# Changelog
## [0.4.0-alpha.2]
- Inspector updates:
- Added `inspect --max-cache SIZE` (default `200 MiB`) and switched preview caching to size-aware LRU with approximate byte budgeting.
- Save to markdown file now includes properties in yaml header
- More properties visible in preview pane (select and multi-select had been omitted)
- select and multi-select are looked up and display as Name and Id instead of Name
- Added new keymaps
- ctrl-e open markdown in editor
- ctrl-c copy object id to clipboard
- ctrl-o open object in anytype app (using registered url scheme anytype://)
- single-line edit keys for editing save-as path (ctrl-a, ctrl-e, ctrl-k)
- Fixed inspector preview performance regression by reusing cached snapshot/markdown data instead of repeatedly re-reading and re-rendering the same objects.
## [0.3.0 - alpha] - anyback - 2026-02-16
- Flattened CLI: removed `archive` subcommand group and `info` command.
- `archive inspect` -> `list` (with `--brief`, `--expanded`, `--files`)
- `archive cmp` -> `diff`
- `archive cp` -> `extract` (positional: `<archive> <id> <output>`)
- `info` -> `list` (for summary/object IDs) or `manifest` (for manifest JSON)
- New `manifest` command for standalone manifest output.
- `list` includes summary and object IDs by default; use `--brief` for summary only.
- Legacy `archive`, `info` commands rejected with migration guidance.
- Changed backup/export archive output defaults:
- ZIP is now the default archive format for backup/export.
- removed the `--zip` CLI flag from `backup`/`export`.
- directory-layout archive support remains in archive reader/import paths.
- Added an `anyback_reader::markdown` module that converts protobuf snapshots
(`objects/<id>.pb`) to markdown directly from archive data, including text
block styles, inline marks, links/bookmarks, file/image links, and markdown
table rendering (`Table` / `TableRows` / `TableColumns` blocks).
- Added `anyback archive cp OBJID FILE [--archive ARCHIVE]`:
- saves document-like objects as markdown
- saves file/image objects as raw payload bytes
- Inspector updates:
- Preview panel now attempts real markdown rendering from protobuf snapshots
- Added `w` save-as flow for selected object with remembered output directory
- Renamed primary commands:
- `backup create` -> `backup`
- `restore apply` -> `restore`
- legacy forms are now rejected with guidance to use the simplified commands
- Replaced `restore plan` preflight flow with direct restore/import preflight:
- added `restore --dry-run` and `import --dry-run`
- removed restore plan/apply split from primary CLI flow
- Added backup zip output controls:
- new `--zip` flag on `backup`
- `--dest PATH` now auto-enables zip mode when `PATH` ends in `.zip`
- manifest metadata is written to sidecar `<archive>.manifest.json` (sibling path)
- inspect/restore still read legacy in-archive `manifest.json` for compatibility
- Added `archive cmp ARCHIVE1 ARCHIVE2`:
- compares directory/zip archives
- emits `ARCHIVE1 only`, `ARCHIVE2 only`, and `Changed` sections
- supports `pb` vs `pb`, `pb-json` vs `pb-json`, and cross-compare `pb` vs `pb-json`
- Restore now supports archives without `manifest.json` (including desktop-generated backups), and infers object IDs for reporting when possible.
- `backup create` output args updated:
- renamed parent directory option to `--dir`
- added `--dest PATH` for explicit archive output path
- added `--prefix` for generated archive naming
- added conflict and path validation (`--dest` vs `--dir/--prefix`, existing/missing paths)
- Added milestone-02 e2e coverage for full backup + restore in same space.
- Added nightly integrity fuzz scaffold test (`tests/integrity_nightly.rs`) with profile/env controls for object count, body size, elapsed time, and total bytes.
- Restore transport compatibility update:
- reintroduced path-based restore/import via `Rpc.Object.Import.Request.params.PbParams.path` for full archive apply
- moved snapshot restore code paths behind cargo feature `snapshot-import` (opt-in)
- when built without `snapshot-import`, `--objects` restore/import returns actionable guidance to rebuild with that feature
- Added integrity failure artifact capture in `tests/integrity_nightly.rs`:
- persists failed batch metadata, backup/restore command output, archive file listing, and copied archive under `/tmp/anyback-integrity-failures/...`
- adds `ANYBACK_INTEGRITY_TYPES` and `ANYBACK_INTEGRITY_FORMAT=pb|pb-json` controls
- On Linux/macOS, call `nix::unistd::sync()` after writing backup archive metadata to reduce transient filesystem visibility races in follow-up restore/integrity runs.
- Added `archive inspect` command with optional JSON output (`--json`) and optional file listing (`--files`).
