skills.rest
mathiasbourgoin
mathiasbourgoinCOMMUNITY
Explore all skills by this author →
GitHub

kb-search

Search LanceDB knowledge base chunks using hybrid vector and BM25 rankings.

2|2|Updated Mar 17, 2026
Search Tags:#lancedb#semantic-search#knowledge-base#kb-search#hybrid-search#keyword-search#documentation-search
One-click install
npx skills add https://github.com/mathiasbourgoin/roster --skill kb-search-mathiasbourgoin
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kb-search
Source: https://github.com/mathiasbourgoin/roster/tree/main/.opencode/skills/kb-search
Command: npx skills add https://github.com/mathiasbourgoin/roster --skill kb-search-mathiasbourgoin

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the tedious manual work of loading and sifting through entire knowledge base (KB) files to find specific concepts, terms, or constraints, drastically reducing the time spent on documentation lookups.

Core Features & Use Cases

  • Hybrid Search Algorithm: Combines semantic vector search and BM25 keyword search to return the most relevant KB chunks, even for vague or highly specific queries.
  • Contextual, Ranked Results: Each result includes the source file, section, relevance score, and content status for full context.
  • Smart Default Filtering: Automatically excludes outdated historical and superseded chunks by default, and flags low-confidence results when no strong matches are found.
  • Use Case: A developer implementing a new API endpoint can search for "API error response format" to instantly pull the exact relevant constraint from the KB spec without reading through unrelated documentation.

Quick Start

Use the kb-search skill to find all KB content related to database transaction requirements, returning the top 5 most relevant live results.

Frequently Asked Questions about kb-search

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

FAQPage Schema
How does hybrid semantic and keyword search over a knowledge base improve documentation retrieval?▼

Hybrid knowledge base search combines semantic vector search and BM25 keyword rankings to return highly relevant chunks with source attribution, eliminating manual file scanning for precise concept lookup.

What is the best way to search a LanceDB index for specific project requirements?▼

The best way to search a LanceDB index for specific requirements is using hybrid search, which fuses vector and BM25 rankings to locate exact constraints and terms with source file attribution and relevance scores.

How do I filter outdated historical content from knowledge base search results?▼

To filter outdated content from knowledge base search results, use smart default filtering that automatically excludes superseded and historical chunks, returning only live documentation while flagging low-confidence matches.

Does kb-search work for compliance checks and requirements verification in development workflows?▼

Yes, kb-search works for compliance checks and requirements verification by providing filtered, ranked retrieval of knowledge base chunks with status metadata, ensuring developers reference only current, live specifications.

Why does my knowledge base search return low-confidence flagged results for vague queries?▼

Knowledge base search returns low-confidence flagged results for vague queries when no strong semantic or BM25 matches are found, indicating the retrieved chunks may lack sufficient relevance for direct compliance reference.

Related Skills

Explore More Software Engineering Skills→
mav-rikmav-rik
12

kb

Maintain a persistent markdown wiki with hybrid semantic and keyword search.

Community
Advanced
skwuwuskwuwu

gvm_audit

Verify GVM audit log integrity and export compliance reports.

Community
Intermediate
Mercurium-AnalyticsMercurium-Analytics

pg-search-bm25

Index BM25 text searches in Postgres with the pg_search extension.

Official
Advanced
NITISH-gitbitNITISH-gitbit

qmd

Combine BM25 keyword matching, vector semantic search, and LLM reranking for local Markdown file retrieval.

Community
Intermediate