What problem does it solve? Choosing the right embedding model and chunking strategy for vector search is difficult, and poor choices lead to weak retrieval quality in RAG pipelines. This Skill provides model comparisons, ready-to-use embedding templates, and evaluation metrics to make informed decisions. ## Core Features & Use Cases - Model Selection Guidance: Compare Voyage AI, OpenAI, and open-source embedding models by dimensions, token limits, and domain fit. - Embedding Templates: Production-ready Python code for Voyage AI, OpenAI (with Matryoshka dimension reduction), and local sentence-transformers models. - Chunking Strategies: Token-based, sentence-based, semantic-section, and recursive character splitting implementations. - Quality Evaluation: Compute precision@k, recall@k, MRR, and NDCG to compare embedding models on retrieval tasks. - Use Case: When building a RAG system for legal documents, use this Skill to select voyage-law-2, implement token-based chunking with overlap, and validate retrieval quality with evaluation metrics. ## Quick Start Ask the AI to help you choose an embedding model and set up a chunking and embedding pipeline for your RAG application.
