Spaces:
Sleeping
Sleeping
File size: 36,753 Bytes
0ab6c82 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 | """
Knowledge Graph & Explainable AI System
This module implements:
1. Knowledge Graph - Extract and connect entities from emails
2. Relationship Tracking - Sender-topic-context relationships
3. Explainable AI - Decision explanations with confidence
4. Context Propagation - Use historical context for better decisions
5. Entity Recognition - Extract people, organizations, topics, dates
6. Reasoning Chain - Show step-by-step decision process
"""
import re
import math
from datetime import datetime, timedelta
from typing import Dict, List, Any, Optional, Set, Tuple
from dataclasses import dataclass, field
from enum import Enum
from collections import defaultdict
class EntityType(Enum):
"""Types of entities that can be extracted."""
PERSON = "person"
ORGANIZATION = "organization"
EMAIL_ADDRESS = "email_address"
PHONE = "phone"
DATE = "date"
TIME = "time"
MONEY = "money"
URL = "url"
TOPIC = "topic"
PROJECT = "project"
PRODUCT = "product"
LOCATION = "location"
DEADLINE = "deadline"
ACTION_ITEM = "action_item"
class RelationType(Enum):
"""Types of relationships between entities."""
SENT_BY = "sent_by"
SENT_TO = "sent_to"
MENTIONS = "mentions"
ABOUT = "about"
WORKS_FOR = "works_for"
RELATED_TO = "related_to"
RESPONDS_TO = "responds_to"
DEADLINE_FOR = "deadline_for"
ASSIGNED_TO = "assigned_to"
REQUESTED_BY = "requested_by"
class DecisionFactor(Enum):
"""Factors that influence decisions."""
SENDER_REPUTATION = "sender_reputation"
CONTENT_KEYWORDS = "content_keywords"
HISTORICAL_PATTERN = "historical_pattern"
ENTITY_CONTEXT = "entity_context"
URGENCY_INDICATORS = "urgency_indicators"
SPAM_SIGNALS = "spam_signals"
RELATIONSHIP_STRENGTH = "relationship_strength"
TIME_SENSITIVITY = "time_sensitivity"
DOMAIN_TRUST = "domain_trust"
THREAD_CONTEXT = "thread_context"
@dataclass
class Entity:
"""An entity extracted from an email."""
id: str
type: EntityType
value: str
normalized_value: str
confidence: float
source_email_id: str
context: str
extracted_at: datetime = field(default_factory=datetime.now)
def to_dict(self) -> Dict[str, Any]:
return {
"id": self.id,
"type": self.type.value,
"value": self.value,
"normalized_value": self.normalized_value,
"confidence": self.confidence,
"source_email_id": self.source_email_id,
"context": self.context,
"extracted_at": self.extracted_at.isoformat()
}
@dataclass
class Relationship:
"""A relationship between two entities."""
id: str
source_entity_id: str
target_entity_id: str
relation_type: RelationType
strength: float # 0-1
evidence: List[str]
email_ids: List[str]
created_at: datetime = field(default_factory=datetime.now)
last_seen: datetime = field(default_factory=datetime.now)
def to_dict(self) -> Dict[str, Any]:
return {
"id": self.id,
"source": self.source_entity_id,
"target": self.target_entity_id,
"type": self.relation_type.value,
"strength": self.strength,
"evidence": self.evidence,
"email_ids": self.email_ids,
"created_at": self.created_at.isoformat(),
"last_seen": self.last_seen.isoformat()
}
@dataclass
class ReasoningStep:
"""A step in the reasoning chain."""
step_number: int
factor: DecisionFactor
observation: str
inference: str
confidence: float
supporting_evidence: List[str]
def to_dict(self) -> Dict[str, Any]:
return {
"step": self.step_number,
"factor": self.factor.value,
"observation": self.observation,
"inference": self.inference,
"confidence": self.confidence,
"evidence": self.supporting_evidence
}
@dataclass
class Decision:
"""An explainable decision."""
decision_id: str
email_id: str
decision_type: str # categorize, prioritize, etc.
recommendation: str
confidence: float
reasoning_chain: List[ReasoningStep]
alternative_options: List[Dict[str, Any]]
timestamp: datetime = field(default_factory=datetime.now)
def to_dict(self) -> Dict[str, Any]:
return {
"decision_id": self.decision_id,
"email_id": self.email_id,
"decision_type": self.decision_type,
"recommendation": self.recommendation,
"confidence": self.confidence,
"reasoning_chain": [r.to_dict() for r in self.reasoning_chain],
"alternatives": self.alternative_options,
"timestamp": self.timestamp.isoformat()
}
class KnowledgeGraph:
"""
Knowledge Graph for email intelligence.
