Skills vector-index-tuning

📦

vector-index-tuning

Name: vector-index-tuning
Author: wshobson

Content revision r1 Safe ⚙️ External commands🌐 Network access📁 Filesystem access

Optimize Vector Search Indexes

Vector search systems often trade recall, latency, and memory without clear tuning guidance. This skill provides parameter templates, quantization choices, and monitoring patterns for production indexes.

Supports: Claude Codex Code(CC)

🥉 77 Bronze

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "vector-index-tuning" from https://skillstore.io/skills/wshobson-vector-index-tuning.md and its manifest at https://skillstore.io/api/skills/wshobson-vector-index-tuning/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/wshobson-vector-index-tuning.md Signed manifest GET /api/skills/wshobson-vector-index-tuning/manifest Signed lockfile GET /api/skills/wshobson-vector-index-tuning/lockfile

Test it

Using "vector-index-tuning". Recommend HNSW settings for two million vectors with 768 dimensions, 95 percent recall, and 20 ms latency.

Expected outcome:

Start with M around 48, efConstruction near 256, and efSearch near 128. Benchmark efSearch values from 64 to 256 against real queries before rollout.

Using "vector-index-tuning". How can we reduce memory for a large semantic search index?

Expected outcome:

Evaluate INT8 scalar quantization first, then product quantization if memory remains too high. Measure recall loss, rescoring quality, and p95 latency.

Using "vector-index-tuning". What should we monitor after changing index parameters?

Expected outcome:

Track recall against ground truth, p50, p95, and p99 latency, query throughput, build time, index size, and drift in representative queries.

Security Audit

Safe

v7 • 7/8/2026 Open versioned report

The static findings are false positives caused by Markdown code fences, reference links, and Qdrant memmap configuration examples. No prompt injection, data exfiltration intent, or unauthorized execution behavior was found in SKILL.md.

Files scanned

522

Lines analyzed

Review items

False positives ignored

Risk Factors

⚙️ External commands (11)

SKILL.md:23-30 SKILL.md:30-42 SKILL.md:42-48 SKILL.md:48-54 SKILL.md:54-155 SKILL.md:155-159 SKILL.md:159-285 SKILL.md:285-289 SKILL.md:289-405 SKILL.md:405-409 SKILL.md:409-500

🌐 Network access (3)

SKILL.md:519 SKILL.md:520 SKILL.md:521

📁 Filesystem access (4)

SKILL.md:339 SKILL.md:343 SKILL.md:347 SKILL.md:351

No confirmed security findings were detected by the latest completed static and semantic audit. This does not prove the skill has no side effects.

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/wshobson-vector-index-tuning/audits/7?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/wshobson-vector-index-tuning/security.svg)](https://skillstore.io/skills/wshobson-vector-index-tuning?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/wshobson-vector-index-tuning?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/wshobson-vector-index-tuning/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/wshobson-vector-index-tuning.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

wshobson. (2026). vector-index-tuning security audit report (audit version 7) [Author version unspecified]. Skillstore. https://skillstore.io/skills/wshobson-vector-index-tuning/audits/7

BibTeX citation

@techreport{wshobson-wshobson-vector-index-tuning-2026,
  author = {wshobson},
  title = {vector-index-tuning security audit report (audit version 7)},
  institution = {Skillstore},
  year = {2026},
  number = {7},
  url = {https://skillstore.io/skills/wshobson-vector-index-tuning/audits/7},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "vector-index-tuning security audit report (audit version 7)"
version: "unspecified"
type: report
authors:
  - name: "wshobson"
date-released: "2026-07-08"
url: "https://skillstore.io/skills/wshobson-vector-index-tuning/audits/7"
identifiers:
  - type: other
    value: "skillstore:wshobson-vector-index-tuning:audit:7"
    description: "Skillstore immutable audit report identifier"

Compare variants

2 installable variants

Each author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.

Why this variant is first

Highest Skillstore Score

Author / source Skillstore Score Evidence Confidence Skillstore usage Updated

sickn33 Recommended

sickn33-vector-index-tuning

Skillstore Score 79

Evidence Confidence High

Skillstore usage 10

Updated

2026-08-21

wshobson Current

wshobson-vector-index-tuning

Skillstore Score 77

Evidence Confidence High

Skillstore usage 16

Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: High

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Tune Production Recall

Balance efSearch, M, and quantization choices against a target recall and latency budget.

Plan Vector Memory Capacity

Estimate storage needs for embedding dimensions, vector count, index overhead, and compression strategy.

Compare Database Settings

Create Qdrant collection profiles for speed, recall, balanced performance, or reduced memory use.

Try These Prompts

Choose an Index Type

Recommend a vector index type for my dataset. I have [vector count], [dimensions], [target recall], [latency target], and [memory budget].

Tune HNSW Settings

Suggest HNSW M, efConstruction, and efSearch values for [vector count], [recall target], [update rate], and [maximum query latency].

Evaluate Quantization

Compare FP32, FP16, INT8 scalar, product quantization, and binary quantization for [dimensions], [vector count], and [recall tolerance].

Design a Benchmark Plan

Create a benchmark plan for our vector search index using real queries, ground truth results, latency percentiles, recall metrics, and memory measurements.

Best Practices

Benchmark with production-like queries and known ground truth before changing index settings.
Tune one major parameter at a time so recall and latency changes are attributable.
Monitor recall, latency percentiles, throughput, memory, and rebuild cost after deployment.

Avoid

Optimizing for low latency without measuring recall loss.
Choosing aggressive quantization before testing representative embeddings and queries.
Ignoring index build time, warmup, and maintenance windows for large collections.

Frequently Asked Questions

What vector index sizes does this skill cover?

It covers small flat indexes, HNSW for medium and large datasets, and quantized approaches for larger deployments.

Can it tune Qdrant collections?

Yes. It includes Qdrant examples for HNSW, quantization, optimizer settings, and search parameters.

Does it run benchmarks automatically?

No. It provides benchmark templates and measurement guidance that users can adapt to their own environment.

How should I choose efSearch?

Start from the recall target, test several values, and select the lowest value that meets recall and latency goals.

When should I use quantization?

Use quantization when memory, storage, or cache pressure limits performance, then validate recall impact with real queries.

Is this skill limited to one database?

No. Qdrant examples are included, but the HNSW, quantization, and monitoring guidance applies broadly.

Developer Details

Author

wshobson

License

MIT

Skillstore revision

Version notice

The author did not declare a version.

Repository

https://github.com/wshobson/agents/tree/main/plugins/llm-application-dev/skills/vector-index-tuning

Ref

64ca8af0f54a325752f08bd54e52151061ea659a

Maintenance freshness

7/20/2026

Usage

14 downloads · 203 views

File structure

📄 SKILL.md