Quadfecta Search System

What is Quadfecta?

Quadfecta is Nexus's 4-layer intelligent search that combines multiple search strategies for superior results:

Layer 1: Keyword Search (L1)

• Traditional keyword/term matching
• BM25 ranking algorithm
• Fast exact matches
• Good for: Known terms, IDs, exact phrases

Layer 2: Vector Search (L2)

• Semantic similarity using embeddings
• Model: all-MiniLM-L6-v2 (384 dimensions)
• Understands meaning, not just words
• Good for: "Find things like X", conceptual search

Layer 3: Graph Search (L3)

• Relationship-based discovery
• FalkorDB (Redis + Graph)
• OpenCypher queries
• Good for: "What's connected to X?", entity relationships

Layer 4: Temporal Search (L4)

• Time-aware scoring
• Recency decay (recent = higher score)
• Access frequency tracking
• Good for: "What did we discuss yesterday?", recent context

Combined Scoring

Each layer contributes to final score:

Score = (K * keyword) + (V * vector) + (G * graph) + (T * temporal)

Default weights: - Keyword: 0.30 - Vector: 0.40 - Graph: 0.20 - Temporal: 0.10

Cross-Environment Search

Quadfecta searches across ALL environments: - Context, Track, Contact, Documents, Transcripts, Web, etc. - Single query returns results from any relevant environment - Results ranked by combined Quadfecta score

Search Tools

• search.all - Search all environments
• search.context - Search Context only
• search.track - Search Track only
• search.temporal - Search by date range
• (etc. for each environment)