Skip to main content

What is vcal-core?

vcal-core is a lightweight Rust library that provides a fast semantic cache / vector index built on HNSW.
Use it to store embeddings, search nearest neighbors, and implement cache-like reuse in your LLM apps.

Highlights​

  • HNSW index (Cosine / Dot)
  • Insert / Upsert / Delete (soft tombstone)
  • TTL and capacity-based eviction (optional)
  • Batch search
  • Snapshots (save/load with serde)
  • Optional SIMD acceleration

When to use vcal-core​

  • You want an in-process vector index with zero network hops.
  • You need predictable performance and control over lifecycle (TTL/eviction).
  • You don’t need a distributed vector DB for this component.

See Quickstart → minimal example in Rust.