# Vector Index Tuning

> Optimize vector index performance for latency, recall, and memory. Use when tuning HNSW parameters, selecting quantization strategies, or scaling vector search infrastructure.

Source: https://skillsagentes.com/skills/wshobson/agents/vector-index-tuning
Repository: https://github.com/wshobson/agents
Author: wshobson
License: MIT
Updated: hace 2 meses
Context cost: 44 tok installed, 581 tok once triggered, 3.9k tok with every bundled file
Bundle: 2 files, 15 KB
Permissions requested: none declared

## Install

```bash
npx -y skills add wshobson/agents --skill vector-index-tuning --agent claude-code
```

## What it does

- Guía para elegir tipo de índice vectorial según tamaño de datos
- Explica parámetros HNSW (M, efConstruction, efSearch) y su efecto
- Detalla estrategias de cuantización (FP32, FP16, INT8, PQ, binaria)
- Referencia plantillas y ejemplos detallados en references/details.md

## Use it when

- Ajustar parámetros HNSW
- Implementar cuantización
- Optimizar uso de memoria
- Reducir latencia de búsqueda o balancear recall vs velocidad

## What triggers it

- "Ayúdame a tunear los parámetros HNSW de mi índice"
- "¿Qué tipo de cuantización uso para 50M vectores?"
- "Cómo reducir la latencia de búsqueda vectorial"
- "Necesito escalar mi búsqueda vectorial a miles de millones de vectores"

## Files

- SKILL.md — 2 KB
- references/details.md — 13 KB

## SKILL.md

Reproduced verbatim from wshobson/agents under MIT. This section is the upstream document and is in English.

# Vector Index Tuning

Guide to optimizing vector indexes for production performance.

## When to Use This Skill

- Tuning HNSW parameters
- Implementing quantization
- Optimizing memory usage
- Reducing search latency
- Balancing recall vs speed
- Scaling to billions of vectors

## Core Concepts

### 1. Index Type Selection

```
Data Size           Recommended Index
────────────────────────────────────────
< 10K vectors  →    Flat (exact search)
10K - 1M       →    HNSW
1M - 100M      →    HNSW + Quantization
> 100M         →    IVF + PQ or DiskANN
```

### 2. HNSW Parameters

| Parameter          | Default | Effect                                               |
| ------------------ | ------- | ---------------------------------------------------- |
| **M**              | 16      | Connections per node, ↑ = better recall, more memory |
| **efConstruction** | 100     | Build quality, ↑ = better index, slower build        |
| **efSearch**       | 50      | Search quality, ↑ = better recall, slower search     |

### 3. Quantization Types

```
Full Precision (FP32): 4 bytes × dimensions
Half Precision (FP16): 2 bytes × dimensions
INT8 Scalar:           1 byte × dimensions
Product Quantization:  ~32-64 bytes total
Binary:                dimensions/8 bytes
```

## Templates and detailed worked examples

Full template library and detailed worked examples live in `references/details.md`. Read that file when you need the concrete templates.

## Best Practices

### Do's

- **Benchmark with real queries** - Synthetic may not represent production
- **Monitor recall continuously** - Can degrade with data drift
- **Start with defaults** - Tune only when needed
- **Use quantization** - Significant memory savings
- **Consider tiered storage** - Hot/cold data separation

### Don'ts

- **Don't over-optimize early** - Profile first
- **Don't ignore build time** - Index updates have cost
- **Don't forget reindexing** - Plan for maintenance
- **Don't skip warming** - Cold indexes are slow

---

Skills Agentes — https://skillsagentes.com/skills/wshobson/agents/vector-index-tuning