mirror of
https://github.com/allaunthefox/Research-Stack.git
synced 2026-07-31 03:05:21 +00:00
357 lines
18 KiB
Python
357 lines
18 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
Ingest: GCCL-GEC Spec — Full Compression Architecture
|
||
======================================================
|
||
Geometric-Cognitive Compression Law / Glyph Eigen Codec
|
||
Byte-exact compression via lawful callable glyph kernels.
|
||
"""
|
||
|
||
import json, time
|
||
from pathlib import Path
|
||
|
||
RESEARCH_STACK = Path("/home/allaun/Documents/Research Stack")
|
||
|
||
GCCL_GEC = {
|
||
"id": "gccl-gec-spec-v1",
|
||
"source": "USER formal spec — complete compression architecture",
|
||
"title": "GCCL-GEC: Geometric-Cognitive Compression Law / Glyph Eigen Codec — Full Specification",
|
||
"date": "2026-05-07",
|
||
|
||
"purpose": (
|
||
"Compress a byte corpus by finding the smallest lawful executable "
|
||
"glyph/eigen/manifold program that reconstructs the exact original bytes. "
|
||
"Do not store the text. Store the cheapest lawful generator of the byte projection."
|
||
),
|
||
|
||
"archive_structure": {
|
||
"formula": "A = D ⊕ 𝔊 ⊕ Χ ⊕ Τ ⊕ 𝕌 ⊕ Γ ⊕ Θ ⊕ Ε ⊕ R",
|
||
"components": {
|
||
"D": {"name": "Deterministic Decompressor", "role": "Loads profile, interprets packets, emits exact bytes"},
|
||
"𝔊": {"name": "GlyphBook", "role": "Maps printable codepoints to callable reconstruction kernels"},
|
||
"Χ": {"name": "ChiralityBook", "role": "Maps chirality vectors to law-axes for each glyph"},
|
||
"Τ": {"name": "TypeBook", "role": "Maps datatype witnesses to structural generative laws"},
|
||
"𝕌": {"name": "EigenBook", "role": "Stores reusable eigenbasis/spectrum/coefficient descriptors"},
|
||
"Γ": {"name": "Glyph Packet Stream", "role": "Atomic compression units — not characters, but kernel invocations"},
|
||
"Θ": {"name": "Parameter Stream", "role": "Side-stream of integer/arithmetic-coded payload data"},
|
||
"Ε": {"name": "Residual Stream", "role": "Exact byte repair — honesty layer where speculative compressors die"},
|
||
"R": {"name": "Receipt/Checksum/Audit", "role": "SHA256 verification + audit trail"}
|
||
},
|
||
"compact_equation": "C = Π_B(Bind_GCCL(𝔊, Χ, Τ, 𝕌, Γ, Θ, Ε))",
|
||
"compact_meaning": "Corpus = byte projection of GCCL-bound kernel composition"
|
||
},
|
||
|
||
"fundamental_packet": {
|
||
"formula": "Γᵢ = γᵢ ⊗ χᵢ ⊗ κᵢ ⊗ τᵢ ⊗ UᵢΛᵢaᵢ ⊗ θᵢ ⊗ εᵢ",
|
||
"fields": {
|
||
"γᵢ": {"name": "visible_glyph", "domain": "emoji / math symbol / PUA codepoint / Unicode printable", "meaning": "Invokes a specific reconstruction kernel"},
|
||
"χᵢ": {"name": "chirality_vector", "domain": "⟨G_geo, G_comp, G_load, G_spec, G_topo, G_arith⟩", "meaning": "Which law-axis the glyph operates on"},
|
||
"κᵢ": {"name": "local_context / manifold_coordinate", "domain": "position in n-D semantic manifold", "meaning": "Where in the field this packet applies"},
|
||
"τᵢ": {"name": "datatype_witness", "domain": "TypeBook entry", "meaning": "Structural law to import (biography, math article, citation graph...)"},
|
||
"UᵢΛᵢaᵢ": {"name": "eigen_descriptor", "domain": "EigenBook entry", "meaning": "Reusable geometric basis + spectrum + sparse coefficients"},
|
||
"θᵢ": {"name": "parameters", "domain": "side-stream encoded integers", "meaning": "Mode selectors, eigenbook indices, residual class tags"},
|
||
"εᵢ": {"name": "residual", "domain": "byte repair data", "meaning": "Exact correction to generated prediction"}
|
||
},
|
||
"core_rule": "glyph ≠ symbol. glyph = callable compression kernel."