Extracts entities from emails, builds relationships, and provides
explainable AI decisions based on accumulated knowledge.
"""
def __init__(self):
"""Initialize the knowledge graph."""
self.entities: Dict[str, Entity] = {}
self.relationships: Dict[str, Relationship] = {}
self.entity_index: Dict[EntityType, Set[str]] = defaultdict(set)
self.email_entities: Dict[str, List[str]] = defaultdict(list)
self.decisions: List[Decision] = []
# Pattern matchers for entity extraction
self.patterns = {
EntityType.EMAIL_ADDRESS: re.compile(r'[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}'),
EntityType.PHONE: re.compile(r'[\+]?[(]?[0-9]{1,3}[)]?[-\s\.]?[0-9]{3}[-\s\.]?[0-9]{4,6}'),
EntityType.URL: re.compile(r'https?://[^\s<>"{}|\\^`\[\]]+'),
EntityType.MONEY: re.compile(r'\$[\d,]+(?:\.\d{2})?|\d+(?:,\d{3})*(?:\.\d{2})?\s*(?:USD|EUR|GBP|dollars?|euros?)', re.IGNORECASE),
EntityType.DATE: re.compile(r'\b(?:\d{1,2}[/-]\d{1,2}[/-]\d{2,4}|\d{4}[/-]\d{1,2}[/-]\d{1,2}|(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)[a-z]*\s+\d{1,2}(?:,?\s+\d{4})?)\b', re.IGNORECASE),
EntityType.TIME: re.compile(r'\b\d{1,2}:\d{2}(?::\d{2})?\s*(?:AM|PM|am|pm)?\b'),
}
# Topic keywords
self.topic_keywords = {
"meeting": ["meeting", "sync", "standup", "huddle", "call", "conference"],
"deadline": ["deadline", "due", "by EOD", "by COB", "ASAP", "urgent", "immediately"],
"project": ["project", "initiative", "sprint", "milestone", "roadmap"],
"sales": ["deal", "opportunity", "prospect", "pipeline", "revenue", "contract"],
"support": ["issue", "problem", "bug", "error", "help", "assist", "ticket"],
"billing": ["invoice", "payment", "subscription", "renewal", "charge"],
"hiring": ["candidate", "interview", "hire", "recruit", "resume", "position"],
"security": ["security", "vulnerability", "breach", "compliance", "audit"],
}
# Urgency indicators
self.urgency_indicators = [
"urgent", "asap", "immediately", "critical", "emergency",
"time-sensitive", "deadline", "overdue", "p1", "p0",
"production down", "outage", "blocked", "escalation"
]
# Spam signals
self.spam_signals = [
"winner", "lottery", "congratulations", "click here",
"act now", "limited time", "free money", "100% free",
"claim your", "urgent action required", "account suspended",
"verify your", "password expired", "lottery",
r"\.xyz", r"\.ru", r"\.cn", "nigerian prince"
]
# Sender reputation cache
self.sender_reputation: Dict[str, Dict[str, Any]] = {}
# Domain trust scores
self.domain_trust: Dict[str, float] = {
"company.com": 0.95,
"gmail.com": 0.7,
"outlook.com": 0.7,
"yahoo.com": 0.65,
}
self._entity_counter = 0
self._relationship_counter = 0
self._decision_counter = 0
def _generate_entity_id(self) -> str:
"""Generate a unique entity ID."""
self._entity_counter += 1
return f"entity_{self._entity_counter}"
def _generate_relationship_id(self) -> str:
"""Generate a unique relationship ID."""
