_tags=tuple(payload.get("morphology_tags", [])),
|
||||
semantic_domains=tuple(payload.get("semantic_domains", [])),
|
||||
manifold_point_checksum=payload.get("manifold_point_checksum"),
|
||||
provenance_ids=tuple(payload.get("provenance_ids", [])),
|
||||
)
|
||||
|
||||
|
||||
def load_pack(pack_id: str) -> tuple[LanguagePackManifest, VocabManifold]:
|
||||
pack_dir = Path(__file__).parent / "data" / pack_id
|
||||
manifest_path = pack_dir / "manifest.json"
|
||||
lexicon_path = pack_dir / "lexicon.jsonl"
|
||||
|
||||
manifest_payload = json.loads(manifest_path.read_text(encoding="utf-8"))
|
||||
lexicon_bytes = lexicon_path.read_bytes()
|
||||
checksum = hashlib.sha256(lexicon_bytes).hexdigest()
|
||||
if checksum != manifest_payload["checksum"]:
|
||||
raise ValueError(f"Checksum mismatch for {pack_id}: {checksum} != {manifest_payload['checksum']}")
|
||||
|
||||
entries = load_pack_entries(pack_id)
|
||||
|
||||
manifest = LanguagePackManifest(
|
||||
pack_id=manifest_payload["pack_id"],
|
||||
language=manifest_payload["language"],
|
||||
role=LanguageRole(manifest_payload["role"]),
|
||||
script=manifest_payload["script"],
|
||||
normalization_policy=manifest_payload["normalization_policy"],
|
||||
source_manifest=manifest_payload["source_manifest"],
|
||||
determinism_class=manifest_payload["determinism_class"],
|
||||
checksum=manifest_payload["checksum"],
|
||||
version=manifest_payload.get("version", "1.0.0"),
|
||||
gate_engaged=manifest_payload.get("gate_engaged", False),
|
||||
oov_policy=OOVPolicy(manifest_payload.get("oov_policy", OOVPolicy.FAIL_CLOSED.value)),
|
||||
)
|
||||
return manifest, compile_entries_to_manifold(entries)
|
||||
|
||||
|
||||
def load_pack_entries(pack_id: str) -> list[LexicalEntry]:
|
||||
pack_dir = Path(__file__).parent / "data" / pack_id
|
||||
lexicon_path = pack_dir / "lexicon.jsonl"
|
||||
entries: list[LexicalEntry] = []
|
||||
for line in lexicon_path.read_text(encoding="utf-8").splitlines():
|
||||
if line.strip():
|
||||
entries.append(_parse_entry(json.loads(line)))
|
||||
return entries
|
||||
60
language_packs/data/en_minimal_v1/lexicon.jsonl
Normal file
60
language_packs/data/en_minimal_v1/lexicon.jsonl
Normal file
|
|
@ -0,0 +1,60 @@
|
|||
{"entry_id":"en-001","surface":"yes","lemma":"yes","language":"en","pos":"PART","semantic_domains":["affirmation","logic"],"morphology_tags":["particle"]}
|
||||
{"entry_id":"en-002","surface":"no","lemma":"no","language":"en","pos":"PART","semantic_domains":["negation","logic"],"morphology_tags":["particle"]}
|
||||
{"entry_id":"en-003","surface":"not","lemma":"not","language":"en","pos":"PART","semantic_domains":["negation","logic"],"morphology_tags":["adverbial"]}
|
||||
{"entry_id":"en-004","surface":"if","lemma":"if","language":"en","pos":"SCONJ","semantic_domains":["condition","logic"],"morphology_tags":["conjunction"]}
|
||||
{"entry_id":"en-005","surface":"then","lemma":"then","language":"en","pos":"ADV","semantic_domains":["sequence","logic"],"morphology_tags":["adverbial"]}
|
||||
{"entry_id":"en-006","surface":"because","lemma":"because","language":"en","pos":"SCONJ","semantic_domains":["causality","logic"],"morphology_tags":["conjunction"]}
|
||||
{"entry_id":"en-007","surface":"can","lemma":"can","language":"en","pos":"AUX","semantic_domains":["ability","modality"],"morphology_tags":["auxiliary"]}
|
||||
{"entry_id":"en-008","surface":"cannot","lemma":"cannot","language":"en","pos":"AUX","semantic_domains":["negation","modality"],"morphology_tags":["auxiliary","negative"]}
|
||||
{"entry_id":"en-009","surface":"I","lemma":"I","language":"en","pos":"PRON","semantic_domains":["self","agent"],"morphology_tags":["pronoun","first_person"]}
