Personal project
RATS
A work dashboard for planning tasks, running coding agents, and handling decisions across persistent sessions.
Personal adaptation and continued development
- Python
- FastAPI
- Jinja
- HTMX
- SQLite
- tmux
Problem
Parallel coding sessions create work to track: task context, execution plans, questions, and results.
My contribution
- Adapted the dashboard for a personal Linux server and integrated GitHub task tracking.
- Built workflows for persistent coding sessions, execution plans, and decisions that need human input.
- Connected Claude and Codex sessions to a shared view of task progress.
How it works
Workflow diagram
Define a task
Keep the request and its acceptance criteria with the work item.
Plan the work
Review the execution plan before starting implementation.
Run the session
Follow persistent agent sessions and respond when a decision is needed.
Review the result
Inspect the change and verification before accepting the work.
Engineering decisions
Persistent sessions
Sessions continue independently of the browser view, so navigation does not discard ongoing work.
Visible decisions
Questions and approval requests appear alongside task state rather than staying hidden in terminal output.
Result and limits
A private dashboard brings planning, agent sessions, and review into one workflow.
- This case study covers the personal Linux adaptation and continued development. The source repository is private.