self._relationship_counter += 1
return f"rel_{self._relationship_counter}"
def _generate_decision_id(self) -> str:
"""Generate a unique decision ID."""
self._decision_counter += 1
return f"decision_{self._decision_counter}"
def extract_entities(self, email: Dict[str, Any]) -> List[Entity]:
"""Extract entities from an email."""
entities = []
email_id = email.get("id", "unknown")
text = f"{email.get('subject', '')} {email.get('body', '')}"
# Extract pattern-based entities
for entity_type, pattern in self.patterns.items():
for match in pattern.finditer(text):
value = match.group()
context = text[max(0, match.start()-30):min(len(text), match.end()+30)]
entity = Entity(
id=self._generate_entity_id(),
type=entity_type,
value=value,
normalized_value=value.lower().strip(),
confidence=0.9,
source_email_id=email_id,
context=context
)
entities.append(entity)
self._add_entity(entity)
# Extract sender as entity
sender = email.get("sender", "")
if sender:
sender_entity = Entity(
id=self._generate_entity_id(),
type=EntityType.EMAIL_ADDRESS,
value=sender,
normalized_value=sender.lower(),
confidence=1.0,
source_email_id=email_id,
context=f"From: {sender}"
)
entities.append(sender_entity)
self._add_entity(sender_entity)
# Extract person from sender name
sender_name = email.get("sender_name", "")
if sender_name:
person_entity = Entity(
id=self._generate_entity_id(),
type=EntityType.PERSON,
value=sender_name,
normalized_value=sender_name.lower(),
confidence=0.95,
source_email_id=email_id,
context=f"Sender: {sender_name}"
)
entities.append(person_entity)
self._add_entity(person_entity)
# Extract topics
text_lower = text.lower()
for topic, keywords in self.topic_keywords.items():
for keyword in keywords:
if keyword in text_lower:
topic_entity = Entity(
id=self._generate_entity_id(),
type=EntityType.TOPIC,
value=topic,
normalized_value=topic,
confidence=0.7,
source_email_id=email_id,
context=f"Contains keyword: {keyword}"
)
entities.append(topic_entity)
self._add_entity(topic_entity)
break
# Extract deadlines
deadline_patterns = [
(r'by\s+(EOD|COB|end of day|close of business)', 0.85),
(r'deadline\s*:?\s*([^.\n]+)', 0.8),
(r'due\s+(by|on|before)\s+([^.\n]+)', 0.8),
(r'need\s+(by|before)\s+([^.\n]+)', 0.75),
]
for pattern, confidence in deadline_patterns:
match = re.search(pattern, text, re.IGNORECASE)
if match:
deadline_entity = Entity(
id=self._generate_entity_id(),
type=EntityType.DEADLINE,
value=match.group(0),
normalized_value=match.group(0).lower(),
confidence=confidence,
source_email_id=email_id,
context=text[max(0, match.start()-20):min(len(text), match.end()+20)]
)
entities.append(deadline_entity)
self._add_entity(deadline_entity)
# Extract action items
action_patterns = [
(r'please\s+([^.]+)', 0.7),
(r'could you\s+([^?]+)', 0.7),
(r'action\s*:?\s*([^.\n]+)', 0.85),
(r'todo\s*:?\s*([^.\n]+)', 0.85),
(r'need\s+(?:you\s+to\s+)?([^.\n]+)', 0.65),
]
for pattern, confidence in action_patterns:
for match in re.finditer(pattern, text, re.IGNORECASE):
action_entity = Entity(
id=self._generate_entity_id(),
type=EntityType.ACTION_ITEM,
value=match.group(0)[:100],
normalized_value=match.group(0).lower()[:100],
confidence=confidence,
source_email_id=email_id,
context=text[max(0, match.start()-10):min(len(text), match.end()+10)]
)
entities.append(action_entity)
self._add_entity(action_entity)
return entities
def _add_entity(self, entity: Entity) -> None:
"""Add an entity to the graph."""
self.entities[entity.id] = entity
self.entity_index[entity.type].add(entity.id)
self.email_entities[entity.source_email_id].append(entity.id)
def create_relationship(
self,
source_id: str,
target_id: str,
relation_type: RelationType,
evidence: str,
email_id: str
) -> Optional[Relationship]:
"""Create a relationship between two entities."""
