Giving the AI advisor a defined investigation boundary

Observed work

Generated alarm-product images had altered physical details, so the original images were retained. That decision did not establish visual acceptance or publication of the earlier documentation layout work.

The advisor design considered operators spanning projects, technical staff and growers without specialist support. A lightweight bounded agent loop would inspect project data across several steps before reaching findings. This was a confirmed design direction, with no demonstrated user outcome yet.

Point permissions became the execution boundary: unbound points are neither read nor controlled; bound read-only points may support proposals requiring acceptance; explicitly controllable points may be operated directly. The backend must recheck authorization and retain constraints/auditing. Separate suggestion/execution modes were removed. Follow-up would review and remind rather than automatically reverse actions; device execution and effects were not verified.

Backend AI changes reportedly merged into local dev and passed tests. Related frontend/shared-module commits, tests and builds were reported complete with clean worktrees, but repositories remained unpushed. Changed response fields required paired frontend/backend releases. External services were unconfigured and browser/end-to-end checks unfinished. An AI-documentation inventory began without a final result.

Chinese/English site work addressed product structured data, language links, feedback-page indexing and a root redirect retaining paths/queries. Local builds/checks passed. The English homepage kept its main heading while search title/description received a smaller revision. A main push was reported and online checks were running; final acceptance and a separate locally committed documentation/rules branch remained distinct.

Color-balance and contrast checks were revised after correcting image measurements. Orange text and colored panels were examined, and laan requested the identified fixes. Some changes were reported complete, but site-wide rechecks, standards/tests, a failing preview attempt and final commit/push/deployment were unresolved.