Store live photos as their image and their video (closes #107)

A live photo, which Ente stores as one ZIP, is unpacked as it downloads
into its image and its video, each `<fileID>.<ext>` with its extension
from the ZIP, beside `<fileID>.livephoto.json`, which names the two.
Both are checked against the recorded hash and renamed into place only
when both are complete. The backup and the content cache count a live
photo as stored only with both files, album folders link both,
`quak get` writes both, and the content result gives the video as
`videoPath`. A ZIP an earlier version stored is replaced.

Model: opus-5-5

Dockerfile

@@ -31,6 +31,12 @@ RUN script/bootstrap
 
 COPY . .
 
+# Unlike the template, the suite runs as the image's non-root `node` user:
+# root ignores directory permissions, so the tests of a destination that is
+# not writable would otherwise fail. vitest writes into /app.
+RUN chown -R node:node /app
+USER node
+
 RUN timeout 90 yarn run vitest run --reporter=dot || \
     { echo "--- Rerunning with verbose for details ---"; \
       timeout 90 yarn run vitest run --reporter=verbose; exit 1; }

README.md

@@ -214,11 +214,23 @@ quak/
         auth/           login flow (SRP + email OTP + TOTP), key unwrap
         model/          decrypted Collection, File, Metadata types + decrypt fns
         download/       streaming file/thumbnail download + decryption
+        library/        the cache-backed Library: metadata store, read
+                        surface, records, content cache, precache, ML data
+                        and search, request pools
         backup.ts       resilient full-account backup with dedup
+        metadata-backup.ts
+                        backup-metadata: all decrypted metadata as JSON
+        mldata-fetch.ts fetch + decrypt per-file ML data
+        filename.ts     safe file names from server metadata
         errors.ts       error types shared across layers
         retry.ts        retry classifier + exponential backoff with jitter
         thumbnails.ts   detect + regenerate missing thumbnails
         client.ts       high-level Client class assembled from the above
+        cli-commands.ts the CLI's commands as functions returning exit codes
+        cli-output.ts   how the CLI prints a file's title and time
+        cli-read.ts     fresh reads for the CLI's read commands
+        cli-run.ts      run a command, print its error, exit with its code
+        cli-session.ts  read the saved session file back into a Client
         index.ts        public library exports
     bin/
         quak.ts         CLI entrypoint (commander.js)
@@ -296,6 +308,7 @@ Endpoints used:
   encrypted token plus key attributes.
 - `POST /users/ott` and `POST /users/verify-email`: email OTP fallback path.
 - `POST /users/two-factor/verify`: TOTP second factor.
+- `POST /users/logout`: end the calling token's session (`quak logout`).
 - `GET /collections/v2?sinceTime=<usec>`: list collections changed since
   microsecond timestamp; pass 0 for a full enumeration.
 - `GET /collections/v2/diff?collectionID=<id>&sinceTime=<usec>`: list files in a
@@ -410,7 +423,9 @@ whatever else fits their use case. `Client.fromJSON(snapshot)` restores a
 working client from that snapshot without re-authenticating; it checks every
 field and each key's length first, and throws an error naming the bad field.
 `client.logout()` clears the token and zeroes the key buffers in place; every
-later call on that client throws.
+later call on that client throws. It does not contact the server, so the token
+stays valid there and in any saved snapshot; `await client.logoutOnServer()`
+first ends the session on the server (`POST /users/logout`).
 
 The CLI stores the snapshot at the platform-appropriate data directory via
 `env-paths`: `~/Library/Application Support/quak/session.json` on macOS,
@@ -420,13 +435,21 @@ you would treat the password itself. A missing file is reported as "not logged
 in"; a file that exists but is corrupt is reported as such, naming the bad
 field. Both exit with status 1.
 
+`quak logout` ends the session on the server, so the token in `session.json`
+stops working even in a copy of the file, and then deletes the file. If the
+server call fails (or the file is corrupt), the file is still deleted, the
+command says the server session could not be ended, and it exits with status 1.
+It does not delete the cache: it prints the account's cache directory and says
+it still holds decrypted data (file keys in `metadata.json`, cached originals
+and thumbnails), for the user to delete if they want it gone.
+
 ### CLI surface
 
 ```
 quak [--cache-dir <path>] <command>             global: local metadata/content cache location
 quak login                                      interactive or QUAK_EMAIL/QUAK_PASSWORD
 quak whoami                                     print logged-in account as JSON
-quak logout                                     delete saved session
+quak logout                                     end the session, delete it
 quak collections [--json]                       list all collections
 quak files --collection <id> [--json]           list files in a collection
 quak get <fileID> [--out path] [--collection]   download and decrypt a file
@@ -438,22 +461,40 @@ quak helper fix-missing-thumbnails [--file ids] generate + upload missing thumbn
 ```
 
 Every command runs on the same cache-backed library. The read commands —
-`collections`, `files`, `get`, and `get-thumb` — force a fresh server round-trip
+`collections`, `files`, `get`, `get-thumb`, `backup-metadata`,
-before they answer, so they report current account state rather than whatever
+`helper list-missing-thumbnails` and `helper fix-missing-thumbnails` — force a
-the cache last held. `--cache-dir` overrides where the cache lives; without it
+fresh server round-trip before they answer, so they report current account state
-each account gets its own directory under the per-user cache path.
+rather than whatever the cache last held. If that round-trip fails, the command
+prints the error on one line and exits 1. `--cache-dir` overrides where the
+cache lives; without it each account gets its own directory under the per-user
+cache path.
 
 `get` and `get-thumb` resolve the file by ID directly, so `--collection` is
-accepted for backward compatibility but ignored. `backup-metadata --exif` (alias
+accepted for backward compatibility but ignored. For a live photo, `get` writes
-`--all`) additionally downloads each file to extract full EXIF/IPTC/XMP
+its image and its video, each named after the title with its own extension, as
-metadata. The listing and backup commands support `--json` for machine-readable
+Ente's clients name them (`IMG_0001.heic` and `IMG_0001.mov`). With
-output.
+`--out PATH`, the image is written to `PATH` and the video beside it, with
+`PATH`'s name and the video's extension; a `PATH` with the video's extension is
+refused. `backup-metadata --exif` (alias `--all`) additionally downloads each
+file to extract full EXIF/IPTC/XMP metadata. The listing and backup commands
+support `--json` for machine-readable output.
+
+`backup-metadata` fetches ML data in requests of up to 200 files. When a request
+still fails after its retries, the error is logged, each of its files is written
+with the reason in an `mlDataError` field instead of `mlData`, and the dump goes
+on. The exit code is non-zero if any ML data request failed.
 
 `helper fix-missing-thumbnails` regenerates thumbnails for baseline JPEG images
 only, because the bundled decoder (`jpeg-js`) decodes only JPEG. A non-JPEG
 image (PNG, HEIC) or a video is reported as `skipped` (unsupported format), kept
 distinct from a `failed` repair, and does not affect the exit code; a genuine
-failure still exits non-zero.
+failure still exits non-zero. The server accepts a new thumbnail only from the
+file's owner and only when it is no larger than the thumbnail size it records
+for the file. So a file another account owns, in an album shared with you, is
+skipped by both thumbnail helpers without being fetched, and the fixer skips a
+file whose recorded thumbnail size is 0 or unknown. Otherwise the fixer lowers
+the quality and size of the thumbnail until it fits, and skips the file if even
+the smallest does not.
 
 ### Backup layout
 
@@ -462,26 +503,65 @@ failure still exits non-zero.
 ```
 <dir>/
     originals/
-        <fileID>.<ext>          actual file content (one per unique file)
+        <fileID>.<ext>          actual file content (one per unique file,
+                                two for a live photo: see below)
         <fileID>.json           all decrypted metadata for that file
+        <fileID>.livephoto.json which of a live photo's two files is which
     collections/
         <name>/
             <title> -> ../../originals/<fileID>.<ext>   (symlink)
         <name>.json             collection metadata + file list
+    failures.json               files that failed and have not yet succeeded
 ```
 
-Each file is downloaded exactly once regardless of how many collections it
+`failures.json` records each failed file with the kind of failure, how many
-appears in. On subsequent runs, existing originals are skipped. If a download
+times it has been tried and when it was last tried. A file leaves it once it
-fails, the error is logged and the backup continues with the next file. The exit
+succeeds, or once it is no longer in the library or in the backup's scope. The
-code is non-zero if any files failed.
+library's `lib.backup({ includeThumbnails: true })` also writes
+`thumbnails/<fileID>.jpg` beside `originals/`; `quak backup` does not.
 
-Each original is copied to a temporary file named
+A collection's directory and JSON are named after the collection, and a symlink
-`.quak-backup-<fileID>.<ext>-<pid>-<random>.tmp` in the same directory, synced
+after the file's title, both with unsafe characters replaced. When two
-to disk, and renamed into place, so an original is either complete or absent,
+collections would get the same name, or two symlinks in one collection the same
-even after a power cut. A run that is killed can leave one of these temporary
+name (ignoring case in both), each of them gets its ID added: two albums named
-files behind; the next backup deletes those whose process is no longer running.
+`Trip` become `Trip (10)/` and `Trip (11)/`, and two files titled `IMG_0001.JPG`
-Downloads and the content cache use the same scheme with `.quak-<random>.tmp`
+become `IMG_0001 (12345).JPG` and `IMG_0001 (12346).JPG`. IDs never change, so a
-names. The rename replaces whatever was at the destination rather than writing
+name stays the same from run to run until such a clash appears or goes away.
+
+A live photo, which Ente stores as one ZIP of its image and its video, is stored
+as those two files, which a photo viewer can open: each is
+`originals/<fileID>.<ext>` with the extension it has inside the ZIP (for example
+`12345.heic` and `12345.mov`), and `<fileID>.livephoto.json` names the two. The
+live photo counts as stored only when both files are present and not empty. Its
+album folder links both, each named after the title with that file's extension
+(`IMG_0001.heic` and `IMG_0001.mov`). A live photo that an earlier version of
+quak stored as the ZIP, under the image's name, is replaced by its two files on
+the next run, and the ZIP and its link are removed.
+
+Each run removes the symlinks into `originals/` that no longer belong in their
+collection's directory, and the directories (and JSON) of collections that were
+deleted or renamed. Nothing else in `collections/` is touched: a file or a
+symlink you put there stays, and a directory that still holds one after its
+symlinks are removed stays too, with its JSON.
+
+Each file is downloaded exactly once regardless of how many collections it
+appears in, and written once: straight into `originals/`, with no copy left in
+the cache. An original the cache already held is copied from there instead. On
+subsequent runs, existing originals are skipped. If a download fails, the error
+is logged and the backup continues with the next file. The exit code is non-zero
+if any files failed. `quak backup` opens its library with the thumbnail and
+originals precache off, so it fetches only what the backup stores.
+
+Each original is written to a temporary file in the same directory, synced to
+disk, and renamed into place, so an original is either complete or absent, even
+after a power cut. A downloaded original's temporary file is named
+`.quak-<pid>-<random>.tmp`, one copied from the cache
+`.quak-backup-<fileID>.<ext>-<pid>-<random>.tmp`. A run that is killed can leave
+one of these temporary files behind; the next backup deletes those whose process
+is no longer running. The content cache uses the same scheme, and opening a
+library deletes the temporary files in the cache whose process is no longer
+running, so a download another process has in progress in the same cache is left
+alone. The rename replaces whatever was at the destination rather than writing
 through it: a symlink there is replaced, not followed, and the new file has the
 temporary file's permissions, not those of the file it replaced.
 
@@ -491,7 +571,12 @@ temporary file's permissions, not those of the file it replaced.
       errors
 - [x] Update the API reference section below to match the current implementation
 - [x] `make docker` green
-- [ ] Tag `v1.0.0`
+- [x] Store live photos in a form a photo viewer can open
+      (https://git.eeqj.de/sneak/quak/issues/107): unpacked into the image and
+      the video
+
+Tagging and releases are decided by sneak alone, and happen only when he
+declares one.
 
 Future (desktop client, separate repo):
 
@@ -511,10 +596,11 @@ test suite is the canonical, executable documentation — `test/library/` and
 ### Opening a library
 
 `Library.open(options)` loads the on-disk cache, starts the background refresh
-loop, and resolves to a `Library`. On an empty cache it awaits the first refresh
+loop, and resolves to a `Library`. On an empty cache it awaits the first
-so it never opens onto empty data; on an existing cache it returns immediately
+refresh, so it opens onto the account's data whenever the server is reachable;
-and refreshes in the background, so an unreachable server does not block
+if that refresh fails, it opens with no data and records the error in
-opening.
+`lib.status()`. On an existing cache it returns immediately and refreshes in the
+background, so an unreachable server does not block opening.
 
 `LibraryOptions`:
 
@@ -579,7 +665,9 @@ An `Album` exposes its record fields and `album.photos.list()` → `Photo[]`
 (newest first). A `Photo` exposes its record fields, `photo.record()` →
 `PhotoRecord`, and two content methods:
 
-- `await photo.original(opts?)` → `{ path, bytes }` — the full-resolution file.
+- `await photo.original(opts?)` → `{ path, bytes, videoPath? }` — the
+  full-resolution file. For a live photo, `path` and `bytes` are its image's and
+  `videoPath` is its video.
 - `await photo.thumbnail(opts?)` → `{ path, bytes }`.
 
 Both serve from the on-disk content cache when the bytes are present and
@@ -599,7 +687,7 @@ The GUI-facing records hold no key material and no binary, so they survive
 - `PhotoRecord`: `fileID`, `albumIDs`, `title`, `takenAt` (milliseconds),
   `fileType`, optional `caption` / `width` / `height` / `latitude` /
   `longitude`, `isArchived`, `isHidden`, and `thumbnailPath` / `originalPath`
-  once the bytes are cached.
+  once the bytes are cached (for a live photo, `originalPath` is its image).
 - `AlbumRecord`: `collectionID`, `name`, `type`, `isShared`, `updationTime`, and
   `fileIDs` (newest first).
 - `LibrarySnapshot`: `{ albums, photos, takenAt }`.
@@ -625,12 +713,15 @@ photos newest first). `lib.subscribe({ onChange })` delivers a `LibraryChange`
   default limit 20). quak bundles no text encoder, so `searchByEmbedding` takes
   a query vector the caller produced elsewhere.
 - `await lib.backup(opts?)` → `BackupResult`. It refreshes, fetches every
-  in-scope original (and, with `includeThumbnails`, thumbnails) through the
+  in-scope original not already in the backup (and, with `includeThumbnails`,
-  content cache, and rebuilds the on-disk backup tree with a durable failure
+  thumbnails) through the content cache, and rebuilds the on-disk backup tree
-  ledger. `BackupOptions`: `downloadDirectory` (falls back to the one `open()`
+  with a durable failure ledger. A fetched original is written straight into the
-  was given), `includeOriginals` (default `true`), `includeThumbnails` (default
+  backup's `originals/` and not into the cache, which then counts it as present;
-  `false`), `onlyAlbumNames`, and `onProgress`. See Backup layout above for the
+  one the cache already held is copied from there. `BackupOptions`:
-  tree it writes.
+  `downloadDirectory` (falls back to the one `open()` was given),
+  `includeOriginals` (default `true`), `includeThumbnails` (default `false`),
+  `onlyAlbumNames`, and `onProgress`. See Backup layout above for the tree it
+  writes.
 
