Rag Pipeline skills for AI agents

Advanced RAG patterns beyond basic retrieve-and-generate. Covers multi-hop RAG, agentic RAG with tool use, graph RAG (knowledge graphs + vector retrieval), recursive retrieval, self-querying retrievers, query decomposition, citation extraction, and corrective RAG. Includes implementation patterns and guidance on when each advanced technique is warranted.

Comprehensive guide to document chunking strategies for RAG pipelines. Covers fixed-size, semantic, recursive character, sentence-based, parent-child, markdown-aware, and code-aware chunking. Includes chunk size optimization, overlap strategies, and practical benchmarks for choosing the right approach based on document type and retrieval quality.

Guide to selecting, using, and optimizing text embedding models for RAG pipelines. Covers commercial models (OpenAI text-embedding-3, Cohere embed-v3, Voyage AI) and open-source options (BGE, E5, Nomic Embed). Includes dimensionality selection, batch processing, embedding caching, fine-tuning for domain-specific retrieval, and cost analysis.

Evaluating RAG systems end-to-end. Covers retrieval metrics (context precision, context recall, MRR), generation metrics (faithfulness, answer relevance, hallucination detection), the RAGAS framework, human evaluation protocols, A/B testing retrieval strategies, building evaluation datasets, and continuous monitoring in production.

Teaches the foundational architecture of Retrieval-Augmented Generation (RAG) systems. Covers why RAG outperforms fine-tuning for most knowledge-grounding use cases, the three core stages (indexing, retrieval, generation), component design, latency budgets, and evaluation metrics including faithfulness, relevance, and hallucination rate. Use when building or explaining any RAG system from scratch.

Production-grade RAG deployment patterns. Covers caching strategies (semantic and exact), streaming responses, token budget management, fallback strategies for retrieval failures, monitoring retrieval quality, cost optimization, incremental indexing, multi-tenancy, and operational best practices for running RAG systems at scale.

Building RAG pipelines with LangChain and LangGraph. Covers document loaders, text splitters, vector stores, retrievers, chains, and agents. Includes practical patterns for conversational RAG, multi-source retrieval, streaming, and LangGraph-based agentic RAG workflows.

Building RAG systems with LlamaIndex (formerly GPT Index). Covers data connectors, node parsers, index types (vector, keyword, knowledge graph, summary), query engines, response synthesizers, and advanced patterns like sub-question queries and recursive retrieval. Practical code for production LlamaIndex RAG pipelines.

Covers retrieval strategies for RAG pipelines: dense retrieval, sparse retrieval (BM25), hybrid search, re-ranking with cross-encoders and Cohere Rerank, Maximal Marginal Relevance (MMR), contextual retrieval, and Hypothetical Document Embeddings (HyDE). Includes practical implementation patterns and guidance on when to use each strategy.

Practical guide to vector databases for RAG systems. Covers Pinecone, Qdrant, Weaviate, ChromaDB, pgvector, and Milvus with setup, indexing, querying, metadata filtering, hybrid search, and scaling considerations. Includes selection criteria, performance benchmarks, and production deployment patterns.