status: per volume×destination coverage + durability panel (CLI + TUI)

[mbertschler]

Summary

The two standing questions no surface could answer now have one answer,
built once and consumed twice:

• "Is every configured target caught up?" (F16) — a per-(volume ×
  target) grid replaces the single LAST SYNC cell that hid a week-behind
  target behind a fresh ✓ earned by any other target.
• "What is durable where, and what could I offload?" (F17/F23) — vector
  coverage, verify method, evidence age, and offloadable bytes, surfaced
  directly instead of only through offload --dry-run side effects.

This is read-only introspection — a question per ux-principle 2. It opens
no run, reads no disk bytes, validates no marker, and mutates nothing.

Changes

• New status package — the shared query layer. status.Build(ctx, s, cfg) produces, per configured volume and per (volume × target):
  • last successful sync + age, coloured against that pair's own
    sync_every (late is relative to the pair's cadence, not a global
    constant);
  • the runs: every failure and refusal becomes a run row or a standing state #157 standing states: alarm (latched verify mismatch, red),
    refused (a gate that regressed from a prior success, red), and
    needs-bootstrap (a first-use --init refusal on a pair that never
    succeeded — amber, an expected human step, not a fault);
  • durability: whether the target's vector covers this node's local
    content origin, the verify method behind it, and evidence age vs
    offload_max_evidence_age (stale flagged);
  • offload readiness: "N GB offloadable now" per volume.
    Each cell carries a Level (neutral/ok/amber/red); the report's worst
    level is the single "am I safe?" answer. Neutral text labels live in the
    package so both surfaces render identical words.
• offload.Readiness — reuses the real gate (loadGate +
  selectCandidates + gate.check), so the offloadable totals match what
  an offload would move; reads only the index.
• Step 1 — squirrel status [volume] (cmd/squirrel/status.go):
  renders the grid and exits with the worst level (0 green / 1 amber / 2
  red) so the same command scripts a health check. A small exitCodeError
  carries the code out of cobra without printing.
• Step 2 — TUI dashboard: the top panel is now this grid, built from the
  same query layer, with the STATE and DURABLE cells coloured per level.
  The old formatLast/latestByVol path (single LAST SYNC cell) is gone.

Schema

None. This is a query/display layer over existing tables — no migration,
no SchemaVersion bump.

Testing

• status package: end-to-end Build tests (fully-durable green,
  needs-bootstrap amber, latched alarm red, never-indexed amber, relayed
  required target amber) plus table-driven unit tests for the pure logic
  (cadence colouring, standing classification, durability level, evidence
  staleness, exit codes, byte formatting, target ordering, state labels).
• CLI: never-synced → amber + exit 1 with cobra's error line silenced,
  post-sync → green + exit 0 (skips without rclone), unknown-volume and
  missing-config errors.
• TUI: dashboard run-partitioning test updated to the new signature.
• go vet ./..., go test ./... green; golangci-lint run clean on the
  changed packages (linter rebuilt with the module's go1.26.1 toolchain —
  the packaged binary targeted an older Go).

Closes #159

Stacked on #157 (PR #174). Base is #157's branch so the diff shows only this
delta; retargets to main once #157 merges.

🤖 Generated with Claude Code

https://claude.ai/code/session_014Skm2Qm6wWUB7dD4CpPTz7

---

Generated by Claude Code

[Copilot]

Pull request overview

Adds a shared, read-only status query layer that answers “am I safe?” per (volume × target) and reuses it across a new squirrel status CLI command and the TUI dashboard, including durability/evidence details and offload-readiness totals.

Changes:

• Introduces new status package with per-volume/target coverage, standing states, durability/evidence age, and worst-level aggregation.
• Adds squirrel status [volume] CLI with scriptable exit codes (0/1/2) driven by the report’s worst level.
• Reworks the TUI dashboard to render the per-(volume × target) grid via the shared status layer (replacing the old single “LAST SYNC” per volume).

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tui/tui.go	Wires config into dashboard model construction.
tui/dashboard.go	Replaces volumes table with shared coverage/durability grid; loads coverage via status.Build.
tui/dashboard_test.go	Updates dashboard data-load test for new coverage behavior when config is nil.
status/status.go	Defines report model: levels, standing states, and core data structures.
status/runs.go	Helpers for retrieving run ages/outcomes and classifying targets against config.
status/labels.go	Shared neutral text labels for CLI/TUI rendering.
status/durability.go	Computes per-target durability coverage, evidence age, and staleness.
status/build.go	Builds the full report from config + store, including offload readiness.
status/logic_test.go	Unit tests for pure logic (cadence coloring, labels, ordering, etc.).
status/build_test.go	End-to-end Build tests for key scenarios (durable, bootstrap-needed, alarm, etc.).
offload/readiness.go	Adds a read-only offload readiness tally matching the real offload gate logic.
cmd/squirrel/status.go	Adds the status command and grid rendering.
cmd/squirrel/status_test.go	CLI tests for exit codes, output, and config/volume error handling.
cmd/squirrel/root.go	Registers the new status subcommand.
cmd/squirrel/main.go	Adds exitCodeError handling to allow non-1 exit codes from cobra commands.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.