|
||||
{"entry_id":"en-010","surface":"you","lemma":"you","language":"en","pos":"PRON","semantic_domains":["other","agent"],"morphology_tags":["pronoun","second_person"]}
|
||||
{"entry_id":"en-011","surface":"it","lemma":"it","language":"en","pos":"PRON","semantic_domains":["object","reference"],"morphology_tags":["pronoun","third_person"]}
|
||||
{"entry_id":"en-012","surface":"we","lemma":"we","language":"en","pos":"PRON","semantic_domains":["group","agent"],"morphology_tags":["pronoun","first_person_plural"]}
|
||||
{"entry_id":"en-013","surface":"they","lemma":"they","language":"en","pos":"PRON","semantic_domains":["group","other"],"morphology_tags":["pronoun","third_person_plural"]}
|
||||
{"entry_id":"en-014","surface":"am","lemma":"be","language":"en","pos":"AUX","semantic_domains":["being","identity"],"morphology_tags":["present","first_person"]}
|
||||
{"entry_id":"en-015","surface":"is","lemma":"be","language":"en","pos":"AUX","semantic_domains":["being","identity"],"morphology_tags":["present","third_person_singular"]}
|
||||
{"entry_id":"en-016","surface":"are","lemma":"be","language":"en","pos":"AUX","semantic_domains":["being","identity"],"morphology_tags":["present","plural"]}
|
||||
{"entry_id":"en-017","surface":"know","lemma":"know","language":"en","pos":"VERB","semantic_domains":["knowledge","cognition"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-018","surface":"think","lemma":"think","language":"en","pos":"VERB","semantic_domains":["cognition","reason"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-019","surface":"mean","lemma":"mean","language":"en","pos":"VERB","semantic_domains":["semantics","reason"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-020","surface":"say","lemma":"say","language":"en","pos":"VERB","semantic_domains":["speech","communication"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-021","surface":"ask","lemma":"ask","language":"en","pos":"VERB","semantic_domains":["question","communication"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-022","surface":"answer","lemma":"answer","language":"en","pos":"NOUN","semantic_domains":["response","communication"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-023","surface":"question","lemma":"question","language":"en","pos":"NOUN","semantic_domains":["question","reason"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-024","surface":"word","lemma":"word","language":"en","pos":"NOUN","semantic_domains":["logos","language"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-025","surface":"truth","lemma":"truth","language":"en","pos":"NOUN","semantic_domains":["logos","epistemic"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-026","surface":"light","lemma":"light","language":"en","pos":"NOUN","semantic_domains":["logos","illumination"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-027","surface":"life","lemma":"life","language":"en","pos":"NOUN","semantic_domains":["logos","vitality"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-028","surface":"beginning","lemma":"beginning","language":"en","pos":"NOUN","semantic_domains":["logos","origin"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-029","surface":"create","lemma":"create","language":"en","pos":"VERB","semantic_domains":["logos","formation"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-030","surface":"spirit","lemma":"spirit","language":"en","pos":"NOUN","semantic_domains":["logos","breath"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-031","surface":"order","lemma":"order","language