rag-architect

What problem does it solve?

This Skill provides a comprehensive toolkit and knowledge base for designing, implementing, and optimizing production-grade Retrieval Augmented Generation (RAG) pipelines, enabling the creation of scalable, efficient, and accurate retrieval systems.

Core Features & Use Cases

• Document Processing & Chunking: Strategies for splitting documents into manageable pieces (fixed-size, sentence-based, semantic, etc.).
• Embedding Model Selection: Guidance on choosing embedding models based on dimensionality, speed, quality, and domain.
• Vector Database Options: Overview of popular vector databases like Pinecone, Weaviate, Qdrant, Chroma, and pgvector.
• Retrieval Strategies: Explores dense, sparse, and hybrid retrieval methods, along with reranking techniques.
• Query Transformation: Techniques like HyDE, Multi-Query Generation, and Step-Back Prompting to improve retrieval.
• Context Window Optimization: Methods for dynamic context assembly and compression.
• Evaluation Frameworks: Metrics and tools for assessing RAG system performance (faithfulness, relevance, RAGAS).
• Production Best Practices: Strategies for caching, streaming, graceful degradation, and cost optimization.
• Guardrails & Safety: Content filtering, query security, and response safety measures.
• Use Case: A developer needs to build a RAG system for customer support documentation. This Skill guides them through selecting the right chunking strategy, embedding model, vector database, and evaluation metrics to ensure accurate and relevant responses.

Quick Start

Use the rag-architect skill to explore strategies for optimizing document chunking in RAG pipelines.