|
||
},
|
||
|
||
"chirality_book": {
|
||
"description": "Same glyph means different lawful things depending on chirality vector.",
|
||
"vector_axes": {
|
||
"G_geo": "geometric primitive",
|
||
"G_comp": "mathematics article/domain macro",
|
||
"G_spec": "eigenbasis selector",
|
||
"G_topo": "incidence/angle graph operator",
|
||
"G_arith": "numeric constraint kernel",
|
||
"G_load": "expensive/fallback region marker"
|
||
},
|
||
"example": {
|
||
"📐_geo": "geometric primitive",
|
||
"📐_comp": "mathematics article/domain macro",
|
||
"📐_spec": "eigenbasis selector",
|
||
"📐_topo": "incidence/angle graph operator",
|
||
"📐_arith": "numeric constraint kernel",
|
||
"📐_load": "expensive/fallback region marker"
|
||
},
|
||
"power": "Finite glyph set becomes combinatorially huge through chirality rotation"
|
||
},
|
||
|
||
"type_book": {
|
||
"core_rule": "The datatype is the engine. UTF-8 is only the exhaust.",
|
||
"example_types": [
|
||
"WikiArticle<Mathematics>",
|
||
"WikiArticle<Biography>",
|
||
"WikiArticle<Geography>",
|
||
"Infobox<Person>",
|
||
"CitationGraph",
|
||
"SectionTree",
|
||
"HistoricalTimeline",
|
||
"BranchTaxonomy",
|
||
"TableMatrix",
|
||
"SameReferentCluster",
|
||
"FormulaRegion",
|
||
"ListRegion",
|
||
"MarkupRegion"
|
||
],
|
||
"compression_mass": "WikiArticle<Biography> already implies title, lead, infobox, birth/death fields, occupation, chronology, categories, citation patterns, linking conventions. The datatype itself carries structure.",
|
||
"value": "A type imports generative structure without storing every instance explicitly."
|
||
},
|
||
|
||
"eigen_book": {
|
||
"formula": "Gᵢ = ⟨τᵢ, Uᵢ, Λᵢ, aᵢ, εᵢ⟩",
|
||
"components": {
|
||
"τᵢ": "manifold/type class",
|
||
"Uᵢ": "eigenbasis / local frame (column vectors)",
|
||
"Λᵢ": "eigenvalue spectrum / scale-pressure",
|
||
"aᵢ": "sparse coefficients (activation weights)",
|
||
"εᵢ": "residual bytes (perturbation from ideal eigenstate)"
|
||
},
|
||
"example_math_article": {
|
||
"U": ["u_definition", "u_taxonomy", "u_history", "u_notation", "u_application", "u_philosophy", "u_reference"],
|
||
"meaning": "A mathematics page ≈ sparse activation of these eigenmodes + residual repair"
|
||
},
|
||
"keeper": "A wiki page is a sparse eigenstate of a typed reconstruction manifold, plus apology bytes."
|
||
},
|
||
|
||
"parameter_encoding": {
|
||
"channels": [
|
||
"side streams",
|
||
"variation selectors",
|
||
"combining marks",
|
||
"PUA suffixes",
|
||
"integer-coded payloads",
|
||
"arithmetic-coded payloads"
|
||
],
|
||
"example": "📐︖︉ → MathDomainKernel(mode=22, eigenbook=9, residual_class=26)",
|
||
"rule": "Decompressor reads codepoints, not what fonts display."
|
||
},
|
||
|
||
"residual_stream": {
|
||
"role": "The most important honesty layer.",
|
||
"core_rule": "Decode(generative_model) ⊕ ε = exact original bytes",
|
||
"forms": [
|
||
"literal patch",
|
||
"XOR patch",
|
||
"edit script",
|
||
"structural diff",
|
||
"entropy-coded correction",
|
||
"markup repair",
|
||
"serialization repair"
|
||
],
|
||
"diagnostic": "ρ = |ε| / |raw_span|. ρ < 0.1 = excellent. ρ ≈ 0.5 = maybe useful. ρ ≈ 1.0 = generator failed."