 ### Request pools
 
@@ -647,6 +738,8 @@ Under `cacheDirectory`:
 <cacheDirectory>/
     metadata.json                 decrypted account state + refresh cursor
     originals/<fileID>.<ext>      cached full-resolution files
+    originals/<fileID>.livephoto.json
+                                  which of a live photo's two files is which
     thumbnails/<fileID>.jpg       cached thumbnails
     mldata/
         <fileID>.json             one decrypted ML payload per file
@@ -654,11 +747,26 @@ Under `cacheDirectory`:
         fetched.json              per-file fetch bookkeeping
 ```
 
+When `metadata.json` belongs to a different account than the client's,
+`Library.open` deletes it and `mldata/` and starts from an empty cache. Cached
+originals and thumbnails are kept; they are reached only through the files the
+current account's records name.
+
+A live photo's original is cached as in the backup: its image and its video,
+each `originals/<fileID>.<ext>` with its own extension, and
+`originals/<fileID>.livephoto.json` naming them; the two are evicted together. A
+live photo that an earlier version cached as its ZIP is not served, and is
+replaced by its two files the next time it is read.
+
 A stored file appears only via an atomic temp-then-rename, so its presence means
-it is complete. The design also calls for a content-hash comparison against
+it is complete. Every downloaded original (by `quak get`, the cache, or
-`FileMetadata.hash` on each fetched original; that check is deferred (issue
+`backup`) whose metadata records a content hash (`FileMetadata.hash`) is hashed
-https://git.eeqj.de/sneak/quak/issues/68) because the exact hash construction
+as it is written: unkeyed BLAKE2b with a 64-byte output, standard base64. A live
-cannot yet be confirmed against the repo's fixtures.
+photo arrives as a ZIP and is unpacked as it is written; its image and its video
+are hashed separately and joined as `<imageHash>:<videoHash>`, and neither is
+stored unless both are complete and match. A mismatch stores nothing and fails
+the download with an error naming the file ID. An original with no recorded
+hash, from a very old client, is stored unchecked.
 
 ### Key types by source file
 
@@ -719,8 +827,10 @@ documents:
   markdown. Use `make fmt` to format. Use `yarn` not `npm`.
 
 - **Testing:** vitest. Tests go in `test/` mirroring the `src/` structure.
-  `make test` must complete in under 20 seconds. Use `mkdtempSync` for temporary
+  `make test` must finish in under 60 seconds (the hard cap) and should finish
-  directories, never manual timestamp paths.
+  in under 20. The 90-second `timeout` in the `test` phase of the `Dockerfile`
+  is a backstop that catches a hung test, not the time limit. Use `mkdtempSync`
+  for temporary directories, never manual timestamp paths.
 
 - **Code style:** `const` for everything, `let` if reassignment is needed, never
   `var`. Avoid unnecessary comments. No hand-rolled crypto. The

TODO.md

@@ -14,16 +14,136 @@ pre-1.0
 
 # Next Step
 
-Tag v1.0.0.
+None: every issue still open is done on `next` or `next2` and waits for it to
+reach `main`.
+
+Tagging and releases are decided by sneak alone, and happen only when he
+declares one.
 
 # Completed Steps
 
+- 2026-09-23: Live photos are stored as their image and their video (issue 107).
+  A live photo, which Ente stores as one ZIP, is unpacked as it downloads into
+  `<fileID>.<ext>` for the image and for the video, each with its extension from
+  the ZIP, beside `<fileID>.livephoto.json`, which names the two. Both are
+  checked against the recorded hash and renamed into place only when both are
+  complete. The backup and the content cache count the live photo as stored only
+  with both files, the album folder links both, `quak get` writes both, and the
+  content result gives the video as `videoPath`. A ZIP an earlier version stored
+  is replaced on the next backup run, and in the cache when the photo is next
+  read.
+
+- 2026-09-23: Settled the package metadata (issue 6). quak is not published, so
+  `package.json` is marked `"private": true` and the `files` field is gone.
+  `engines.node` is `>=22`, the major version `script/bootstrap` and the
+  `Dockerfile` use. An `exports` map makes `.` and `./package.json` the only
+  importable paths; `runMetadataBackup`, the thumbnail helpers and their types
+  stay internal to the CLI.
+
+- 2026-09-23: Brought the README and this file in line with the tree (issue
+  111). The layout lists `src/library/` and the other source files, the backup
+  layout names `failures.json` and the optional `thumbnails/`, "Opening a
+  library" says what happens when the first refresh fails, the Testing section
+  gives the 60-second hard cap and 20-second target for `make test` and names
+  the 90-second `timeout` in the `Dockerfile` as the backstop for a hung test,
+  and "Tag v1.0.0" is no longer listed as the next step.
+
+- 2026-09-23: Tested `quak login` and `backup-metadata --exif` (issue 110).
+  `loginCommand` takes its login function and its prompts from `CliContext`, and
+  `bin/quak.ts` passes `Client.login` and the terminal prompts. Tests cover a
+  login from `QUAK_EMAIL` and `QUAK_PASSWORD` with no prompt, the TOTP prompt, a
+  failed login, and the saved session's modes, and show that `--exif` and
+  `--all` each turn on EXIF extraction and that it is off without them.
+
+- 2026-09-23: `quak backup` writes each original once and no longer fills the
+  cache (issue 106). An original fetched for a backup is written by the download
+  writer straight into the backup's `originals/`, and the content cache records
+  it there instead of keeping its own copy; one the cache already held is still
+  copied. `quak backup` opens its library with the thumbnail and originals
+  precache off.
+
+- 2026-09-23: `backup-metadata`, `helper list-missing-thumbnails` and
+  `helper fix-missing-thumbnails` refresh before they answer (issue 100). Each
+  awaits `lib.fresh()` before reading, so a file added since the cache was
+  written is included, and a failed refresh prints one line and exits 1 instead
+  of answering from a stale or empty cache. The README lists them among the
+  commands that refresh first.
+
+- 2026-09-23: `quak backup` waits for the server refresh and fails when it fails
+  (issue 99). `lib.backup()` joins a refresh already running or starts one, as
+  `fresh()` does, and rejects before touching any file when it fails, leaving
+  `failures.json` as it was, so `quak backup` prints the error as one line and
+  exits 1 instead of backing up the previous run's file list, or nothing, and
+  exiting 0.
+
+- 2026-09-23: CLI errors print a message instead of a stack trace (issue 102).
+  An error a command throws is printed as one `quak: MESSAGE` line on stderr and
+  the CLI exits 1 once output has drained. The wrapper that does this moved from
+  `bin/quak.ts` to `src/cli-run.ts`, and `bin/quak.ts` now awaits
+  `program.parseAsync()`.
+
+- 2026-09-23: Opening a library no longer deletes another process's download in
+  progress (issue 105). The download writer's temp files are named
+  `.quak-<pid>-<random>.tmp`, and `removeLeftoverTempFiles`, moved from the
+  backup into the download module, deletes a `.quak-*.tmp` file only when the
+  process ID in its name is no longer running. The content cache calls it at
+  `open()` for `originals/` and `thumbnails/`, the backup as before.
+
 - 2026-09-23: Re-vendored the lint and test setup from the template (issue 96).
   Linting and testing are the `lint` and `test` phases of the `Dockerfile`;
   `script/lint` and `script/test` each build one with `--no-cache`, and the last
   stage compiles and depends on both, so `script/cibuild` is one build.
   `Dockerfile.lint`, `CHECK_EPOCH`, `LINT_EPOCH` and the tests that checked them
   are gone; `REPO_POLICIES.md` is re-copied.
+
+- 2026-09-23: Stopped `helper fix-missing-thumbnails` retrying files the server
+  always refuses (issue 109). Both thumbnail helpers skip a file another account
+  owns without fetching it. The fixer skips a file whose recorded thumbnail size
+  is 0 or unknown before downloading it, and otherwise tries smaller encodings
+  (720 px quality 50 down to 160 px quality 20) until the encrypted thumbnail is
+  no larger than that size, skipping the file if none fits.
+
+- 2026-09-23: Tested the live-photo hash check's error paths (issue 117). Tests
+  download a live photo whose ZIP names an unknown compression method, one whose
+  ZIP has no image entry and one with no video entry, and check that nothing is
+  stored and the error names the file ID; the unreadable one is not retried.
+
+- 2026-09-23: `quak logout` ends the session on the server (issue 108). It calls
+  `POST /users/logout` through the new `Client.logoutOnServer()`, then deletes
+  `session.json` even when that call fails, says so and exits 1. It prints the
+  account's cache directory and says it still holds decrypted data. The default
+  cache path is now `defaultCacheDirectory()` in the library, shared with
+  `Library.open`.
+
+- 2026-09-23: Fixed the backup's per-collection folders (issue 103). Two files
+  in one collection with the same title, and two collections with the same name,
+  each get their ID added to the name (`IMG_0001 (12345).JPG`, `Trip (10)/`), so
+  none replaces another's symlink or JSON. Each run removes symlinks into
+  `originals/` for files no longer in the collection, and the folders of deleted
+  or renamed collections, leaving anything else in `collections/` alone. The
+  README backup layout states the naming rule.
+
+- 2026-09-23: Checked downloaded originals against their recorded content hash
+  (issue 68). `downloadFile`, which `quak get`, the content cache and backup all
+  use, hashes the decrypted bytes (unkeyed BLAKE2b-512, standard base64) and
+  stores nothing on a mismatch, failing with an error naming the file ID. A live
+  photo ZIP is unpacked as it streams with `fflate` and its image and video
+  hashed separately as `<imageHash>:<videoHash>`. `decryptFile` reads older
+  clients' `imageHash` and `videoHash` fields for live photos. A file with no
+  recorded hash is stored unchecked.
+
+- 2026-09-23: Kept one account's cache from mixing with another's (issue 104).
+  When `metadata.json` in the cache directory was written for a different,
+  non-zero user ID than the client's, `Library.open` deletes it and `mldata/`
+  and starts empty, so the first refresh enumerates from 0. This only happens
+  with `--cache-dir` or an explicit `cacheDirectory`; the default path already
+  includes the user ID. A test opens one account's cache as another account.
+
+- 2026-09-23: `backup-metadata` no longer stops on one failed ML data request
+  (issue 101). Each request of up to 200 files is tried on its own; a failed one
+  is logged, its files are written with the reason in `mlDataError`, and the
+  command exits 1 once the dump is complete. `fetchMLData`, which only this
+  command used, is gone; the command calls `fetchMLDataBatch` per batch.
 - 2026-09-23: Single-sourced the version string (issue 5). `package.json` is the
   only place it is written: `src/index.ts` imports it for `VERSION` and
   `bin/quak.ts` passes `VERSION` to commander. tsc copies `package.json` to

bin/quak.ts

@@ -1,6 +1,7 @@
 #!/usr/bin/env node
 
 import { stdout, stderr } from "node:process";
+import { input, password } from "@inquirer/prompts";
 import { Command } from "commander";
 import envPaths from "env-paths";
 import { init } from "../src/crypto/index.js";
@@ -18,7 +19,9 @@ import {
     listMissingThumbnailsCommand,
     fixMissingThumbnailsCommand,
 } from "../src/cli-commands.js";
+import { run as runCommand } from "../src/cli-run.js";
 import { loadSession } from "../src/cli-session.js";
+import { Client } from "../src/client.js";
 import { VERSION } from "../src/index.js";
 
 const paths = envPaths("quak", { suffix: "" });
@@ -41,25 +44,13 @@ const context = (): CliContext => ({
     sessionDir: paths.data,
     cacheDir: program.opts<{ cacheDir?: string }>().cacheDir,
     loadSession,
+    login: (opts) => Client.login(opts),
+    prompt: (message) => input({ message }),
+    promptSecret: (message) => password({ message, mask: true }),
 });
 
-// Run a command and exit with its code once stdout/stderr have drained.
+const run = (command: Promise<number>): Promise<void> =>
-// Exiting before the drain can truncate piped output, and the library can keep
+    runCommand(command, stdout, stderr, (code) => process.exit(code));
-// the event loop alive after a command returns, so a plain return could hang.
-const run = async (command: Promise<number>): Promise<void> => {
-    process.exitCode = await command;
-    const pending = [stdout, stderr].filter((s) => s.writableLength > 0);
-    if (pending.length === 0) {
-        process.exit();
-        return;
-    }
-    let remaining = pending.length;
-    for (const s of pending) {
-        s.once("drain", () => {
-            if (--remaining === 0) process.exit();
-        });
-    }
-};
 
 program
     .command("login")
@@ -73,7 +64,7 @@ program
 
 program
     .command("logout")
-    .description("Delete the saved session")
+    .description("End the session on the server and delete the saved session")
     .action(() => run(logoutCommand(context())));
 
 program
@@ -169,4 +160,4 @@ helper
     );
 
 await init();
-program.parse();
+await program.parseAsync();

package.json

@@ -9,17 +9,23 @@
         "type": "git",
         "url": "https://git.eeqj.de/sneak/quak.git"
     },
+    "private": true,
     "type": "module",
+    "engines": {
+        "node": ">=22"
+    },
     "main": "./dist/src/index.js",
     "types": "./dist/src/index.d.ts",
+    "exports": {
+        ".": {
+            "types": "./dist/src/index.d.ts",
+            "import": "./dist/src/index.js"
+        },
+        "./package.json": "./package.json"
+    },
     "bin": {
         "quak": "./dist/bin/quak.js"
     },
-    "files": [
-        "dist/",
-        "README.md",
-        "LICENSE"
-    ],
     "scripts": {
         "build": "script/build",
         "quak": "node ./dist/bin/quak.js",
@@ -42,6 +48,7 @@
         "env-paths": "4.0.0",
         "exif-reader": "2.0.3",
         "fast-srp-hap": "2.0.4",
+        "fflate": "0.8.3",
         "jpeg-js": "0.4.4",
         "libsodium-wrappers-sumo": "0.8.4"
     }

src/backup.ts

@@ -1,9 +1,11 @@
 // The backup command, rebuilt on the library API (issue #51).
 //
-// `lib.backup()` refreshes the library, then, for every file in scope, gets its
+// `lib.backup()` waits for a completed refresh of the library (a failed one
-// original bytes onto disk under `downloadDirectory` and rebuilds the derived
+// fails the backup before any file is touched), then, for every file in scope,
-// views (per-file sidecars, per-collection symlink trees, per-collection JSON)
+// gets its original bytes onto disk under `downloadDirectory` and rebuilds the
-// from the model. The on-disk layout is the historical one, unchanged:
+// derived views (per-file sidecars, per-collection symlink trees,
+// per-collection JSON) from the model. The on-disk layout is the historical
+// one:
 //
 //     <downloadDirectory>/
 //         originals/<fileID>.<ext>     the decrypted bytes
@@ -12,12 +14,18 @@
 //         collections/<name>.json      per-collection metadata
 //         failures.json                durable ledger of unresolved failures
 //
+// A live photo's original is its image and its video, `<fileID>.<ext>` each
+// with its own extension, and `originals/<fileID>.livephoto.json` naming them;
+// its album folders link both.
+//
 // Crash-safety rests on two properties. Bytes are present-means-complete: an
 // original appears under `originals/` only via the content layer's atomic
 // temp-then-rename, so a file that exists is whole and is never re-fetched — an
 // interrupted run resumes by listing the directory. The derived views hold no
 // unique state, so they are rebuilt every run; that repairs stale sidecars and
-// missing or broken symlinks left by an earlier crash.
+// missing or broken symlinks left by an earlier crash. A rebuild also removes
+// the symlinks into originals/ that no longer belong to an album, and the
+// directories of albums that no longer exist.
 //
 // Resilience (issue #8): no per-file condition aborts the run. A failed
 // download or a failed symlink is caught, recorded in `failures.json` with a
@@ -34,16 +42,22 @@ import {
     readdirSync,
     readFileSync,
     readlinkSync,
+    rmdirSync,
     rmSync,
     statSync,
     symlinkSync,
     writeFileSync,
 } from "node:fs";
 import { copyFile, rename, rm } from "node:fs/promises";
-import { basename, dirname, join, relative } from "node:path";
+import { basename, dirname, extname, join, relative } from "node:path";
 
-import { fsyncPath } from "./download/index.js";
+import { fsyncPath, removeLeftoverTempFiles } from "./download/index.js";
-import { safeExtension, sanitizeFileName } from "./filename.js";
+import { sanitizeFileName, withExtension } from "./filename.js";
+import {
+    originalName,
+    storedOriginal,
+    writeLivePhoto,
+} from "./library/content.js";
 import type { Collection, EnteFile } from "./model/types.js";
 
 export type ProgressCallback = (message: string) => void;
@@ -92,8 +106,14 @@ export interface BackupLibrary {
     listCollections(): Collection[];
     listFiles(collectionID: number): EnteFile[];
     // Get an original's bytes onto disk through the content cache/pools,
-    // returning where they landed (the cache, or a prior backup).
+    // returning where they landed: `destination` when they were fetched now,
-    original(fileID: number): Promise<{ path: string }>;
+    // otherwise wherever they already were (the cache, or a prior backup). A
+    // live photo lands as its image and its video, fetched now beside
+    // `destination`.
+    original(
+        fileID: number,
+        destination: string,
+    ): Promise<{ path: string; videoPath?: string }>;
     thumbnail(fileID: number): Promise<{ path: string }>;
 }
 
@@ -110,12 +130,6 @@ interface FailureEntry {
 
 const LEDGER_VERSION = 1;
 
-// The originals/ filename for a file: `<id><ext>`, the extension taken from the
-// title (or `.bin`). Matches the content cache's own naming so a present check
-// lines up with what a fetch would write.
-const originalName = (file: EnteFile): string =>
-    `${file.id}${safeExtension(file.metadata.title)}`;
-
 // A regular file with content is treated as complete. A zero-byte file is not:
 // it is the shape an aborted write leaves and must be re-fetched.
 const isPresent = (path: string): boolean => {
@@ -181,34 +195,29 @@ const copyAtomic = async (src: string, dest: string): Promise<void> => {
     }
 };
 
-// A process-ID check: signal 0 delivers nothing and only reports whether the
+// Put an original the library returned at `dest` in originals/, where a fresh
-// process exists. EPERM means it exists but belongs to another user.
+// fetch already wrote it. A live photo's image and video go beside `dest`: when
-const isRunning = (pid: number): boolean => {
+// they came from the cache they are copied, after removing whatever was at
-    try {
+// `dest` (an earlier version's ZIP of the two). Then the JSON file naming them
-        process.kill(pid, 0);
+// is written, which is what makes the live photo count as stored.
-        return true;
+const placeOriginal = async (
-    } catch (err) {
+    file: EnteFile,
-        return (err as NodeJS.ErrnoException).code === "EPERM";
+    dest: string,
-    }
+    got: { path: string; videoPath?: string },
-};
+): Promise<void> => {
-
+    if (got.videoPath === undefined) {
-// Delete the temp files `copyAtomic` leaves behind when a backup is killed
+        await copyAtomic(got.path, dest);
-// before its rename. Only files whose process is no longer running are
-// removed, so a backup running at the same time keeps its own. A reused
-// process ID can only keep a leftover a while longer, never remove a live one.
-const removeLeftoverTempFiles = (dir: string): void => {
-    let names: string[];
-    try {
-        names = readdirSync(dir);
-    } catch {
         return;
     }
-    for (const name of names) {
+    const originalsDir = dirname(dest);
-        const match = /^\.quak-backup-.*-(\d+)-[0-9a-z]*\.tmp$/.exec(name);
+    const path = join(originalsDir, basename(got.path));
-        if (match && !isRunning(Number(match[1]))) {
+    const videoPath = join(originalsDir, basename(got.videoPath));
-            rmSync(join(dir, name), { force: true });
+    if (got.path !== path) {
-        }
+        await rm(dest, { force: true });
+        await copyAtomic(got.path, path);
+        await copyAtomic(got.videoPath, videoPath);
     }
+    await writeLivePhoto(originalsDir, file.id, { path, videoPath });
 };
 
 // Ensure `linkPath` is a symlink to `target`, rebuilding a missing, wrong, or
@@ -227,6 +236,111 @@ const rebuildSymlink = (linkPath: string, target: string): void => {
     symlinkSync(target, linkPath);
 };
 
+// The on-disk names for the entries of one directory, in entry order. Each
+// name is used as is unless another entry would get the same name, ignoring
+// case (two names that differ only in case are one entry on a case-insensitive
+// file system); then every entry sharing it gets ` (<id>)`, before the
+// extension when `beforeExtension` is set. A name with an ID added can match
+// another entry's own name (`IMG (6).JPG`), so this repeats until no name is
+// shared. IDs are stable, so the names are too.
+const uniqueNames = (
