An open-source, local first RAG pipeline optimized for AMD Instinct GPUs via ROCm and vLLM, delivering zero-data-leakage semantic search over enterprise documents.
soms3r is an open-source, local-first RAG engine built on AMD ROCm and vLLM. It delivers high-throughput, private semantic search and citation-backed question answering over enterprise documents on AMD Instinct cloud GPUs.