":"en","pos":"NOUN","semantic_domains":["logos","structure"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-032","surface":"reason","lemma":"reason","language":"en","pos":"NOUN","semantic_domains":["logos","cognition"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-033","surface":"name","lemma":"name","language":"en","pos":"NOUN","semantic_domains":["identity","language"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-034","surface":"time","lemma":"time","language":"en","pos":"NOUN","semantic_domains":["temporal","sequence"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-035","surface":"place","lemma":"place","language":"en","pos":"NOUN","semantic_domains":["spatial","location"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-036","surface":"thing","lemma":"thing","language":"en","pos":"NOUN","semantic_domains":["object","reference"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-037","surface":"begin","lemma":"begin","language":"en","pos":"VERB","semantic_domains":["origin","sequence"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-038","surface":"end","lemma":"end","language":"en","pos":"NOUN","semantic_domains":["boundary","sequence"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-039","surface":"good","lemma":"good","language":"en","pos":"ADJ","semantic_domains":["value","evaluation"],"morphology_tags":["adjective","positive"]}
|
||||
{"entry_id":"en-040","surface":"evil","lemma":"evil","language":"en","pos":"ADJ","semantic_domains":["value","evaluation"],"morphology_tags":["adjective","negative"]}
|
||||
{"entry_id":"en-041","surface":"love","lemma":"love","language":"en","pos":"NOUN","semantic_domains":["affect","relation"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-042","surface":"peace","lemma":"peace","language":"en","pos":"NOUN","semantic_domains":["affect","order"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-043","surface":"power","lemma":"power","language":"en","pos":"NOUN","semantic_domains":["ability","agency"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-044","surface":"will","lemma":"will","language":"en","pos":"NOUN","semantic_domains":["agency","modality"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-045","surface":"hear","lemma":"hear","language":"en","pos":"VERB","semantic_domains":["perception","audio"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-046","surface":"see","lemma":"see","language":"en","pos":"VERB","semantic_domains":["perception","vision"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-047","surface":"speak","lemma":"speak","language":"en","pos":"VERB","semantic_domains":["speech","communication"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-048","surface":"listen","lemma":"listen","language":"en","pos":"VERB","semantic_domains":["attention","audio"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-049","surface":"walk","lemma":"walk","language":"en","pos":"VERB","semantic_domains":["motion","body"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-050","surface":"path","lemma":"path","language":"en","pos":"NOUN","semantic_domains":["motion","sequence"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-051","surface":"field","lemma":"field","language":"en","pos":"NOUN","semantic_domains":["geometry","state"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-052","surface":"form","lemma":"form","language":"en","pos":"NOUN","semantic_domains":["geometry","structure"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-053","surface":"memory","lemma":"memory","language":"en","pos":"NOUN","semantic_domains":["storage","cognition"],"morphology_tags":["noun","singular"]}