if source_id not in self.entities or target_id not in self.entities:
return None
# Check for existing relationship
for rel in self.relationships.values():
if (rel.source_entity_id == source_id and
rel.target_entity_id == target_id and
rel.relation_type == relation_type):
# Strengthen existing relationship
rel.strength = min(1.0, rel.strength + 0.1)
rel.evidence.append(evidence)
rel.email_ids.append(email_id)
rel.last_seen = datetime.now()
return rel
# Create new relationship
relationship = Relationship(
id=self._generate_relationship_id(),
source_entity_id=source_id,
target_entity_id=target_id,
relation_type=relation_type,
strength=0.5,
evidence=[evidence],
email_ids=[email_id]
)
self.relationships[relationship.id] = relationship
return relationship
def build_relationships(self, email: Dict[str, Any]) -> List[Relationship]:
"""Build relationships from email entities."""
relationships = []
email_id = email.get("id", "unknown")
entities = [self.entities[eid] for eid in self.email_entities.get(email_id, [])]
# Find sender entity
sender_entities = [e for e in entities if e.type == EntityType.EMAIL_ADDRESS and "From:" in e.context]
for sender in sender_entities:
# Sender SENT_BY relationship
for topic in [e for e in entities if e.type == EntityType.TOPIC]:
rel = self.create_relationship(
sender.id, topic.id, RelationType.ABOUT,
f"Email about {topic.value}", email_id
)
if rel:
relationships.append(rel)
# Sender to person relationships
for person in [e for e in entities if e.type == EntityType.PERSON and e.id != sender.id]:
rel = self.create_relationship(
sender.id, person.id, RelationType.MENTIONS,
f"Email mentions {person.value}", email_id
)
if rel:
relationships.append(rel)
# Deadline relationships
for deadline in [e for e in entities if e.type == EntityType.DEADLINE]:
rel = self.create_relationship(
sender.id, deadline.id, RelationType.DEADLINE_FOR,
f"Deadline: {deadline.value}", email_id
)
if rel:
relationships.append(rel)
return relationships
def explain_categorization(self, email: Dict[str, Any]) -> Decision:
"""Generate an explainable categorization decision."""
email_id = email.get("id", "unknown")
reasoning_chain = []
step = 0
text = f"{email.get('subject', '')} {email.get('body', '')}".lower()
sender = email.get("sender", "")
sender_info = email.get("sender_info", {})
# Step 1: Analyze sender reputation
step += 1
trust_score = sender_info.get("trust_score", 0.5)
sender_type = sender_info.get("sender_type", "unknown")
if trust_score >= 0.9:
inference = "High trust sender - likely legitimate email"
confidence = 0.9
elif trust_score < 0.3:
inference = "Low trust sender - potential spam or unknown source"
confidence = 0.7
else:
inference = "Normal trust level - evaluate content"
confidence = 0.5
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.SENDER_REPUTATION,
observation=f"Sender {sender} has trust score {trust_score:.2f} ({sender_type})",
inference=inference,
confidence=confidence,
supporting_evidence=[
f"Previous emails: {sender_info.get('previous_emails', 0)}",
f"Avg response time: {sender_info.get('avg_response_time_hours', 'N/A')}h"
]
))
# Step 2: Check spam signals
step += 1
spam_matches = []
for signal in self.spam_signals:
if re.search(signal, text, re.IGNORECASE):
spam_matches.append(signal)
if spam_matches:
spam_confidence = min(0.95, 0.4 + len(spam_matches) * 0.15)
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.SPAM_SIGNALS,
observation=f"Found {len(spam_matches)} spam indicators",
inference="Strong spam signals detected",
confidence=spam_confidence,
supporting_evidence=spam_matches[:5]
))
else:
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.SPAM_SIGNALS,
observation="No spam indicators found",
inference="Content appears legitimate",
confidence=0.7,
supporting_evidence=["Clean content analysis"]
