CORE GUIDE
Knowledge graph structure
A knowledge graph represents entities and typed relationships explicitly so retrieval and reasoning can follow meaningful connections that are difficult to recover from text similarity alone.
Mental model
Think of a knowledge graph as a structured map of claims about relationships. Nodes identify things; edges state how they are connected; provenance and scope determine when those connections can be trusted.
Why it matters
Some questions are inherently relational: which service depends on this API, which policy supersedes another, which customer owns this asset, or which concept is a prerequisite. Flattening all of that into chunks can make multi-hop or constraint-heavy questions brittle. Graph structure gives the application a separate retrieval primitive rather than replacing text retrieval.
01
Model stable entities and relationships that change decisions
Start with a small ontology tied to concrete queries. Give entities stable identifiers, define edge types and direction, attach provenance where relationships can change, and keep graph updates synchronized with source systems. Use graph traversal to produce candidate facts or neighborhoods, then combine them with text evidence when explanation or unstructured detail is needed.
02
Example: dependency impact is invisible to vector search
An engineering assistant must answer which applications are affected by removing an internal API. Documentation chunks mention services separately and have weak semantic overlap with the question. A graph with service→depends-on→API edges can enumerate impacted applications, while linked documents provide the human-readable evidence for each dependency.
Common failure modes
- Building a large ontology before proving a relational query needs it.
- Treating every co-occurrence as a meaningful graph edge.
- Returning graph paths without provenance or the text evidence needed to validate them.
Engineering heuristics
- Create graph structure around stable entities and decision-relevant relationships.
- Keep edge semantics narrow enough that traversal has a clear meaning.
- Combine structured traversal with source evidence when users need justification.
Takeaways
- 01Knowledge graphs add a relational retrieval primitive, not a replacement for all RAG.
- 02Typed edges matter more than raw connection count.
- 03Graph answers need provenance and lifecycle just like document evidence.
Reading evidence
This records actions you actually took; it does not claim mastery, proficiency, or certification.
Used in
This Concept is reused across these canonical learning paths.
Related concepts from the Knowledge Graph
These relationships come from the canonical graph, not a separate Guide taxonomy.