+    entries: { id: number; name: string }[],
+    beforeExtension: boolean,
+): string[] => {
+    const withID = (id: number, name: string): string => {
+        const ext = beforeExtension ? extname(name) : "";
+        const stem = name.slice(0, name.length - ext.length);
+        return `${stem} (${id})${ext}`;
+    };
+    const names = entries.map((e) => e.name);
+    const suffixed = new Set<number>();
+    for (;;) {
+        const counts = new Map<string, number>();
+        for (const name of names) {
+            const key = name.toLowerCase();
+            counts.set(key, (counts.get(key) ?? 0) + 1);
+        }
+        let changed = false;
+        for (const [i, { id, name }] of entries.entries()) {
+            if (suffixed.has(i)) continue;
+            if (counts.get(name.toLowerCase()) === 1) continue;
+            names[i] = withID(id, name);
+            suffixed.add(i);
+            changed = true;
+        }
+        if (!changed) return names;
+    }
+};
+
+// The links a file gets in its album's folder: one named after its title, to
+// its original if that is stored. A stored live photo gets two, to its image
+// and its video, each named after the title with that file's extension.
+const linksFor = (
+    file: EnteFile,
+    stored: { path: string; videoPath?: string } | undefined,
+): { id: number; name: string; file: EnteFile; target?: string }[] => {
+    const name = sanitizeFileName(file.metadata.title, `file-${file.id}`);
+    if (stored?.videoPath === undefined) {
+        return [{ id: file.id, name, file, target: stored?.path }];
+    }
+    return [stored.path, stored.videoPath].map((target) => ({
+        id: file.id,
+        name: withExtension(name, extname(target)),
+        file,
+        target,
+    }));
+};
+
+// Remove the symlinks in the album directory `dir` that point into
+// `originalsDir` and are not named in `keep`. Nothing else in the directory
+// is touched: anything else there was put there by the user.
+const removeStaleLinks = (
+    dir: string,
+    keep: Set<string>,
+    originalsDir: string,
+): void => {
+    const target = relative(dir, originalsDir);
+    for (const name of readdirSync(dir)) {
+        if (keep.has(name)) continue;
+        const path = join(dir, name);
+        if (
+            lstatSync(path).isSymbolicLink() &&
+            dirname(readlinkSync(path)) === target
+        ) {
+            rmSync(path);
+        }
+    }
+};
+
+// Remove the directories under `collectionsDir` that an earlier run wrote for
+// an album that is gone or renamed: a directory not named in `current` with a
+// `<name>.json` beside it holding an album ID, which is what a run writes. Its
+// symlinks into originals/ are removed; if that leaves it empty, it and its
+// JSON are deleted, otherwise both stay for what the user put there.
+const removeStaleAlbumDirs = (
+    collectionsDir: string,
+    current: Set<string>,
+    originalsDir: string,
+): void => {
+    for (const entry of readdirSync(collectionsDir, { withFileTypes: true })) {
+        if (!entry.isDirectory() || current.has(entry.name)) continue;
+        const jsonPath = join(collectionsDir, `${entry.name}.json`);
+        try {
+            const album = JSON.parse(readFileSync(jsonPath, "utf-8")) as {
+                id?: unknown;
+            };
+            if (typeof album.id !== "number") continue;
+        } catch {
+            continue;
+        }
+        const dir = join(collectionsDir, entry.name);
+        removeStaleLinks(dir, new Set(), originalsDir);
+        if (readdirSync(dir).length > 0) continue;
+        rmdirSync(dir);
+        rmSync(jsonPath);
+    }
+};
+
 const loadLedger = (path: string): Map<number, FailureEntry> => {
     const ledger = new Map<number, FailureEntry>();
     try {
@@ -303,9 +417,10 @@ export const runBackup = async (
 
     // Collections in scope, and the distinct files across them (a file shared
     // by two albums is one original).
-    const collections = lib
+    const allCollections = lib.listCollections();
-        .listCollections()
+    const collections = allCollections.filter((c) =>
-        .filter((c) => (only ? only.has(c.name) : true));
+        only ? only.has(c.name) : true,
+    );
     const collectionName = new Map<number, string>();
     for (const c of collections) collectionName.set(c.id, c.name);
 
@@ -355,15 +470,21 @@ export const runBackup = async (
     // tree; a present file is left as is.
     if (includeOriginals) {
         for (const [fileID, file] of distinct) {
-            const dest = join(originalsDir, originalName(file));
+            if (storedOriginal(originalsDir, file) !== undefined) {
-            if (isPresent(dest)) {
                 skipped++;
                 continue;
             }
+            const dest = join(originalsDir, originalName(file));
             try {
                 log(`Fetching original ${file.metadata.title} (${fileID})...`);
-                const { path } = await lib.original(fileID);
+                // A fetched original is written straight to `dest` (a live
-                await copyAtomic(path, dest);
+                // photo beside it); only one that was already cached elsewhere
+                // is copied.
+                await placeOriginal(
+                    file,
+                    dest,
+                    await lib.original(fileID, dest),
+                );
                 downloaded++;
             } catch (err) {
                 log(
@@ -399,38 +520,67 @@ export const runBackup = async (
     // every present original (this repairs stale ones).
     if (includeOriginals) {
         for (const [fileID, file] of distinct) {
-            const orig = join(originalsDir, originalName(file));
+            if (storedOriginal(originalsDir, file) !== undefined) {
-            if (isPresent(orig)) {
                 writeSidecar(join(originalsDir, `${fileID}.json`), file);
             }
         }
     }
 
-    // Then the per-collection symlink trees and JSON.
+    // Then the per-collection symlink trees and JSON. Directory names are
+    // chosen across every album, not just those in scope, so a scoped run
+    // names an album the same as a full one and never takes the directory of
+    // an album it skipped. Stale entries are removed before anything is
+    // rebuilt, so on a case-insensitive file system removing an old name can
+    // never remove the new one.
+    const dirNames = uniqueNames(
+        allCollections.map((c) => ({
+            id: c.id,
+            name: sanitizeFileName(c.name, `collection-${c.id}`),
+        })),
+        false,
+    );
+    const albumDirNames = new Map(
+        allCollections.map((c, i) => [c.id, dirNames[i]!]),
+    );
+    try {
+        removeStaleAlbumDirs(collectionsDir, new Set(dirNames), originalsDir);
+    } catch (err) {
+        log(`FAILED removing old album directories: ${errorMessage(err)}`);
+    }
+
     for (const c of collections) {
-        const colDirName = sanitizeFileName(c.name, `collection-${c.id}`);
+        const colDirName = albumDirNames.get(c.id)!;
         const colDir = join(collectionsDir, colDirName);
         mkdirSync(colDir, { recursive: true });
 
         const files = filesByCollection.get(c.id) ?? [];
-        const metaFiles: { id: number; metadata: EnteFile["metadata"] }[] = [];
+        const links = files.flatMap((f) =>
-        for (const file of files) {
+            linksFor(f, storedOriginal(originalsDir, f)),
-            metaFiles.push({ id: file.id, metadata: file.metadata });
-            if (!includeOriginals) continue;
-            const orig = join(originalsDir, originalName(file));
-            if (!isPresent(orig)) continue;
-            const linkName = sanitizeFileName(
-                file.metadata.title,
-                `file-${file.id}`,
         );
-            const linkPath = join(colDir, linkName);
+        const linkNames = uniqueNames(links, true);
         try {
-                rebuildSymlink(linkPath, relative(colDir, orig));
+            removeStaleLinks(colDir, new Set(linkNames), originalsDir);
+        } catch (err) {
+            log(`FAILED removing old links in ${c.name}: ${errorMessage(err)}`);
+        }
+
+        const metaFiles = files.map((f) => ({
+            id: f.id,
+            metadata: f.metadata,
+        }));
+        for (const [i, link] of links.entries()) {
+            if (!includeOriginals || link.target === undefined) continue;
+            const linkName = linkNames[i]!;
+            try {
+                rebuildSymlink(
+                    join(colDir, linkName),
+                    relative(colDir, link.target),
+                );
             } catch (err) {
                 log(
                     `FAILED symlink ${c.name}/${linkName}: ${errorMessage(err)}`,
                 );
-                recordFailure(file, c.name, err);
+                recordFailure(link.file, c.name, err);
             }
         }
 

src/cli-commands.ts

@@ -2,22 +2,30 @@
 //
 // Each command takes its options and a `CliContext` and resolves to the exit
 // code; a thrown error is left to the caller. Nothing here calls
-// `process.exit`: `bin/quak.ts` wires these to the command line and exits with
+// `process.exit`: `bin/quak.ts` wires these to the command line, and `run` in
-// the returned code once output has drained. Output must stay byte-identical
+// `cli-run.ts` prints a thrown error as one line and exits once output has
-// (see `cli-output.ts`).
+// drained. Output must stay byte-identical (see `cli-output.ts`).
 
-import { input, password as passwordPrompt } from "@inquirer/prompts";
 import {
     copyFileSync,
     existsSync,
     mkdirSync,
+    statSync,
     unlinkSync,
     writeFileSync,
 } from "node:fs";
-import { join } from "node:path";
+import { extname, join } from "node:path";
-import { Client, type ClientSnapshot } from "./client.js";
+import {
+    type Client,
+    type ClientSnapshot,
+    type LoginOptions,
+} from "./client.js";
 import { init } from "./crypto/index.js";
-import { Library, type LibraryClient } from "./library/index.js";
+import {
+    defaultCacheDirectory,
+    Library,
+    type LibraryClient,
+} from "./library/index.js";
 import {
     fileListRow,
     fileListLine,
@@ -25,6 +33,7 @@ import {
     thumbnailName,
 } from "./cli-output.js";
 import { freshCollections, freshFiles, freshFile } from "./cli-read.js";
+import { withExtension } from "./filename.js";
 import { runMetadataBackup } from "./metadata-backup.js";
 import { listMissingThumbnails, fixMissingThumbnails } from "./thumbnails.js";
 
@@ -39,6 +48,12 @@ export interface CliContext {
     // Reads the session file into a client, or null when there is none. The
     // CLI passes `loadSession` from `cli-session.ts`; tests pass a fake client.
     loadSession: (path: string) => Client | null;
+    // Used by `login` only. The CLI passes `Client.login` and terminal
+    // prompts; tests pass fakes.
+    login: (opts: LoginOptions) => Promise<Client>;
+    prompt: (message: string) => Promise<string>;
+    // Like `prompt`, but the answer is masked as it is typed.
+    promptSecret: (message: string) => Promise<string>;
 }
 
 const sessionPath = (ctx: CliContext): string =>
@@ -104,24 +119,19 @@ const openReadLibrary = (ctx: CliContext, client: Client): Promise<Library> =>
         precacheOriginals: false,
     });
 
-const prompt = async (message: string): Promise<string> => input({ message });
-
-const promptSecret = async (message: string): Promise<string> =>
-    passwordPrompt({ message, mask: true });
-
 export const loginCommand = async (ctx: CliContext): Promise<number> => {
     await init();
-    const email = process.env.QUAK_EMAIL ?? (await prompt("Email"));
+    const email = process.env.QUAK_EMAIL ?? (await ctx.prompt("Email"));
     const password =
-        process.env.QUAK_PASSWORD ?? (await promptSecret("Password"));
+        process.env.QUAK_PASSWORD ?? (await ctx.promptSecret("Password"));
 
     ctx.stderr.write("Authenticating...\n");
     try {
-        const client = await Client.login({
+        const client = await ctx.login({
             email,
             password,
-            totp: async () => prompt("TOTP code: "),
+            totp: async () => ctx.prompt("TOTP code: "),
-            emailOTP: async () => prompt("Email verification code: "),
+            emailOTP: async () => ctx.prompt("Email verification code: "),
         });
 
         saveSession(ctx.sessionDir, client.toJSON());
@@ -146,14 +156,43 @@ export const whoamiCommand = async (ctx: CliContext): Promise<number> => {
     return 0;
 };
 
+// Ends the session on the server, then deletes the session file even when that
+// failed, and exits 1 if it did. The cache is left in place; the user is told
+// where it is.
 export const logoutCommand = async (ctx: CliContext): Promise<number> => {
-    if (existsSync(sessionPath(ctx))) {
+    const path = sessionPath(ctx);
-        unlinkSync(sessionPath(ctx));
+    if (!existsSync(path)) {
-        ctx.stderr.write("Session deleted.\n");
-    } else {
         ctx.stderr.write("No session found.\n");
-    }
         return 0;
+    }
+    await init();
+    let cacheDir = ctx.cacheDir;
+    let failure: string | undefined;
+    try {
+        const client = ctx.loadSession(path);
+        if (client) {
+            cacheDir ??= defaultCacheDirectory(client.whoami().userID);
+            await client.logoutOnServer();
+            client.logout();
+        }
+    } catch (err) {
+        failure = err instanceof Error ? err.message : String(err);
+    }
+    unlinkSync(path);
+    if (failure === undefined) {
+        ctx.stderr.write("Session ended on the server.\n");
+    } else {
+        ctx.stderr.write(
+            `Could not end the session on the server: ${failure}\n`,
+        );
+    }
+    ctx.stderr.write("Session deleted.\n");
+    if (cacheDir !== undefined) {
+        ctx.stderr.write(
+            `Cache directory ${cacheDir} still holds decrypted data; delete it to remove that data.\n`,
+        );
+    }
+    return failure === undefined ? 0 : 1;
 };
 
 export const collectionsCommand = async (
@@ -268,9 +307,30 @@ export const getCommand = async (
         // Default name is the file's own title, as the pre-library CLI used
         // (not the editedName-preferring projection title) (issue #52).
         const outPath = opts.out ?? originalName(file);
+        if (result.videoPath === undefined) {
             copyFileSync(result.path, outPath);
             ctx.stderr.write(`${result.bytes} bytes -> ${outPath}\n`);
             return 0;
+        }
+        // A live photo is written as its image and its video, each named after
+        // the title with its own extension, as Ente's clients name them. With
+        // --out, the image goes there and the video beside it.
+        const imageOut =
+            opts.out ?? withExtension(outPath, extname(result.path));
+        const videoOut = withExtension(outPath, extname(result.videoPath));
+        if (imageOut.toLowerCase() === videoOut.toLowerCase()) {
+            ctx.stderr.write(
+                `File ${fileID} is a live photo, and its video would also be written to ${imageOut}\n`,
+            );
+            return 1;
+        }
+        copyFileSync(result.path, imageOut);
+        copyFileSync(result.videoPath, videoOut);
+        ctx.stderr.write(
+            `${result.bytes} bytes -> ${imageOut}\n` +
+                `${statSync(videoOut).size} bytes -> ${videoOut}\n`,
+        );
+        return 0;
     } finally {
         await lib.close();
     }
@@ -323,11 +383,14 @@ export const backupMetadataCommand = async (
     if (!client) return 1;
     const lib = await openReadLibrary(ctx, client);
     try {
-        await runMetadataBackup(lib, client, dir, {
+        // Refresh first so the dump holds current account state, not what the
+        // cache last held; a failed refresh throws.
+        await lib.fresh();
+        const { failedMLBatches } = await runMetadataBackup(lib, client, dir, {
             exif: opts.exif || opts.all,
             onProgress: (msg) => ctx.stderr.write(msg + "\n"),
         });
-        return 0;
+        return failedMLBatches > 0 ? 1 : 0;
     } finally {
         await lib.close();
     }
@@ -343,10 +406,14 @@ export const backupCommand = async (
     if (!client) return 1;
 
     ctx.stderr.write("Starting backup...\n");
+    // The precache is off: the backup fetches what it needs, and must not
+    // also fill the cache with every thumbnail and the recent originals.
     const lib = await Library.open({
         client,
         downloadDirectory: dir,
         cacheDirectory: ctx.cacheDir,
+        precacheThumbnails: false,
+        precacheOriginals: false,
     });
     try {
         const result = await lib.backup({
@@ -389,6 +456,9 @@ export const listMissingThumbnailsCommand = async (
     if (!client) return 1;
     const lib = await openReadLibrary(ctx, client);
     try {
+        // Refresh first so files added since the cache was written are
+        // checked; a failed refresh throws.
+        await lib.fresh();
         const missing = await listMissingThumbnails(lib, client, (msg) => {
             if (!opts.json) ctx.stderr.write(msg + "\n");
         });
@@ -424,6 +494,9 @@ export const fixMissingThumbnailsCommand = async (
     if (!client) return 1;
     const lib = await openReadLibrary(ctx, client);
     try {
+        // Refresh first so files added since the cache was written are found;
+        // a failed refresh throws.
+        await lib.fresh();
         let fileIDs: number[];
         if (opts.file && opts.file.length > 0) {
             fileIDs = opts.file.map(Number).filter(Number.isFinite);
@@ -462,7 +535,7 @@ export const fixMissingThumbnailsCommand = async (
             ctx.stderr.write(`  Skipped: ${skipped}\n`);
             ctx.stderr.write(`  Failed:  ${failed}\n`);
             if (skipped > 0) {
-                ctx.stderr.write("\nSkipped (unsupported format):\n");
+                ctx.stderr.write("\nSkipped:\n");
                 for (const r of results.filter((r) => r.status === "skipped")) {
                     ctx.stderr.write(
                         `  ${r.fileID}\t${r.title}\t${r.reason}\n`,

src/cli-run.ts

@@ -0,0 +1,36 @@
+// Runs one CLI command for `bin/quak.ts` and exits with its code.
+
+import type { Writable } from "node:stream";
+
+// Run a command and exit with its code once stdout/stderr have drained.
+// Exiting before the drain can truncate piped output, and the library can keep
+// the event loop alive after a command returns, so a plain return could hang.
+// An error the command throws is printed as one `quak: MESSAGE` line, without
+// the stack trace, and exits 1.
+export const run = async (
+    command: Promise<number>,
+    stdout: Writable,
+    stderr: Writable,
+    exit: (code: number) => void,
+): Promise<void> => {
+    let code: number;
+    try {
+        code = await command;
+    } catch (err) {
+        stderr.write(
+            `quak: ${err instanceof Error ? err.message : String(err)}\n`,
+        );
+        code = 1;
+    }
+    const pending = [stdout, stderr].filter((s) => s.writableLength > 0);
+    if (pending.length === 0) {
+        exit(code);
+        return;
+    }
+    let remaining = pending.length;
+    for (const s of pending) {
+        s.once("drain", () => {
+            if (--remaining === 0) exit(code);
+        });
+    }
+};

src/client.ts

@@ -217,6 +217,14 @@ export class Client {
         };
     }
 
+    // Ends this client's session on the server (`POST /users/logout`), so the
+    // token stops working everywhere, including in any saved copy of it. This
+    // client is left as it was; call `logout()` to clear it.
+    async logoutOnServer(): Promise<void> {
+        this.assertLoggedIn();
+        await this.api.postJSON("/users/logout", {});
+    }
+
     // Zeroes the key buffers in place, so any copy of the reference held
     // elsewhere is wiped too. Every method checks `assertLoggedIn` before
     // touching the keys, so nothing decrypts with the zeroed keys.

src/crypto/hash.ts

@@ -0,0 +1,22 @@
+import sodium, { type StateAddress } from "libsodium-wrappers-sumo";
+import { toBase64 } from "./encoding.js";
+
+// The content hash an uploading client records in a file's metadata: unkeyed
+// BLAKE2b with a 64-byte output over the original's bytes, fed in chunks, as
+// standard base64 with padding. Named after the upstream client's functions.
+// The output length is read at call time for the same reason as
+// `streamTagFinal` in stream.ts: libsodium sets its constants only once ready.
+
+export const chunkHashInit = (): StateAddress =>
+    sodium.crypto_generichash_init(null, sodium.crypto_generichash_BYTES_MAX);
+
+export const chunkHashUpdate = (state: StateAddress, chunk: Uint8Array): void =>
+    sodium.crypto_generichash_update(state, chunk);
+
+export const chunkHashFinal = (state: StateAddress): string =>
+    toBase64(
+        sodium.crypto_generichash_final(
+            state,
+            sodium.crypto_generichash_BYTES_MAX,
+        ),
+    );

src/crypto/index.ts

@@ -7,6 +7,7 @@ export {
 } from "./encoding.js";
 export { deriveKEK, deriveLoginSubkey } from "./kdf.js";
 export { decryptBox, decryptSealed } from "./box.js";
+export { chunkHashFinal, chunkHashInit, chunkHashUpdate } from "./hash.js";
 export {
     decryptBlob,
     encryptBlob,

src/download/index.ts

@@ -1,8 +1,13 @@
-import { randomUUID } from "node:crypto";
+import { randomBytes } from "node:crypto";
+import { readdirSync, rmSync } from "node:fs";