))
# Step 3: Analyze content keywords for category
step += 1
category_scores = defaultdict(float)
# Customer support signals
support_keywords = ["issue", "problem", "help", "not working", "broken", "complaint", "refund"]
for kw in support_keywords:
if kw in text:
category_scores["customer_support"] += 0.15
# Sales signals
sales_keywords = ["demo", "pricing", "enterprise", "interested in", "partnership", "opportunity"]
for kw in sales_keywords:
if kw in text:
category_scores["sales"] += 0.15
# Technical signals
tech_keywords = ["bug", "error", "api", "integration", "code", "deploy", "server"]
for kw in tech_keywords:
if kw in text:
category_scores["technical"] += 0.15
# Internal signals
internal_keywords = ["team", "meeting", "sync", "pto", "vacation", "internal"]
for kw in internal_keywords:
if kw in text:
category_scores["internal"] += 0.15
# Newsletter signals
newsletter_keywords = ["unsubscribe", "newsletter", "digest", "weekly update", "product update"]
for kw in newsletter_keywords:
if kw in text:
category_scores["newsletter"] += 0.2
# Billing signals
billing_keywords = ["invoice", "payment", "billing", "subscription", "charge", "receipt"]
for kw in billing_keywords:
if kw in text:
category_scores["billing"] += 0.15
top_categories = sorted(category_scores.items(), key=lambda x: -x[1])[:3]
if top_categories and top_categories[0][1] > 0:
best_cat, best_score = top_categories[0]
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.CONTENT_KEYWORDS,
observation=f"Keyword analysis: {dict(top_categories)}",
inference=f"Best category match: {best_cat}",
confidence=min(0.9, best_score),
supporting_evidence=[f"{cat}: {score:.2f}" for cat, score in top_categories]
))
else:
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.CONTENT_KEYWORDS,
observation="No strong category signals",
inference="May need human review",
confidence=0.3,
supporting_evidence=["Ambiguous content"]
))
# Step 4: Check urgency
step += 1
urgency_matches = [ind for ind in self.urgency_indicators if ind in text]
if urgency_matches:
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.URGENCY_INDICATORS,
observation=f"Found {len(urgency_matches)} urgency indicators",
inference="High priority suggested",
confidence=min(0.9, 0.5 + len(urgency_matches) * 0.15),
supporting_evidence=urgency_matches[:5]
))
# Step 5: Domain trust
step += 1
domain = sender.split("@")[-1] if "@" in sender else ""
domain_trust = self.domain_trust.get(domain, 0.5)
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.DOMAIN_TRUST,
observation=f"Domain {domain} has trust score {domain_trust:.2f}",
inference="Internal domain" if domain_trust > 0.8 else "External domain",
confidence=domain_trust,
supporting_evidence=[f"Known domain: {domain in self.domain_trust}"]
))
# Make final recommendation
if spam_matches and len(spam_matches) >= 2:
recommendation = "spam"
final_confidence = min(0.95, 0.5 + len(spam_matches) * 0.15)
elif top_categories and top_categories[0][1] > 0.3:
recommendation = top_categories[0][0]
final_confidence = min(0.9, top_categories[0][1] + trust_score * 0.2)
else:
recommendation = "internal" if domain_trust > 0.8 else "personal"
final_confidence = 0.5
# Generate alternatives
alternatives = []
for cat, score in top_categories[:3]:
if cat != recommendation:
alternatives.append({
"category": cat,
"confidence": score,
"reason": f"Alternative based on keyword score {score:.2f}"
})
decision = Decision(
decision_id=self._generate_decision_id(),
email_id=email_id,
decision_type="categorize",
recommendation=recommendation,
confidence=final_confidence,
reasoning_chain=reasoning_chain,
alternative_options=alternatives
)
self.decisions.append(decision)
return decision
def explain_priority(self, email: Dict[str, Any]) -> Decision:
"""Generate an explainable priority decision."""
