File size: 17,064 Bytes
23a5cce | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 | """
STAT7-RAG Bridge: Realm-Agnostic Hybrid Scoring for Document Retrieval
Bridges RAG documents with STAT7 addressing coordinates for intelligent,
multi-dimensional hybrid scoring that combines semantic similarity with
STAT7 entanglement resonance.
Supports any realm type (game, system, faculty, pattern, data, business, concept, etc.)
and scales deterministically to 10K+ documents.
Author: The Seed Phase 1 Integration
Status: Production-ready validation bridge
"""
from dataclasses import dataclass, field
from typing import Dict, Any, List, Tuple, Optional
import math
import random
# ============================================================================
# Data Structures: Realm-Agnostic STAT7 Addressing
# ============================================================================
@dataclass
class Realm:
"""Flexible realm definition for any relationship domain."""
type: str # e.g. "game", "system", "faculty", "pattern", "data", "narrative", "business", "concept"
label: str # human-readable name
@dataclass
class STAT7Address:
"""
STAT7 coordinate system: 7 dimensions for unique, multidimensional addressing.
- realm: Domain/context (flexible type + label)
- lineage: Version/generation (int >= 0)
- adjacency: Graph connectivity score (0.0-1.0)
- horizon: Zoom level / lifecycle stage (logline, outline, scene, panel, etc.)
- luminosity: Clarity/coherence/activity (0.0-1.0)
- polarity: Tension/contrast/resonance (0.0-1.0)
- dimensionality: Complexity/thread count (1-7 or bucketed)
"""
realm: Realm
lineage: int
adjacency: float
horizon: str
luminosity: float
polarity: float
dimensionality: int
def __post_init__(self):
"""Validate STAT7 constraints."""
assert 0.0 <= self.adjacency <= 1.0, f"adjacency must be [0,1], got {self.adjacency}"
assert 0.0 <= self.luminosity <= 1.0, f"luminosity must be [0,1], got {self.luminosity}"
assert 0.0 <= self.polarity <= 1.0, f"polarity must be [0,1], got {self.polarity}"
assert self.lineage >= 0, f"lineage must be >= 0, got {self.lineage}"
assert 1 <= self.dimensionality <= 7, f"dimensionality must be [1,7], got {self.dimensionality}"
def to_dict(self) -> Dict[str, Any]:
"""Export as dictionary for serialization."""
return {
"realm": {"type": self.realm.type, "label": self.realm.label},
"lineage": self.lineage,
"adjacency": self.adjacency,
"horizon": self.horizon,
"luminosity": self.luminosity,
"polarity": self.polarity,
"dimensionality": self.dimensionality,
}
@dataclass
class RAGDocument:
"""RAG document enhanced with STAT7 addressing."""
id: str
text: str
embedding: List[float]
stat7: STAT7Address
metadata: Dict[str, Any] = field(default_factory=dict)
def __post_init__(self):
"""Validate document structure."""
assert len(self.embedding) > 0, f"embedding must not be empty for {self.id}"
# ============================================================================
# Scoring Functions: Semantic + STAT7 Hybrid
# ============================================================================
def cosine_similarity(a: List[float], b: List[float]) -> float:
"""
Compute cosine similarity between two embedding vectors.
Range: [-1, 1], typically [0, 1] for normalized embeddings.
"""
if not a or not b:
return 0.0
dot = sum(x * y for x, y in zip(a, b))
norm_a = math.sqrt(sum(x * x for x in a))
norm_b = math.sqrt(sum(y * y for y in b))
denom = norm_a * norm_b + 1e-12 # Avoid division by zero
return dot / denom
def stat7_resonance(query_stat7: STAT7Address, doc_stat7: STAT7Address) -> float:
"""
Compute STAT7 resonance between query and document addresses.
This is the "entanglement score" — how well-aligned are the 7 dimensions?
