CRITICAL: At the start of every session, you MUST read these files in order:
.ralph/STATE.md— Current phase, what's done, what's next, key context.ralph/agent/learnings.md— Mistakes to avoid, patterns to follow.ralph/agent/decisions.md— Decisions made so far with rationale.ralph/tasks/addr-process.md— Task tracking and progress
Before ending a session or when significant progress is made, UPDATE .ralph/STATE.md.
When making decisions, record them in .ralph/agent/decisions.md following the existing format.
When learning from mistakes, record them in .ralph/agent/learnings.md following the existing format.
Do not duplicate decision entries in .ralph/STATE.md; keep STATE.md as a handoff/status document and reference .ralph/agent/decisions.md as the single decision source.
Treat .ralph/STATE.md as single-agent/single-active-plan context. List non-active tracks as baseline/reference unless an explicit manager-agent coordination layer exists.
This project uses the ADDR (Align-Define-Design-Refine) API design methodology by James Higginbotham.
- Prompts guide:
design/addr-ai-prompts.md— follow these prompts for each phase - Design artifacts:
design/{domain}/addr/{phase}/— Align, Define, Design, Refine deliverables per domain - Domain: Mechatronic product ecommerce (drones, motors, ESCs, flight controllers, sensors, frames, batteries, FPV gear)
- Remote:
origin→ GitHub (cjjohansen/drone-web) - Push requires token auth via
GITHUB_PERSONAL_ACCESS_TOKENenv var - Commit and push after each completed ADDR phase