 import { open, rename, rm } from "node:fs/promises";
 import type { FileHandle } from "node:fs/promises";
 import { dirname, join } from "node:path";
+import { Unzip, UnzipInflate } from "fflate";
 import {
+    chunkHashFinal,
+    chunkHashInit,
+    chunkHashUpdate,
     fromBase64,
     initStreamPull,
     pullStreamChunk,
@@ -11,14 +16,18 @@ import {
     streamTagFinal,
 } from "../crypto/index.js";
 import { TruncatedStreamError } from "../errors.js";
-import { sanitizeFileName } from "../filename.js";
+import { safeExtension, sanitizeFileName, withExtension } from "../filename.js";
 import { withRetry } from "../retry.js";
 import type { ApiClient } from "../api/client.js";
 import type { EnteFile } from "../model/types.js";
 
 export interface DownloadResult {
+    // Where the file was written. A live photo is written as two files, its
+    // image here and its video at `videoPath` (see `decryptLivePhoto`).
     path: string;
+    // The decrypted length; for a live photo, that of the ZIP it arrives as.
     bytesWritten: number;
+    videoPath?: string;
 }
 
 // Fired as decrypted plaintext accumulates, with the running total of
@@ -40,9 +49,9 @@ const ENC_CHUNK_SIZE = STREAM_CHUNK_SIZE + STREAM_CHUNK_OVERHEAD;
 // new: a body cut short still decrypts and authenticates up to its last whole
 // chunk, so the absence of TAG_FINAL is the sole evidence it was cut short, and
 // this throws rather than let a caller keep a short file. The sink has already
-// seen those chunks by then; the caller (`decryptToTemp`) stages them in a temp
+// seen those chunks by then; the callers (`decryptToTemp`, `decryptLivePhoto`)
-// file that is renamed into place only on a clean return, so a throw leaves
+// stage them in temp files that are renamed into place only on a clean return,
-// nothing on disk.
+// so a throw leaves nothing on disk.
 const streamDecrypt = async (
     stream: ReadableStream<Uint8Array>,
     header: Uint8Array,
@@ -176,6 +185,45 @@ export const fsyncPath = async (path: string): Promise<void> => {
     }
 };
 
+// A process-ID check: signal 0 delivers nothing and only reports whether the
+// process exists. EPERM means it exists but belongs to another user.
+const isRunning = (pid: number): boolean => {
+    try {
+        process.kill(pid, 0);
+        return true;
+    } catch (err) {
+        return (err as NodeJS.ErrnoException).code === "EPERM";
+    }
+};
+
+// Delete the temp files a killed process left in `dir`: the writer's
+// `.quak-<pid>-<random>.tmp` and the backup copy's
+// `.quak-backup-<name>-<pid>-<random>.tmp`. Only files whose process is no
+// longer running are removed, so another process writing into the same
+// directory keeps its own. A reused process ID can only keep a leftover a while
+// longer, never remove a live one.
+export const removeLeftoverTempFiles = (dir: string): void => {
+    let names: string[];
+    try {
+        names = readdirSync(dir);
+    } catch {
+        return;
+    }
+    for (const name of names) {
+        const match = /^\.quak-(?:.*-)?(\d+)-[0-9a-z]*\.tmp$/.exec(name);
+        if (match && !isRunning(Number(match[1]))) {
+            rmSync(join(dir, name), { force: true });
+        }
+    }
+};
+
+// A new temp file name in `dir`. The random suffix keeps concurrent downloads
+// of the same destination from stepping on each other's temporary file; the
+// process ID lets `removeLeftoverTempFiles` tell a leftover from a write in
+// progress.
+const tempPathIn = (dir: string): string =>
+    join(dir, `.quak-${process.pid}-${randomBytes(16).toString("hex")}.tmp`);
+
 // Stage a write to `destination` atomically and durably, then rename it into
 // place. `fill` writes the contents into the open temp file handle — either the
 // whole buffer at once (`writeAtomic`) or chunk by chunk as they decrypt
@@ -200,9 +248,7 @@ const stageAtomic = async (
     fill: (handle: FileHandle) => Promise<void>,
 ): Promise<void> => {
     const dir = dirname(destination);
-    // The random suffix keeps concurrent downloads of the same destination
+    const tmpPath = tempPathIn(dir);
-    // from stepping on each other's temporary file.
-    const tmpPath = join(dir, `.quak-${randomUUID()}.tmp`);
     try {
         const handle = await open(tmpPath, "w");
         try {
@@ -237,42 +283,201 @@ export const writeAtomic = async (
 ): Promise<void> =>
     stageAtomic(destination, (handle) => handle.writeFile(plaintext));
 
+// Refuse an original whose bytes do not hash to what its uploader recorded.
+// The error is not retried.
+const checkHash = (file: EnteFile, actual: string): void => {
+    if (actual !== file.metadata.hash) {
+        throw new Error(
+            `download: file ${file.id}: content hash ${actual} does not match the hash its uploader recorded, ${file.metadata.hash}`,
+        );
+    }
+};
+
 // Decrypt `stream` straight to `destination`, one plaintext chunk at a time,
 // under the atomic writer's temp-then-rename discipline. Memory stays bounded
 // by the chunk size: each decrypted chunk is written to the temp file and
 // dropped. The rename happens only after the stream authenticates as terminated
 // on TAG_FINAL; a truncated stream throws and leaves the destination untouched.
 // Returns the plaintext length written.
+//
+// `original` is the file whose original this is (none for a thumbnail, which
+// has no recorded hash). When its metadata has a hash, the decrypted bytes are
+// hashed as they stream and must match it, or nothing is stored.
 const decryptToTemp = async (
     destination: string,
     stream: ReadableStream<Uint8Array>,
     header: Uint8Array,
     key: Uint8Array,
     onProgress?: ProgressCallback,
+    original?: EnteFile,
 ): Promise<number> => {
+    const hash =
+        original?.metadata.hash === undefined ? undefined : chunkHashInit();
     let bytesWritten = 0;
-    try {
     await stageAtomic(destination, async (handle) => {
         bytesWritten = await streamDecrypt(
             stream,
             header,
             key,
             async (plaintext) => {
+                if (hash !== undefined) chunkHashUpdate(hash, plaintext);
                 await handle.write(plaintext);
             },
             onProgress,
         );
+        if (original !== undefined && hash !== undefined) {
+            checkHash(original, chunkHashFinal(hash));
+        }
     });
+    return bytesWritten;
+};
+
+// One of the two parts of a live photo being unpacked: the ZIP entry whose
+// name starts with `kind`, written to its own temp file.
+interface LivePhotoPart {
+    kind: "image" | "video";
+    tmpPath: string;
+    handle: FileHandle;
+    hash: ReturnType<typeof chunkHashInit>;
+    // Decompressed bytes not yet written.
+    pending: Uint8Array[];
+    // The entry's extension, set once all of the entry has been read.
+    ext?: string;
+}
+
+const openPart = async (
+    kind: "image" | "video",
+    dir: string,
+): Promise<LivePhotoPart> => {
+    const tmpPath = tempPathIn(dir);
+    const handle = await open(tmpPath, "w");
+    return { kind, tmpPath, handle, hash: chunkHashInit(), pending: [] };
+};
+
+// A live photo arrives as a ZIP of its image and its video. Ente's clients
+// name the entries `image.<ext>` and `video.<ext>`, and like the upstream
+// client's decoder this takes the first entries whose names start with `image`
+// and `video`. It is written unpacked: each part is named `destination` with
+// the extension replaced by its own entry's, and the two must differ ignoring
+// case. When the file records a hash, `<imageHash>:<videoHash>` must match it,
+// each over that part's own bytes. Only then is whatever was at `destination`
+// removed and the image, then the video, renamed into place; on any failure
+// neither is stored.
+//
+// The ZIP is chosen by its uploader and may expand enormously, so each part is
+// written as it decompresses and never held. fflate's `Unzip` inflates each
+// push in one piece before `push` returns, and deflate expands at most about
+// 1000-fold, so the ZIP is pushed in 4 KiB slices, keeping each decompressed
+// piece near 4 MiB, one plaintext chunk, and each piece is written before the
+// next slice is pushed. Every entry is started, even one that is not kept,
+// because fflate keeps an unstarted entry's data in memory.
+const decryptLivePhoto = async (
+    destination: string,
+    stream: ReadableStream<Uint8Array>,
+    header: Uint8Array,
+    key: Uint8Array,
+    onProgress: ProgressCallback | undefined,
+    file: EnteFile,
+): Promise<DownloadResult> => {
+    const sliceSize = 4096;
+    const dir = dirname(destination);
+    const fail = (message: string, cause?: unknown): Error =>
+        new Error(`download: file ${file.id}: ${message}`, { cause });
+    const parts: LivePhotoPart[] = [];
+    try {
+        const image = await openPart("image", dir);
+        parts.push(image);
+        const video = await openPart("video", dir);
+        parts.push(video);
+
+        const claimed = new Set<LivePhotoPart>();
+        const unzip = new Unzip((entry) => {
+            const part = parts.find(
+                (p) => !claimed.has(p) && entry.name.startsWith(p.kind),
+            );
+            if (part !== undefined) claimed.add(part);
+            entry.ondata = (err, data, final) => {
+                if (err) throw err;
+                if (part === undefined) return;
+                chunkHashUpdate(part.hash, data);
+                part.pending.push(data);
+                if (final) part.ext = safeExtension(entry.name);
+            };
+            entry.start();
+        });
+        unzip.register(UnzipInflate);
+        const push = async (
+            data: Uint8Array,
+            final: boolean,
+        ): Promise<void> => {
+            // fflate reports a bad ZIP by throwing, sometimes a TypeError,
+            // which the retry would take for a network failure; a bad ZIP is
+            // never retried.
+            try {
+                unzip.push(data, final);
             } catch (err) {
-        // Cancel the body so its connection is closed now rather than held
+                throw fail("live photo is not a readable ZIP", err);
-        // until the stream is garbage collected. A backup run carries on past
+            }
-        // a failed file, so without this every failure would hold a socket.
+            for (const part of parts) {
-        // This covers every failure, including a temp file that cannot be
+                for (const piece of part.pending)
-        // opened and a header that is rejected before the body is read.
+                    await part.handle.write(piece);
-        await stream.cancel(err).catch(() => undefined);
+                part.pending = [];
+            }
+        };
+
+        const bytesWritten = await streamDecrypt(
+            stream,
+            header,
+            key,
+            async (plaintext) => {
+                for (let i = 0; i < plaintext.length; i += sliceSize) {
+                    await push(plaintext.subarray(i, i + sliceSize), false);
+                }
+            },
+            onProgress,
+        );
+        await push(new Uint8Array(0), true);
+
+        if (image.ext === undefined || video.ext === undefined) {
+            throw fail(
+                "live photo ZIP does not hold both an image and a video",
+            );
+        }
+        if (file.metadata.hash !== undefined) {
+            checkHash(
+                file,
+                `${chunkHashFinal(image.hash)}:${chunkHashFinal(video.hash)}`,
+            );
+        }
+        if (image.ext.toLowerCase() === video.ext.toLowerCase()) {
+            throw fail(
+                `live photo's image and video have the same extension, ${video.ext}`,
+            );
+        }
+        const path = withExtension(destination, image.ext);
+        const videoPath = withExtension(destination, video.ext);
+        for (const part of parts) {
+            await part.handle.sync();
+            await part.handle.close();
+        }
+        await rm(destination, { force: true });
+        await rename(image.tmpPath, path);
+        try {
+            await rename(video.tmpPath, videoPath);
+        } catch (err) {
+            await rm(path, { force: true }).catch(() => undefined);
+            throw err;
+        }
+        await fsyncPath(dir);
+        return { path, bytesWritten, videoPath };
+    } catch (err) {
+        // Best-effort cleanup, as in `stageAtomic`.
+        for (const part of parts) {
+            await part.handle.close().catch(() => undefined);
+            await rm(part.tmpPath, { force: true }).catch(() => undefined);
+        }
         throw err;
     }
-    return bytesWritten;
 };
 
 // Fetch a stream and decrypt it to `destination`, retrying the whole sequence.
@@ -302,12 +507,45 @@ const fetchAndDecrypt = async (
     key: Uint8Array,
     destination: string,
     onProgress?: ProgressCallback,
-): Promise<number> =>
+    original?: EnteFile,
+): Promise<DownloadResult> =>
     withRetry(async () => {
         const stream = await openStream();
-        return decryptToTemp(destination, stream, header, key, onProgress);
+        try {
+            if (original?.metadata.fileType === "livePhoto") {
+                return await decryptLivePhoto(
+                    destination,
+                    stream,
+                    header,
+                    key,
+                    onProgress,
+                    original,
+                );
+            }
+            const bytesWritten = await decryptToTemp(
+                destination,
+                stream,
+                header,
+                key,
+                onProgress,
+                original,
+            );
+            return { path: destination, bytesWritten };
+        } catch (err) {
+            // Cancel the body so its connection is closed now rather than held
+            // until the stream is garbage collected. A backup run carries on
+            // past a failed file, so without this every failure would hold a
+            // socket. This covers every failure, including a temp file that
+            // cannot be opened and a header that is rejected before the body
+            // is read.
+            await stream.cancel(err).catch(() => undefined);
+            throw err;
+        }
     }, api.getRetryOptions());
 
+// Write `file`'s original to `outPath`. A live photo is written as its image
+// and its video beside `outPath` instead, and whatever was at `outPath` is
+// removed (see `decryptLivePhoto`).
 export const downloadFile = async (
     api: ApiClient,
     file: EnteFile,
@@ -319,15 +557,15 @@ export const downloadFile = async (
     const resolvedPath =
         outPath ?? sanitizeFileName(file.metadata.title, `file-${file.id}`);
     const header = fromBase64(file.file.decryptionHeader);
-    const bytesWritten = await fetchAndDecrypt(
+    return fetchAndDecrypt(
         api,
         () => api.getFileStream(file.id, { retry: false }),
         header,
         file.key,
         resolvedPath,
         onProgress,
+        file,
     );
-    return { path: resolvedPath, bytesWritten };
 };
 
 export const downloadThumbnail = async (
@@ -340,7 +578,7 @@ export const downloadThumbnail = async (
         outPath ??
         `thumb_${sanitizeFileName(file.metadata.title, `file-${file.id}`)}`;
     const header = fromBase64(file.thumbnail.decryptionHeader);
-    const bytesWritten = await fetchAndDecrypt(
+    return fetchAndDecrypt(
         api,
         () => api.getThumbnailStream(file.id, { retry: false }),
         header,
@@ -348,5 +586,4 @@ export const downloadThumbnail = async (
         resolvedPath,
         onProgress,
     );
-    return { path: resolvedPath, bytesWritten };
 };

src/filename.ts

@@ -35,3 +35,7 @@ export const safeExtension = (title: string): string => {
     const ext = extname(title);
     return /^\.[A-Za-z0-9]+$/.test(ext) ? ext : ".bin";
 };
+
+// `name` with its extension, if it has one, replaced by `ext` (".mov").
+export const withExtension = (name: string, ext: string): string =>
+    name.slice(0, name.length - extname(name).length) + ext;

src/library/content.ts

@@ -1,11 +1,13 @@
 // The on-disk content and thumbnail cache keyed by fileID (issue #46).
 //
 // Layout under `cacheDirectory`: `originals/<fileID>.<ext>` and
-// `thumbnails/<fileID>.<ext>`, flat directories at 0700 with files at 0600.
+// `thumbnails/<fileID>.<ext>`, flat directories at 0700 with files at 0600. A
-// Content appears only by the streaming atomic writer's rename (the download
+// live photo's original is two files, its image and its video, with
-// layer, #40), so a file that exists is whole — "present means complete". The
+// `originals/<fileID>.livephoto.json` naming them. Content appears only by the
-// directory listing taken at `open()` is the record of what is cached, and the
+// streaming atomic writer's rename (the download layer, #40), so a file that
-// orphan temp files a crashed write may have left are reaped there.
+// exists is whole — "present means complete". The directory listing taken at
+// `open()` is the record of what is cached, and the orphan temp files a crashed
+// write may have left are reaped there.
 //
 // A fetch goes through the shared request pools (#45): the content pool for
 // originals, the thumbnail pool for thumbnails. The pool limits concurrency,
@@ -15,14 +17,14 @@
 // Integrity. The reused streaming decrypt is the enforced guarantee: every
 // chunk is authenticated and the writer renames the file into place only once
 // the stream ends on TAG_FINAL, so a truncated or corrupt fetch throws and
-// nothing is stored. On top of that this module refuses to record a stored file
+// nothing is stored. For an original whose metadata records a content hash
-// that came out empty. The design also asks for a content-hash comparison
+// (`FileMetadata.hash`), the writer also hashes the decrypted bytes and stores
-// against `FileMetadata.hash` (with a `fileSize` fallback); that is deferred —
+// nothing if they differ, failing the fetch with an error naming the file. An
-// see the PR — because the exact hash construction cannot be confirmed against
+// original with no recorded hash is stored unchecked, as the upstream client
-// the repo's fixtures and `FileBlob.size` is the encrypted object size, not the
+// does; thumbnails have none. On top of that this module refuses to record a
-// decrypted length this layer has.
+// stored file that came out empty.
 
-import { existsSync, statSync } from "node:fs";
+import { existsSync, readFileSync, statSync } from "node:fs";
 import {
     chmod,
     mkdir,
@@ -32,13 +34,15 @@ import {
     statfs,
     utimes,
 } from "node:fs/promises";
-import { dirname, extname, join } from "node:path";
+import { basename, dirname, extname, join } from "node:path";
 
 import type { ApiClient } from "../api/client.js";
 import {
     downloadFile,
     downloadThumbnail,
     type ProgressCallback,
+    removeLeftoverTempFiles,
+    writeAtomic,
 } from "../download/index.js";
 import { safeExtension } from "../filename.js";
 import type { EnteFile } from "../model/types.js";
@@ -46,8 +50,6 @@ import type { Priority, RequestPools } from "./pools.js";
 
 const DIR_MODE = 0o700;
 const FILE_MODE = 0o600;
-const TEMP_PREFIX = ".quak-";
-const TEMP_SUFFIX = ".tmp";
 const GIB = 1024 * 1024 * 1024;
 // Owner ruling (#36): bound the originals cache at 100 GiB, but back off when
 // the volume has under 50 GiB free so the cache never crowds the disk.
@@ -78,6 +80,8 @@ const poolPriorityOf = (priority: ThumbnailPriority): Priority =>
 export interface ContentResult {
     path: string;
     bytes: number;
+    // A live photo's video. `path` and `bytes` are then its image's.
+    videoPath?: string;
 }
 
 // Progress for a single `original`/`thumbnail` call. A present file emits one
@@ -128,11 +132,14 @@ export interface ThumbnailsAPI {
 // stand-in so the cache logic runs with no crypto and no network. Pool routing,
 // dedup, present-checks and integrity live in the cache, not here.
 export interface ContentSource {
+    // Writes the original at `destination`. A live photo is written beside it
+    // as its image and its video instead, and their paths are returned, as
+    // `downloadFile` does.
     original(args: {
         file: EnteFile;
         destination: string;
         onProgress?: ProgressCallback;
-    }): Promise<{ bytesWritten: number }>;
+    }): Promise<{ bytesWritten: number; path?: string; videoPath?: string }>;
     thumbnail(args: {
         file: EnteFile;
         destination: string;
@@ -210,7 +217,9 @@ class AbortDrop extends Error {
     }
 }
 
-const originalName = (file: EnteFile): string =>
+// The originals/ name for a file: `<fileID><ext>`, the extension taken from the
+// title (or `.bin`). A backup names its originals the same way.
+export const originalName = (file: EnteFile): string =>
     `${file.id}${safeExtension(file.metadata.title)}`;
 
 // The fileID a cache filename encodes, or undefined when the name is not one
@@ -232,6 +241,73 @@ const fileSize = (path: string): number | undefined => {
     }
 };
 
+// Whether `path` is a regular file with content. A zero-byte file is the shape
+// an aborted write leaves, so it does not count.
+const hasContent = (path: string | undefined): boolean =>
+    path !== undefined && (fileSize(path) ?? 0) > 0;
+
+// A live photo's image and video are named with the extensions from inside its
+// ZIP, so their names alone do not say which is which. Wherever the cache or a