Scoring strategy:
- Realm match (type > label): 1.0 if type matches, 0.85 if not; +0.1 if label matches
- Horizon alignment: 1.0 if same, 0.9 if adjacent, 0.7 if different
- Lineage proximity: decay by generation distance (±1 best)
- Signal alignment: how close are luminosity/polarity? (0.0-1.0)
- Adjacency/Dimensionality: connectivity and complexity bonuses
Returns: [0.0, 1.0] resonance score
"""
# Realm match (type is primary, label is secondary boost)
realm_score = 1.0 if query_stat7.realm.type == doc_stat7.realm.type else 0.85
if query_stat7.realm.label == doc_stat7.realm.label:
realm_score += 0.1
realm_score = min(realm_score, 1.0) # Cap at 1.0
# Horizon alignment: scale by distance
horizon_levels = {"logline": 1, "outline": 2, "scene": 3, "panel": 4}
h_query = horizon_levels.get(query_stat7.horizon, 3)
h_doc = horizon_levels.get(doc_stat7.horizon, 3)
h_distance = abs(h_query - h_doc)
if h_distance == 0:
horizon_score = 1.0
elif h_distance == 1:
horizon_score = 0.9
else:
horizon_score = 0.7
# Lineage proximity: prefer ±0-1 generation distance
lineage_distance = abs(query_stat7.lineage - doc_stat7.lineage)
lineage_score = max(0.7, 1.0 - 0.05 * lineage_distance)
# Signal alignment: luminosity + polarity
luminosity_diff = abs(query_stat7.luminosity - doc_stat7.luminosity)
polarity_diff = abs(query_stat7.polarity - doc_stat7.polarity)
signal_score = 1.0 - 0.5 * (luminosity_diff + polarity_diff)
signal_score = max(0.0, signal_score)
# Adjacency/Dimensionality bonus: connectivity + complexity
adj_bonus = doc_stat7.adjacency # Prefer well-connected docs
dim_bonus = min(1.0, doc_stat7.dimensionality / 7.0) # Normalize to [0,1]
adj_dim_score = 0.5 * adj_bonus + 0.5 * dim_bonus
# Combine all scores (multiplicative for strict alignment, additive bonus for complexity)
resonance = realm_score * horizon_score * lineage_score * signal_score
resonance *= (0.8 + 0.2 * adj_dim_score) # 20% bonus from connectivity/complexity
return max(0.0, min(resonance, 1.0)) # Clamp to [0,1]
def hybrid_score(
query_embedding: List[float],
doc: RAGDocument,
query_stat7: STAT7Address,
weight_semantic: float = 0.6,
weight_stat7: float = 0.4,
) -> float:
"""
Hybrid scoring: combine semantic similarity with STAT7 resonance.
Args:
query_embedding: Query embedding vector
doc: RAG document with embedding and STAT7 address
query_stat7: Query STAT7 address
weight_semantic: Weight for semantic similarity (default 0.6)
weight_stat7: Weight for STAT7 resonance (default 0.4)
Returns: [0.0, 1.0] hybrid score
"""
assert weight_semantic + weight_stat7 == 1.0, "Weights must sum to 1.0"
semantic_sim = cosine_similarity(query_embedding, doc.embedding)
stat7_res = stat7_resonance(query_stat7, doc.stat7)
hybrid = (weight_semantic * semantic_sim) + (weight_stat7 * stat7_res)
return max(0.0, min(hybrid, 1.0)) # Clamp to [0,1]
# ============================================================================
# Retrieval: Hybrid RAG Search
# ============================================================================
def retrieve(
documents: List[RAGDocument],
query_embedding: List[float],
query_stat7: STAT7Address,
k: int = 10,
weight_semantic: float = 0.6,
weight_stat7: float = 0.4,
) -> List[Tuple[str, float]]:
"""
Retrieve top-k documents using hybrid (semantic + STAT7) scoring.
