skills.rest
Muvon
MuvonOFFICIAL
Explore all skills by this author →
Website𝕏@muvonteamGitHub

ai-rag-patterns

Design and troubleshoot production RAG pipelines with chunking, hybrid retrieval, reranking, and evaluation.

3|Updated Mar 14, 2026
Search Tags:#rag#vector database#evaluation#retrieval#hybrid search#chunking#reranking
One-click install
npx skills add https://github.com/Muvon/octomind-tap --skill ai-rag-patterns
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ai-rag-patterns
Source: https://github.com/Muvon/octomind-tap/tree/main/skills/ai-rag-patterns
Command: npx skills add https://github.com/Muvon/octomind-tap --skill ai-rag-patterns

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you design, tune, and troubleshoot production Retrieval-Augmented Generation (RAG) pipelines so retrieval failures get fixed at the correct layer with measurable results.

Core Features & Use Cases

  • Chunking strategy selection: Choose fixed/recursive, semantic, late chunking, Anthropic Contextual Retrieval, and parent-document approaches to improve what can be retrieved.
  • Retrieval and fusion tuning: Combine BM25 and dense retrieval with reciprocal rank fusion and related hybrid methods to improve candidate recall.
  • Reranking and evaluation: Apply cross-encoders or rerankers on top candidates and evaluate with RAGAS, TruLens RAG Triad, or DeepEval to quantify impact and diagnose the seven common RAG failure modes.
  • Agentic and multi-modal RAG patterns: Use agentic retrieval loops and multi-modal retrieval approaches like ColPali for document-image-heavy corpora.

Quick Start

Use the ai-rag-patterns skill to diagnose why my RAG agent is missing relevant information and propose a production-grade chunking, hybrid retrieval, reranking, and evaluation plan with cited metrics.

Frequently Asked Questions about ai-rag-patterns

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

FAQPage Schema
How do I fix RAG retrieval failures where my agent misses relevant information?▼

Diagnose RAG retrieval failures by applying failure-mode diagnostics to identify root causes, then implement targeted chunking, hybrid retrieval, and reranking strategies to reduce failure rates measurably.

What's the best way to combine BM25 and dense retrieval for hybrid search?▼

Combine BM25 and dense retrieval using reciprocal rank fusion to improve candidate recall, then apply cross-encoders or rerankers on top candidates to maximize retrieval precision.

How do I evaluate RAG pipelines with RAGAS, TruLens, or DeepEval?▼

Evaluate RAG pipelines using RAGAS, TruLens RAG Triad, or DeepEval to quantify context grounding, measure retrieval precision, and diagnose failure modes with cited metrics.

Does contextual retrieval and late chunking work for long document corpora?▼

Contextual retrieval and late chunking work for long document corpora by preserving semantic context during chunking, improving what can be retrieved from complex documents without losing critical information.

When should I use agentic RAG patterns or multi-modal retrieval like ColPali?▼

Use agentic RAG patterns for dynamic retrieval loops or multi-modal retrieval like ColPali when processing document-image-heavy corpora that require visual context alongside text.

Why does my RAG return irrelevant chunks despite using a vector database?▼

Irrelevant chunks occur when chunking strategies or retrieval fusion are misconfigured; apply parent-document approaches, tune hybrid search fusion, and add cross-encoder reranking to filter candidates.

Related Skills

Explore More Data & Analytics Skills→
bethanychamberlainbethanychamberlain
2

rag

Configure LibreChat's RAG pipeline for document ingestion and retrieval.

Community
Advanced
ancolemanancoleman
503

evaluating-llms

Evaluate LLM systems with automated metrics, LLM-as-judge, and benchmarks.

Community
Advanced
JeffallanJeffallan
10.9k

rag-architect

Design and implement retrieval-augmented generation pipelines with vector stores and hybrid search.

Community
Advanced
bugrabilgebugrabilge
10

rag-engineer

Design and optimize RAG systems for LLM applications.

Community
Advanced