Definition
Retrieval-Augmented Generation (RAG) combines large language models with external knowledge retrieval to ground responses in up-to-date or domain-specific documents. Quality hinges on the embedding-models and index design beneath the generator.
Key Points
-
2026-08-04: Built-in RAG primitive highlighted in convex Series B positioning (2026-08-04-convex-57m-series-b-ai-application-backend)
-
Production stacks increasingly use tiered retrieval (cheap high-recall embedder + larger precision model)
-
Agentic systems extend RAG into multi-hop agentic-retrieval and long-context agent memory
-
2026-07-16: nvidia nemotron-3-embed open collection tops rteb; marketed for RAG, code search, and agent memory (nvidia-nemotron-3-embed-rteb)
Related
- llm
- llm-applications
- document-processing
- embedding-models
- rteb
- agentic-retrieval
- nemotron-3-embed
- mistral