A post-retrieval temporal layer designed to improve RAG systems by addressing time-blindness in vector searches. This library implements validity filtering, document kind classification, and exponential decay scoring to ensure retrieved information is fresh and accurate. It functions downstream of existing vector search systems without requiring re-indexing or new infrastructure.
This pull request adds initial support for reranking to libllama, llama-embeddings, and llama-server using two models: BAAI/bge-reranker-v2-m3 and jinaai/jina-reranker-v1-tiny-en. The reranking is implemented as a classification head added to the model graph. Testing and benchmarking were performed with server integration.
