[{"type":"thinking","thinking":"The user asked me to read RH-WEBSITE.md. I've read it. Per memory: read ban lifted, but research commits exclude those paths; website edits are the website agent's job. I should just summarize what this file is. Keep it brief.","signature":""},{"type":"text","text":"Read. This is the website ops document — key points:\n\n- **Hard rule (§0):** the website touches only `web/` and `RH-WEBSITE.md`; all research files are read-only (SQLite `mode=ro`, text `'r'`). Commits stage only those two paths.\n- **Stack:** Django (`rhweb`/`dashboard`) in `.venv`, served at **127.0.0.1:8001** (8000 reserved), vanilla JS + vendored Plotly/MathJax, 3 s polling of `/api/status/`.\n- **13 pages** incl. Dashboard, Roadmap, Ledger, Experiments (charts), Journal, Resources atlas, Live Runs, Feedback, Conversations (agent.db browser).\n- **Ops:** session-start = healthcheck `--restart` + watchdog check + review new feedback + one improvement per session (backlog in §7); restart required after any code/template change.\n- **Standing constraints** (consistent with my memory): archive renders **verbatim** — no typography/formula transforms (the 2026-08-20 revert); website work is the website agent's job, not the research agent's.\n\nNote the doc is slightly stale vs. the research side — it says \"37 D-files\" and BUILD STATUS mentions 12 pages, while the archive is now at D-066 and the page list is 13. What would you like to do next?"}]