Local inference is only half the problem. This adds the retrieval layer
that makes your own documents queryable.

WHAT YOU GET

Everything in Local Inference Starter, plus three vector engines running and
connected:

• Qdrant - networked vector store with payload filtering
• Chroma - embedded store, ideal for single-node and prototypes
• FAISS - Meta’s ANN library for in-process similarity search

Plus a retrieval smoke test that proves the index actually returns results,
and documentation explaining which engine to pick and why.

WHY THREE

They solve different problems. Chroma gets you moving without operating a
database. Qdrant is the default when you want a real service with filtering.
FAISS is for maximum recall and speed inside your own process. Shipping all
three means you can choose per workload instead of rebuilding later.

ALL OPEN SOURCE, ALL PERMISSIVE

MIT and Apache-2.0 throughout. Full attribution with pinned commits included.

REQUIREMENTS

Docker Engine 24+ with Compose v2.

Instant delivery after payment. One-time purchase, free updates.