Skip to content

ROADMAP.md

Latest commit

 

History

History
125 lines (109 loc) · 7.08 KB

ROADMAP.md

File metadata and controls

125 lines (109 loc) · 7.08 KB
Raw

TrenchNote — Roadmap notes

Not a promise list — a parking lot for ideas that have a decided shape but no build date. The boundary rules come first: anything here still has to clear the non-goals in CLAUDE.md and the free/paid line of ADR 0011 (the ledger is free; insight about the ledger can be paid, and lives outside this repo).

TrenchNote is post-v1.0 and in parking-lot mode. Committed near-term work, when there is any, is broken into task files under docs/tasks/; incident-driven product ideas live in docs/BACKLOG.md. This file holds the shaped-but-undated ideas and the boundary that governs all of them.

Future — hosted / premium tier (outside this repo, per ADR 0011)

  • Email digests of upcoming/overdue inspections (ADR 0014). The free core answers "what needs attention" with the dashboard Inspections panel and the CSV export; a scheduled "here's your week" email is office intelligence, needs operated infrastructure (schedules, SMTP, retries), and belongs to the hosted tier. The data it would read is already fully public API contract — any subscriber-built script can do the same today.
  • Equipment timecard generator (ADR 0012) — already underway as the bindery-trenchnote sidecar.

Hardening / debt (core, not features)

Small correctness/robustness items — not new capability. They clear the ethos trivially (they make the existing thing more reliable) and can be picked up as docs/tasks/ files when convenient.

  • Make the off-site move email non-blocking (ADR 0012, pb_hooks/main.pb.js). Today the notify-email is sent synchronously in the movement write path, so an unreachable SMTP server can delay a field write (recorded under "Known limitations" in docs/current-state.md). Mail failure already cannot undo the write; the goal is that mail slowness can't stall the crew either. On a dirt lot, field execution comes first.

Explicitly not planned

Decided nos, recorded so they are not re-proposed. See also the non-goals in CLAUDE.md.

  • No workflow (inspection assignments/approvals/escalations), no vendor integrations, no accounting, no multi-tenant instances (CLAUDE.md non-goals).
  • The monthly equipment report stays export-only, forever (docs/BACKLOG.md item 7). It may summarise the ledger as CSV for accounting's existing billing process; it must never compute charges, apply rates, or do fractional-cost job-splits. Rates stay in the premium sidecar (ADR 0015). This is the item closest to the billing wall — hold the wall.
  • No Docker / container self-host. ADR 0003 chose boring ops (a single binary + systemd, or a bare Pi) precisely so a $5 VPS or a trailer Pi is enough. Containers add a dependency and a build/runtime layer the target self-hoster should not need.
  • No rich in-core reporting, dashboards, burn-rate, or email digests. Office intelligence is the premium sidecar's job (ADR 0011). The core answers what/where/who and exports raw data; analysis about that data is paid and lives outside this repo.

Decided 2026-07-21

  • Promote the equipment-timecard handoff to the first real ecosystem contract (movements + readings → bindery-trenchnote). This is committed direction now, not a maybe. The core keeps BACKLOG item 7 minimal (it exposes clean append-only data); the sidecar composes the timecard + rates. First concrete step — a contract ADR + example fixtures in this repo, per the promotion path in docs/ecosystem-contracts.md — was broken out as docs/tasks/010-timecard-handoff-contract.md and is done: ADR 0022 is accepted on the producer side (2026-08-02), with fixtures in docs/contracts/timecard-handoff/.
    • Step 4 is complete on both halves (2026-08-02). The consumer record — specified here as docs/tasks/030-timecard-handoff-consumer-adr.md and executed in the sidecar, which is why it was filed BLOCKED here — was accepted as that repo's decision 0009. Of note for this side: core ships no generator, so the consumer assembles the manifest itself from documented contract reads (which ADR 0022 permits) and therefore assigns its own revision, by content hash. Replacement by (manifest_public_id, revision) is preserved as ADR 0022 defined it.
    • Next: steps 5–6 — a published versioned contract with compatibility tests. Unblocked, not yet broken out as a task. Two things belong to it: the fixtures in docs/contracts/timecard-handoff/ still carry full timestamps in arrived_at/departed_at and are due regeneration against moved_at, and ecosystem-contracts.md suggests publishing contract source separately from private implementation once more than one repo consumes it — a placement decision to make before writing it, not after.
    • The gate on freezing a version is cleared. ADR 0022 open issue 3 (movements had no client-set observation date, so an offline move synced late landed its segment boundary on the sync day) is resolved by ADR 0023 via docs/tasks/020-movement-observation-date.md. Open issues 1, 2, 4 and 5 remain as documented caveats.
  • Contributions are gated by a DCO sign-off, not a CLA — the maintainer is an individual, not a company. See ADR 0011 and CONTRIBUTING.md.
  • trenchnote is the only public Level Books repo. The premium sidecar (bindery-trenchnote) stays private; publishing is irreversible and there is no rush.

Open questions

  • Whether to state the copyright holder's name explicitly in LICENSE / README (today the LICENSE is stock AGPL text with no name). Maintainer's call; not blocking anything.

Ops follow-up — domain verification (added 2026-07-21)

  • Verify trenchnote.com under the levelbookdev org once the Porkbun registrar transfer completes. The domain was verified under the old personal account (mds08011) during the org migration, not the org, so GitHub's protected_domain_state is still null. This is takeover-hardening only — the site is already live with HTTPS enforced. Deferred because the registration is mid-transfer from Namecheap to Porkbun as of 2026-07-21.
    • First, keep the site from going dark: a registrar transfer does not carry DNS records. If the nameservers switch to Porkbun, recreate the GitHub Pages apex A records there — 185.199.108.153, 185.199.109.153, 185.199.110.153, 185.199.111.153 (optional www CNAME → levelbookdev.github.io). Leave the repo's docs/CNAME (= trenchnote.com) untouched.
    • Then verify: org → Settings → Pages → Add a domain → add the challenge TXT record (_github-pages-challenge-levelbookdev) at Porkbun → Verify domain.