email_id = email.get("id", "unknown")
reasoning_chain = []
step = 0
text = f"{email.get('subject', '')} {email.get('body', '')}".lower()
sender_info = email.get("sender_info", {})
time_in_inbox = email.get("time_in_inbox_hours", 0)
sla_priority = email.get("sla_priority")
category = email.get("category", "")
priority_score = 0.0
# Step 1: SLA check
step += 1
if sla_priority:
if sla_priority == "critical":
priority_score += 0.5
inference = "Critical SLA - highest priority"
elif sla_priority == "high":
priority_score += 0.3
inference = "High SLA priority"
else:
inference = f"SLA priority: {sla_priority}"
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.TIME_SENSITIVITY,
observation=f"SLA priority is {sla_priority}",
inference=inference,
confidence=0.95,
supporting_evidence=[f"SLA defined: {sla_priority}"]
))
# Step 2: Time in inbox
step += 1
if time_in_inbox > 8:
priority_score += 0.3
inference = "Email has been waiting too long"
elif time_in_inbox > 4:
priority_score += 0.15
inference = "Email getting stale"
else:
inference = "Email is recent"
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.TIME_SENSITIVITY,
observation=f"Email in inbox for {time_in_inbox:.1f} hours",
inference=inference,
confidence=0.8,
supporting_evidence=[f"Age: {time_in_inbox:.1f}h"]
))
# Step 3: Urgency keywords
step += 1
urgency_matches = [ind for ind in self.urgency_indicators if ind in text]
if urgency_matches:
priority_score += min(0.4, len(urgency_matches) * 0.1)
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.URGENCY_INDICATORS,
observation=f"Found urgency words: {', '.join(urgency_matches[:3])}",
inference="Content indicates urgency",
confidence=min(0.9, 0.5 + len(urgency_matches) * 0.1),
supporting_evidence=urgency_matches
))
# Step 4: Sender importance
step += 1
trust_score = sender_info.get("trust_score", 0.5)
is_vip = trust_score >= 0.9
if is_vip:
priority_score += 0.2
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.SENDER_REPUTATION,
observation=f"VIP sender (trust: {trust_score:.2f})",
inference="Prioritize VIP senders",
confidence=0.9,
supporting_evidence=["High trust score", "Known sender"]
))
# Step 5: Category-based priority
step += 1
category_priorities = {
"customer_support": 0.3,
"security": 0.5,
"billing": 0.2,
"technical": 0.2,
"sales": 0.15,
"spam": -0.5,
"newsletter": -0.3,
}
cat_boost = category_priorities.get(category, 0)
priority_score += cat_boost
if category:
reasoning_chain.append(ReasoningStep(
step_number=step,
factor=DecisionFactor.ENTITY_CONTEXT,
observation=f"Category is {category}",
inference=f"Category priority boost: {cat_boost:+.2f}",
confidence=0.85,
supporting_evidence=[f"Category: {category}"]
))
# Determine final priority
if priority_score >= 0.6:
recommendation = "urgent"
final_confidence = min(0.95, priority_score)
elif priority_score >= 0.4:
recommendation = "high"
final_confidence = min(0.9, priority_score + 0.3)
elif priority_score >= 0.2:
recommendation = "normal"
final_confidence = 0.7
else:
recommendation = "low"
final_confidence = 0.7
decision = Decision(
decision_id=self._generate_decision_id(),
email_id=email_id,
decision_type="prioritize",
recommendation=recommendation,
confidence=final_confidence,
reasoning_chain=reasoning_chain,
alternative_options=[
{"priority": "urgent", "threshold": "score >= 0.6"},
{"priority": "high", "threshold": "score >= 0.4"},
{"priority": "normal", "threshold": "score >= 0.2"},
{"priority": "low", "threshold": "score < 0.2"},
]
)
self.decisions.append(decision)
return decision
def get_context_for_email(self, email: Dict[str, Any]) -> Dict[str, Any]:
"""Get accumulated context for an email from the knowledge graph."""
