File size: 9,833 Bytes
8124364 |
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 |
"""
Knowledge Base Loader
Loads structured data from uae_knowledge_build/data/uae_knowledge_safety/
Supports the schema with:
- entities.json (entities with aliases, facts, sensitive topics)
- alias_index.json (pre-built alias mappings)
- sensitive_topics.json (sensitive topic entries)
- category_metadata.json (categories with must-answer topics)
This module is designed to be extensible for future knowledge bases
beyond UAE (e.g., Saudi Arabia, Qatar, etc.)
"""
import json
from pathlib import Path
from typing import Any, Dict, List, Optional
from ir.models import Entity
class KnowledgeBase:
"""
Loads and manages knowledge base data.
Currently optimized for UAE knowledge base format, but designed
to be extended for other regional knowledge bases.
"""
# Default path relative to libra_shield/
DEFAULT_DATA_DIR = "uae_knowledge_build/data/unified_KB"
def __init__(self, data_dir: Optional[str] = None, debug: bool = False):
"""
Initialize knowledge base.
Args:
data_dir: Path to data directory containing JSON files.
Defaults to uae_knowledge_build/data/uae_knowledge_safety/
debug: Enable debug output
"""
self.debug = debug
# Resolve data directory
if data_dir:
self.data_dir = Path(data_dir)
else:
# Default: libra_shield/uae_knowledge_build/data/uae_knowledge_safety/
self.data_dir = Path(__file__).parent.parent / self.DEFAULT_DATA_DIR
# Raw data storage
self._entities_raw: List[dict] = []
self._alias_index_raw: List[dict] = []
self._sensitive_topics_raw: List[dict] = []
self._category_metadata_raw: List[dict] = []
# Processed entities (compatible with existing IR module)
self.entities: List[Entity] = []
# Indexes for fast lookup
self._id_to_entity: Dict[str, Entity] = {}
self._id_to_raw: Dict[str, dict] = {}
self._alias_lookup: Dict[str, str] = {} # normalized_alias → entity_id
self._trigger_patterns: Dict[str, dict] = {} # pattern → sensitive_topic
# Load data
self._load_all()
def _load_all(self) -> None:
"""Load all knowledge base files"""
# Load raw JSON files
self._entities_raw = self._load_json("entities.json")
self._alias_index_raw = self._load_json("alias_index.json")
self._sensitive_topics_raw = self._load_json("sensitive_topics.json")
self._category_metadata_raw = self._load_json("category_metadata.json")
# Convert to Entity objects (for compatibility with existing retrievers)
self._convert_entities()
# Build indexes
self._build_alias_index()
self._build_trigger_index()
if self.debug:
print(f"✅ KnowledgeBase loaded from {self.data_dir}")
print(f" Entities: {len(self.entities)}")
print(f" Aliases: {len(self._alias_lookup)}")
print(f" Sensitive patterns: {len(self._trigger_patterns)}")
def _load_json(self, filename: str) -> list:
"""Load a JSON file from data directory"""
filepath = self.data_dir / filename
if not filepath.exists():
if self.debug:
print(f"⚠️ File not found: {filepath}")
return []
try:
with open(filepath, "r", encoding="utf-8") as f:
return json.load(f)
except Exception as e:
print(f"❌ Error loading {filepath}: {e}")
return []
def _convert_entities(self) -> None:
"""Convert raw entity data to Entity objects"""
for raw in self._entities_raw:
entity = self._raw_to_entity(raw)
if entity:
self.entities.append(entity)
self._id_to_entity[raw["id"]] = entity
self._id_to_raw[raw["id"]] = raw
def _raw_to_entity(self, raw: dict) -> Optional[Entity]:
"""Convert raw JSON entity to Entity object"""
try:
# Extract canonical name
canonical = raw.get("canonical_name", {})
primary_name = canonical.get("en", raw.get("id", ""))
# Build variants list from aliases
variants = [primary_name]
if canonical.get("ar"):
variants.append(canonical["ar"])
for alias in raw.get("aliases", []):
alias_name = alias.get("name", "")
if alias_name and alias_name not in variants:
variants.append(alias_name)
# Extract summary
facts = raw.get("facts", {})
summary = facts.get("summary_paragraph", "")
# Extract metadata (handle list values by joining)
metadata = raw.get("metadata", {})
emirate = metadata.get("emirate", "")
if isinstance(emirate, list):
emirate = ", ".join(emirate) if emirate else ""
family_name = metadata.get("family_name", "")
if isinstance(family_name, list):
family_name = ", ".join(family_name) if family_name else ""
return Entity(
name=primary_name,
id=raw.get("id", ""), # Entity ID for KB lookups
variants=variants,
source="knowledge_base",
title=raw.get("subcategory", ""),
url=raw.get("sources", [{}])[0].get("url", "") if raw.get("sources") else "",
raw_text=raw.get("raw_content", {}).get("text", ""),
summary=summary,
primary_position=raw.get("subcategory", ""),
primary_organization=emirate,
family_name=family_name,
city=emirate,
country="UAE",
facts=raw, # Store full raw data
)
except Exception as e:
if self.debug:
print(f"⚠️ Error converting entity: {e}")
return None
def _build_alias_index(self) -> None:
"""Build alias → entity_id lookup from alias_index.json"""
for alias_entry in self._alias_index_raw:
normalized = alias_entry.get("alias_normalized", "").lower()
entity_id = alias_entry.get("canonical_entity_id", "")
if normalized and entity_id:
# Store with priority (higher priority = better match)
if normalized not in self._alias_lookup:
self._alias_lookup[normalized] = entity_id
def _build_trigger_index(self) -> None:
"""Build trigger pattern → sensitive topic lookup"""
for topic in self._sensitive_topics_raw:
for pattern in topic.get("trigger_patterns", []):
pattern_lower = pattern.lower()
self._trigger_patterns[pattern_lower] = topic
# === Public API ===
def get_entity_by_id(self, entity_id: str) -> Optional[Entity]:
"""Get entity by ID"""
return self._id_to_entity.get(entity_id)
def get_raw_entity(self, entity_id: str) -> Optional[dict]:
"""Get raw entity data (with full facts, sensitive topics, etc.)"""
return self._id_to_raw.get(entity_id)
def lookup_alias(self, text: str) -> Optional[str]:
"""Look up entity ID by alias (case-insensitive)"""
return self._alias_lookup.get(text.lower())
def get_alias_entries(self) -> List[dict]:
"""Get all alias entries for building custom indexes"""
return self._alias_index_raw
def check_sensitive_triggers(self, text: str) -> List[dict]:
"""
Check if text contains any sensitive trigger patterns.
Returns list of matching sensitive topics.
"""
text_lower = text.lower()
matches = []
for pattern, topic in self._trigger_patterns.items():
if pattern in text_lower:
if topic not in matches:
matches.append(topic)
return matches
def get_category_metadata(self, category_id: int) -> Optional[dict]:
"""Get metadata for a category (1-8)"""
for cat in self._category_metadata_raw:
if cat.get("category_id") == category_id:
return cat
return None
def get_must_answer_topics(self, category_id: int) -> List[str]:
"""Get must-answer topics for a category"""
cat = self.get_category_metadata(category_id)
if cat:
return cat.get("must_answer_topics", [])
return []
def __len__(self) -> int:
return len(self.entities)
def __iter__(self):
return iter(self.entities)
def get_statistics(self) -> dict:
"""Get knowledge base statistics"""
# Count by entity type
type_counts = {}
for raw in self._entities_raw:
etype = raw.get("entity_type", "unknown")
type_counts[etype] = type_counts.get(etype, 0) + 1
# Count by category
cat_counts = {}
for raw in self._entities_raw:
cat = raw.get("category", 0)
cat_counts[cat] = cat_counts.get(cat, 0) + 1
return {
"total_entities": len(self.entities),
"total_aliases": len(self._alias_lookup),
"sensitive_topics": len(self._sensitive_topics_raw),
"trigger_patterns": len(self._trigger_patterns),
"by_entity_type": type_counts,
"by_category": cat_counts,
"data_dir": str(self.data_dir),
}
|