+// backup stores one, this JSON file beside them names both.
+const livePhotoFileName = (fileID: number): string =>
+    `${fileID}.livephoto.json`;
+
+// The image and video that the live photo's JSON file in `dir` names, or
+// undefined when there is none. Only names of the form the cache writes are
+// taken, so the file cannot point outside `dir`.
+const readLivePhoto = (
+    dir: string,
+    fileID: number,
+): { path: string; videoPath: string } | undefined => {
+    const valid = (name: unknown): name is string =>
+        typeof name === "string" && name === `${fileID}${safeExtension(name)}`;
+    try {
+        const { image, video } = JSON.parse(
+            readFileSync(join(dir, livePhotoFileName(fileID)), "utf-8"),
+        );
+        if (valid(image) && valid(video)) {
+            return { path: join(dir, image), videoPath: join(dir, video) };
+        }
+    } catch {
+        // No such file, or not one the cache wrote.
+    }
+    return undefined;
+};
+
+// Write the JSON file naming a live photo's image and video, both in `dir`.
+export const writeLivePhoto = (
+    dir: string,
+    fileID: number,
+    stored: { path: string; videoPath: string },
+): Promise<void> =>
+    writeAtomic(
+        join(dir, livePhotoFileName(fileID)),
+        new TextEncoder().encode(
+            JSON.stringify({
+                image: basename(stored.path),
+                video: basename(stored.videoPath),
+            }),
+        ),
+    );
+
+// The original of `file` as the cache or a backup stored it in `dir`, when all
+// of it is there: `<fileID><ext>`, or a live photo's image and video.
+export const storedOriginal = (
+    dir: string,
+    file: EnteFile,
+): { path: string; videoPath?: string } | undefined => {
+    if (file.metadata.fileType !== "livePhoto") {
+        const path = join(dir, originalName(file));
+        return hasContent(path) ? { path } : undefined;
+    }
+    const stored = readLivePhoto(dir, file.id);
+    return stored !== undefined &&
+        hasContent(stored.path) &&
+        hasContent(stored.videoPath)
+        ? stored
+        : undefined;
+};
+
 export class ContentCache implements PhotoContent, ThumbnailsAPI {
     private readonly pools: RequestPools;
     private readonly source: ContentSource;
@@ -239,10 +315,17 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
     private readonly getFile: (fileID: number) => EnteFile | undefined;
     private readonly originalsDir: string;
     private readonly thumbnailsDir: string;
-    // fileID -> absolute path of the cached bytes, seeded from the directory
+    // fileID -> absolute path of the cached bytes, and for a live photo's
-    // listing at open() and extended as fetches store new files.
+    // original its video's, seeded from the directory listing at open() and
-    private readonly originals = new Map<number, string>();
+    // extended as fetches store new files.
-    private readonly thumbnails = new Map<number, string>();
+    private readonly originals = new Map<
+        number,
+        { path: string; videoPath?: string }
+    >();
+    private readonly thumbnails = new Map<
+        number,
+        { path: string; videoPath?: string }
+    >();
     private readonly maxOriginalsBytes: number;
     private readonly freeBelowBytes: number;
     private readonly isPinned: (fileID: number) => boolean;
@@ -302,9 +385,9 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
     pathsFor(fileID: number): CachedPaths {
         const out: CachedPaths = {};
         const original = this.originals.get(fileID);
-        if (original !== undefined) out.originalPath = original;
+        if (original !== undefined) out.originalPath = original.path;
         const thumbnail = this.thumbnails.get(fileID);
-        if (thumbnail !== undefined) out.thumbnailPath = thumbnail;
+        if (thumbnail !== undefined) out.thumbnailPath = thumbnail.path;
         return out;
     }
 
@@ -322,6 +405,27 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
         return this.get(fileID, "thumbnail", "on-demand", opts?.onProgress);
     }
 
+    // Get an original for a backup. One not present anywhere is written
+    // straight to `destination` and recorded there, so no second copy lands
+    // in the cache; one already present is returned where it is.
+    async backupOriginal(
+        fileID: number,
+        destination: string,
+    ): Promise<ContentResult> {
+        const result = await this.acquire(
+            fileID,
+            "original",
+            "on-demand",
+            undefined,
+            { destination },
+        );
+        return {
+            path: result.path,
+            bytes: result.bytes,
+            videoPath: result.videoPath,
+        };
+    }
+
     async ensure(args: EnsureOptions): Promise<EnsureResult[]> {
         return this.ensureThumbnails(args);
     }
@@ -418,51 +522,72 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
                 ? { status: "skipped", bytes: result.bytes }
                 : { status: "done", bytes: result.bytes },
         );
-        return { path: result.path, bytes: result.bytes };
+        return {
+            path: result.path,
+            bytes: result.bytes,
+            videoPath: result.videoPath,
+        };
     }
 
     // The core: return the cached path if present, else fetch through the pool,
     // store, and return it. `cached` distinguishes a present hit (no network,
-    // no download event) from a fresh fetch.
+    // no download event) from a fresh fetch. A fetched original is stored at
+    // `opts.destination` when given, instead of in `originalsDir`. A live
+    // photo's original is present only with its video, and is returned with
+    // it.
     private async acquire(
         fileID: number,
         kind: Kind,
         priority: Priority,
         signal: AbortSignal | undefined,
-        opts?: { onByte?: ProgressCallback },
+        opts?: { onByte?: ProgressCallback; destination?: string },
-    ): Promise<{ path: string; bytes: number; cached: boolean }> {
+    ): Promise<{
+        path: string;
+        bytes: number;
+        videoPath?: string;
+        cached: boolean;
+    }> {
         const file = this.getFile(fileID);
         if (!file) throw new Error(`content cache: unknown file ${fileID}`);
+        const isLivePhoto =
+            kind === "original" && file.metadata.fileType === "livePhoto";
 
         const known = kind === "original" ? this.originals : this.thumbnails;
         const cached = known.get(fileID);
         if (cached !== undefined) {
-            const size = fileSize(cached);
+            const size = fileSize(cached.path);
-            if (size !== undefined && size > 0) {
+            if (
+                size !== undefined &&
+                size > 0 &&
+                (!isLivePhoto || hasContent(cached.videoPath))
+            ) {
                 // Returning an original's path is a use: bump its mtime so LRU
                 // order reflects it and survives a restart with no ledger.
                 if (
                     kind === "original" &&
-                    dirname(cached) === this.originalsDir
+                    dirname(cached.path) === this.originalsDir
                 )
-                    await this.touch(cached);
+                    await this.touch(cached.path);
-                return { path: cached, bytes: size, cached: true };
+                return { ...cached, bytes: size, cached: true };
             }
-            // A recorded file that has since gone re-fetches below.
+            // A recorded file that has since gone, or a live photo an earlier
+            // version stored as one ZIP, re-fetches below.
             known.delete(fileID);
         }
 
         // An original a backup already stored counts as present.
         if (kind === "original" && this.downloadDirectory !== undefined) {
-            const backupPath = join(
+            const stored = storedOriginal(
-                this.downloadDirectory,
+                join(this.downloadDirectory, "originals"),
-                "originals",
+                file,
-                originalName(file),
             );
-            const size = fileSize(backupPath);
+            if (stored !== undefined) {
-            if (size !== undefined && size > 0) {
+                this.originals.set(fileID, stored);
-                this.originals.set(fileID, backupPath);
+                return {
-                return { path: backupPath, bytes: size, cached: true };
+                    ...stored,
+                    bytes: fileSize(stored.path) ?? 0,
+                    cached: true,
+                };
             }
         }
 
@@ -470,7 +595,7 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
             kind === "original" ? this.originalsDir : this.thumbnailsDir;
         const dest =
             kind === "original"
-                ? join(dir, originalName(file))
+                ? (opts?.destination ?? join(dir, originalName(file)))
                 : join(dir, `${fileID}${THUMBNAIL_EXT}`);
         const pool =
             kind === "original" ? this.pools.content : this.pools.thumbnails;
@@ -491,21 +616,39 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
                         ? this.beginOriginalWrite(fileID)
                         : null;
                 try {
-                    await this.download(file, dest, kind, opts?.onByte);
+                    const stored = await this.download(
-                    await chmod(dest, FILE_MODE);
+                        file,
-                    const size = (await stat(dest)).size;
+                        dest,
-                    if (size === 0) {
+                        kind,
+                        opts?.onByte,
+                    );
+                    for (const path of [stored.path, stored.videoPath]) {
+                        if (path === undefined) continue;
+                        await chmod(path, FILE_MODE);
+                        if ((await stat(path)).size === 0) {
                             throw new Error(
                                 `content cache: ${kind} ${fileID} stored empty`,
                             );
                         }
-                    known.set(fileID, dest);
+                    }
+                    // A backup records its own live photos.
+                    if (
+                        stored.videoPath !== undefined &&
+                        opts?.destination === undefined
+                    ) {
+                        await writeLivePhoto(dir, fileID, {
+                            path: stored.path,
+                            videoPath: stored.videoPath,
+                        });
+                    }
+                    known.set(fileID, stored);
                     // A fresh original may have crossed the limit; make room by
                     // evicting least-recently-used originals. An over-budget
                     // fetch keeps the file it returns, and no overlapping
                     // sibling is evicted. Thumbnails are never bounded.
                     if (write) await this.enforceOriginalsLimit(write);
-                    return { path: dest, bytes: size, cached: false };
+                    const size = (await stat(stored.path)).size;
+                    return { ...stored, bytes: size, cached: false };
                 } finally {
                     if (write) this.inFlightOriginals.delete(write);
                 }
@@ -514,18 +657,24 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
         );
     }
 
+    // Fetch into `destination`, returning where the bytes landed: there, or
+    // for a live photo, its image and video beside it.
     private async download(
         file: EnteFile,
         destination: string,
         kind: Kind,
         onProgress: ProgressCallback | undefined,
-    ): Promise<number> {
+    ): Promise<{ path: string; videoPath?: string }> {
         const args = { file, destination, onProgress };
-        const result =
+        if (kind === "thumbnail") {
-            kind === "original"
+            await this.source.thumbnail(args);
-                ? await this.source.original(args)
+            return { path: destination };
-                : await this.source.thumbnail(args);
+        }
-        return result.bytesWritten;
+        const result = await this.source.original(args);
+        return {
+            path: result.path ?? destination,
+            videoPath: result.videoPath,
+        };
     }
 
     // Best-effort bump of a file's mtime to now; a failed touch must never fail
@@ -536,13 +685,14 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
     }
 
     // Every stored original that lives under `originalsDir` (a backup-directory
-    // hit recorded in the map is excluded), with its size and mtime. Entries
+    // hit recorded in the map is excluded), with its size and mtime; a live
-    // whose file has vanished are dropped from the map. Backups and thumbnails
+    // photo's size includes its video. Entries whose file has vanished are
-    // are never counted.
+    // dropped from the map. Backups and thumbnails are never counted.
     private async measureOriginals(): Promise<{
         entries: {
             fileID: number;
             path: string;
+            videoPath?: string;
             size: number;
             mtimeMs: number;
         }[];
@@ -551,21 +701,28 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
         const entries: {
             fileID: number;
             path: string;
+            videoPath?: string;
             size: number;
             mtimeMs: number;
         }[] = [];
         let used = 0;
-        for (const [fileID, path] of this.originals) {
+        for (const [fileID, { path, videoPath }] of this.originals) {
             if (dirname(path) !== this.originalsDir) continue;
             try {
                 const s = await stat(path);
+                const size =
+                    s.size +
+                    (videoPath === undefined
+                        ? 0
+                        : (await stat(videoPath)).size);
                 entries.push({
                     fileID,
                     path,
-                    size: s.size,
+                    videoPath,
+                    size,
                     mtimeMs: s.mtimeMs,
                 });
-                used += s.size;
+                used += size;
             } catch {
                 this.originals.delete(fileID);
             }
@@ -629,6 +786,18 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
                 for (const e of evictable) {
                     if (remaining <= limit) break;
                     await rm(e.path, { force: true });
+                    // A live photo goes whole: its video and the JSON file
+                    // naming the two go with its image.
+                    if (e.videoPath !== undefined) {
+                        await rm(e.videoPath, { force: true });
+                        await rm(
+                            join(
+                                this.originalsDir,
+                                livePhotoFileName(e.fileID),
+                            ),
+                            { force: true },
+                        );
+                    }
                     this.originals.delete(e.fileID);
                     remaining -= e.size;
                 }
@@ -653,23 +822,30 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
         await chmod(dir, DIR_MODE);
     }
 
-    private async scan(dir: string, into: Map<number, string>): Promise<void> {
+    private async scan(
+        dir: string,
+        into: Map<number, { path: string; videoPath?: string }>,
+    ): Promise<void> {
+        // Another process sharing this cache may still be writing its temp
+        // files, so only those whose process has exited are removed.
+        removeLeftoverTempFiles(dir);
         let entries: string[];
         try {
             entries = await readdir(dir);
         } catch {
             return;
         }
+        const names = new Set(entries);
         for (const name of entries) {
-            if (name.startsWith(TEMP_PREFIX) && name.endsWith(TEMP_SUFFIX)) {
-                await rm(join(dir, name), { force: true }).catch(
-                    () => undefined,
-                );
-                continue;
-            }
             const id = fileIDFromName(name);
             const path = join(dir, name);
-            if (id !== undefined && existsSync(path)) into.set(id, path);
+            if (id === undefined || !existsSync(path)) continue;
+            // A live photo's image and video are one entry, as the JSON file
+            // beside them names them.
+            const livePhoto = names.has(livePhotoFileName(id))
+                ? readLivePhoto(dir, id)
+                : undefined;
+            into.set(id, livePhoto ?? { path });
         }
     }
 }

src/library/index.ts

@@ -26,6 +26,7 @@
 // store marked unsaved until a later save actually lands, so a stuck disk is
 // never masked by a subsequent empty refresh.
 
+import { rm } from "node:fs/promises";
 import { join } from "node:path";
 import envPaths from "env-paths";
 
@@ -97,6 +98,11 @@ export {
 
 export const DEFAULT_REFRESH_INTERVAL_SECONDS = 3;
 
+// The account's cache directory when `cacheDirectory` is not given: the
+// env-paths cache directory plus the user id, so each account has its own.
+export const defaultCacheDirectory = (userID: number): string =>
+    join(envPaths("quak", { suffix: "" }).cache, String(userID));
+
 // Project a metadata store into by-id records, filling each record's cache
 // paths from the content cache when one is given. Shared by the live read
 // projection and the precache's initial seeding at open().
@@ -343,11 +349,21 @@ export class Library {
     static async open(opts: LibraryOptions): Promise<Library> {
         const { userID } = opts.client.whoami();
         const cacheDirectory =
-            opts.cacheDirectory ??
+            opts.cacheDirectory ?? defaultCacheDirectory(userID);
-            join(envPaths("quak", { suffix: "" }).cache, String(userID));
+        const metadataPath = join(cacheDirectory, "metadata.json");
-        const store = await MetadataStore.load(
+        let store = await MetadataStore.load(metadataPath);
-            join(cacheDirectory, "metadata.json"),
+        // A cache directory given explicitly can hold another account's cache.
-        );
+        // Its records and cursor are not this account's, so delete it and the
+        // ML data beside it and start empty. A user ID of 0 means the cache
+        // was never refreshed and so holds nothing to discard.
+        if (store.userID !== 0 && store.userID !== userID) {
+            await rm(metadataPath, { force: true });
+            await rm(join(cacheDirectory, "mldata"), {
+                recursive: true,
+                force: true,
+            });
+            store = await MetadataStore.load(metadataPath);
+        }
         const intervalMs =
             (opts.refreshIntervalSeconds ?? DEFAULT_REFRESH_INTERVAL_SECONDS) *
             1000;
@@ -522,11 +538,13 @@ export class Library {
     }
 
     // Back up every in-scope file to `downloadDirectory` in the historical
-    // on-disk layout, with a durable failure ledger (issue #51). Refreshes
+    // on-disk layout, with a durable failure ledger (issue #51). Waits for a
-    // first, fetches pending originals (and optional thumbnails) through the
+    // completed refresh first, as `fresh()` does, joining one already running,
-    // content cache and pools, then rebuilds the derived symlink/JSON views
+    // and rejects before touching any file when it fails. Then fetches pending
-    // from the model. Throws before any network work when no download directory
+    // originals (and optional thumbnails) through the content cache and pools,
-    // is available or no content cache backs the originals it must fetch.
+    // and rebuilds the derived symlink/JSON views from the model. Throws before
+    // any network work when no download directory is available or no content
+    // cache backs the originals it must fetch.
     backup(opts?: BackupOptions): Promise<BackupResult> {
         const downloadDirectory =
             opts?.downloadDirectory ?? this.downloadDirectory;
@@ -550,10 +568,11 @@ export class Library {
         const cache = this.cache;
         return runBackup(
             {
-                refresh: () => this.runRefresh(),
+                refresh: () => this.refreshNow(),
                 listCollections: () => this.store.listCollections(),
                 listFiles: (id) => this.store.listFiles(id),
-                original: (fileID) => cache!.original(fileID),
+                original: (fileID, destination) =>
+                    cache!.backupOriginal(fileID, destination),
                 thumbnail: (fileID) => cache!.thumbnail(fileID),
             },
             { ...opts, downloadDirectory },

src/library/read.ts

@@ -80,7 +80,8 @@ export class Photo {
     }
 
     // Fetch and cache the full-resolution original, returning its on-disk path
-    // and byte length. Served from the cache (or the backup download directory)
+    // and byte length; for a live photo, its image's, and its video's path as
+    // `videoPath`. Served from the cache (or the backup download directory)
     // when already present, otherwise fetched through the content pool.
     async original(opts?: ContentOptions): Promise<ContentResult> {
