Knowledge Engine
Knowledge Engine Architecture
Last Updated: 2026-03-06 Status: Active high-level reference Layer: AI context / retrieval augmentation
This document describes the current conceptual role of the knowledge layer without overclaiming historical implementations.
For current architecture, see current-architecture.md.
Overview
The knowledge layer enriches Jorvis beyond raw schema retrieval. It provides business-context augmentation and retrieval-adjacent signals that can improve query generation and explanation quality.
Current high-level areas:
- glossary-driven context
- metric definitions / reusable business semantics
- schema-aware retrieval support
- query examples / retrieval support
- optional document-oriented context depending on environment
Current Positioning
The knowledge layer should be read as context augmentation, not as proof that every historical submodule remains active exactly as originally designed.
In particular:
- do not treat historical Apache AGE-based graph-linking designs as current architecture truth
- do not use old orphan-module-era docs as proof of current DI reachability
Current Architectural Truth
| Area | Current Reading |
|---|---|
| Glossary / business terms | Feature-flagged context layer; verify current activation per environment |
| Metric definitions | Conceptually supported; verify concrete runtime usage before claiming prod default |
| GraphRAG / schema graph behavior | Current production story is plain-SQL / graph-edges oriented, not Apache AGE based |
| Document linking / deeper semantic graphing | Treat as conditional / evolutionary, not unconditional current prod truth |
Reading Rules
When interpreting older docs about the knowledge layer:
- distinguish between historical design intent and current runtime behavior
- prefer current architecture docs for activation claims
- prefer current architecture docs for product/runtime claims