|
||||
{"entry_id":"en-054","surface":"recall","lemma":"recall","language":"en","pos":"VERB","semantic_domains":["memory","cognition"],"morphology_tags":["verb","present"]}
|
||||
{"entry_id":"en-055","surface":"with","lemma":"with","language":"en","pos":"ADP","semantic_domains":["relation","association"],"morphology_tags":["preposition"]}
|
||||
{"entry_id":"en-056","surface":"in","lemma":"in","language":"en","pos":"ADP","semantic_domains":["relation","location"],"morphology_tags":["preposition"]}
|
||||
{"entry_id":"en-057","surface":"from","lemma":"from","language":"en","pos":"ADP","semantic_domains":["relation","origin"],"morphology_tags":["preposition"]}
|
||||
{"entry_id":"en-058","surface":"to","lemma":"to","language":"en","pos":"PART","semantic_domains":["relation","direction"],"morphology_tags":["particle"]}
|
||||
{"entry_id":"en-059","surface":"what","lemma":"what","language":"en","pos":"PRON","semantic_domains":["question","reference"],"morphology_tags":["pronoun","interrogative"]}
|
||||
{"entry_id":"en-060","surface":"who","lemma":"who","language":"en","pos":"PRON","semantic_domains":["question","agent"],"morphology_tags":["pronoun","interrogative"]}
|
||||
13
language_packs/data/en_minimal_v1/manifest.json
Normal file
13
language_packs/data/en_minimal_v1/manifest.json
Normal file
|
|
@ -0,0 +1,13 @@
|
|||
{
|
||||
"pack_id": "en_minimal_v1",
|
||||
"language": "en",
|
||||
"role": "operational_base",
|
||||
"script": "Latin",
|
||||
"normalization_policy": "unitize_versor",
|
||||
"source_manifest": "en_minimal_v1.lexicon.jsonl",
|
||||
"determinism_class": "D0",
|
||||
"checksum": "3b2804670b37fe51e2c93209a048df0f1fb1a426ea0ed3580a6fd08021b2bc70",
|
||||
"version": "1.0.0",
|
||||
"gate_engaged": true,
|
||||
"oov_policy": "tagged_fallback"
|
||||
}
|
||||
7
language_packs/data/grc_logos_micro_v1/lexicon.jsonl
Normal file
7
language_packs/data/grc_logos_micro_v1/lexicon.jsonl
Normal file
|
|
@ -0,0 +1,7 @@
|
|||
{"entry_id":"grc-001","surface":"λόγος","lemma":"λόγος","language":"grc","pos":"NOUN","semantic_domains":["logos","word"],"morphology_tags":["noun","nominative","singular"]}
|
||||
{"entry_id":"grc-002","surface":"ἀρχή","lemma":"ἀρχή","language":"grc","pos":"NOUN","semantic_domains":["logos","beginning"],"morphology_tags":["noun","nominative","singular"]}
|
||||
{"entry_id":"grc-003","surface":"φῶς","lemma":"φῶς","language":"grc","pos":"NOUN","semantic_domains":["logos","light"],"morphology_tags":["noun","nominative","singular"]}
|
||||
{"entry_id":"grc-004","surface":"ζωή","lemma":"ζωή","language":"grc","pos":"NOUN","semantic_domains":["logos","life"],"morphology_tags":["noun","nominative","singular"]}
|
||||
{"entry_id":"grc-005","surface":"ἀλήθεια","lemma":"ἀλήθεια","language":"grc","pos":"NOUN","semantic_domains":["logos","truth"],"morphology_tags":["noun","nominative","singular"]}
|
||||
{"entry_id":"grc-006","surface":"πνεῦμα","lemma":"πνεῦμα","language":"grc","pos":"NOUN","semantic_domains":["logos","spirit"],"morphology_tags":["noun","nominative","singular"]}
|
||||
{"entry_id":"grc-007","surface":"κτίζω","lemma":"κτίζω","language":"grc","pos":"VERB","semantic_domains":["logos","create"],"morphology_tags":["verb","present","active"]}
|
||||
13
language_packs/data/grc_logos_micro_v1/manifest.json
Normal file
13
language_packs/data/grc_logos_micro_v1/manifest.json
Normal file
|
|
@ -0,0 +1,13 @@
|
|||
{
|
||||
"pack_id": "grc_logos_micro_v1",
|
||||
"language": "grc",
|
||||
"role": "depth_relation",
|
||||
"script": "Greek",
|
||||
"normalization_policy": "unitize_versor",
|
||||
"source_manifest": "grc_logos_micro_v1.lexicon.jsonl",