         return this.contentOrThrow().original(this.rec.fileID, opts);

src/library/records.ts

@@ -42,6 +42,7 @@ export interface PhotoRecord {
     isArchived: boolean;
     isHidden: boolean;
     // Local cache paths, set once a later phase caches the bytes; unset here.
+    // A live photo's `originalPath` is its image.
     thumbnailPath?: string;
     originalPath?: string;
 }

src/metadata-backup.ts

@@ -5,7 +5,11 @@ import exifReader from "exif-reader";
 import type { Client } from "./client.js";
 import type { Library, Photo } from "./library/index.js";
 import { sanitizeFileName } from "./filename.js";
-import { fetchMLData } from "./mldata-fetch.js";
+import {
+    fetchMLDataBatch,
+    MLDATA_BATCH_SIZE,
+    type MLData,
+} from "./mldata-fetch.js";
 import type { EnteFile } from "./model/types.js";
 
 export type ProgressCallback = (message: string) => void;
@@ -124,7 +128,8 @@ export const extractImageMetadata = (
 // Read a file's original bytes through the library's content cache and extract
 // its embedded image metadata. The bytes come from `photo.original()` — the
 // same on-disk cache the rest of the library fills — rather than a fresh
-// per-call download to a throwaway temp file.
+// per-call download to a throwaway temp file. For a live photo, its `path` is
+// the image.
 const extractExif = async (
     photo: Photo,
 ): Promise<Record<string, unknown> | undefined> => {
@@ -136,14 +141,15 @@ const extractExif = async (
 // Dump every decrypted metadata layer the account holds into a directory tree
 // of plain JSON: account, per-collection, and per-file records including the
 // private and public magic metadata and (by default) the ML data. Collections
-// and files are enumerated from the library's cache rather than a fresh server
+// and files are enumerated from the library's cache, which the caller refreshes
-// scan; the ML fetch and EXIF extraction are unchanged.
+// first. Returns how many ML data requests failed; their files are still
+// written, with `mlDataError` in place of `mlData`.
 export const runMetadataBackup = async (
     lib: Library,
     client: Client,
     outDir: string,
     opts?: MetadataBackupOptions,
-): Promise<void> => {
+): Promise<{ failedMLBatches: number }> => {
     const log = opts?.onProgress ?? (() => {});
     const wantExif = opts?.exif ?? false;
 
@@ -208,12 +214,31 @@ export const runMetadataBackup = async (
         }
     }
 
+    // One failed request (retries exhausted) must not end the dump: its files
+    // get the reason in `mlDataError` and the other batches go on.
     log("Fetching ML data (face detections, CLIP embeddings)...");
-    const mlDataMap = await fetchMLData(
+    const mlDataMap = new Map<number, MLData>();
+    const mlDataErrors = new Map<number, string>();
+    let failedMLBatches = 0;
+    const fileIDs = [...fileKeys.keys()];
+    for (let i = 0; i < fileIDs.length; i += MLDATA_BATCH_SIZE) {
+        const batch = fileIDs.slice(i, i + MLDATA_BATCH_SIZE);
+        try {
+            const result = await fetchMLDataBatch(
                 client.getApiClient(),
-        [...fileKeys.keys()],
+                batch,
                 fileKeys,
             );
+            for (const [id, payload] of result) mlDataMap.set(id, payload);
+        } catch (err) {
+            const reason = err instanceof Error ? err.message : String(err);
+            failedMLBatches++;
+            log(
+                `ML data request for ${batch.length} file(s) failed: ${reason}`,
+            );
+            for (const id of batch) mlDataErrors.set(id, reason);
+        }
+    }
     log(`Got ML data for ${mlDataMap.size} file(s)`);
 
     const writtenFileIDs = new Set<number>();
@@ -233,6 +258,8 @@ export const runMetadataBackup = async (
 
         const ml = mlDataMap.get(file.id);
         if (ml) fileMeta.mlData = ml;
+        const mlError = mlDataErrors.get(file.id);
+        if (mlError) fileMeta.mlDataError = mlError;
 
         if (wantExif && !writtenFileIDs.has(file.id)) {
             log(`[${file.metadata.title}] Extracting EXIF...`);
@@ -253,4 +280,5 @@ export const runMetadataBackup = async (
     }
 
     log("Metadata backup complete.");
+    return { failedMLBatches };
 };

src/mldata-fetch.ts

@@ -5,9 +5,8 @@
 // comes back encrypted under the file's own key and gzipped; decrypting and
 // gunzipping yields the JSON payload
 // `{ face: { faces: [...] }, clip: { embedding } }`. Ente caps a request at 200
-// ids, so `fetchMLData` batches for callers that want many at once while
+// ids, so callers that want many at once split them into batches of
-// `fetchMLDataBatch` is the single-request unit the library submits to its
+// `MLDATA_BATCH_SIZE` and call `fetchMLDataBatch` once per batch.
-// request pool.
 
 import { gunzipSync } from "node:zlib";
 
@@ -69,25 +68,3 @@ export const fetchMLDataBatch = async (
     }
     return result;
 };
-
-// Fetch ML data for arbitrarily many ids, batching at `MLDATA_BATCH_SIZE`. Used
-// by the one-shot metadata backup; the library fetches through its request pool
-// with `fetchMLDataBatch` instead.
-export const fetchMLData = async (
-    api: ApiClient,
-    fileIDs: number[],
-    fileKeys: Map<number, Uint8Array>,
-): Promise<Map<number, MLData>> => {
-    const result = new Map<number, MLData>();
-    for (let i = 0; i < fileIDs.length; i += MLDATA_BATCH_SIZE) {
-        const batch = fileIDs.slice(i, i + MLDATA_BATCH_SIZE);
-        for (const [id, payload] of await fetchMLDataBatch(
-            api,
-            batch,
-            fileKeys,
-        )) {
-            result.set(id, payload);
-        }
-    }
-    return result;
-};

src/model/decrypt.ts

@@ -34,6 +34,28 @@ const FILE_TYPE_MAP: Record<number, FileType> = {
 
 const parseFileType = (n: number): FileType => FILE_TYPE_MAP[n] ?? "unknown";
 
+// The hash the uploading client recorded for the original's bytes, read the
+// way the upstream client's `metadataHash` reads it: `hash` if present,
+// otherwise, for a live photo from an older client that wrote the two parts
+// separately, `<imageHash>:<videoHash>`. A field that is not a non-empty
+// string counts as absent, and a file with no hash at all is normal.
+const expectedHash = (json: Record<string, unknown>): string | undefined => {
+    const text = (v: unknown): string | undefined =>
+        typeof v === "string" && v !== "" ? v : undefined;
+    const hash = text(json.hash);
+    if (hash !== undefined) return hash;
+    const imageHash = text(json.imageHash);
+    const videoHash = text(json.videoHash);
+    if (
+        json.fileType === 2 &&
+        imageHash !== undefined &&
+        videoHash !== undefined
+    ) {
+        return `${imageHash}:${videoHash}`;
+    }
+    return undefined;
+};
+
 export const decryptCollection = (
     raw: RawCollection,
     keys: KeyMaterial,
@@ -115,7 +137,7 @@ export const decryptFile = (
         modificationTime: metadataJSON.modificationTime ?? 0,
         latitude: metadataJSON.latitude,
         longitude: metadataJSON.longitude,
-        hash: metadataJSON.hash,
+        hash: expectedHash(metadataJSON),
     };
 
     const magicMetadata = decryptMagicMetadata(raw.magicMetadata, key);

src/model/types.ts

@@ -29,6 +29,9 @@ export interface FileMetadata {
     modificationTime: Microseconds;
     latitude?: number;
     longitude?: number;
+    // The content hash the uploader recorded (see `expectedHash` in
+    // decrypt.ts); `downloadFile` refuses an original that does not match it.
+    // Absent for files from very old clients.
     hash?: string;
 }
 

src/thumbnails.ts

@@ -7,8 +7,21 @@ import { ApiError } from "./api/client.js";
 import { encryptBlob, toBase64 } from "./crypto/index.js";
 import type { EnteFile } from "./model/types.js";
 
-const THUMB_MAX_DIMENSION = 720;
+// The server refuses a thumbnail larger than the one it already records for the
-const THUMB_JPEG_QUALITY = 50;
+// file (`thumbnail.size`, the encrypted size), so these encodings are tried
+// from largest to smallest and the first that fits is uploaded.
+const THUMB_ENCODINGS = [
+    { maxDimension: 720, quality: 50 },
+    { maxDimension: 720, quality: 30 },
+    { maxDimension: 480, quality: 30 },
+    { maxDimension: 320, quality: 20 },
+    { maxDimension: 160, quality: 20 },
+];
+
+// The server accepts a new thumbnail only from the file's owner, so files other
+// people own in albums shared with this account are never checked or repaired.
+const NOT_OWNED_REASON =
+    "owned by another account (only the owner can replace its thumbnail)";
 
 export interface MissingThumbnailInfo {
     fileID: number;
@@ -19,11 +32,12 @@ export interface MissingThumbnailInfo {
 
 // Three outcomes, not two. "fixed": a thumbnail was generated and uploaded.
 // "failed": something went wrong (download, encode, upload) and the file still
-// has no thumbnail. "skipped": the file is a format this helper cannot
+// has no thumbnail. "skipped": the server would refuse any thumbnail for the
-// regenerate — a video, or an image that is not a baseline JPEG. Skipped is a
+// file or this helper cannot regenerate it — a file another account owns, a
-// deliberate, expected outcome, not an error (issue #17): the repair path is
+// recorded thumbnail size nothing fits within, a video, or an image that is
-// JPEG-only because `jpeg-js` is, and a PNG or HEIC is left for a format-aware
+// not a baseline JPEG. Skipped is a deliberate, expected outcome, not an error
-// tool rather than reported as a failure.
+// (issue #17): the repair path is JPEG-only because `jpeg-js` is, and a PNG or
+// HEIC is left for a format-aware tool rather than reported as a failure.
 export type ThumbnailFixStatus = "fixed" | "skipped" | "failed";
 
 export interface ThumbnailFixResult {
@@ -45,6 +59,7 @@ export type ProgressCallback = (message: string) => void;
 // exists, so it is logged and the file is left unreported. That distinction is
 // what stops `fix-missing-thumbnails` from regenerating and uploading over
 // thumbnails that were fine all along while the CDN was briefly returning 500s.
+// Files another account owns are logged as skipped and not checked.
 export const listMissingThumbnails = async (
     lib: Library,
     client: Client,
@@ -52,6 +67,7 @@ export const listMissingThumbnails = async (
 ): Promise<MissingThumbnailInfo[]> => {
     const log = onProgress ?? (() => {});
     const api = client.getApiClient();
+    const { userID } = client.whoami();
     const missing: MissingThumbnailInfo[] = [];
     const seen = new Set<number>();
 
@@ -60,6 +76,13 @@ export const listMissingThumbnails = async (
         for (const photo of album.photos.list()) {
             if (seen.has(photo.fileID)) continue;
             seen.add(photo.fileID);
+            const file = lib.getFile(album.collectionID, photo.fileID);
+            if (file && file.ownerID !== userID) {
+                log(
+                    `[${album.name}] Skipping ${photo.title}: ${NOT_OWNED_REASON}`,
+                );
+                continue;
+            }
             try {
                 const stream = await api.getThumbnailStream(photo.fileID);
                 const reader = stream.getReader();
@@ -135,17 +158,13 @@ const resizeRGBA = (
     return dst;
 };
 
-const generateThumbnail = (fileBytes: Uint8Array): Uint8Array => {
+const generateThumbnail = (
-    const decoded = jpeg.decode(fileBytes, {
+    decoded: { data: Uint8Array; width: number; height: number },
-        useTArray: true,
+    maxDimension: number,
-        formatAsRGBA: true,
+    quality: number,
-    });
+): Uint8Array => {
     const { width: srcW, height: srcH } = decoded;
-    const scale = Math.min(
+    const scale = Math.min(maxDimension / srcW, maxDimension / srcH, 1);
-        THUMB_MAX_DIMENSION / srcW,
-        THUMB_MAX_DIMENSION / srcH,
-        1,
-    );
     const dstW = Math.round(srcW * scale);
     const dstH = Math.round(srcH * scale);
 
@@ -158,7 +177,7 @@ const generateThumbnail = (fileBytes: Uint8Array): Uint8Array => {
 
     const encoded = jpeg.encode(
         { data: pixels, width: dstW, height: dstH },
-        THUMB_JPEG_QUALITY,
+        quality,
     );
     return new Uint8Array(encoded.data);
 };
@@ -171,14 +190,35 @@ const generateThumbnail = (fileBytes: Uint8Array): Uint8Array => {
 const isJpeg = (bytes: Uint8Array): boolean =>
     bytes.length >= 2 && bytes[0] === 0xff && bytes[1] === 0xd8;
 
-// The reason a file cannot have a JPEG thumbnail regenerated for it from its
+// The reason a file cannot have a JPEG thumbnail regenerated for it, known from
-// metadata alone, before any bytes are fetched, or undefined when it might. A
+// its record alone before any bytes are fetched, or undefined when it might. A
-// non-image (video, live photo) is unsupported outright; a still image still
+// still image still has to be checked against its actual bytes once
-// has to be checked against its actual bytes once downloaded.
+// downloaded.
-const unsupportedByType = (file: EnteFile): string | undefined => {
+const reasonToSkip = (file: EnteFile, userID: number): string | undefined => {
+    if (file.ownerID !== userID) {
+        return NOT_OWNED_REASON;
+    }
     if (file.metadata.fileType !== "image") {
         return `unsupported file type: ${file.metadata.fileType} (only JPEG images can be regenerated)`;
     }
+    if (!file.thumbnail.size) {
+        return `recorded thumbnail size is ${file.thumbnail.size ?? "unknown"} (the server refuses a thumbnail larger than the one it records)`;
+    }
+    return undefined;
+};
+
+// Encrypt the largest encoding of the decoded image whose ciphertext is no
+// larger than `maxSize`, or return undefined when even the smallest is larger.
+const encryptThumbnailWithin = (
+    decoded: { data: Uint8Array; width: number; height: number },
+    key: Uint8Array,
+    maxSize: number,
+): { header: Uint8Array; ciphertext: Uint8Array } | undefined => {
+    for (const { maxDimension, quality } of THUMB_ENCODINGS) {
+        const thumbJpeg = generateThumbnail(decoded, maxDimension, quality);
+        const encrypted = encryptBlob(thumbJpeg, key);
+        if (encrypted.ciphertext.length <= maxSize) return encrypted;
+    }
     return undefined;
 };
 
@@ -197,6 +237,7 @@ export const fixMissingThumbnails = async (
     const log = onProgress ?? (() => {});
     const results: ThumbnailFixResult[] = [];
     const api = client.getApiClient();
+    const { userID } = client.whoami();
 
     // Resolve each requested fileID to its file record and owning album by
     // enumerating the library, each file taken from the first album that holds
@@ -237,18 +278,19 @@ export const fixMissingThumbnails = async (
         const { file, collectionName } = entry;
         const title = file.metadata.title;
 
-        const typeReason = unsupportedByType(file);
+        const skipReason = reasonToSkip(file, userID);
-        if (typeReason) {
+        if (skipReason) {
-            log(`[${collectionName}] Skipping ${title}: ${typeReason}`);
+            log(`[${collectionName}] Skipping ${title}: ${skipReason}`);
             results.push({
                 fileID,
                 title,
                 collection: collectionName,
                 status: "skipped",
-                reason: typeReason,
+                reason: skipReason,
             });
             continue;
         }
+        const maxSize = file.thumbnail.size!;
 
         try {
             const photo = lib.photos.byID({ fileID });
@@ -277,12 +319,28 @@ export const fixMissingThumbnails = async (
             }
 
             log(`[${collectionName}] Generating thumbnail for ${title}...`);
-            const thumbJpeg = generateThumbnail(fileBytes);
+            const decoded = jpeg.decode(fileBytes, {
+                useTArray: true,
+                formatAsRGBA: true,
+            });
+            const fitting = encryptThumbnailWithin(decoded, file.key, maxSize);
+            if (!fitting) {
+                const reason = `no thumbnail encoding fits the recorded thumbnail size of ${maxSize} bytes`;
+                log(`[${collectionName}] Skipping ${title}: ${reason}`);
+                results.push({
+                    fileID,
+                    title,
+                    collection: collectionName,
+                    status: "skipped",
+                    reason,
+                });
+                continue;
+            }
+            const { header, ciphertext } = fitting;
 
             log(
-                `[${collectionName}] Encrypting and uploading thumbnail (${thumbJpeg.length} bytes)...`,
+                `[${collectionName}] Uploading thumbnail (${ciphertext.length} bytes)...`,
             );
-            const { header, ciphertext } = encryptBlob(thumbJpeg, file.key);
             const md5 = createHash("md5").update(ciphertext).digest("base64");
             const { objectKey, url } = await api.getUploadURL(
                 ciphertext.length,

test/cli/backup.test.ts

@@ -40,6 +40,7 @@ import {
     readFileSync,
     readlinkSync,
     rmSync,
+    symlinkSync,
     writeFileSync,
 } from "node:fs";
 import { spawnSync } from "node:child_process";
@@ -47,10 +48,18 @@ import { join } from "node:path";
 import { tmpdir } from "node:os";
 import { describe, it, expect, beforeEach, afterEach, vi } from "vitest";
 
+import { runBackup, type BackupLibrary } from "../../src/backup.js";
 import { Library } from "../../src/library/index.js";
 import type { ContentSource } from "../../src/library/content.js";
 import type { CollectionsPage, FilesPage } from "../../src/client.js";
 import type { Collection, EnteFile } from "../../src/model/types.js";
+import {
+    asLivePhoto,
+    cdnSource,
+    IMAGE,
+    livePhotoZip,
+    VIDEO,
+} from "../live-photo.js";
 
 // `open` and `rename` are wrapped to record, in order, every fsync and rename,
 // so a test can pin the sequence "fsync the temp file, rename, fsync the
@@ -560,11 +569,39 @@ describe("lib.backup", () => {
         lib.close();
     });
 
-    it("fsyncs a copied original before the rename and its directory after", async () => {
+    it("fetches each original once and writes it only into the backup", async () => {
+        // A backup of a 500 GB account must write 500 GB, not a copy in the
+        // cache as well: an original fetched for the backup goes straight
+        // into its originals/, and the cache records it there.
+        const source = stubSource();
+        const lib = await openLibrary(source);
+        const outDir = join(root, "backup");
+
+        const result = await lib.backup({ downloadDirectory: outDir });
+
+        expect(result.downloaded).toBe(3);
+        expect(source.originalCalls).toBe(3);
+        expect(readdirSync(join(root, "cache", "originals"))).toEqual([]);
+        const stored = readdirSync(join(outDir, "originals")).filter(
+            (name) => !name.endsWith(".json"),
+        );
+        expect(stored.sort()).toEqual(["100.jpg", "101.jpg", "200.png"]);
+        // The cache counts the backup's copy as present: reading the
+        // original afterwards fetches nothing and answers with that copy.
+        const read = await lib.photos.byID({ fileID: 100 })!.original();
+        expect(read.path).toBe(join(outDir, "originals", "100.jpg"));
+        expect(source.originalCalls).toBe(3);
+        await lib.close();
+    });
+
+    it("fsyncs an original copied from the cache before the rename and its directory after", async () => {
         const lib = await openLibrary(stubSource());
         const outDir = join(root, "backup");
         const originals = join(outDir, "originals");
         const dest = join(originals, "100.jpg");
+        // Only an original already in the cache is copied into the backup;
+        // one fetched for the backup is written there by the download writer.
+        await lib.photos.byID({ fileID: 100 })!.original();
         fsEvents.length = 0;
 
         await lib.backup({ downloadDirectory: outDir });
@@ -622,3 +659,592 @@ describe("lib.backup", () => {
         lib.close();
     });
 });
+
+// Every refresh fails, as with an expired session or no network.
+class FailingClient extends MockClient {
+    override async collectionsSince(): Promise<CollectionsPage> {
+        throw new Error("HTTP 401 from server");
+    }
+}
+
+// Holds its refresh open until `release()` is called, then reports a third
+// album, so a backup can be started while that refresh is still running.
+class HeldClient extends MockClient {
+    release!: () => void;
+    private held = new Promise<void>((resolve) => {
+        this.release = resolve;
+    });
+    override async collectionsSince(): Promise<CollectionsPage> {
+        await this.held;
+        return {
+            collections: [collection(3, "Later")],
+            deleted: [],
+            cursor: 2,
+        };
+    }
+    override async filesSince(args: {
+        collectionID: number;
+    }): Promise<FilesPage> {
+        if (args.collectionID !== 3) return super.filesSince(args);
+        return { files: [file(300, 3, "late.jpg")], deleted: [], cursor: 2 };
+    }
+}
+
+// Fill the library cache on disk, so the next open starts its refresh in the
+// background instead of waiting for it.
+const fillCache = async (): Promise<void> => {
+    const lib = await openLibrary(stubSource());
+    await lib.close();
+};
+
+describe("the refresh before a backup", () => {
+    it("waits for a refresh already running and backs up what it found", async () => {
+        await fillCache();
+        const client = new HeldClient();
