We present DS-Serve, a framework that transforms large-scale text datasets, comprising half a trillion tokens, into a high-performance neural retrieval system. DS-Serve offers both a web interface and API endpoints, achieving low latency with modest memory overhead on a single node. The framework also supports inference-time trade-offs between latency, accuracy, and result diversity. We anticipate that DS-Serve will be broadly useful for a range of applications, including large-scale retrieval-augmented generation (RAG), training data attribution, training search agents, and beyond.
@article{arxiv.2602.22224,
title = {DS SERVE: A Framework for Efficient and Scalable Neural Retrieval},
author = {Jinjian Liu and Yichuan Wang and Xinxi Lyu and Rulin Shao and Joseph E. Gonzalez and Matei Zaharia and Sewon Min},
journal= {arXiv preprint arXiv:2602.22224},
year = {2026}
}