skills.rest
rizaldiem
rizaldiemCOMMUNITY
Explore all skills by this author →
GitHub

chunking-strategy

Break large documents into semantically meaningful segments for retrieval and embedding.

Updated Apr 5, 2026
Search Tags:#embeddings#rag#chunking#semantic-search#document-processing#text-splitting
One-click install
npx skills add https://github.com/rizaldiem/digital-invitation-web_V2 --skill chunking-strategy-rizaldiem
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: chunking-strategy
Source: https://github.com/rizaldiem/digital-invitation-web_V2/tree/main/.windsurf/skills/chunking-strategy
Command: npx skills add https://github.com/rizaldiem/digital-invitation-web_V2 --skill chunking-strategy-rizaldiem

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Breaks large documents into semantically coherent chunks to preserve context for embeddings and improve retrieval quality in RAG pipelines, reducing irrelevant or fragmented search results.

Core Features & Use Cases

  • Multiple Strategies: Fixed-size, recursive character, structure-aware, semantic boundary detection, sliding window, and late-chunking approaches with practical parameter recommendations.
  • Pipeline Integration: Guidance for pre-processing, strategy selection, embedding generation, vector indexing, and evaluation metrics (precision, recall, MRR, MAP).
  • Evaluation & Optimization: Tools and methods for measuring retrieval quality, end-to-end RAG accuracy, processing time, and memory trade-offs; use cases include semantic search, QA over long documents, and multi-modal content handling.
  • Use Case Example: Convert a mixed-format technical manual into optimized chunks, generate embeddings, and store them in a vector DB to improve analytical and factoid query retrieval.

Quick Start

Chunk the provided document into 512-token chunks with 10% overlap using semantic-aware chunking and validate chunk coherence and retrieval performance.

Frequently Asked Questions about chunking-strategy

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

FAQPage Schema
What is the best way to split documents for RAG and semantic search?▼

Document chunking splits large texts into semantically meaningful segments using fixed-size, recursive, structure-aware, or semantic boundary detection methods to preserve context and improve RAG retrieval quality.

How do I choose the right chunk size and overlap for my vector database?▼

Configurable chunk size and overlap parameters allow balancing processing time and memory trade-offs; setting 512-token chunks with 10% overlap using semantic-aware chunking validates coherence and optimizes embedding generation for vector indexing.

Can I use late chunking instead of recursive character splitting for long documents?▼

Late chunking and recursive character splitting are both supported strategies; late chunking processes embeddings differently to retain broader context, while structure-aware splitting respects document formatting for multi-modal content handling.

Why does my RAG pipeline return fragmented or irrelevant search results?▼

Fragmented RAG search results often occur when chunking fails to preserve semantic boundaries; applying semantic boundary detection and evaluating chunk coherence with precision and recall metrics improves end-to-end retrieval accuracy.

How do I evaluate chunking performance for question answering over technical manuals?▼

Evaluate chunking performance by measuring retrieval quality metrics like precision, recall, MRR, and MAP, alongside processing time and memory trade-offs, to validate end-to-end RAG accuracy for factoid and analytical queries.

Related Skills

Explore More Software Engineering Skills→
lanejlanej
40

lancer

Ingest and search text and images in LanceDB using the lancer CLI.

Community
Intermediate
christian-bromannchristian-bromann
3

langchain-embeddings-vectorstores

Convert text into embeddings and index them into vector stores for semantic search.

Community
Advanced
jillescajillesca
3

langchain-rag

Build end-to-end RAG pipelines with LangChain document loaders, embeddings, and vector stores.

Community
Advanced
findinfinitelabsfindinfinitelabs

intelligent-text-chunking

Split long texts into coherent chunks while preserving semantic structure.

Community
Intermediate