Vector Search

Vector Search

Two distinct vector-search paths live in this plugin. Pick the right one — they're not interchangeable.

The "12,500×" headline applies to the large-scale embeddings_search path. The WASM router is not that path.

When to use

Standard search

1. Check status — mcp__claude-flow__embeddings_status to verify the embedding engine. 2. Initialize — mcp__claude-flow__embeddings_init if not active. 3. Generate — mcp__claude-flow__embeddings_generate for text input. 4. Search — mcp__claude-flow__embeddings_search with the query. 5. Compare — mcp__claude-flow__embeddings_compare to measure similarity. 6. Unified search — mcp__claude-flow__memory_search_unified for cross-namespace.

Quantized search (32× memory reduction)

For corpora ≥5,000 vectors and/or memory-constrained environments, use the RaBitQ 1-bit quantization workflow. Below 5,000 vectors the rebuild cost outweighs the savings — use the standard path instead.

Note: embeddings_rabitq_search returns candidate IDs only — the rerank in step 4 is the user's responsibility (mirrors the docstring at embeddings-tools.ts:911). Without rerank, results are approximate; with rerank, you get full-precision quality at 32× lower memory.

Tuning

HNSW exposes three knobs that trade recall against latency. The "12,500×" headline assumes defaults; tune deliberately for your workload:

efSearch is passed via ruvllm_hnsw_create (ruvllm-tools.ts:64). M is registry-level today; raise as a follow-up if it should be MCP-tunable. efConstruction defaults to 200 in the lite index (hnsw-index.ts:537).

HNSW pattern router (WASM, ≤11 patterns)

For routing a small number of high-priority patterns:

• mcp__claude-flow__ruvllm_hnsw_create — create the WASM index (cap ~11)
• mcp__claude-flow__ruvllm_hnsw_add — add a pattern
• mcp__claude-flow__ruvllm_hnsw_route — route an incoming query

This is not a corpus index. Treat it as a fast classifier over a curated set of patterns.

Hyperbolic embeddings

For hierarchical data (code trees, org charts), use mcp__claude-flow__embeddings_hyperbolic which maps to Poincare ball space. Distance is geodesic, not cosine.

CLI alternative

npx @claude-flow/cli@latest embeddings search --query "authentication patterns"
npx @claude-flow/cli@latest embeddings init
npx @claude-flow/cli@latest memory search --query "your query"

Performance

Measured numbers (source: scripts/benchmark-intelligence.mjs, ruvector NAPI backend; recall@10 ≈ 0.99). The older "150×–12,500×" figures were brute-force-fallback artifacts and have been retired — see project CLAUDE.md "V3 Performance Targets".