|
||
},
|
||
|
||
"gccl_gain_test": {
|
||
"formula": "ΔGCL(Γᵢ) = ΔG_geo + ΔG_comp + ΔG_spec + ΔG_topo + ΔG_arith - G_load - L(εᵢ) - L(θᵢ) - amortized_decoder_cost",
|
||
"accept_rule": "ΔGCL(Γᵢ) > 0",
|
||
"practical_form": "gain(Γᵢ) = literal_cost(span) - encoded_cost(γᵢ, χᵢ, κᵢ, τᵢ, UΛa, θᵢ, εᵢ)",
|
||
"principle": "No vibes. No 'semantic compression' handwaving. Only: shorter, deterministic, byte-exact, auditable."
|
||
},
|
||
|
||
"decode_pipeline": [
|
||
"Load decompressor profile D",
|
||
"Load GlyphBook 𝔊",
|
||
"Load ChiralityBook Χ",
|
||
"Load TypeBook Τ",
|
||
"Load EigenBook 𝕌",
|
||
"Read region index I",
|
||
"For each Γᵢ: resolve glyph γᵢ, chirality χᵢ, type τᵢ, eigen descriptor UᵢΛᵢaᵢ, parameters θᵢ",
|
||
"Generate predicted byte span ŝᵢ",
|
||
"Apply residual εᵢ",
|
||
"Emit exact span sᵢ",
|
||
"Concatenate spans",
|
||
"Verify checksum / receipt"
|
||
],
|
||
|
||
"encode_pipeline": [
|
||
"Segment corpus into candidate spans (pages, sections, infoboxes, tables, citations, formulas, markup regions)",
|
||
"Infer candidate types τ",
|
||
"Fit candidate geometric model (choose U, Λ, a)",
|
||
"Choose glyph kernel γ and chirality χ",
|
||
"Generate predicted bytes",
|
||
"Compute residual ε",
|
||
"Score: gain = literal_cost - encoded_cost",
|
||
"Keep candidates with gain > 0",
|
||
"Solve covering problem: choose packet set Γ* covering C with minimum total cost",
|
||
"Emit archive",
|
||
"Decode immediately and verify exact byte equality"
|
||
],
|
||
|
||
"model_families": {
|
||
"A_Wiki_structural": {
|
||
"kernels": ["WikiArticle", "Infobox", "SectionTree", "CitationGraph", "CategoryList", "InternalLinkGraph", "ReferenceList", "TableMatrix"],
|
||
"value": "Highest practical value. Structural redundancy in encyclopedic corpora is massive."
|
||
},
|
||
"B_Same_referent": {
|
||
"kernels": ["SameReferentVariation", "EntityAliasCluster", "PronounEpithetChain"],
|
||
"value": "Handles elegant-variation / synonym-heavy text where literal repetition is low."
|
||
},
|
||
"C_Arithmetic_date": {
|
||
"kernels": ["Year", "DateInterval", "Coordinate", "PopulationTable", "UnitExpression", "Ranking", "Ordinal"],
|
||
"value": "Numbers are dense but highly structured. Very reliable wins."
|
||
},
|
||
"D_Fractal_generator": {
|
||
"kernels": ["Mandelbrot", "L-system", "CellularAutomaton", "ProceduralImage", "ParametricCurve"],
|
||
"value": "Only useful if byte projection matches generator closely. SVG serialization cost usually dominates."
|
||
},
|
||
"E_Eigenfield": {
|
||
"kernels": ["ArticleEigenfield", "CitationEigenfield", "MarkupEigenfield", "SemanticDensityField"],
|
||
"value": "Region-level reconstruction. Connects to density-field encoding theory."