+        const lib = await openLibrary(stubSource(), client);
+        const outDir = join(root, "backup");
+
+        const backup = lib.backup({ downloadDirectory: outDir });
+        client.release();
+        const result = await backup;
+
+        expect(result.totalFiles).toBe(4);
+        expect(existsSync(join(outDir, "originals", "300.jpg"))).toBe(true);
+        await lib.close();
+    });
+
+    it("fails before any download when the refresh fails, leaving failures.json as it was", async () => {
+        await fillCache();
+        const outDir = join(root, "backup");
+        seedLedger(outDir, 100, "beach.jpg");
+        const ledgerPath = join(outDir, "failures.json");
+        const ledgerBefore = readFileSync(ledgerPath, "utf-8");
+        const source = stubSource();
+        const lib = await openLibrary(source, new FailingClient());
+
+        await expect(lib.backup({ downloadDirectory: outDir })).rejects.toThrow(
+            "HTTP 401 from server",
+        );
+
+        expect(source.originalCalls).toBe(0);
+        expect(readFileSync(ledgerPath, "utf-8")).toBe(ledgerBefore);
+        expect(existsSync(join(outDir, "originals"))).toBe(false);
+        await lib.close();
+    });
+
+    it("fails when the refresh fails on an empty cache, instead of backing up nothing", async () => {
+        const source = stubSource();
+        const lib = await openLibrary(source, new FailingClient());
+        const outDir = join(root, "backup");
+
+        await expect(lib.backup({ downloadDirectory: outDir })).rejects.toThrow(
+            "HTTP 401 from server",
+        );
+
+        expect(source.originalCalls).toBe(0);
+        expect(existsSync(outDir)).toBe(false);
+        await lib.close();
+    });
+});
+
+// Every entry under collections/, one level of directories deep, with each
+// symlink's target.
+const tree = (outDir: string): string[] => {
+    const lines: string[] = [];
+    const list = (dir: string, prefix: string): void => {
+        for (const name of readdirSync(dir).sort()) {
+            const path = join(dir, name);
+            const st = lstatSync(path);
+            if (st.isSymbolicLink()) {
+                lines.push(`${prefix}${name} -> ${readlinkSync(path)}`);
+            } else if (st.isDirectory() && prefix === "") {
+                lines.push(`${name}/`);
+                list(path, `${name}/`);
+            } else {
+                lines.push(`${prefix}${name}`);
+            }
+        }
+    };
+    list(join(outDir, "collections"), "");
+    return lines;
+};
+
+// The album folders under collections/, driven through `runBackup` with a
+// stand-in library whose albums a test changes between runs.
+describe("backup album folders", () => {
+    interface Album {
+        collection: Collection;
+        files: EnteFile[];
+    }
+
+    const libraryOf = (albums: Album[]): BackupLibrary => ({
+        refresh: async () => {},
+        listCollections: () => albums.map((a) => a.collection),
+        listFiles: (id) =>
+            albums.find((a) => a.collection.id === id)?.files ?? [],
+        original: async (fileID) => {
+            const path = join(root, `source-${fileID}`);
+            writeFileSync(path, `original ${fileID}`);
+            return { path };
+        },
+        thumbnail: async () => {
+            throw new Error("no thumbnails in this stand-in");
+        },
+    });
+
+    const albumID = (outDir: string, jsonName: string): number =>
+        JSON.parse(readFileSync(join(outDir, "collections", jsonName), "utf-8"))
+            .id;
+
+    it("gives every file and every album its own name when names repeat", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [
+                    file(1, 10, "IMG_0001.JPG"),
+                    file(2, 10, "IMG_0001.JPG"),
+                    file(4, 10, "img_0001.jpg"),
+                    file(3, 10, "other.jpg"),
+                ],
+            },
+            {
+                collection: collection(11, "Trip"),
+                files: [file(3, 11, "other.jpg")],
+            },
+        ]);
+
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        expect(result.failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Trip (10)/",
+            "Trip (10)/IMG_0001 (1).JPG -> ../../originals/1.JPG",
+            "Trip (10)/IMG_0001 (2).JPG -> ../../originals/2.JPG",
+            "Trip (10)/img_0001 (4).jpg -> ../../originals/4.jpg",
+            "Trip (10)/other.jpg -> ../../originals/3.jpg",
+            "Trip (10).json",
+            "Trip (11)/",
+            "Trip (11)/other.jpg -> ../../originals/3.jpg",
+            "Trip (11).json",
+        ]);
+        expect(albumID(outDir, "Trip (10).json")).toBe(10);
+        expect(albumID(outDir, "Trip (11).json")).toBe(11);
+    });
+
+    it("keeps names unique when a name with an ID added is another entry's own name", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [
+                    file(5, 10, "IMG (6).JPG"),
+                    file(6, 10, "IMG.JPG"),
+                    file(7, 10, "IMG.JPG"),
+                ],
+            },
+            {
+                collection: collection(11, "Trip"),
+                files: [file(8, 11, "a.jpg")],
+            },
+            {
+                collection: collection(12, "Trip (11)"),
+                files: [file(9, 12, "b.jpg")],
+            },
+        ]);
+
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        expect(result.failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Trip (10)/",
+            "Trip (10)/IMG (6) (5).JPG -> ../../originals/5.JPG",
+            "Trip (10)/IMG (6).JPG -> ../../originals/6.JPG",
+            "Trip (10)/IMG (7).JPG -> ../../originals/7.JPG",
+            "Trip (10).json",
+            "Trip (11)/",
+            "Trip (11)/a.jpg -> ../../originals/8.jpg",
+            "Trip (11) (12)/",
+            "Trip (11) (12)/b.jpg -> ../../originals/9.jpg",
+            "Trip (11) (12).json",
+            "Trip (11).json",
+        ]);
+        expect(albumID(outDir, "Trip (10).json")).toBe(10);
+        expect(albumID(outDir, "Trip (11).json")).toBe(11);
+        expect(albumID(outDir, "Trip (11) (12).json")).toBe(12);
+    });
+
+    it("changes nothing on a second run over an unchanged account", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [
+                    file(1, 10, "IMG_0001.JPG"),
+                    file(2, 10, "IMG_0001.JPG"),
+                ],
+            },
+            {
+                collection: collection(11, "Trip"),
+                files: [file(3, 11, "other.jpg")],
+            },
+        ]);
+
+        await runBackup(lib, { downloadDirectory: outDir });
+        const before = tree(outDir);
+        const second = await runBackup(lib, { downloadDirectory: outDir });
+
+        expect(second.downloaded).toBe(0);
+        expect(second.failed).toBe(0);
+        expect(tree(outDir)).toEqual(before);
+    });
+
+    it("leaves the albums an onlyAlbumNames run skips as they were", async () => {
+        const outDir = join(root, "backup");
+        // "trip" is skipped by the scoped run but its name clashes with the
+        // in-scope "Trip", so "Trip" must keep its ID suffix.
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [file(1, 10, "a.jpg")],
+            },
+            {
+                collection: collection(11, "trip"),
+                files: [file(2, 11, "b.jpg")],
+            },
+            {
+                collection: collection(12, "Work"),
+                files: [file(3, 12, "c.jpg")],
+            },
+        ]);
+        const json = (name: string): string =>
+            readFileSync(join(outDir, "collections", name), "utf-8");
+
+        await runBackup(lib, { downloadDirectory: outDir });
+        const before = tree(outDir);
+        const skippedJSON = [json("trip (11).json"), json("Work.json")];
+        const scoped = await runBackup(lib, {
+            downloadDirectory: outDir,
+            onlyAlbumNames: ["Trip"],
+        });
+
+        expect(scoped.failed).toBe(0);
+        expect(before).toEqual([
+            "Trip (10)/",
+            "Trip (10)/a.jpg -> ../../originals/1.jpg",
+            "Trip (10).json",
+            "Work/",
+            "Work/c.jpg -> ../../originals/3.jpg",
+            "Work.json",
+            "trip (11)/",
+            "trip (11)/b.jpg -> ../../originals/2.jpg",
+            "trip (11).json",
+        ]);
+        expect(tree(outDir)).toEqual(before);
+        expect([json("trip (11).json"), json("Work.json")]).toEqual(
+            skippedJSON,
+        );
+    });
+
+    it("removes links and album folders that are gone, and nothing the user added", async () => {
+        const outDir = join(root, "backup");
+        const albums: Album[] = [
+            {
+                collection: collection(10, "Trip"),
+                files: [
+                    file(1, 10, "IMG_0001.JPG"),
+                    file(2, 10, "IMG_0001.JPG"),
+                    file(3, 10, "other.jpg"),
+                ],
+            },
+            {
+                collection: collection(12, "Work"),
+                files: [file(5, 12, "a.jpg")],
+            },
+            {
+                collection: collection(13, "Old"),
+                files: [file(5, 13, "a.jpg")],
+            },
+        ];
+        const lib = libraryOf(albums);
+        await runBackup(lib, { downloadDirectory: outDir });
+
+        // What the user put in the tree: a note and a symlink of their own in
+        // an album, a note in an album about to be renamed, and a folder quak
+        // did not create.
+        const collectionsDir = join(outDir, "collections");
+        writeFileSync(join(collectionsDir, "Trip", "notes.txt"), "mine");
+        symlinkSync("../elsewhere", join(collectionsDir, "Trip", "mine"));
+        writeFileSync(join(collectionsDir, "Work", "keep.txt"), "mine");
+        mkdirSync(join(collectionsDir, "Mine"));
+        writeFileSync(join(collectionsDir, "Mine", "keep.txt"), "mine");
+
+        // File 2 leaves Trip, Work is renamed Office, Old is deleted.
+        albums[0]!.files.splice(1, 1);
+        albums[1]!.collection = collection(12, "Office");
+        albums.splice(2, 1);
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        expect(result.failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Mine/",
+            "Mine/keep.txt",
+            "Office/",
+            "Office/a.jpg -> ../../originals/5.jpg",
+            "Office.json",
+            "Trip/",
+            "Trip/IMG_0001.JPG -> ../../originals/1.JPG",
+            "Trip/mine -> ../elsewhere",
+            "Trip/notes.txt",
+            "Trip/other.jpg -> ../../originals/3.jpg",
+            "Trip.json",
+            "Work/",
+            "Work/keep.txt",
+            "Work.json",
+        ]);
+    });
+
+    // One album backed up, then a folder the user made beside it holding a
+    // symlink into originals/, with `json` (if given) as its sibling JSON.
+    const backupWithUserFolder = async (
+        json: string | undefined,
+    ): Promise<{ outDir: string; failed: number }> => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [file(1, 10, "a.jpg")],
+            },
+        ]);
+        await runBackup(lib, { downloadDirectory: outDir });
+        const collectionsDir = join(outDir, "collections");
+        mkdirSync(join(collectionsDir, "Mine"));
+        symlinkSync(
+            "../../originals/1.jpg",
+            join(collectionsDir, "Mine", "a.jpg"),
+        );
+        if (json !== undefined) {
+            writeFileSync(join(collectionsDir, "Mine.json"), json);
+        }
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+        return { outDir, failed: result.failed };
+    };
+
+    it("leaves a user folder with no JSON beside it as it was", async () => {
+        const { outDir, failed } = await backupWithUserFolder(undefined);
+
+        expect(failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Mine/",
+            "Mine/a.jpg -> ../../originals/1.jpg",
+            "Trip/",
+            "Trip/a.jpg -> ../../originals/1.jpg",
+            "Trip.json",
+        ]);
+    });
+
+    it("leaves a user folder whose JSON has no album ID as it was", async () => {
+        const json = '{"name":"Mine"}';
+        const { outDir, failed } = await backupWithUserFolder(json);
+
+        expect(failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Mine/",
+            "Mine/a.jpg -> ../../originals/1.jpg",
+            "Mine.json",
+            "Trip/",
+            "Trip/a.jpg -> ../../originals/1.jpg",
+            "Trip.json",
+        ]);
+        expect(
+            readFileSync(join(outDir, "collections", "Mine.json"), "utf-8"),
+        ).toBe(json);
+    });
+});
+
+// A live photo, which Ente stores as one ZIP, is backed up as its image and
+// its video, which a photo viewer can open, beside a JSON file naming them,
+// and its album folder links both. These tests download a live photo ZIP
+// through the real download layer (test/live-photo.ts).
+describe("backup of live photos", () => {
+    // An account of one album, Trip (10), holding `files`.
+    class TripClient extends MockClient {
+        constructor(private readonly files: EnteFile[]) {
+            super();
+        }
+        override async collectionsSince(): Promise<CollectionsPage> {
+            return {
+                collections: [collection(10, "Trip")],
+                deleted: [],
+                cursor: 1,
+            };
+        }
+        override async filesSince(): Promise<FilesPage> {
+            return { files: this.files, deleted: [], cursor: 1 };
+        }
+    }
+
+    const open = (files: EnteFile[], bodies: Map<number, Uint8Array>) =>
+        openLibrary(cdnSource(bodies), new TripClient(files));
+
+    // What an earlier version stored for live photo 500: the ZIP under the
+    // image's name, and its link.
+    const earlierZIP = (outDir: string): void => {
+        mkdirSync(join(outDir, "originals"), { recursive: true });
+        mkdirSync(join(outDir, "collections", "Trip"), { recursive: true });
+        writeFileSync(join(outDir, "originals", "500.HEIC"), livePhotoZip());
+        symlinkSync(
+            "../../originals/500.HEIC",
+            join(outDir, "collections", "Trip", "IMG_0500.HEIC"),
+        );
+    };
+
+    const stored = ["500.heic", "500.json", "500.livephoto.json", "500.mov"];
+    const linked = [
+        "Trip/",
+        "Trip/IMG_0500.heic -> ../../originals/500.heic",
+        "Trip/IMG_0500.mov -> ../../originals/500.mov",
+        "Trip.json",
+    ];
+
+    it("stores a live photo as its image and its video and links both", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(500, 10, "IMG_0500.HEIC"),
+        );
+        const lib = await open([live], new Map([[500, body]]));
+        const outDir = join(root, "backup");
+        const originals = join(outDir, "originals");
+
+        const result = await lib.backup({ downloadDirectory: outDir });
+
+        expect(result).toMatchObject({ downloaded: 1, failed: 0 });
+        expect(readdirSync(originals).sort()).toEqual(stored);
+        expect(readFileSync(join(originals, "500.heic"))).toEqual(
+            Buffer.from(IMAGE),
+        );
+        expect(readFileSync(join(originals, "500.mov"))).toEqual(
+            Buffer.from(VIDEO),
+        );
+        expect(tree(outDir)).toEqual(linked);
+
+        // Both parts are there, so the next run fetches nothing.
+        const second = await lib.backup({ downloadDirectory: outDir });
+        expect(second).toMatchObject({ downloaded: 0, skipped: 1, failed: 0 });
+        await lib.close();
+    });
+
+    it("gives both links of each live photo their own names when titles clash", async () => {
+        const a = await asLivePhoto(file(500, 10, "IMG_0001.HEIC"));
+        const b = await asLivePhoto(file(501, 10, "IMG_0001.HEIC"));
+        const lib = await open(
+            [a.file, b.file],
+            new Map([
+                [500, a.body],
+                [501, b.body],
+            ]),
+        );
+        const outDir = join(root, "backup");
+
+        await lib.backup({ downloadDirectory: outDir });
+
+        expect(tree(outDir)).toEqual([
+            "Trip/",
+            "Trip/IMG_0001 (500).heic -> ../../originals/500.heic",
+            "Trip/IMG_0001 (500).mov -> ../../originals/500.mov",
+            "Trip/IMG_0001 (501).heic -> ../../originals/501.heic",
+            "Trip/IMG_0001 (501).mov -> ../../originals/501.mov",
+            "Trip.json",
+        ]);
+        await lib.close();
+    });
+
+    it("replaces the ZIP an earlier version stored, and its link", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(500, 10, "IMG_0500.HEIC"),
+        );
+        const outDir = join(root, "backup");
+        earlierZIP(outDir);
+        const lib = await open([live], new Map([[500, body]]));
+
+        const result = await lib.backup({ downloadDirectory: outDir });
+
+        expect(result).toMatchObject({ downloaded: 1, failed: 0 });
+        expect(readdirSync(join(outDir, "originals")).sort()).toEqual(stored);
+        expect(tree(outDir)).toEqual(linked);
+        await lib.close();
+    });
+
+    it("stores nothing for a live photo that fails its hash, and keeps what was there", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(500, 10, "IMG_0500.HEIC"),
+            livePhotoZip(),
+            "not:the recorded hash",
+        );
+        const outDir = join(root, "backup");
+        earlierZIP(outDir);
+        const lib = await open([live], new Map([[500, body]]));
+
+        const result = await lib.backup({ downloadDirectory: outDir });
+
+        expect(result).toMatchObject({ downloaded: 0, failed: 1 });
+        expect(result.errors.map((e) => e.fileID)).toEqual([500]);
+        expect(Object.keys(readLedger(outDir).files)).toEqual(["500"]);
+        expect(readdirSync(join(outDir, "originals"))).toEqual(["500.HEIC"]);
+        expect(tree(outDir)).toEqual([
+            "Trip/",
+            "Trip/IMG_0500.HEIC -> ../../originals/500.HEIC",
+            "Trip.json",
+        ]);
+        await lib.close();
+    });
+
+    it("copies both parts of a live photo the cache already holds", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(500, 10, "IMG_0500.HEIC"),
+        );
+        const lib = await open([live], new Map([[500, body]]));
+        const cached = await lib.photos.byID({ fileID: 500 })!.original();
+        const outDir = join(root, "backup");
+
+        const result = await lib.backup({ downloadDirectory: outDir });
+
+        expect(result).toMatchObject({ downloaded: 1, failed: 0 });
+        expect(readdirSync(join(outDir, "originals")).sort()).toEqual(stored);
+        expect(readFileSync(join(outDir, "originals", "500.mov"))).toEqual(
+            Buffer.from(VIDEO),
+        );
+        expect(tree(outDir)).toEqual(linked);
+        // The cache keeps its own copy.
+        expect(existsSync(cached.videoPath!)).toBe(true);
+        await lib.close();
+    });
+
+    it("serves a live photo the backup stored to a library reading the backup", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(500, 10, "IMG_0500.HEIC"),
+        );
+        const lib = await open([live], new Map([[500, body]]));
+        const outDir = join(root, "backup");
+        await lib.backup({ downloadDirectory: outDir });
+        await lib.close();
+
+        // Another cache over the same backup, whose server has nothing.
+        const reader = await Library.open({
+            client: new TripClient([live]),
+            cacheDirectory: join(root, "other-cache"),
+            downloadDirectory: outDir,
+            contentSource: cdnSource(new Map()),
+            refreshIntervalSeconds: 3600,
+            precacheThumbnails: false,
+            precacheOriginals: false,
+        });
+        const read = await reader.photos.byID({ fileID: 500 })!.original();
+
+        expect(read).toEqual({
+            path: join(outDir, "originals", "500.heic"),
+            videoPath: join(outDir, "originals", "500.mov"),
+            bytes: IMAGE.length,
+        });
+        await reader.close();
+    });
+});

test/cli/commands.test.ts

@@ -12,7 +12,9 @@
 
 import {
     existsSync,
+    mkdirSync,
     mkdtempSync,
+    readdirSync,
     readFileSync,
     rmSync,
     statSync,
@@ -20,11 +22,22 @@ import {
 } from "node:fs";
 import { join } from "node:path";
 import { tmpdir } from "node:os";
-import { describe, it, expect, beforeAll, beforeEach, afterEach } from "vitest";
+import { PassThrough } from "node:stream";
+import * as jpegJs from "jpeg-js";
+import {
+    describe,
+    it,
+    expect,
+    vi,
+    beforeAll,
+    beforeEach,
+    afterEach,
+} from "vitest";
 
 import {
     type CliContext,
     saveSession,
+    loginCommand,
     whoamiCommand,
     logoutCommand,
     collectionsCommand,
@@ -32,13 +45,25 @@ import {
     getCommand,
     getThumbCommand,
     backupCommand,
+    backupMetadataCommand,
     listMissingThumbnailsCommand,
+    fixMissingThumbnailsCommand,
 } from "../../src/cli-commands.js";
+import { run } from "../../src/cli-run.js";
 import { loadSession } from "../../src/cli-session.js";
-import type { Client, ClientSnapshot } from "../../src/client.js";
+import type { Client, ClientSnapshot, LoginOptions } from "../../src/client.js";
 import type { ContentSource } from "../../src/library/content.js";
 import type { Collection, EnteFile } from "../../src/model/types.js";
-import { init } from "../../src/crypto/index.js";
+import { init, toBase64 } from "../../src/crypto/index.js";
+import { defaultCacheDirectory } from "../../src/library/index.js";
