Spruce cuts private RAG retrieval from ~90 GB per query to something you'd actually run.

arXiv 2609.03376 targets organizations outsourcing vector indexes to untrusted clouds, where each query touches corpus-scale state, so a naive secure implementation costs minutes and about 90 GB of communication per query at million-document scale, and recent optimized systems still need 10-22 seconds. Spruce learns compact binary codes preserving candidates for full-precision reranking, turning corpus-wide embedding scoring into Hamming-distance computation under two-server MPC, with a corpus-calibrated fixed-radius protocol avoiding multi-round candidate selection. Across four corpora of 383K to 5.42M documents it preserves original search quality (arXiv).