Memora: Self-Managing Memory Agent for LLMs | Qwen Cloud Hackathon

Memora is a self-managing memory agent built for the Global AI Hackathon Series with Qwen Cloud, Track 1: MemoryAgent. The problem: LLMs are stateless. They forget everything after a session, while many basic memory systems store outdated, irrelevant, or contradictory memories. Memora solves this with: - Persistent user memory - Conflict resolution between old and new memories - Decay-based forgetting for stale information - Smart retrieval using Qwen Cloud models In this demo, I show the Streamlit UI, stored memories, memory update behavior, decay checking, benchmark results against a naive baseline, architecture, API endpoints, and Alibaba Cloud deployment. Tech stack: Qwen Cloud, FastAPI, Streamlit, ChromaDB, Docker, Alibaba Cloud ECS Built by Ahmad Hassan for the Qwen Cloud MemoryAgent track.