We've just open-sourced *CRoM*, a context-efficiency layer for Retrieval-Augmented Generation (RAG) pipelines.

Why this exists: RAG pipelines often suffer from "context rot" — long prompts or noisy retrieval degrading answer quality. CRoM focuses on efficient, relevant injection.

What it does: - Token-aware prompt packing under strict token limits - Cross-encoder fallback for low-confidence answers - FastAPI-based orchestration - Modular & ready for integration with existing RAG stacks

GitHub: https://github.com/Flamehaven/CRoM-Context-Rot-Mitigation--E...

Still early — feedback & suggestions welcome.