What is cai?

CAI, short for Context Aware Infrastructure, is a personal memory and automation system for developers who use AI coding assistants. It is built in Python and is meant to give an AI assistant a persistent memory that survives across separate coding sessions, instead of starting fresh every time you open a project. The system stores information in several layers. A vector database called LanceDB holds text turned into searchable embeddings, so related past notes can be found by meaning rather than exact keywords. A second storage layer, called Mycelium, keeps a plain markdown file archive that grows and branches over time like a small file based graph. These layers sit underneath a set of small background services, described as MCP servers, that handle jobs like reading and writing memory, managing the filesystem archive, watching for changes, and relaying messages through a service called Threema. There is also a scheduled batch process that reviews the last one hundred stored memory entries once a day, looks for repeating patterns using a local language model, and can turn those patterns into new reusable tools that get loaded automatically. The project ships with a full VS Code setup, including recommended extensions and launch configurations, and includes its own test suite. Setup requires Python 3.10 or newer, the Ollama tool for running local models, and pulling two specific models for embeddings and analysis before installing Python dependencies from a requirements file. Some automation pieces, such as the autostart component, are written for Windows specifically. The project is licensed under the MIT license, which allows free use, including commercial use, as long as the copyright notice is kept. It has a small number of stars so far and appears to be an individual developer's personal infrastructure project rather than a widely adopted tool.