skills.rest
svssdeva
svssdevaCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

rag-architect

Design retrieval-augmented generation systems with hybrid retrieval and reranking.

1|1|Updated Apr 20, 2026
Search Tags:#embeddings#rag#vector database#hybrid search#chunking#reranking#retrieval evaluation
One-click install
npx skills add https://github.com/svssdeva/agentic-skills --skill rag-architect-svssdeva
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: rag-architect
Source: https://github.com/svssdeva/agentic-skills/tree/main/ai/rag-architect
Command: npx skills add https://github.com/svssdeva/agentic-skills --skill rag-architect-svssdeva

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you design production-grade retrieval-augmented generation (RAG) systems that return high-quality, relevant context instead of relying on ad-hoc chunking or unmeasured retrieval.

Core Features & Use Cases

  • Chunking strategy and document preparation: Defines chunking approaches and validation checkpoints to preserve semantic boundaries and metadata (source, timestamps, section context).
  • Embeddings, vector store design, and hybrid retrieval: Guides embedding model selection, vector database/schema choices, hybrid search (dense + keyword), and reranking of top-k results.
  • Evaluation and iteration loop: Provides retrieval and grounding metrics (precision/recall/MRR/NDCG plus faithfulness/relevance), with pass/fail thresholds to prevent slow quality regressions in production.

Quick Start

Use the rag-architect skill to plan a complete RAG system for your document corpus, including ingestion (chunking + embeddings), retrieval (hybrid + reranking), and an evaluation plan with quality targets.

Frequently Asked Questions about rag-architect

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

FAQPage Schema
How do I design a RAG retrieval pipeline for production?▼

Design a production RAG retrieval pipeline by defining document chunking, embedding model selection, vector store configuration, and hybrid retrieval with reranking. Include explicit metadata enrichment and idempotent ingestion with deduplication to maintain data quality.

What is hybrid search and reranking in retrieval-augmented generation?▼

Hybrid search in retrieval-augmented generation combines dense vector similarity with keyword matching to improve recall, while reranking top-k results reorders them by relevance. This ensures the retrieved context is both semantically accurate and highly targeted.

How do I evaluate retrieval quality using precision, recall, and MRR?▼

Evaluate retrieval quality using precision, recall, MRR, and NDCG to measure search accuracy, plus groundedness and relevance metrics for generated output. Establish pass/fail thresholds to prevent slow quality regressions in production systems.

What's the best way to chunk documents for semantic search without losing context?▼

The best way to chunk documents for semantic search is to apply chunking strategies with validation checkpoints that preserve semantic boundaries. Enrich each chunk with metadata like source, timestamps, and section context to maintain structural awareness.

Can I use vector database schema design for multi-tenant or domain-aware systems?▼

Yes, you can use vector database schema design for multi-tenant or domain-aware systems requiring measurable retrieval quality. Configure the schema to support hybrid search and metadata filtering for precise, tenant-specific context retrieval.

Why does my RAG system return irrelevant context and how can I fix it?▼

A RAG system returns irrelevant context due to ad-hoc chunking, unmeasured retrieval, or missing reranking. Fix it by implementing hybrid vector and keyword retrieval, reranking top-k results, and applying retrieval evaluation metrics like precision and recall.

Related Skills

Explore More Data & Analytics Skills→
vaaynevaayne
52

kreuzberg

Extract text, tables, metadata, and images from 91+ document formats.

Community
Advanced
JeffallanJeffallan
10.9k

prompt-engineer

Design and optimize LLM prompts with structured outputs and evaluation frameworks.

Community
Advanced
EstomEstom

rag-architect

Design retrieval-augmented generation systems with vector databases, chunking, and reranking pipelines.

Community
Advanced
simplysmartaisimplysmartai

hybrid-search-implementation

Combines vector similarity and keyword retrieval with Reciprocal Rank Fusion for hybrid search systems.

Community
Advanced