|
||||
"determinism_class": "D0",
|
||||
"checksum": "cf8930df2a4faf1b8d87916fb707d44fa3034685a7f14b56fcde1bef794547c8",
|
||||
"version": "1.0.0",
|
||||
"gate_engaged": false,
|
||||
"oov_policy": "fail_closed"
|
||||
}
|
||||
7
language_packs/data/he_logos_micro_v1/lexicon.jsonl
Normal file
7
language_packs/data/he_logos_micro_v1/lexicon.jsonl
Normal file
|
|
@ -0,0 +1,7 @@
|
|||
{"entry_id":"he-001","surface":"דבר","lemma":"דבר","language":"he","pos":"NOUN","semantic_domains":["logos","word"],"morphology_tags":["root:DBR","noun"]}
|
||||
{"entry_id":"he-002","surface":"ראשית","lemma":"ראשית","language":"he","pos":"NOUN","semantic_domains":["logos","beginning"],"morphology_tags":["root:RASH","noun"]}
|
||||
{"entry_id":"he-003","surface":"אור","lemma":"אור","language":"he","pos":"NOUN","semantic_domains":["logos","light"],"morphology_tags":["root:AWR","noun"]}
|
||||
{"entry_id":"he-004","surface":"חיים","lemma":"חיים","language":"he","pos":"NOUN","semantic_domains":["logos","life"],"morphology_tags":["root:CHYY","noun"]}
|
||||
{"entry_id":"he-005","surface":"אמת","lemma":"אמת","language":"he","pos":"NOUN","semantic_domains":["logos","truth"],"morphology_tags":["root:AMT","noun"]}
|
||||
{"entry_id":"he-006","surface":"רוח","lemma":"רוח","language":"he","pos":"NOUN","semantic_domains":["logos","spirit"],"morphology_tags":["root:RWH","noun"]}
|
||||
{"entry_id":"he-007","surface":"ברא","lemma":"ברא","language":"he","pos":"VERB","semantic_domains":["logos","create"],"morphology_tags":["root:BRA","verb","qal"]}
|
||||
13
language_packs/data/he_logos_micro_v1/manifest.json
Normal file
13
language_packs/data/he_logos_micro_v1/manifest.json
Normal file
|
|
@ -0,0 +1,13 @@
|
|||
{
|
||||
"pack_id": "he_logos_micro_v1",
|
||||
"language": "he",
|
||||
"role": "depth_root",
|
||||
"script": "Hebrew",
|
||||
"normalization_policy": "unitize_versor",
|
||||
"source_manifest": "he_logos_micro_v1.lexicon.jsonl",
|
||||
"determinism_class": "D0",
|
||||
"checksum": "ef21e0cf2b15663603e64f94bdd42634acfde969280e556e915c1fbacc055921",
|
||||
"version": "1.0.0",
|
||||
"gate_engaged": false,
|
||||
"oov_policy": "fail_closed"
|
||||
}
|
||||
|
|
@ -94,7 +94,9 @@ class LexicalEntry:
|
|||
lemma: str
|
||||
language: str
|
||||
part_of_speech: str | None = None
|
||||
pos: str | None = None
|
||||
morphology_id: str | None = None
|
||||
morphology_tags: tuple[str, ...] = field(default_factory=tuple)
|
||||
semantic_domains: tuple[str, ...] = field(default_factory=tuple)
|
||||
manifold_point_checksum: str | None = None
|
||||
provenance_ids: tuple[str, ...] = field(default_factory=tuple)
|
||||
|
|
|
|||
31
probe/repl.py
Normal file
31
probe/repl.py
Normal file
|
|
@ -0,0 +1,31 @@
|
|||
from __future__ import annotations
|
||||
|
||||
from language_packs import load_pack
|
||||
|
||||
|
||||
def field_walk(seed: str, steps: int = 4) -> list[str]:
|
||||
_, vocab = load_pack("en_minimal_v1")
|
||||
F = vocab.get_versor(seed)
|
||||
walk = [seed]
|
||||
idx = -1
|
||||
for _ in range(steps - 1):
|
||||
word, idx = vocab.nearest(F, exclude_idx=idx)
|
||||
walk.append(word)
|
||||
F = vocab.get_versor(word)
|
||||
return walk
|
||||
|
||||
|
||||
def main() -> None:
|
||||
while True:
|
||||
text = input("> ").strip()
|
||||
if text in {"quit", "exit"}:
|
||||
break
|
||||
try:
|
||||
chain = field_walk(text)
|
||||
print(f"[field walk: {' → '.join(chain)}]")
|
||||
except KeyError:
|
||||
print("[unknown token]")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
|
@ -7,6 +7,8 @@ requires-python = ">=3.11"
|
|||
dependencies = [
|
||||
"numpy>=1.26",
|
||||
