# Changelog

All notable changes to this project are documented here.
The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
and the project follows [Semantic Versioning](https://semver.org/).

## [Unreleased]

## [1.2.2] - 2026-08-22

### Added

- Added audited single-job and bulk recovery controls for jobs that are no
  longer running on their agents but remain stuck server-side. Agents now
  acknowledge unknown cancellation targets so orphaned jobs become terminal,
  while host ownership checks prevent cross-host updates. ([#56])

### Fixed

- Source-group JSON create and update requests now accept encrypted pre- and
  post-hooks, preserve hooks when their fields are omitted, allow explicit
  clearing, and expose only boolean presence indicators. ([#54])
- Generic backup-job requests without paths are rejected before a broken job
  is created. Configured run-now backups use the source-group endpoint. ([#55])

## [1.2.1] - 2026-08-22

### Fixed

- Made the Agent updates host-selection screen stable and responsive: removed
  the redundant typed-count confirmation, corrected name/version/eligibility
  filtering, and stopped replacing idle or terminal controls through polling.
  Progress polling now runs only while a rollout is actively running. ([#50])

## [1.2.0] - 2026-08-22

### Added

- Fleet-wide agent updates are now discoverable from Settings and the
  dashboard, with arbitrary host subsets, name/tag/version/eligibility
  filtering, explicit exclusion reasons, server-side membership validation,
  and a canary-first pause after the first verified reconnect. Halted rolls
  expose retry and resume actions. ([#43])
- Running jobs with no recent persisted activity are detected server-side and
  surfaced through deduplicated alerts. Kind-aware thresholds protect normal
  long-running work, terminal jobs self-resolve, and Prometheus exports stuck
  job count and oldest inactivity age. ([#41])
- Snapshot projections now retain host-level refresh timestamps—including
  authoritative empty reports—expose a derived stale flag, refresh after
  backup, forget, and prune, and support an explicit operator reconciliation
  endpoint. ([#40])

### Fixed

- Raised the bounded WebSocket read limit on both peers so ordinary large
  snapshot reports and restic events no longer disconnect otherwise healthy
  agents. Regression coverage exercises payloads beyond the library's former
  32 KiB default in both directions. ([#44])
- Fleet-update timeout verification now performs a final authoritative read of
  the agent version delivered by the reconnect `hello`, avoiding a poll/deadline
  race and reporting the last observed version when verification fails. ([#43])

### Changed

- CI runner images and release images use the anonymous-pull `public`
  namespace at `docker.dcglab.co.uk`; registry authentication remains required
  only for publishing.

## [1.1.1] - 2026-08-22

### Fixed

- Prevented agents from panicking when a WebSocket connection ends. The
  WebSocket library transfers ownership of a successful upgrade stream to
  the connection and leaves the HTTP response body nil; attempting to close
  that body caused agents to restart and left forget jobs permanently stuck
  in `running`. ([#37])
- Manual forget jobs now receive the same per-source-group retention policies
  as scheduled forget jobs. The API also supports a validated `--dry-run`
  option through the complete server-to-agent-to-restic path, and rejects
  hosts without configured retention before creating a job. ([#36])
- Corrected the admin-credentials help text to reflect that forget uses normal
  append-only credentials and blank admin credentials do not provide a
  fallback for prune. ([#34])

## [1.1.0] - 2026-06-15

### Added

- **Always-On vs intermittent host mode.** A host can now be marked as
  not always-on — for laptops/workstations that legitimately sleep,
  travel, or shut down outside hours. An intermittent host no longer
  raises "agent offline" alerts when it disappears; instead it shows a
  calm "asleep" state in the UI ("asleep · last seen … · will catch up
  on return") and is covered by a longer-horizon staleness alert (raised
  only when it has an enabled schedule and no successful backup in 7
  days). When such a host reconnects, the server waits a short settle
  window and then automatically dispatches any scheduled backup whose
  window elapsed while it was asleep. Toggle per host from the host
  detail page (operator-band, audited as `host.mode_updated`). New and
  existing hosts default to always-on, so current fleets are unaffected.

### Changed

- Host-detail header redesign: tags and presence are grouped into
  labelled, boxed pills with click-to-edit; presence shows a `24x7` /
  `Free` chip; the agent "out of date" indicator is simplified (the full
  version detail remains in the Agent-update panel and on hover).
- Relative timestamps ("2h ago") now tick client-side, so a tab left
  open no longer shows a stale value as wall-clock time moves on.
- Release and CI container images are now published to and pulled from
  the zot OCI registry (`docker.dcglab.co.uk`).

