Does it support vector search optimization?

Yes, it includes specific playbooks for vector ANN parameters, helping balance the trade-off between recall and latency.

What metrics can I improve with this skill?

It targets search latency (p50/p95), indexing throughput, memory usage (RSS), and HTTP ingestion rates.

Does this skill work with any search engine?

This skill is specifically designed for the Searchlite embedded search engine, utilizing its specific Rust API, profiling tools, and pruning algorithms like WAND and BMW.

How does it ensure data isn't corrupted during optimization?

The skill includes non-negotiable rules that forbid weakening durability. It ensures WAL, manifest files, and fsync operations remain intact while tuning the hot path.

Search Performance Optimization

Name: Search Performance Optimization
Author: davidkelley

bydavidkelley

0•

Database Management

Optimizes Searchlite search engine performance through methodical profiling, bottleneck identification, and targeted code enhancements.

This skill provides a rigorous, measurement-first workflow for enhancing the performance of Searchlite, an embedded Rust search engine. It guides users through a structured discovery phase using built-in profiling tools like BM25 and WAND pruning to identify hotspots in CPU usage, memory allocation, or I/O overhead. By enforcing strict 'measure first' rules and preserving durability requirements, it enables developers to safely improve indexing throughput, query latency, and memory efficiency while ensuring that optimizations never compromise data correctness or system stability.

Key Features

01Strict durability preservation for WAL, manifests, and atomic write ordering

02Structured discovery phase for identifying true performance hotspots

03Comprehensive validation workflow using benchmarks and regression testing

04Targeted optimization playbooks for CPU, memory churn, and I/O efficiency

05Support for built-in query profiling tools including BM25, WAND, and BMW

060 GitHub stars

Use Cases

01Reducing search query p95 latency in production environments

02Increasing document indexing throughput and reducing commit overhead

03Minimizing memory usage and heap allocations in resource-constrained embedded systems

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add davidkelley/searchlite performance-improvements

For use in Claude.ai and ChatGPT

Download Skill