openJiuwen AgentMemory Launches multimodal memory capability OmniMem-Lens, with long-video question-answering latency reduced by up to 97%

Techub News news,openJiuwen AgentMemory officially launches multimodal memory capability OmniMem-Lens, aimed at solving the problems of fragmented memory and low retrieval efficiency when agents process long videos. The technology uses dual-stream linkage of vision and hearing to segment videos according to real event boundaries and builds a three-layer memory structure. In some evaluations, its memory recall accuracy improves by up to 45.7%, and online question-answering latency is reduced by up to 97%. OmniMem-Lens is already available in openJiuwen AgentMemory 0.2.0 version,WorkSwarm and will soon support integration to use this feature. (QbitAI)