The pairing of startsession and endsession is what makes it stick — closing a session is a structured prompt for achievements and task changes, so context that usually evaporates gets written down as a side effect of stopping work. Later, the decision entities answer the question that costs teams the most time: why is it built this way.
A knowledge graph for software projects, kept between sessions. It models projects, components, features, issues, tasks, developers, technologies, decisions, milestones and environments, and the relationships between them, then serves that back when you start work again.
- A session opened with recent sessions, active projects, high-priority tasks and upcoming milestones already in front of you (startsession)
- Full context for any entity — a component's purpose, dependencies, related issues and tasks — loaded and recorded as part of the session (loadcontext)
- A structured close: summary, achievements, task status changes, new tasks and project status, written into the graph (endsession)
- Entities, relations and observations created or removed as the project changes (buildcontext, deletecontext)
- Queries over the graph in several forms — the whole graph, search, specific nodes, related entities, decisions, or a milestone's progress (advancedcontext)
- A dependency model that answers what breaks if you change a component, and a decision history recording who decided what and when
Node.js. Two file paths hold the graph and the session log; both default inside the working directory and can be pointed elsewhere as environment settings. A container image is also available.
One command — npx -y github:tejpalvirk/developer