skills.rest
JaviMontano
JaviMontanoCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

genai-architecture

Designs RAG pipelines, LLM orchestration, agent workflows, and vector database architectures for generative AI systems.

Updated Jun 28, 2026
Search Tags:#embeddings#rag#vector database#agent workflow#hallucination reduction#llm orchestration#genai architecture
One-click install
npx skills add https://github.com/JaviMontano/claude-plugins --skill genai-architecture-javimontano
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: genai-architecture
Source: https://github.com/JaviMontano/claude-plugins/tree/main/plugins/claude-native-toolkit/skills/genai-architecture
Command: npx skills add https://github.com/JaviMontano/claude-plugins --skill genai-architecture-javimontano

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and assets (resource) components.

What problem does it solve? Designing production generative AI systems requires coordinated decisions across retrieval pipelines, model routing, agent tooling, vector storage, and quality controls; this Skill turns those decisions into a structured, evidence-tagged architecture deliverable. ## Core Features & Use Cases - RAG Architecture Design: Covers query processing, hybrid retrieval, chunking strategies, re-ranking, context assembly, and response validation, including GraphRAG and agentic RAG variants. - LLM Orchestration & Multi-Model Tiering: Defines tiered model routing with complexity classification, confidence-based escalation, and cost- and latency-aware routing. - Vector Database Selection: Compares Pinecone, Qdrant, Weaviate, Milvus, Chroma, and pgvector against scale, latency, filtering, and cost criteria, plus embedding model selection. - GenAI Quality Assurance: Specifies hallucination reduction, RAGAS-based evaluation metrics, guardrails architecture, and a continuous improvement loop. - Use Case: Architect an enterprise RAG assistant over policy, CRM, and ticketing knowledge with model routing, connector permission boundaries, grounding, and a quality monitoring loop. ## Quick Start Ask the assistant to design a RAG architecture for your knowledge base, including vector database selection, model routing, and quality guardrails.

Frequently Asked Questions about genai-architecture

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

FAQPage Schema
How do I design a RAG architecture for an enterprise knowledge base?▼

Design the full pipeline: query processing, hybrid retrieval (vector plus keyword with reciprocal rank fusion), re-ranking, context assembly, generation, and validation. Select chunking and embedding strategies based on content type, and add hallucination detection with citation enforcement before delivery.

Which vector database should I choose for RAG: Pinecone, Qdrant, or pgvector?▼

Choose based on scale, latency, metadata filtering, update frequency, and cost. Pinecone suits managed billion-scale deployments, Qdrant and Milvus fit self-hosted or data-residency requirements, and pgvector works well for moderate scale within existing PostgreSQL infrastructure.

When should I use CAG instead of RAG?▼

Use context-aware generation (CAG) when the knowledge base is small enough to fit in the context window, stable, and latency-sensitive. Use RAG for large or frequently updated knowledge bases, or when source attribution and citation are required; hybrid approaches combine both.

How do I reduce hallucinations in LLM-generated responses?▼

Ground every claim against retrieved context, enforce citations to sources, and add confidence scoring per response. Combine input guardrails (prompt injection detection, PII masking) with output guardrails (hallucination detection, format validation) and track faithfulness metrics with frameworks like RAGAS.

When should I not use this generative AI architecture approach?▼

Do not use it for traditional non-generative ML systems such as feature engineering, model training, or drift detection for classifiers; those belong to AI pipeline or AI software architecture. It also assumes the use case genuinely benefits from generative AI rather than simpler methods.

Related Skills

Explore More Software Engineering Skills→
DisposableByDefaultDisposableByDefault
1

Review Ticket

Validates implementation tickets for scope, dependency, and quality issues before development begins.

Official
Advanced
Kingdaddy007Kingdaddy007
2

DEVOPS & INFRASTRUCTURE

Govern infrastructure provisioning, CI/CD pipelines, monitoring, and disaster recovery.

Community
Advanced
do360nowdo360now

prompt-injection

Analyze language model systems for prompt injection vulnerabilities.

Community
Advanced
sunbluesomesunbluesome

ml-system-design

Design production ML systems to prevent training-serving skew and data leakage.

Community
Advanced