skills.rest
NOMARJ
NOMARJOFFICIAL
Explore all skills by this author →
WebsiteGitHub

vector-index-tuning

Tune HNSW parameters and quantization strategies for vector index performance.

Updated Aug 23, 2026
Search Tags:#scalability#performance tuning#similarity search#quantization#HNSW#vector index
One-click install
npx skills add https://github.com/NOMARJ/nomark-method --skill vector-index-tuning-nomarj
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: vector-index-tuning
Source: https://github.com/NOMARJ/nomark-method/tree/main/claude/skills/llm/vector-index-tuning
Command: npx skills add https://github.com/NOMARJ/nomark-method --skill vector-index-tuning-nomarj

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires hnswlib, sklearn, qdrant-client, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill addresses the challenge of optimizing vector index performance, ensuring efficient and accurate similarity searches in large-scale datasets.

Core Features & Use Cases

  • Parameter Tuning: Adjust HNSW parameters (M, efConstruction, efSearch) for optimal recall and latency.
  • Quantization Strategies: Implement scalar (INT8) or product quantization to reduce memory footprint.
  • Index Selection: Guidance on choosing the right index type (Flat, HNSW, IVF, DiskANN) based on data size.
  • Use Case: When deploying a large-scale recommendation system, use this Skill to fine-tune the vector index for billions of user embeddings, balancing search speed with recommendation accuracy.

Quick Start

Use the vector-index-tuning skill to recommend HNSW parameters for 10 million vectors targeting 95% recall.

Frequently Asked Questions about vector-index-tuning

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I tune HNSW parameters for optimal vector search recall and latency?▼

Tune HNSW parameters by adjusting M, efConstruction, and efSearch to balance vector search recall against latency. This Skill provides specific recommendations for configuring these parameters to achieve targeted performance goals like 95% recall.

What is the best way to reduce memory footprint for large-scale similarity search?▼

Reduce memory footprint for similarity search by implementing scalar (INT8) or product quantization strategies. This Skill guides the selection and application of quantization methods to compress vector data while maintaining search accuracy.

How do I choose the right vector index type for my dataset size?▼

Choose a vector index type by evaluating Flat, HNSW, IVF, and DiskANN against your specific data size. This Skill provides selection guidance to match your dataset scale with the most efficient index structure for your infrastructure.

Can I use this to optimize vector search for billions of user embeddings?▼

Yes, you can optimize vector search for billions of user embeddings. This Skill is designed for large-scale recommendation systems, fine-tuning vector indexes to balance search speed with recommendation accuracy at massive scale.

Does this Skill support tuning vector indexes with Qdrant and hnswlib?▼

Yes, this Skill supports tuning vector indexes using Qdrant and hnswlib. It relies on qdrant-client, hnswlib, and sklearn to apply parameter adjustments, quantization strategies, and scaling infrastructure configurations.

Why does my vector search latency increase when targeting higher recall?▼

Vector search latency increases with higher recall targets because HNSW parameters like efSearch must scan more nodes. This Skill helps you tune these parameters to find the optimal trade-off between search speed and accuracy.

Related Skills

Explore More Software Engineering Skills→
dotrurudotruru

umap-learn

Reduce high-dimensional data to low-dimensional embeddings for visualization and preprocessing.

Community
Advanced
FalkorDBFalkorDB
22

Create and query vector indexes

Create and query HNSW vector indexes in FalkorDB graph databases.

Official
Intermediate
sourcegraphsourcegraph
32

ir-analysis

Compute precision, recall, MRR, nDCG, and MAP for file retrieval comparisons.

Official
Advanced
nidhi-subrahnidhi-subrah

AgentDB Performance Optimization

Configure AgentDB with quantization, HNSW indexing, caching, and batch operations.

Community
Advanced