"mlx>=0.18; sys_platform == 'darwin'",
|
||||
"pytest>=9.0.3",
|
||||
"pytest-asyncio>=1.3.0",
|
||||
]
|
||||
|
||||
[project.optional-dependencies]
|
||||
|
|
|
|||
60
tests/test_holonomy_resonance.py
Normal file
60
tests/test_holonomy_resonance.py
Normal file
|
|
@ -0,0 +1,60 @@
|
|||
from __future__ import annotations
|
||||
|
||||
import numpy as np
|
||||
|
||||
from algebra.cga import cga_inner
|
||||
from algebra.holonomy import holonomy_encode, holonomy_similarity
|
||||
from language_packs import load_pack
|
||||
|
||||
|
||||
def _encode(manifold, tokens: list[str]) -> np.ndarray:
|
||||
return holonomy_encode([manifold.get_versor(t) for t in tokens])
|
||||
|
||||
|
||||
def test_aligned_clauses_have_higher_similarity_than_unrelated():
|
||||
_, en = load_pack("en_minimal_v1")
|
||||
_, he = load_pack("he_logos_micro_v1")
|
||||
_, grc = load_pack("grc_logos_micro_v1")
|
||||
|
||||
en_clause = _encode(en, ["word", "beginning", "with", "truth"])
|
||||
he_clause = _encode(he, ["דבר", "ראשית", "אמת"])
|
||||
grc_clause = _encode(grc, ["λόγος", "ἀρχή", "ἀλήθεια"])
|
||||
grc_unrelated = _encode(grc, ["ζωή", "πνεῦμα", "κτίζω"])
|
||||
|
||||
aligned = (holonomy_similarity(en_clause, he_clause) + holonomy_similarity(en_clause, grc_clause)) / 2.0
|
||||
unrelated = holonomy_similarity(en_clause, grc_unrelated)
|
||||
assert aligned > unrelated
|
||||
|
||||
|
||||
def test_triple_alignment_closer_than_other_triples():
|
||||
_, en = load_pack("en_minimal_v1")
|
||||
_, he = load_pack("he_logos_micro_v1")
|
||||
_, grc = load_pack("grc_logos_micro_v1")
|
||||
|
||||
word_trip = [
|
||||
en.get_versor("word"),
|
||||
he.get_versor("דבר"),
|
||||
grc.get_versor("λόγος"),
|
||||
]
|
||||
aligned_score = np.mean(
|
||||
[
|
||||
cga_inner(en.get_versor("word"), he.get_versor("דבר")),
|
||||
cga_inner(en.get_versor("word"), grc.get_versor("λόγος")),
|
||||
cga_inner(he.get_versor("דבר"), grc.get_versor("λόγος")),
|
||||
]
|
||||
)
|
||||
misaligned_score = np.mean(
|
||||
[
|
||||
cga_inner(en.get_versor("word"), he.get_versor("ראשית")),
|
||||
cga_inner(en.get_versor("word"), grc.get_versor("πνεῦμα")),
|
||||
cga_inner(he.get_versor("דבר"), grc.get_versor("ἀρχή")),
|
||||
]
|
||||
)
|
||||
assert aligned_score > misaligned_score
|
||||
|
||||
|
||||
def test_word_order_permutation_changes_holonomy():
|
||||
_, en = load_pack("en_minimal_v1")
|
||||
a = _encode(en, ["word", "truth", "light", "life"])
|
||||
b = _encode(en, ["life", "word", "truth", "light"])
|
||||
assert abs(holonomy_similarity(a, b) - holonomy_similarity(a, a)) > 1e-4
|
||||
26
tests/test_language_pack_runtime.py
Normal file
26
tests/test_language_pack_runtime.py
Normal file
|
|
@ -0,0 +1,26 @@
|
|||
from __future__ import annotations
|
||||
|
||||
from chat.runtime import ChatRuntime
|
||||
from language_packs import load_pack
|
||||
|
||||
|
||||
def test_load_pack_and_vocab_size():
|
||||
manifest, manifold = load_pack("en_minimal_v1")
|
||||
assert manifest.pack_id == "en_minimal_v1"
|
||||
assert len(manifold) == 60
|
||||
|
||||
|
||||
def test_light_neighborhood_not_negation():
|
||||
_, manifold = load_pack("en_minimal_v1")
|
||||
light = manifold.get_versor("light")
|
||||
word, _ = manifold.nearest(light, exclude_idx=25)
|
||||
assert word not in {"no", "cannot"}
|
||||
|
||||
|
||||
def test_chat_runtime_responds_and_varies():
|
||||
runtime = ChatRuntime("en_minimal_v1")
|
||||
a = runtime.respond("what is truth", max_tokens=8)
|
||||
b = runtime.respond("what is light", max_tokens=8)
|
||||
assert a.strip()
|
||||
assert b.strip()
|
||||
assert a != b
|
||||
Loading…
Reference in a new issue