|
||
}
|
||
},
|
||
|
||
"stress_test_lesson": {
|
||
"mandelbrot_svg": "generator cost ≈ tiny, serialized artifact cost ≈ huge",
|
||
"conclusion": "z ↦ z² + c generates the image, but not the exact SVG file. Residual = ε_serialize.",
|
||
"best_diagnostic": "ρ = |ε| / |raw_span|"
|
||
},
|
||
|
||
"implementation_phases": {
|
||
"Phase_1": {
|
||
"name": "Byte-exact toy codec",
|
||
"scope": "GlyphBook + TypeBook + ResidualStream",
|
||
"kernels": ["CitationGraphKernel", "InfoboxKernel", "SectionTreeKernel"],
|
||
"goal": "generated_span + residual = original_span"
|
||
},
|
||
"Phase_2": {
|
||
"name": "Add arithmetic/date kernels",
|
||
"scope": "Years, dates, coordinates, measurements, rankings, table values",
|
||
"value": "Reliable wins on dense numeric data"
|
||
},
|
||
"Phase_3": {
|
||
"name": "Add same-referent variation",
|
||
"scope": "EntityClusterKernel, AliasEmitter, CoreferenceSurfaceFormKernel",
|
||
"value": "Attacks low-repetition text"
|
||
},
|
||
"Phase_4": {
|
||
"name": "Add eigen descriptors",
|
||
"scope": "UΛa for region classes",
|
||
"caution": "Only after above works. Use as descriptor reuse, not magic semantic compression."
|
||
},
|
||
"Phase_5": {
|
||
"name": "Add PUA glyph acceleration",
|
||
"rule": "promotion_gain = repeated_invocation_savings - glyph_definition_cost. Promote only if positive."
|
||
}
|
||
},
|
||
|
||
"prototype_archive": {
|
||
"magic": "GEC1",
|
||
"struct_fields": ["decoder_profile", "glyphbook", "typebook", "packets", "params", "residuals", "sha256"],
|
||
"packet_fields": ["glyph_id: u32", "chirality: u8", "type_id: u16", "region_start: u64", "region_len: u32", "eigen_id: Option<u16>", "param_ref", "residual_ref"],
|
||
"decode_invariant": "sha256(decode(archive)) == sha256(original)"
|
||
},
|
||
|
||
"stack_integration": {
|
||
"density_field_encoding": {
|
||
"role": "REGION-LEVEL MODEL FAMILY E. The density field IS an eigenfield kernel.",
|
||
"mapping": "ρ(x⃗) = SemanticDensityField kernel. Topological skeleton = GlyphBook + EigenBook. Perturbation = ResidualStream."
|
||
},
|
||
"s3c_shells": {
|
||
"role": "LOCAL COORDINATE κᵢ. Shell index k = semantic distance from peak (e.g., article title). a = intra-cluster angular position.",
|
||
"mapping": "κᵢ encoded as S3C shell coordinates: cheap integer manifold position."
|
||
},
|
||
"oac": {
|
||
"role": "GLYPH KERNEL LAZINESS. A glyph is a callable kernel stored in the OAC. It only materializes when touched by the decoder pipeline.",
|
||
"mapping": "GlyphBook = library of OAC touch-manifestable kernels. Chirality = which touch interpretation."
|
||
},
|
||
"hypercube_rhomboid": {
|
||
"role": "MANIFOLD GEOMETRY OF THE PACKET STREAM. UTF-8 is orthogonal hypercube (independent byte positions). GCCL-GEC is sheared hyper-rhomboid: each packet's meaning depends on neighboring packets through chirality and type context.",
|
||
"mapping": "Shear matrix learned from corpus: eigenvectors = principal semantic directions. Packets lean into each other."
|
||
},
|
||
"famm_delay_lines": {
|
||
"role": "TEMPORAL SEQUENCING OF PACKET STREAM. FAMM preshaped delays impose decode order through the packet stream.",
|
||
"mapping": "Delay profile = path integral through packet dependencies. Fast regions = high structural redundancy (predictable). Slow regions = high residual density (needs more context)."
|
||
},
|
||
"erans_entropy": {
|
||
"role": "RESIDUAL AND PARAMETER STREAM CODING. After glyph prediction, residual bytes and parameters are entropy-coded via enumerative rANS.",
|
||
"mapping": "Histogram of residuals is exact; erans enumerative coding is optimal for exact histograms."
|
||
},
|
||
"radius_ratio_motif": {
|
||
"role": "LOCAL ADMISSIBILITY QUANTIZER FOR PACKET SELECTION. Given a local feature ratio ρ, the radius-ratio rule selects the smallest stable coordination motif (kernel).",
|
||
"mapping": "Local scale ratio → admissible kernel class (WikiArticle, Infobox, CitationGraph...) + residual. Same move: continuous witness → finite motif alphabet."