Args:
documents: List of RAG documents to search
query_embedding: Query embedding vector
query_stat7: Query STAT7 address
k: Number of results to return
weight_semantic: Weight for semantic similarity
weight_stat7: Weight for STAT7 resonance
Returns: List of (doc_id, hybrid_score) tuples, sorted by score (descending)
"""
scores = []
for doc in documents:
score = hybrid_score(query_embedding, doc, query_stat7, weight_semantic, weight_stat7)
scores.append((doc.id, score))
# Sort by score descending, return top-k
return sorted(scores, key=lambda x: x[1], reverse=True)[:k]
def retrieve_semantic_only(
documents: List[RAGDocument],
query_embedding: List[float],
k: int = 10,
) -> List[Tuple[str, float]]:
"""
Retrieve top-k documents using semantic similarity only (baseline).
Args:
documents: List of RAG documents to search
query_embedding: Query embedding vector
k: Number of results to return
Returns: List of (doc_id, semantic_score) tuples, sorted by score (descending)
"""
scores = []
for doc in documents:
score = cosine_similarity(query_embedding, doc.embedding)
scores.append((doc.id, score))
return sorted(scores, key=lambda x: x[1], reverse=True)[:k]
# ============================================================================
# Utilities: Document Generation & STAT7 Randomization
# ============================================================================
def generate_random_stat7_address(
realm: Realm,
lineage_range: Tuple[int, int] = (0, 10),
horizon_choices: Optional[List[str]] = None,
seed_offset: int = 0,
) -> STAT7Address:
"""
Generate a random STAT7 address with optional seeding.
Args:
realm: Realm for this address
lineage_range: Min/max for lineage generation
horizon_choices: List of horizon options (default: common levels)
seed_offset: For reproducibility, offset from global random state
Returns: Randomized STAT7Address
"""
if horizon_choices is None:
horizon_choices = ["logline", "outline", "scene", "panel"]
return STAT7Address(
realm=realm,
lineage=random.randint(lineage_range[0], lineage_range[1]),
adjacency=round(random.random(), 2),
horizon=random.choice(horizon_choices),
luminosity=round(random.random(), 2),
polarity=round(random.random(), 2),
dimensionality=random.randint(1, 7),
)
def generate_synthetic_rag_documents(
base_texts: List[str],
realm: Realm,
scale: int,
embedding_fn: callable,
randomize_stat7: bool = False,
seed: Optional[int] = None,
) -> List[RAGDocument]:
"""
Generate synthetic RAG documents with STAT7 addresses.
Args:
base_texts: List of base text templates (will be varied)
realm: Realm for all generated documents
scale: Number of documents to generate
embedding_fn: Function to embed text (e.g., embedding_provider.embed_text)
randomize_stat7: If True, randomize all 7 STAT7 dimensions per doc
seed: Random seed for reproducibility
Returns: List of RAGDocument with embeddings and STAT7 addresses
"""
if seed is not None:
random.seed(seed)
documents = []
for i in range(scale):
# Vary text template
base_idx = i % len(base_texts)
base_text = base_texts[base_idx]
text = f"[Context {i}] {base_text} (instance {i})"
# Embed text
embedding = embedding_fn(text)
# Assign STAT7 address
if randomize_stat7:
stat7 = generate_random_stat7_address(realm, seed_offset=i)
else:
# Deterministic: map index to STAT7 dimensions
stat7 = STAT7Address(
realm=realm,
lineage=i % 10,
adjacency=round((i % 100) / 100.0, 2),
horizon=["logline", "outline", "scene", "panel"][i % 4],
luminosity=round((i % 10) / 10.0, 2),
polarity=round(((i + 5) % 10) / 10.0, 2),
dimensionality=1 + (i % 7),
)
doc = RAGDocument(
id=f"doc-{i:06d}",
text=text,
embedding=embedding,
stat7=stat7,
metadata={
"source": f"pack-{base_idx % 3}",
"category": ["core", "wisdom", "politics"][base_idx % 3],
"generated_index": i,
},
)
documents.append(doc)
return documents
# ============================================================================
# Analysis: Comparison & Diagnostics
# ============================================================================
def compare_retrieval_results(
semantic_results: List[Tuple[str, float]],
hybrid_results: List[Tuple[str, float]],
k: int = 10,
) -> Dict[str, Any]:
"""
Compare semantic-only vs hybrid retrieval results.
