Spaces:
Running
Running
File size: 9,079 Bytes
b32fbe0 e5cdd9c b32fbe0 54ce8e5 7abb8fd e5cdd9c b32fbe0 2b9ab6a b32fbe0 e5cdd9c b32fbe0 b226c60 e5cdd9c 2b9ab6a 7abb8fd b32fbe0 e5cdd9c b32fbe0 54ce8e5 b32fbe0 7abb8fd b32fbe0 e5cdd9c 2b9ab6a e5cdd9c b226c60 e5cdd9c b32fbe0 e5cdd9c 7abb8fd e5cdd9c b32fbe0 7abb8fd b32fbe0 7abb8fd b32fbe0 495ffaa e5cdd9c 495ffaa e5cdd9c 54ce8e5 e5cdd9c 495ffaa e5cdd9c 7abb8fd b32fbe0 7abb8fd b32fbe0 7abb8fd b32fbe0 e5cdd9c b32fbe0 7abb8fd e5cdd9c 54ce8e5 7abb8fd b32fbe0 2b9ab6a b32fbe0 54ce8e5 7abb8fd 2b9ab6a b32fbe0 495ffaa 2549784 7abb8fd e5cdd9c 7abb8fd e5cdd9c b32fbe0 54ce8e5 b32fbe0 7abb8fd e5cdd9c 7abb8fd e5cdd9c 7abb8fd 2b9ab6a b32fbe0 7abb8fd e5cdd9c 7abb8fd 2b9ab6a e5cdd9c b32fbe0 | 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 | """
Knowledge Universe β Knowledge Decay Engine (Core IP, v2.2)
==========================================================
Calculates decay [0.0 - 1.0] and the Penalty Multiplier for ranking.
v2.2 β Enterprise First-Class Fields:
- Added `max_decay_detected` as a first-class field in compute_batch_summary()
- This eliminates adapter-side derivation for enterprise clients (ReconAI, etc.)
- Added `worst_source_id` for graph labeling/tooltips (Dwayne's TrustGraph need)
- Unknown date penalty raised from 0.6 β 0.65
- Added DECAY_ENGINE_VERSION for cache invalidation safety.
- Added decay_velocity and days_until_stale directly into the core object.
"""
import logging
import math
from dataclasses import dataclass, field
from datetime import datetime, timezone
from typing import Dict, List, Optional
logger = logging.getLogger(__name__)
# Version constant to protect cache against silent staleness
DECAY_ENGINE_VERSION = "v2.2"
# Half-lives tuned per platform volatility
HALF_LIVES: Dict[str, int] = {
"arxiv": 1095, # 3 years
"wikipedia": 1460, # 4 years
"openlibrary": 1825, # 5 years
"mit_ocw": 1095, # 3 years
"stackoverflow": 365, # 1 year
"github": 180, # 6 months
"huggingface": 120, # 4 months (ML moves fast)
"kaggle": 365, # 1 year
"youtube": 270, # 9 months
"podcast": 180, # 6 months
"common_crawl": 90, # 3 months
"gharchive": 180, # 6 months
"libgen": 1825, # 5 years
}
DEFAULT_HALF_LIFE = 365
# Unknown date penalty β sources with no date get this multiplier
_UNKNOWN_DATE_PENALTY = 0.65
@dataclass
class DecayReport:
source_id: str
decay_score: float # 0.0 (fresh) β 1.0 (decayed)
freshness: float # 1.0 (fresh) β 0.0 (decayed)
age_days: Optional[int]
half_life: int
source_type: str
label: str
computed_at: str
penalty_multiplier: float # Multiplied against quality score
# Enterprise metrics baked into core IP
decay_velocity: float
days_until_stale: int
version: str = DECAY_ENGINE_VERSION
def as_dict(self) -> dict:
return {
"decay_score": round(self.decay_score, 3),
"freshness": round(self.freshness, 3),
"age_days": self.age_days,
"label": self.label,
"penalty_multiplier": round(self.penalty_multiplier, 3),
"decay_velocity": round(self.decay_velocity, 6),
"days_until_stale": self.days_until_stale,
"version": self.version,
}
class KnowledgeDecayEngine:
"""
Core IP: Penalizes stale content in the ranking pipeline.
