Knowledge Graph Engine

Graph-based knowledge representation where concept nodes are connected by typed, weighted, temporal relationships. Not just 'A relates to B' — each edge captures the relationship type (causal, associative, hierarchical, temporal), its strength (weight accumulated from evidence), and when it was established and last reinforced. Entity nodes carry confidence scores, source attribution, and category classification.

Traversal queries find paths between concepts across multiple hops, expand neighbourhoods around a seed entity, detect clusters of densely connected concepts, and identify trends across temporal data. Path queries answer 'how does X connect to Y' by returning the specific chain of typed relationships with strength scores at each hop. Neighbourhood expansion answers 'what is related to X' with ranked results by relationship strength.

Entity resolution handles the real-world problem of the same concept appearing under different names or spellings across sources. Fuzzy matching and alias tracking consolidate variant references into canonical nodes. Edge weights decay over time for temporal relevance — recent connections carry more weight than stale ones unless reinforced by new evidence.

Shared knowledge structure that multiple systems read from and write to. The content pipeline queries it for idea generation — clusters of well-supported facts that haven't been covered yet. The research hub uses it for lateral exploration — following edges to discover connections the user didn't know existed. The associative memory system uses it for seed expansion during recall. The document intelligence pipeline writes to it as new entities and relationships are extracted from ingested content.