Returns metrics:
- overlap: How many of top-k are shared?
- semantic_avg_score: Average semantic score in top-k
- hybrid_avg_score: Average hybrid score in top-k
- reranking_distance: How much did hybrid rerank results?
"""
semantic_ids = {doc_id for doc_id, _ in semantic_results[:k]}
hybrid_ids = {doc_id for doc_id, _ in hybrid_results[:k]}
overlap = len(semantic_ids & hybrid_ids)
overlap_pct = (overlap / k * 100) if k > 0 else 0.0
semantic_avg = sum(score for _, score in semantic_results[:k]) / k if k > 0 else 0.0
hybrid_avg = sum(score for _, score in hybrid_results[:k]) / k if k > 0 else 0.0
# Measure ranking distance: how far did top-k items move?
semantic_rank = {doc_id: idx for idx, (doc_id, _) in enumerate(semantic_results[:k])}
reranking_distances = []
for idx, (doc_id, _) in enumerate(hybrid_results[:k]):
if doc_id in semantic_rank:
distance = abs(idx - semantic_rank[doc_id])
reranking_distances.append(distance)
avg_reranking_distance = (
sum(reranking_distances) / len(reranking_distances) if reranking_distances else 0.0
)
return {
"overlap_count": overlap,
"overlap_pct": overlap_pct,
"semantic_avg_score": round(semantic_avg, 4),
"hybrid_avg_score": round(hybrid_avg, 4),
"score_improvement": round(hybrid_avg - semantic_avg, 4),
"avg_reranking_distance": round(avg_reranking_distance, 2),
}
# ============================================================================
# STAT7RAGBridge: Wrapper for RetrievalAPI Integration
# ============================================================================
class STAT7RAGBridge:
"""
Bridge class that provides STAT7 functionality for RetrievalAPI integration.
Wraps the module-level STAT7 functions (stat7_resonance, hybrid_score, retrieve)
to provide a consistent interface for the RetrievalAPI's hybrid scoring system.
This allows RetrievalAPI to work with STAT7 coordinates seamlessly through
dependency injection.
"""
def stat7_resonance(self, query_stat7: STAT7Address, doc_stat7: STAT7Address) -> float:
"""
Compute STAT7 resonance between query and document addresses.
Args:
query_stat7: Query STAT7 address
doc_stat7: Document STAT7 address
Returns: [0.0, 1.0] resonance score
"""
return stat7_resonance(query_stat7, doc_stat7)
def hybrid_score(
self,
query_embedding: List[float],
doc: RAGDocument,
query_stat7: STAT7Address,
weight_semantic: float = 0.6,
weight_stat7: float = 0.4,
) -> float:
"""
Compute hybrid score combining semantic similarity with STAT7 resonance.
Args:
query_embedding: Query embedding vector
doc: RAG document with embedding and STAT7 address
query_stat7: Query STAT7 address
weight_semantic: Weight for semantic similarity (default 0.6)
weight_stat7: Weight for STAT7 resonance (default 0.4)
Returns: [0.0, 1.0] hybrid score
"""
return hybrid_score(query_embedding, doc, query_stat7, weight_semantic, weight_stat7)
def retrieve(
self,
documents: List[RAGDocument],
query_embedding: List[float],
query_stat7: STAT7Address,
k: int = 10,
weight_semantic: float = 0.6,
weight_stat7: float = 0.4,
) -> List[Tuple[str, float]]:
"""
Retrieve top-k documents using hybrid (semantic + STAT7) scoring.
Args:
documents: List of RAG documents to search
query_embedding: Query embedding vector
query_stat7: Query STAT7 address
k: Number of results to return
weight_semantic: Weight for semantic similarity
weight_stat7: Weight for STAT7 resonance
Returns: List of (doc_id, hybrid_score) tuples, sorted by score (descending)
"""
return retrieve(documents, query_embedding, query_stat7, k, weight_semantic, weight_stat7) |