Vector Embed

Generates 384-dimensional embeddings using the all-MiniLM-L6-v2 model via ruvector's ONNX runtime, then stores them in an HNSW index for fast semantic search. You get 52,000+ inserts per second and sub-millisecond lookups, which is legitimately fast for local embedding work. The skill walks you through text-only embedding (no batch flags exist in 0.2.25, so you loop yourself), offers an adaptive LoRA variant for domain-specific tuning, and can optionally hook into MCP tools for RAG context and collective brain search. Main gotcha is the ONNX runtime isn't bundled by default, so first-run failures usually mean installing the separate wasm package. Solid choice if you need local embeddings without API calls.

Install to Claude Code

npx -y skills add ruvnet/ruflo --skill vector-embed --agent claude-code