sender = email.get("sender", "")
email_id = email.get("id", "")
# Find related entities
email_entities_list = [
self.entities[eid].to_dict()
for eid in self.email_entities.get(email_id, [])
]
# Find sender history
sender_emails = []
sender_topics = set()
for eid, entity in self.entities.items():
if entity.type == EntityType.EMAIL_ADDRESS and sender.lower() in entity.value.lower():
# Find relationships for this sender
for rel in self.relationships.values():
if rel.source_entity_id == eid:
target = self.entities.get(rel.target_entity_id)
if target and target.type == EntityType.TOPIC:
sender_topics.add(target.value)
# Get related decisions
related_decisions = [
d.to_dict() for d in self.decisions[-10:]
if d.email_id == email_id
]
return {
"email_id": email_id,
"extracted_entities": email_entities_list,
"sender_history": {
"topics": list(sender_topics),
"relationship_count": len([
r for r in self.relationships.values()
if sender.lower() in str(self.entities.get(r.source_entity_id, {}))
])
},
"related_decisions": related_decisions,
"graph_stats": {
"total_entities": len(self.entities),
"total_relationships": len(self.relationships),
"email_entities_count": len(email_entities_list)
}
}
def query_entities(
self,
entity_type: Optional[EntityType] = None,
value_contains: Optional[str] = None,
email_id: Optional[str] = None,
limit: int = 50
) -> List[Dict[str, Any]]:
"""Query entities from the knowledge graph."""
results = []
if entity_type:
entity_ids = self.entity_index.get(entity_type, set())
else:
entity_ids = set(self.entities.keys())
for eid in entity_ids:
if len(results) >= limit:
break
entity = self.entities[eid]
if email_id and entity.source_email_id != email_id:
continue
if value_contains and value_contains.lower() not in entity.value.lower():
continue
results.append(entity.to_dict())
return results
def query_relationships(
self,
source_type: Optional[EntityType] = None,
target_type: Optional[EntityType] = None,
relation_type: Optional[RelationType] = None,
min_strength: float = 0.0,
limit: int = 50
) -> List[Dict[str, Any]]:
"""Query relationships from the knowledge graph."""
results = []
for rel in self.relationships.values():
if len(results) >= limit:
break
if rel.strength < min_strength:
continue
if relation_type and rel.relation_type != relation_type:
continue
source = self.entities.get(rel.source_entity_id)
target = self.entities.get(rel.target_entity_id)
if source_type and (not source or source.type != source_type):
continue
if target_type and (not target or target.type != target_type):
continue
results.append({
**rel.to_dict(),
"source_value": source.value if source else None,
"target_value": target.value if target else None
})
return results
def get_decision_explanation(self, decision_id: str) -> Optional[Dict[str, Any]]:
"""Get a full explanation for a specific decision."""
for decision in self.decisions:
if decision.decision_id == decision_id:
explanation = decision.to_dict()
# Add human-readable summary
summary_parts = []
for step in decision.reasoning_chain:
summary_parts.append(f"• {step.factor.value}: {step.inference}")
explanation["human_readable_summary"] = "\n".join(summary_parts)
explanation["confidence_level"] = (
"High" if decision.confidence >= 0.8 else
"Medium" if decision.confidence >= 0.5 else
"Low"
)
return explanation
return None
def get_statistics(self) -> Dict[str, Any]:
"""Get statistics about the knowledge graph."""
entity_counts = {}
for entity_type in EntityType:
entity_counts[entity_type.value] = len(self.entity_index.get(entity_type, set()))
relationship_counts = {}
for rel in self.relationships.values():
rel_type = rel.relation_type.value
relationship_counts[rel_type] = relationship_counts.get(rel_type, 0) + 1
decision_counts = {}
for decision in self.decisions:
decision_counts[decision.decision_type] = decision_counts.get(decision.decision_type, 0) + 1
avg_confidence = 0.0
if self.decisions:
avg_confidence = sum(d.confidence for d in self.decisions) / len(self.decisions)
return {
"total_entities": len(self.entities),
"total_relationships": len(self.relationships),
"total_decisions": len(self.decisions),
"entity_counts_by_type": entity_counts,
"relationship_counts_by_type": relationship_counts,
"decision_counts_by_type": decision_counts,
"average_decision_confidence": avg_confidence,
"emails_processed": len(self.email_entities)
}
def reset(self) -> None:
"""Reset the knowledge graph."""
self.entities.clear()
self.relationships.clear()
self.entity_index.clear()
self.email_entities.clear()
self.decisions.clear()
self._entity_counter = 0
self._relationship_counter = 0
self._decision_counter = 0
|