- Extended `archive inspect`:
- supports both directory and `.zip` archives without unpacking
- adds `--expanded` mode to parse all `*.pb`/`*.pb.json` snapshot files and emit per-file metadata
- is fault-tolerant: unreadable snapshot files are marked `status="unreadable"` with reason while remaining files continue to parse
- Milestone 1 (restore snapshot transport plan) implementation:
- added shared archive reader module used by restore staging, `archive inspect`, and `archive cmp`
- removed duplicate dir/zip traversal branches in restore staging in favor of one archive-reader path
- added archive reader unit tests for directory and zip listing/read behavior
- Refactored archive reader into a standalone library target:
- `anyback/Cargo.toml` now defines lib crate `anyback_reader` plus CLI bin `anyback`
- archive reader API moved to `anyback_reader::archive` for reuse by external inspector programs
- archive reader code has no keystore dependencies
- CLI-only dependencies are gated behind feature `cli` (enabled by default); library consumers can use `default-features = false`
- Milestone 2 (snapshot transport engine) implementation:
- added snapshot import builder and batch planner for `Rpc.Object.Import.Request.snapshots`
- added chunk limits with env overrides:
- `ANYBACK_IMPORT_MAX_SINGLE_SNAPSHOT_BYTES`
- `ANYBACK_IMPORT_MAX_BATCH_BYTES`
- `ANYBACK_IMPORT_MAX_BATCH_SNAPSHOTS`
- added oversize single-snapshot error path with object id/path/size context
- added unit tests for batch split and oversize enforcement
- added initial transport switch `ANYBACK_RESTORE_TRANSPORT=snapshots` during migration
- snapshot transport currently supports `*.pb`; `*.pb.json` restore returns actionable error
- Milestone 3 (event-driven progress/completion) implementation:
- restore now wraps import lifecycle with `ProcessSubscribe`/`ListenSessionEvents`/`ProcessUnsubscribe`
- restore now waits for `processNew`/`processUpdate`/`processDone` correlation per import dispatch and tracks `importFinish` summaries
- added restore process timeout env overrides:
- `ANYBACK_EVENT_STREAM_CONNECT_TIMEOUT`
- `ANYBACK_PROCESS_START_TIMEOUT`
- `ANYBACK_PROCESS_IDLE_TIMEOUT`
- `ANYBACK_PROCESS_DONE_TIMEOUT`
- restore JSON report now includes `event_progress` summary with process/event counts and final observed process state
- Milestone 3 (path transport hardening) implementation:
- removed restore import staging helpers and manifest-stripping staging branches
- removed path-race retry branch for `"no such file or directory"` in integrity restore harness
- Increased nightly integrity profile complexity in `tests/integrity_nightly.rs`:
- `small` profile now runs more iterations/objects and higher body-byte budgets
- `medium` profile now runs more iterations/objects with larger byte/time caps
- profile parse test now enforces monotonic growth across tiny/small/medium/large
- Expanded nightly integrity semantic coverage in `tests/integrity_nightly.rs`:
- validates restored object type, description property, and per-object markdown markers
- uploads and roundtrips file attachments (with context metadata) on even iterations
- validates file attachment roundtrip via restore JSON success rows (`type=file`, expected file name)
- Restore reliability hardening:
- restore import progress is always enabled at the RPC layer (`no_progress=false`) so process events remain observable in non-interactive runs
- session event stream drop/read failures now trigger reconnect-and-continue within existing restore timeouts
- restore cancellation now uses an `mpsc` channel so signal handlers (SIGINT/SIGTERM) can request cancellation via the same path intended for future TUI cancel actions
- import API errors now include code-based hints for common import failure classes
- Restore/import completion hardening for path-based imports:
- restore/import now fails if process completion events time out instead of silently proceeding
- this prevents early command exit while server-side async import is still running against archive paths
- Test cleanup hardening:
- e2e and nightly integrity tests now register per-test prefix cleanup guards that run on early returns/panics
- cleanup no longer depends solely on end-of-test happy-path execution
- Added chat restore coverage:
- new e2e tests validate chat message roundtrip for chat spaces and regular spaces with chats
- nightly `large` profile now runs chat-message roundtrip checks for both regular spaces-with-chats and chat spaces
- chat validations explicitly assert that sent message tokens are present after restore
## 0.1.0 - 2026-02-10
- Initial proof-of-concept CLI scaffold.
- Added commands: `auth`, `export`, `import`, `info`.
- Implemented object-list export using `ObjectListExport` via `anytype-api` backup wrapper.
- Added archive `manifest.json` generation in export output directory.
- Implemented best-effort import using gRPC `ObjectImport` (`IGNORE_ERRORS`).
- Added JSON import report output via `--log`.
- Added unit tests for object-list parsing and manifest utilities.