## [1.0.1] - 2026-05-09

### Fixed

- Build version is now single-sourced from `internal/version`, and the
  server Dockerfile's ldflags were corrected so docker-built binaries
  report their real version. Previously `internal/version.Version` stayed
  at its "dev" default in docker images, which made every host look
  permanently out-of-date to the update logic.

## [1.0.0] - 2026-05-09

First tagged release. Six development phases brought the project from
empty repo to a self-hostable, multi-tenant restic backup orchestrator
with a web UI, JSON API, and self-updating agent fleet.

### Phase 1 — MVP: enrolment, visibility, on-demand backup

- HTTP server, SQLite store with migrations, AEAD-encrypted
  credentials at rest, Argon2id password hashing, session cookies.
- WebSocket transport between server and agents (heartbeat, hello,
  schedule fan-out, job log streaming).
- Agent install path for Linux (systemd unit + `install.sh`); one-time
  enrolment tokens with embedded repo credentials.
- Run-now backup execution end-to-end, snapshot listing.
- Server-side encrypted repo creds pushed to the agent on hello.

### Phase 2 — Scheduling, retention, repo operations

- Source groups (paths + excludes + pre/post hooks + bandwidth caps)
  decoupled from schedules; a schedule fires a source group.
- Cron-style schedules with retention policies, server-driven
  reconciliation push and ack.
- `restic forget`, `prune`, `check`, `unlock` automation; periodic
  maintenance ticker with per-host stagger.
- Pending-runs queue with backpressure (`max_concurrent_jobs` per
  host).
- Repo stats panel on the host detail page (size, last-check, last-
  prune, stale-lock banner).
- Auto-init of repos on first onboard with credential-failure surface
  on the host detail page.
- Announce-and-approve enrolment path for hosts that don't have a
  pre-minted token (Ed25519 fingerprint, operator approves).
- Windows agent: SCM service integration + `install.ps1` installer.
- Cross-platform alt-enrolment (announce flow on Windows).

### Phase 3 — Restore, alerts, audit

- Restore wizard: pick a snapshot, pick paths, pick a target
  (in-place / new directory), live progress.
- Snapshot diff against parent.
- Alert engine: per-source-group dedup, severity tiers, ack / resolve.
- Live-refresh alerts table with severity cues.
- Audit log UI with filters, sort, CSV export, payload-detail modal.

### Phase 4 — RBAC, OIDC, host tags

- Role-based access control: viewer / operator / admin.
- User management UI (invite, role change, disable, password reset).
- Generic OIDC SSO with JIT user provisioning + role mapping.
- Per-host tags with chip-row filter on the dashboard.

### Phase 5 — OSS readiness

- mdBook-rendered docs site at `docs/book/`.
- Contributor onboarding (CONTRIBUTING.md, security policy, license).
- Docker-only release pipeline + reference deployment compose file.
- Playwright e2e harness covering the smoke runbook.

### Phase 6 — Update delivery + observability

- Agent self-update: server-side channel pin per host, signed binary
  fetch via the WS transport, atomic swap with rollback on failure.
- Fleet-wide update orchestration with per-host stagger and an admin
  pause switch.
- Prometheus `/metrics` endpoint + Grafana dashboard JSON.
- Repo size trend per host (90-day rolling) on the host detail page.

### Cross-cutting

- Live dashboard with column sort, filters, free-text host search,
  background-tab-aware live refresh (5s cadence).
- Pure-Go binary with embedded UI, no Node/CGO at runtime.
- Reproducible `-trimpath -ldflags="-s -w"` builds for
  linux/amd64, linux/arm64, windows/amd64.
- Sharded CI (server-http / store / rest), pre-commit hooks (gofumpt,
  go vet, golangci-lint).
- Threat model published (`docs/threat-model.md`).

[Unreleased]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.2.2...HEAD
[1.2.2]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.2.1...v1.2.2
[1.2.1]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.2.0...v1.2.1
[1.2.0]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.1.1...v1.2.0
[1.1.1]: https://gitea.dcglab.co.uk/steve/restic-manager/compare/v1.1.0...v1.1.1
[1.1.0]: https://gitea.dcglab.co.uk/steve/restic-manager/releases/tag/v1.1.0
[1.0.0]: https://gitea.dcglab.co.uk/steve/restic-manager/releases/tag/v1.0.0

[#37]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/37
[#36]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/36
[#34]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/34
[#40]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/40
[#41]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/41
[#43]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/43
[#44]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/44
[#50]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/50
[#54]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/54
[#55]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/55
[#56]: https://gitea.dcglab.co.uk/steve/restic-manager/issues/56