skills.rest
Sir-chawakorn
Sir-chawakornCOMMUNITY
Explore all skills by this author →
GitHub

build-vector-search

Configure embedding models, ANN indices, and hybrid search pipelines for vector retrieval.

9|3|Updated Jun 13, 2026
Search Tags:#embeddings#rag#retrieval#vector search#hybrid search#ann#hnsw
One-click install
npx skills add https://github.com/Sir-chawakorn/sanook-cli --skill build-vector-search
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: build-vector-search
Source: https://github.com/Sir-chawakorn/sanook-cli/tree/main/skills/build-vector-search
Command: npx skills add https://github.com/Sir-chawakorn/sanook-cli --skill build-vector-search

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill addresses the complexity of building and tuning high-recall vector retrieval systems, preventing common pitfalls like poor index configuration, incorrect embedding usage, and ineffective filtering strategies.

Core Features & Use Cases

  • Retrieval Optimization: Expert guidance on selecting embedding models, distance metrics, and ANN index parameters (HNSW/IVF-PQ) to balance recall, latency, and memory.
  • Hybrid & Reranking: Implementation of RRF-based hybrid search (BM25 + Dense) and cross-encoder reranking to solve semantic mis-ranking.
  • Use Case: If your RAG application returns irrelevant context or suffers from slow search performance, use this skill to tune your HNSW index, implement a reranker, and validate retrieval quality using recall@k metrics.

Quick Start

Use the build-vector-search skill to evaluate my current HNSW index configuration and suggest optimal parameters for a recall target of 0.95.

Frequently Asked Questions about build-vector-search

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

FAQPage Schema
How do I tune HNSW index parameters to improve vector search recall?▼

To improve vector search recall, configure HNSW index parameters by balancing connection limits and search depth to hit your recall target. This tuning process optimizes the tradeoff between retrieval accuracy and query latency.

What is hybrid search and when do I need RRF reranking for RAG?▼

Hybrid search combines BM25 sparse retrieval with dense vector embeddings using Reciprocal Rank Fusion (RRF). You need it when semantic mis-ranking occurs in RAG, ensuring relevant context is retrieved by merging lexical and semantic signals.

How do I evaluate vector retrieval quality using recall@k and nDCG metrics?▼

Evaluate vector retrieval quality by calculating recall@k, nDCG, and MRR metrics against a ground truth dataset. These metrics measure search relevance and rank accuracy, validating that your system meets high-recall requirements.

Why does my RAG application return irrelevant context despite using embeddings?▼

Your RAG application likely returns irrelevant context due to poor index configuration, incorrect embedding usage, or ineffective metadata filtering. Implementing cross-encoder reranking and tuning distance metrics resolves these semantic mis-ranking issues.

Can I use metadata filtering with ANN indices without losing recall performance?▼

Yes, you can apply metadata filtering with ANN indices, but it requires careful pipeline configuration to prevent recall degradation. Effective filtering strategies maintain high-recall retrieval while narrowing down vector search results efficiently.

Related Skills

Explore More Software Engineering Skills→
jhl-labsjhl-labs
50

RAG & Vector Search

Retrieve relevant documents and augment LLM prompts with vector search.

Official
Advanced
floflo777floflo777
33

rag-eval

Evaluate RAG systems using retrieval, generation, and latency metrics.

Community
Intermediate
ArogyaReddyArogyaReddy

hybrid-search-implementation

Combine vector similarity and keyword matching with RRF, Linear, Cross-encoder, or Cascade fusion.

Community
Advanced
mahg-esmahg-es
3

vector-search

Configure HNSW/IVF indices and perform ANN search with vector embeddings.

Official
Advanced