Analytics Overview
The Context Graph analytics subsystem provides intelligent entity suggestions, pattern mining across agent traces, decision precedent ranking, and feedback-driven quality scoring. These services work together to surface insights from the knowledge graph and improve agent performance over time.
Subsections
| Page | Description |
|---|---|
| Entity Suggestions | Auto-completion and typeahead suggestions for entities, tags, and queries |
| Pattern Mining | Sequential and graph-based pattern discovery from agent traces |
| Decision Ranking | Multi-factor ranking of past decisions for precedent search |
| Feedback Scoring | Multi-factor trace quality assessment from explicit and implicit feedback |
How Analytics Fit Together
User Query --> Entity Suggestions (typeahead)
|
Search Results --> Decision Ranking (precedent scoring)
|
Agent Execution --> Pattern Mining (pattern discovery)
|
User Feedback --> Feedback Scoring (quality signals)
|
Improved Rankings and Suggestions (feedback loop)Key Services
| Service | Source | Purpose |
|---|---|---|
EntitySuggestionService | services/entity_suggestion_service.py | Fast prefix-based and semantic suggestions |
PatternMiningService | services/pattern_mining_service.py | Sequential pattern and anomaly detection |
DecisionRankingService | services/decision_ranking_service.py | Multi-factor precedent ranking |
FeedbackScoringService | services/feedback_scoring_service.py | Composite trace quality scoring |