Skip to content
Back to skills

Index Selection And Search

ASecurity

"Guides dense-vector metric selection, index construction,

  • 247 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 8, 2026
toolspythonc++apidatabase

Works with

  • api

Security analysis

A100/100

Pro scans all 5 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill index-selection-and-search --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Index Selection And Search?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Index Selection And Search
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-index-selection-and-search/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-index-selection-and-search)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: index-selection-and-search
description: "Guides dense-vector metric selection, index construction,
  approximate search tuning, and CPU search validation in Faiss."
disable-model-invocation: true
metadata:
  disco-role: operating
license: MIT
---

# Index selection and search

Use this route when the task is to choose or operate a CPU dense-vector index,
compare exact and approximate nearest-neighbor search, or diagnose distance
semantics, result contracts, search tuning, or OpenMP behavior. It covers
float32 dense vectors, L2, inner product, cosine-as-normalized-inner-product,
Flat, IVF-Flat, and graph indexes.

## Route the workflow

1. Normalize the data to a contiguous `float32` matrix of shape `(n, d)` and
   keep database and query dimensions identical. Read
   [the API contract](references/api-reference.md) before calling an index.
2. Choose the smallest index that meets the latency, recall, and memory target.
   Use [the selection and tuning workflows](references/workflows.md) for the
   decision, including the 64D memory-conscious factory case.
3. Build with the lifecycle in the workflow reference: create, train when
   `is_trained` is false, add, then search. Validate against an exact Flat
   baseline before increasing approximation.
4. Use the bundled deterministic helper for a local CPU smoke check:
   `python path/to/skills/disco/faiss/sub-skills/index-selection-and-search/scripts/smoke_search.py --help`.
   It has no downloads and is safe to run from any working directory.
5. For malformed arrays, unavailable methods, empty result ranges, or poor
   recall, follow [troubleshooting](references/troubleshooting.md).

## Boundaries

- Route codecs, binary indexes, PQ/SQ, and compressed-memory choices to
  [training-and-compression](../training-and-compression/SKILL.md).
- Route IDs, selectors, composition, refinement, shards, and replicas to
  [composition-and-filtering](../composition-and-filtering/SKILL.md).
- Route persistence, I/O, clustering, and evaluation tooling to
  [persistence-and-evaluation](../persistence-and-evaluation/SKILL.md).
- Route GPU, cuVS, ROCm, Metal, SVS, and C/C++ interop to
  [accelerated-and-interoperable](../accelerated-and-interoperable/SKILL.md).
- Use the [Faiss root route](../../SKILL.md) for package-wide installation,
  provenance, and cross-workflow routing.

Files in this skill

  • SKILL.md2.3 KB
  • references/api-reference.md8 KB
  • references/troubleshooting.md6.5 KB
  • references/workflows.md7.8 KB
  • scripts/smoke_search.py7.3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.