Overview
- The platform was launched at MWC Barcelona 2026, with Huawei executives Xie Liming and Yuan Yuan presenting the offering.
- It unifies a knowledge base, KV cache and memory bank under a Unified Cache Manager to coordinate retrieval, inference and task memory.
- Huawei cites over 95% retrieval accuracy from multimodal parsing and token-level encoding, plus about a 90% cut in time to first token through KV cache tiering.
- Memory banks capture working and experiential memory with backtracking and multi-agent collaborative learning to improve inference over time.
- Enterprises can deploy an OceanStor A800–based appliance for greenfield rollouts or use AI data engine nodes with OceanStor Dorado to extend existing systems, with results reported by Huawei and not independently verified in these articles.