HAM — Hierarchical Agent Memory

HAM, which stands for Hierarchical Agent Memory, is a system designed to help AI coding assistants work more efficiently. It helps these AI agents use much less of their digital memory, called tokens. This can cut down the number of tokens used by as much as half.

Benefits

Using HAM can save money because fewer tokens mean lower costs for using AI services. It also makes AI agents respond faster, so they can spend more time writing code. Additionally, by using fewer tokens, HAM helps make AI more environmentally friendly, as AI uses a lot of electricity.

Use Cases

HAM is useful for anyone working with AI coding agents. Instead of the AI having to remember everything about a whole project for every single task, HAM helps it focus on the specific part of the project the user is working on. For example, general rules for the whole project might be stored in one file, while specific instructions for a certain section are kept in a file within that section. This means the AI only needs to load a small amount of information, making it quicker and cheaper to use.

Vibes

Users can install HAM by cloning its repository and then running a simple command likego hamin their project directory. The system automatically sets up the necessary memory files. Users can also update HAM easily. It works by storing information in the most relevant directory, loading only the necessary context, and keeping the memory updated as the AI works. There are commands to check savings, carbon impact, and performance. HAM also has a web dashboard that shows how much memory and money have been saved, along with other useful data about the AI's performance. HAM is available under the MIT license.