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.
Security Assessment
About vector-index-tuning
The vector-index-tuning skill is designed to optimize the performance of vector indexes in various applications, focusing on critical parameters such as latency, recall, and memory usage. This skill addresses common challenges faced when working with large datasets and complex vector search infrastructures. By providing guidelines for selecting index types and tuning parameters, it helps users achieve efficient and effective retrieval of information from substantial volumes of vector data.
FAQ
When should I use the vector-index-tuning skill?
Use this skill when tuning HNSW parameters, implementing quantization, optimizing memory usage, reducing search latency, balancing recall versus speed, or scaling to billions of vectors.
What are the recommended index types for different data sizes?
For data sizes under 10K vectors, use Flat (exact search). For 10K to 1M vectors, use HNSW. For 1M to 100M vectors, use HNSW with quantization. For over 100M vectors, use IVF with PQ or DiskANN.
What are the limitations of using this skill?
The effectiveness of the tuning depends on the correct selection of parameters and may require substantial computational resources for large datasets.
Install vector-index-tuning
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
wshobson/agents