How is context window management handled?

The skill provides utilities to truncate and manage context documents to fit within specific token budgets while prioritizing the most relevant information.

What is RAG retrieval?

Retrieval-Augmented Generation (RAG) is a technique that provides LLMs with specific, retrieved documents to ensure answers are grounded in factual data rather than internal training alone.

Does it support hybrid search?

Yes, it includes patterns for combining semantic vector search with traditional keyword search using Reciprocal Rank Fusion (RRF) to balance precision and recall.

How does this skill prevent hallucinations?

It implements context sufficiency checks based on Google Research (2025) to determine if retrieved information is sufficient to answer a query before the LLM attempts to generate a response.

RAG Retrieval Patterns

Name: RAG Retrieval Patterns
Author: yonatangross

byyonatangross

•

데이터 과학 및 ML

Implements production-ready Retrieval-Augmented Generation (RAG) patterns to ground LLM responses in verifiable context and prevent hallucinations.

소개

This skill provides standardized implementations for building reliable RAG pipelines within Claude Code. It covers essential patterns including vector search integration, inline citation management, hybrid search using reciprocal rank fusion, and advanced context sufficiency checks based on the latest research to prevent hallucinations. Whether you are building a simple document QA system or a complex multi-step data pipeline, this skill ensures your AI applications are grounded, accurate, and token-efficient by utilizing proven industry best practices.

주요 기능

Standard and citation-based RAG implementations
Hybrid search patterns combining semantic and keyword retrieval
Automated context window management and token optimization
Pre-generation sufficiency checks to reduce hallucinations
Reciprocal Rank Fusion (RRF) for document ranking
29 GitHub stars

사용 사례

Reducing LLM hallucinations in production knowledge-base applications
Building a document-aware chatbot with verifiable citations
Implementing a hybrid search system to improve retrieval precision

소개

주요 기능

Standard and citation-based RAG implementations
Hybrid search patterns combining semantic and keyword retrieval
Automated context window management and token optimization
Pre-generation sufficiency checks to reduce hallucinations
Reciprocal Rank Fusion (RRF) for document ranking
29 GitHub stars

사용 사례

Reducing LLM hallucinations in production knowledge-base applications
Building a document-aware chatbot with verifiable citations
Implementing a hybrid search system to improve retrieval precision