35 lines
978 B
Python
35 lines
978 B
Python
"""
|
|
sensorium — Modality protocol layer.
|
|
|
|
Every surface signal (text, vision, audio, motor) is converted to a
|
|
(32,) Cl(4,1) multivector by a ProjectionHead before it reaches
|
|
core_ingest/ or ingest/gate.py. The gate is never modified.
|
|
|
|
The ProjectionHead is the Logos-recovery boundary for its modality:
|
|
all inputs reduce to words in the versor manifold. There is no
|
|
multimodal fusion problem because every modality becomes a versor
|
|
before the field sees it. One space.
|
|
|
|
Adding a modality requires:
|
|
1. One file in sensorium/adapters/<modality>.py
|
|
2. A registry entry in sensorium/registry.py
|
|
3. Nothing else — no existing layer is touched.
|
|
"""
|
|
|
|
from sensorium.protocol import (
|
|
Modality,
|
|
ProjectionHead,
|
|
SurfaceDecoder,
|
|
ModalityVocabulary,
|
|
ModalityPack,
|
|
)
|
|
from sensorium.registry import ModalityRegistry
|
|
|
|
__all__ = [
|
|
"Modality",
|
|
"ProjectionHead",
|
|
"SurfaceDecoder",
|
|
"ModalityVocabulary",
|
|
"ModalityPack",
|
|
"ModalityRegistry",
|
|
]
|