Formula: Final Score = Base Quality * Decay Penalty
"""
def compute_from_dict(
self,
source_dict: dict,
customer: Optional[Dict] = None
) -> DecayReport:
platform = source_dict.get("source_platform", "unknown")
half_life = HALF_LIVES.get(platform, DEFAULT_HALF_LIFE)
if customer and isinstance(customer, dict):
overrides = customer.get("half_life_overrides", {})
if platform in overrides:
override_val = int(overrides[platform])
logger.debug(
f"Customer half-life override: {platform} "
f"{half_life}β{override_val} days"
)
half_life = override_val
pub_raw = (
source_dict.get("publication_date")
or source_dict.get("last_updated")
)
if not pub_raw:
return self._unknown_report(
source_dict.get("id", "unknown"), platform, half_life
)
try:
if isinstance(pub_raw, str):
pub_raw = pub_raw.replace("Z", "+00:00")
if len(pub_raw) == 10:
pub_raw += "T00:00:00+00:00"
pub_date = datetime.fromisoformat(pub_raw)
else:
pub_date = pub_raw
if pub_date.tzinfo is None:
pub_date = pub_date.replace(tzinfo=timezone.utc)
now = datetime.now(timezone.utc)
age_days = max(0, (now - pub_date).days)
decay = round(1.0 - math.pow(0.5, age_days / half_life), 4)
decay = min(max(decay, 0.0), 1.0)
freshness = round(1.0 - decay, 4)
if decay <= 0.25:
penalty = round(0.90 + (0.10 * freshness), 4)
elif decay <= 0.50:
penalty = round(0.50 + (0.40 * freshness), 4)
elif decay <= 0.75:
penalty = round(0.20 + (0.45 * freshness), 4)
else:
penalty = round(0.05 + (0.25 * freshness), 4)
if age_days < 90:
penalty = max(penalty, 0.95)
velocity = math.log(2) / half_life
if decay >= 0.50:
days_stale = 0
else:
days_stale = int(max(0, (0.50 - decay) / velocity))
return DecayReport(
source_id=source_dict.get("id", "unknown"),
decay_score=decay,
freshness=freshness,
age_days=age_days,
half_life=half_life,
source_type=platform,
label=self._label(decay),
computed_at=now.isoformat(),
penalty_multiplier=penalty,
decay_velocity=velocity,
days_until_stale=days_stale,
)
except Exception as e:
logger.error(f"Decay computation failed for {source_dict.get('id','?')}: {e}")
return self._unknown_report(
source_dict.get("id", "unknown"), platform, half_life
)
def compute(self, source, customer: Optional[Dict] = None) -> DecayReport:
"""Alias β accepts Source model or dict."""
if hasattr(source, "model_dump"):
return self.compute_from_dict(source.model_dump(), customer=customer)
return self.compute_from_dict(source, customer=customer)
def compute_batch(
self,
sources: List,
customer: Optional[Dict] = None,
) -> Dict[str, dict]:
"""
Compute decay for a list of sources and return the full per-source
map PLUS first-class enterprise fields:
Returns:
{
"per_source": {source_id: decay_dict, ...},
"max_decay_detected": 0.711, β first-class field
"avg_decay_score": 0.234, β for reference
"worst_source_id": "crossref:...",β for TrustGraph tooltip
"stale_count": 2,
"total_sources": 5,
}
"""
per_source: Dict[str, dict] = {}
max_decay = 0.0
worst_source_id = None
decay_sum = 0.0
stale_count = 0
for s in sources:
try:
report = self.compute(s, customer=customer)
r_dict = report.as_dict()
per_source[report.source_id] = r_dict
score = r_dict["decay_score"]
decay_sum += score
if score > max_decay:
max_decay = score
worst_source_id = report.source_id
if r_dict.get("label") in ("stale", "decayed"):
stale_count += 1
except Exception as e:
logger.error(f"compute_batch: failed on source {s}: {e}")
# Fallback: try to get id from source
sid = getattr(s, "id", None) or (s.get("id") if isinstance(s, dict) else "unknown")
per_source[sid] = {"decay_score": 0.4, "label": "unknown", "error": str(e)}
n = len(sources)
avg_decay = round(decay_sum / n, 3) if n else 0.0
return {
"per_source": per_source,
"max_decay_detected": round(max_decay, 3),
"avg_decay_score": avg_decay,
"worst_source_id": worst_source_id,
"stale_count": stale_count,
"total_sources": n,
}
@staticmethod
def _label(decay: float) -> str:
if decay < 0.25: return "fresh"
if decay < 0.50: return "aging"
if decay < 0.75: return "stale"
return "decayed"
def _unknown_report(
self,
sid: str,
platform: str,
half_life: int,
) -> DecayReport:
"""
Sources with no publication date get a neutral penalty.
age_days=None so downstream math never goes negative.
"""
velocity = math.log(2) / half_life
return DecayReport(
source_id=sid,
decay_score=0.4,
freshness=0.6,
age_days=None,
half_life=half_life,
source_type=platform,
label="unknown",
computed_at=datetime.now(timezone.utc).isoformat(),
penalty_multiplier=_UNKNOWN_DATE_PENALTY,
decay_velocity=velocity,
days_until_stale=180,
) |