Tempera is a system that provides persistent memory for Claude Code. Through capturing coding sessions, semantic search, and reinforcement learning, it enables the AI to learn from historical experiences and continuously improve.

An intelligent e-commerce dialogue agent system based on reinforcement learning, integrating ontology reasoning, business toolchains, dialogue memory, and a Gradio interface. It realizes closed-loop learning from data to training and then to deployment through the Stable Baselines3 PPO algorithm, and can autonomously optimize the decision-making strategy of the shopping assistant.