I built NexRAG: a framework-agnostic RAG pipeline SDK.

Short summary

NexRAG is an open-source SDK that eliminates repetitive RAG pipeline scaffolding, reducing 3,000+ lines of custom code to ~300 lines of configuration. Features include pluggable LLM providers, embedding models, chunking strategies, vector stores, and rerankers, plus production-grade observability and LLM-as-a-Judge evaluation. Every component is extensible for custom implementations without core framework modifications.

• •RAG pipeline SDK that reduces 3,000+ lines of code to ~300 lines of configuration
• •Pluggable components: swappable LLMs, embeddings, vector databases, chunking strategies, and rerankers
• •Built for production with observability, evaluation metrics, and extensibility

Generated with AI, which can make mistakes.