Project Dashboard

Managing eight active projects with AI assistance creates a specific organizational problem: each session starts cold, and re-explaining which projects are active, what phase each is in, and what the obvious next step is wastes time and degrades the quality of the context the AI gets. The project dashboard is a local-only Astro SSR app (Tailwind + TypeScript, served via @astrojs/node) that reads the frontmatter from each ~/os/projects/*/README.md (fields like status, next_step, priority, and last_active) and renders them as a live card grid. Opening the dashboard gives an accurate, single-view snapshot of where everything stands without touching any of the individual project files.

Stages 0–4 are complete: project cards with full frontmatter display, inline edit controls that write changes back to the source README files, and a weekly digest view that surfaces what moved and what stalled. The architecture is deliberately local-only. The @astrojs/node adapter runs the SSR layer, reads from the filesystem directly, and is never deployed. There’s no database, no cloud sync, and no multi-user consideration; the point is fast access to current state on the machine where the work actually happens. Stages 5–7 (momentum view, what-to-work-on-next recommendation, Claude token tracking, smart notepad, and a final visual design pass) are still planned.

The dashboard is primarily useful as infrastructure for the personal OS: it’s what keeps the project tracking system honest by making stale next_step fields immediately visible. Building it required wiring Astro’s SSR file-system APIs to a structured data format I designed separately, which made the coupling between the dashboard and the ~/os/projects/ schema explicit and enforced. The repo is at https://github.com/nseluga/project-dashboard; it runs locally only and is not deployed.