|
||
}
|
||
},
|
||
|
||
"keeper_phrases": [
|
||
"Do not store the text. Store the cheapest lawful generator of the byte projection.",
|
||
"glyph ≠ symbol. glyph = callable compression kernel.",
|
||
"The datatype is the engine. UTF-8 is only the exhaust.",
|
||
"A wiki page is a sparse eigenstate of a typed reconstruction manifold, plus apology bytes.",
|
||
"Never trust a glyph until the residual gets smaller.",
|
||
"No vibes. No 'semantic compression' handwaving. Only: shorter, deterministic, byte-exact, auditable.",
|
||
"ρ = |ε| / |raw_span|. This is the only number that matters.",
|
||
"The decompressor reads codepoints, not what fonts display.",
|
||
"A finite glyph set becomes combinatorially huge because each glyph can rotate through many law-axes.",
|
||
"The Morse-Smale complex is the topological skeleton of meaning. GCCL-GEC is the lawful engine that navigates it."
|
||
],
|
||
|
||
"metadata": {
|
||
"ingested_at": time.time(),
|
||
"tags": [
|
||
"gccl-gec", "compression-architecture", "glyph-eigen-codec",
|
||
"byte-exact-compression", "callable-kernel", "chirality-book",
|
||
"type-book", "eigen-book", "residual-stream", "gain-test",
|
||
"hutter-prize", "density-field", "s3c-shells", "oac",
|
||
"hypercube-rhomboid", "famm", "erans", "radius-ratio",
|
||
"geometric-compression"
|
||
]
|
||
}
|
||
}
|
||
|
||
|
||
def ingest():
|
||
germane_dir = RESEARCH_STACK / "shared-data/data/germane/research"
|
||
germane_dir.mkdir(parents=True, exist_ok=True)
|
||
|
||
out_path = germane_dir / "gccl_gec_spec_v1.json"
|
||
with open(out_path, 'w') as f:
|
||
json.dump(GCCL_GEC, f, indent=2)
|
||
|
||
print(f"✓ Ingested: {out_path}")
|
||
|
||
index_path = germane_dir / "research_ingestion_index.json"
|
||
index = []
|
||
if index_path.exists():
|
||
with open(index_path) as f:
|
||
index = json.load(f)
|
||
|
||
index.append({
|
||
"id": GCCL_GEC["id"],
|
||
"title": GCCL_GEC["title"],
|
||
"date": GCCL_GEC["date"],
|
||
"source": GCCL_GEC["source"],
|
||
"ingested_at": GCCL_GEC["metadata"]["ingested_at"],
|
||
"tags": GCCL_GEC["metadata"]["tags"],
|
||
})
|
||
|
||
with open(index_path, 'w') as f:
|
||
json.dump(index, f, indent=2)
|
||
|
||
print(f"✓ Index: {len(index)} entries")
|
||
|
||
print(f"\nArchive components (9):")
|
||
for k, v in GCCL_GEC["archive_structure"]["components"].items():
|
||
print(f" {k} = {v['name']}: {v['role'][:60]}...")
|
||
|
||
print(f"\nPacket fields (7):")
|
||
for field, props in GCCL_GEC["fundamental_packet"]["fields"].items():
|
||
print(f" {field} → {props['name']}: {props['meaning'][:50]}...")
|
||
|
||
print(f"\nModel families (5):")
|
||
for fam, data in GCCL_GEC["model_families"].items():
|
||
print(f" {fam}: {len(data['kernels'])} kernels — {data['value'][:50]}...")
|
||
|
||
print(f"\nStack integration (7):")
|
||
for module, mapping in GCCL_GEC["stack_integration"].items():
|
||
print(f" ↔ {module}: {mapping['role'][:65]}...")
|
||
|
||
print(f"\nImplementation phases (5):")
|
||
for phase, data in GCCL_GEC["implementation_phases"].items():
|
||
print(f" {phase}: {data['name']} — {data.get('goal', data.get('value', data.get('scope', '')))[:50]}...")
|
||
|
||
print(f"\nKeeper phrases ({len(GCCL_GEC['keeper_phrases'])}):")
|
||
for p in GCCL_GEC["keeper_phrases"]:
|
||
print(f" → {p}")
|
||
|
||
|
||
if __name__ == "__main__":
|
||
ingest()
|