+import {
+    asLivePhoto,
+    cdnSource,
+    IMAGE,
+    livePhotoHash,
+    livePhotoZip,
+    VIDEO,
+} from "../live-photo.js";
 
 const USER_ID = 42;
 
@@ -81,8 +106,26 @@ const FILES: Record<number, EnteFile[]> = {
 
 // An original is 7 bytes and a thumbnail 3. `failID` makes that file's
 // original fail; `emptyThumbID` makes the server report that file's
-// thumbnail as empty.
+// thumbnail as empty. `withNewFile` adds new.jpg (102) to Vacation, advancing
-const fakeClient = (opts: { failID?: number; emptyThumbID?: number } = {}) => {
+// the collection's updationTime as the server does, and `refreshError` makes
+// listing collections fail with that message.
+const fakeClient = (
+    opts: {
+        failID?: number;
+        emptyThumbID?: number;
+        withNewFile?: boolean;
+        refreshError?: string;
+    } = {},
+) => {
+    const collections = opts.withNewFile
+        ? [{ ...COLLECTIONS[0], updationTime: 2 }, COLLECTIONS[1]]
+        : COLLECTIONS;
+    const files = opts.withNewFile
+        ? {
+              ...FILES,
+              1: [...FILES[1], { ...file(102, 1, "new.jpg"), updationTime: 2 }],
+          }
+        : FILES;
     const source: ContentSource = {
         original: async ({ file: f, destination }) => {
             if (f.id === opts.failID) throw new Error("HTTP 500 from server");
@@ -96,18 +139,19 @@ const fakeClient = (opts: { failID?: number; emptyThumbID?: number } = {}) => {
     };
     const fake = {
         whoami: () => ({ email: "cli@example.com", userID: USER_ID }),
-        collectionsSince: async () => ({
+        collectionsSince: async () => {
-            collections: COLLECTIONS,
+            if (opts.refreshError) throw new Error(opts.refreshError);
-            deleted: [],
+            return { collections, deleted: [], cursor: 1 };
-            cursor: 1,
+        },
-        }),
         filesSince: async (args: { collectionID: number }) => ({
-            files: FILES[args.collectionID] ?? [],
+            files: files[args.collectionID] ?? [],
             deleted: [],
             cursor: 1,
         }),
         contentSource: () => source,
         getApiClient: () => ({
+            // The ML data request of `backup-metadata`: no file has any.
+            postJSON: async () => ({ data: [] }),
             getThumbnailStream: async (fileID: number) =>
                 new ReadableStream<Uint8Array>({
                     start(controller) {
@@ -141,6 +185,15 @@ const context = (client: Client | null = fakeClient()): CliContext => ({
     sessionDir: join(root, "session"),
     cacheDir: join(root, "cache"),
     loadSession: () => client,
+    login: async () => {
+        throw new Error("login not expected");
+    },
+    prompt: async () => {
+        throw new Error("prompt not expected");
+    },
+    promptSecret: async () => {
+        throw new Error("prompt not expected");
+    },
 });
 
 beforeAll(async () => {
@@ -176,19 +229,6 @@ describe("session file", () => {
         expect(JSON.parse(readFileSync(path, "utf-8"))).toEqual(snapshot);
     });
 
-    it("is removed by logout", async () => {
-        const ctx = context();
-        saveSession(ctx.sessionDir, snapshot);
-        expect(await logoutCommand(ctx)).toBe(0);
-        expect(existsSync(join(ctx.sessionDir, "session.json"))).toBe(false);
-        expect(stderr.text).toBe("Session deleted.\n");
-    });
-
-    it("logout without a session says so and exits 0", async () => {
-        expect(await logoutCommand(context())).toBe(0);
-        expect(stderr.text).toBe("No session found.\n");
-    });
-
     it("a missing session exits 1 with 'Not logged in'", async () => {
         const ctx = { ...context(), loadSession };
         expect(await whoamiCommand(ctx)).toBe(1);
@@ -211,6 +251,184 @@ describe("session file", () => {
     });
 });
 
+// The login function is a fake that hands back a client whose snapshot is
+// `snapshot`; each prompt is recorded and answered with "123456".
+describe("login", () => {
+    const snapshot: ClientSnapshot = {
+        email: "cli@example.com",
+        userID: USER_ID,
+        token: "token",
+        masterKey: "a",
+        secretKey: "b",
+        publicKey: "c",
+    };
+
+    const loggedIn = {
+        whoami: () => ({ email: "cli@example.com", userID: USER_ID }),
+        toJSON: () => snapshot,
+    } as unknown as Client;
+
+    let prompts: string[];
+
+    const loginContext = (
+        login: (opts: LoginOptions) => Promise<Client>,
+    ): CliContext => ({
+        ...context(),
+        login,
+        prompt: async (message) => {
+            prompts.push(message);
+            return "123456";
+        },
+        promptSecret: async (message) => {
+            prompts.push(message);
+            return "123456";
+        },
+    });
+
+    beforeEach(() => {
+        prompts = [];
+        vi.stubEnv("QUAK_EMAIL", "cli@example.com");
+        vi.stubEnv("QUAK_PASSWORD", "hunter2");
+    });
+
+    afterEach(() => {
+        vi.unstubAllEnvs();
+    });
+
+    it("takes the email and password from the environment without a prompt", async () => {
+        const calls: LoginOptions[] = [];
+        const ctx = loginContext(async (opts) => {
+            calls.push(opts);
+            return loggedIn;
+        });
+        expect(await loginCommand(ctx)).toBe(0);
+
+        expect(prompts).toEqual([]);
+        expect(calls).toHaveLength(1);
+        expect(calls[0]!.email).toBe("cli@example.com");
+        expect(calls[0]!.password).toBe("hunter2");
+        const path = join(ctx.sessionDir, "session.json");
+        expect(stderr.text).toBe(
+            "Authenticating...\n" +
+                `Logged in as cli@example.com (user ${USER_ID})\n` +
+                `Session saved to ${path}\n`,
+        );
+    });
+
+    it("saves the session with mode 0600 in a directory with mode 0700", async () => {
+        const ctx = loginContext(async () => loggedIn);
+        expect(await loginCommand(ctx)).toBe(0);
+
+        expect(statSync(ctx.sessionDir).mode & 0o777).toBe(0o700);
+        const path = join(ctx.sessionDir, "session.json");
+        expect(statSync(path).mode & 0o777).toBe(0o600);
+        expect(JSON.parse(readFileSync(path, "utf-8"))).toEqual(snapshot);
+    });
+
+    it("asks for the TOTP code when the account needs one", async () => {
+        let code: string | undefined;
+        const ctx = loginContext(async (opts) => {
+            code = await opts.totp!();
+            return loggedIn;
+        });
+        expect(await loginCommand(ctx)).toBe(0);
+
+        expect(prompts).toEqual(["TOTP code: "]);
+        expect(code).toBe("123456");
+    });
+
+    it("a failed login exits 1, says why and writes no session", async () => {
+        const ctx = loginContext(async () => {
+            throw new Error("HTTP 401 from server");
+        });
+        expect(await loginCommand(ctx)).toBe(1);
+
+        expect(stderr.text).toBe(
+            "Authenticating...\nLogin failed: HTTP 401 from server\n",
+        );
+        expect(existsSync(join(ctx.sessionDir, "session.json"))).toBe(false);
+    });
+});
+
+// These use a real client read from the session file, over a fake API that
+// records each request and answers with `status`.
+describe("logout", () => {
+    const snapshot: ClientSnapshot = {
+        email: "cli@example.com",
+        userID: USER_ID,
+        token: "saved-token",
+        masterKey: toBase64(new Uint8Array(32)),
+        secretKey: toBase64(new Uint8Array(32)),
+        publicKey: toBase64(new Uint8Array(32)),
+    };
+
+    const requests: Request[] = [];
+
+    const logoutContext = (status: number): CliContext => ({
+        ...context(),
+        loadSession: (path) =>
+            loadSession(path, {
+                fetch: async (url, init) => {
+                    requests.push(new Request(url, init));
+                    return new Response(JSON.stringify({}), {
+                        status,
+                        headers: { "content-type": "application/json" },
+                    });
+                },
+            }),
+    });
+
+    beforeEach(() => {
+        requests.length = 0;
+    });
+
+    it("ends the session on the server, then deletes the file", async () => {
+        const ctx = logoutContext(200);
+        saveSession(ctx.sessionDir, snapshot);
+        expect(await logoutCommand(ctx)).toBe(0);
+
+        expect(requests).toHaveLength(1);
+        expect(requests[0]!.method).toBe("POST");
+        expect(new URL(requests[0]!.url).pathname).toBe("/users/logout");
+        expect(requests[0]!.headers.get("X-Auth-Token")).toBe("saved-token");
+        expect(existsSync(join(ctx.sessionDir, "session.json"))).toBe(false);
+        expect(stderr.text).toBe(
+            "Session ended on the server.\n" +
+                "Session deleted.\n" +
+                `Cache directory ${ctx.cacheDir} still holds decrypted data; delete it to remove that data.\n`,
+        );
+    });
+
+    it("still deletes the file when the server call fails, and says so", async () => {
+        const ctx = logoutContext(500);
+        saveSession(ctx.sessionDir, snapshot);
+        expect(await logoutCommand(ctx)).toBe(1);
+
+        expect(requests).toHaveLength(1);
+        expect(existsSync(join(ctx.sessionDir, "session.json"))).toBe(false);
+        expect(stderr.text).toBe(
+            "Could not end the session on the server: HTTP 500\n" +
+                "Session deleted.\n" +
+                `Cache directory ${ctx.cacheDir} still holds decrypted data; delete it to remove that data.\n`,
+        );
+    });
+
+    it("names the account's default cache directory without --cache-dir", async () => {
+        const ctx = { ...logoutContext(200), cacheDir: undefined };
+        saveSession(ctx.sessionDir, snapshot);
+        expect(await logoutCommand(ctx)).toBe(0);
+        expect(stderr.text).toContain(
+            `Cache directory ${defaultCacheDirectory(USER_ID)} still holds decrypted data`,
+        );
+    });
+
+    it("without a session says so, calls nothing and exits 0", async () => {
+        expect(await logoutCommand(logoutContext(200))).toBe(0);
+        expect(requests).toHaveLength(0);
+        expect(stderr.text).toBe("No session found.\n");
+    });
+});
+
 describe("whoami", () => {
     it("prints the account as one line of JSON", async () => {
         expect(await whoamiCommand(context())).toBe(0);
@@ -323,6 +541,120 @@ describe("get and get-thumb", () => {
     });
 });
 
+// A live photo, which Ente stores as one ZIP, is written as its image and its
+// video, which a photo viewer can open. The account here is Vacation holding
+// one live photo, 300, downloaded through the real download layer
+// (test/live-photo.ts).
+describe("a live photo", () => {
+    const livePhotoClient = async (image = IMAGE): Promise<Client> => {
+        const { file: live, body } = await asLivePhoto(
+            file(300, 1, "IMG_0300.HEIC"),
+            livePhotoZip({ "image.heic": image, "video.mov": VIDEO }),
+            livePhotoHash(image, VIDEO),
+        );
+        const fake = {
+            whoami: () => ({ email: "cli@example.com", userID: USER_ID }),
+            collectionsSince: async () => ({
+                collections: [collection(1, "Vacation")],
+                deleted: [],
+                cursor: 1,
+            }),
+            filesSince: async () => ({ files: [live], deleted: [], cursor: 1 }),
+            contentSource: () => cdnSource(new Map([[300, body]])),
+            // The ML data request of `backup-metadata`: no file has any.
+            getApiClient: () => ({ postJSON: async () => ({ data: [] }) }),
+        };
+        return fake as unknown as Client;
+    };
+
+    it("get writes its image and video, named after the title with their own extensions", async () => {
+        const ctx = context(await livePhotoClient());
+        const dir = join(root, "cwd");
+        mkdirSync(dir);
+        const previous = process.cwd();
+        process.chdir(dir);
+        try {
+            expect(await getCommand(ctx, "300", {})).toBe(0);
+        } finally {
+            process.chdir(previous);
+        }
+
+        expect(readdirSync(dir).sort()).toEqual([
+            "IMG_0300.heic",
+            "IMG_0300.mov",
+        ]);
+        expect(readFileSync(join(dir, "IMG_0300.heic"))).toEqual(
+            Buffer.from(IMAGE),
+        );
+        expect(readFileSync(join(dir, "IMG_0300.mov"))).toEqual(
+            Buffer.from(VIDEO),
+        );
+        expect(stderr.text).toBe(
+            `${IMAGE.length} bytes -> IMG_0300.heic\n` +
+                `${VIDEO.length} bytes -> IMG_0300.mov\n`,
+        );
+    });
+
+    it("get --out writes the image there and the video beside it", async () => {
+        const out = join(root, "photo.jpg");
+        const video = join(root, "photo.mov");
+
+        expect(
+            await getCommand(context(await livePhotoClient()), "300", { out }),
+        ).toBe(0);
+
+        expect(readFileSync(out)).toEqual(Buffer.from(IMAGE));
+        expect(readFileSync(video)).toEqual(Buffer.from(VIDEO));
+        expect(stderr.text).toBe(
+            `${IMAGE.length} bytes -> ${out}\n${VIDEO.length} bytes -> ${video}\n`,
+        );
+    });
+
+    it("get exits 1 and writes nothing when --out has the video's extension", async () => {
+        const out = join(root, "photo.MOV");
+
+        expect(
+            await getCommand(context(await livePhotoClient()), "300", { out }),
+        ).toBe(1);
+
+        expect(existsSync(out)).toBe(false);
+        expect(existsSync(join(root, "photo.mov"))).toBe(false);
+        expect(stderr.text).toBe(
+            `File 300 is a live photo, and its video would also be written to ${out}\n`,
+        );
+    });
+
+    it("backup-metadata --exif reads its image", async () => {
+        // A 4x4 JPEG, whose size can only come from reading the image; the
+        // ZIP and the video are not JPEGs.
+        const jpeg = jpegJs.encode(
+            { data: new Uint8Array(4 * 4 * 4), width: 4, height: 4 },
+            50,
+        ).data;
+        const dir = join(root, "dump");
+
+        expect(
+            await backupMetadataCommand(
+                context(await livePhotoClient(new Uint8Array(jpeg))),
+                dir,
+                { exif: true },
+            ),
+        ).toBe(0);
+
+        const record = JSON.parse(
+            readFileSync(
+                join(dir, "collections", "1-Vacation", "300.json"),
+                "utf-8",
+            ),
+        );
+        expect(record.imageMetadata).toMatchObject({
+            format: "jpeg",
+            width: 4,
+            height: 4,
+        });
+    });
+});
+
 describe("backup", () => {
     it("exits 0 and prints a summary when every file is saved", async () => {
         const dir = join(root, "backup");
@@ -337,6 +669,16 @@ describe("backup", () => {
         expect(stdout.text).toBe("");
     });
 
+    // The backup opens its library with the precache off: it fetches the
+    // originals it needs into the backup, and must not also fetch every
+    // thumbnail in the account, or keep originals, in the per-user cache.
+    it("leaves nothing in the cache's originals and thumbnails", async () => {
+        const dir = join(root, "backup");
+        expect(await backupCommand(context(), dir, {})).toBe(0);
+        expect(readdirSync(join(root, "cache", "thumbnails"))).toEqual([]);
+        expect(readdirSync(join(root, "cache", "originals"))).toEqual([]);
+    });
+
     it("exits 1 and lists the file when one download fails", async () => {
         const ctx = context(fakeClient({ failID: 101 }));
         expect(await backupCommand(ctx, join(root, "backup"), {})).toBe(1);
@@ -363,6 +705,34 @@ describe("backup", () => {
         expect(result.errors[0].fileID).toBe(101);
         expect(stderr.text).toBe("Starting backup...\n");
     });
+
+    it("exits 1 with the error on one line when the refresh fails", async () => {
+        const client = {
+            ...fakeClient(),
+            collectionsSince: async () => {
+                throw new Error("HTTP 401 from server");
+            },
+        } as unknown as Client;
+        const dir = join(root, "backup");
+        // Through `run`, as `bin/quak.ts` does, which prints a thrown error.
+        const runStderr = new PassThrough();
+        let runText = "";
+        runStderr.on("data", (chunk: Buffer) => {
+            runText += chunk.toString();
+        });
+        const code = await new Promise<number>((resolve) => {
+            void run(
+                backupCommand(context(client), dir, {}),
+                new PassThrough(),
+                runStderr,
+                resolve,
+            );
+        });
+        expect(code).toBe(1);
+        expect(runText).toBe("quak: HTTP 401 from server\n");
+        expect(stderr.text).toBe("Starting backup...\nRefreshing library...\n");
+        expect(existsSync(join(dir, "originals"))).toBe(false);
+    });
 });
 
 describe("helper list-missing-thumbnails", () => {
@@ -395,3 +765,92 @@ describe("helper list-missing-thumbnails", () => {
         expect(stderr.text).toBe("");
     });
 });
+
+describe("backup-metadata --exif", () => {
+    // Runs the command and returns what it printed to stderr.
+    const backupMetadata = async (opts: { exif?: boolean; all?: boolean }) => {
+        expect(
+            await backupMetadataCommand(context(), join(root, "dump"), opts),
+        ).toBe(0);
+        return stderr.text;
+    };
+
+    it("--exif extracts EXIF", async () => {
+        expect(await backupMetadata({ exif: true })).toContain(
+            "[beach.jpg] Extracting EXIF...\n",
+        );
+    });
+
+    it("--all extracts EXIF", async () => {
+        expect(await backupMetadata({ all: true })).toContain(
+            "[beach.jpg] Extracting EXIF...\n",
+        );
+    });
+
+    it("without either flag extracts no EXIF", async () => {
+        expect(await backupMetadata({})).not.toContain("Extracting EXIF");
+    });
+});
+
+// Each test first runs `collections` so the cache holds the account as it was,
+// then changes the server under it.
+describe("backup-metadata and the thumbnail helpers refresh first", () => {
+    beforeEach(async () => {
+        expect(await collectionsCommand(context(), {})).toBe(0);
+        stdout.text = "";
+        stderr.text = "";
+    });
+
+    it("backup-metadata writes a file added since the cache was written", async () => {
+        const ctx = context(fakeClient({ withNewFile: true }));
+        const dir = join(root, "dump");
+        expect(await backupMetadataCommand(ctx, dir, {})).toBe(0);
+        expect(
+            existsSync(join(dir, "collections", "1-Vacation", "102.json")),
+        ).toBe(true);
+    });
+
+    it("list-missing-thumbnails checks a file added since the cache was written", async () => {
+        const ctx = context(
+            fakeClient({ withNewFile: true, emptyThumbID: 102 }),
+        );
+        expect(await listMissingThumbnailsCommand(ctx, {})).toBe(0);
+        expect(stdout.text).toBe(
+            "102\tnew.jpg\tVacation\tempty thumbnail (0 bytes)\n",
+        );
+    });
+
+    it("fix-missing-thumbnails finds a file added since the cache was written", async () => {
+        const ctx = context(fakeClient({ withNewFile: true }));
+        expect(
+            await fixMissingThumbnailsCommand(ctx, {
+                file: ["102"],
+                json: true,
+            }),
+        ).toBe(0);
+        // Found, then skipped because the server records no thumbnail size
+        // for it; a file missing from the cache would fail as not found.
+        expect(JSON.parse(stdout.text)).toMatchObject([
+            { fileID: 102, title: "new.jpg", status: "skipped" },
+        ]);
+    });
+
+    // `run` in `cli-run.ts` prints a thrown error as one line and exits 1.
+    it("all three throw when the refresh fails", async () => {
+        const ctx = context(
+            fakeClient({ refreshError: "HTTP 503 from server" }),
+        );
+        const dir = join(root, "dump");
+        await expect(backupMetadataCommand(ctx, dir, {})).rejects.toThrow(
+            "HTTP 503 from server",
+        );
+        expect(existsSync(dir)).toBe(false);
+        await expect(listMissingThumbnailsCommand(ctx, {})).rejects.toThrow(
+            "HTTP 503 from server",
+        );
+        await expect(
+            fixMissingThumbnailsCommand(ctx, { file: ["100"] }),
+        ).rejects.toThrow("HTTP 503 from server");
+        expect(stdout.text).toBe("");
+    });
+});

test/cli/metadata-backup.test.ts

@@ -38,7 +38,7 @@ import { join } from "node:path";
 import { tmpdir } from "node:os";
 import sodium from "libsodium-wrappers-sumo";
 import { SRP, SrpServer } from "fast-srp-hap";
-import { beforeAll, afterAll, describe, expect, it } from "vitest";
+import { beforeAll, afterAll, describe, expect, it, vi } from "vitest";
 import {
     init,
     toBase64,
@@ -53,8 +53,16 @@ import {
     runMetadataBackup,
     type MetadataBackupOptions,
 } from "../../src/metadata-backup.js";
+import { backupMetadataCommand } from "../../src/cli-commands.js";
 import type { KeyAttributes } from "../../src/auth/types.js";
 
+// One file per ML data request, so the two files of the mock account are
+// fetched in two requests and one of them can fail on its own.
+vi.mock("../../src/mldata-fetch.js", async (importOriginal) => ({
+    ...(await importOriginal<typeof import("../../src/mldata-fetch.js")>()),
+    MLDATA_BATCH_SIZE: 1,
+}));
+
 const TEST_EMAIL = "metabackup@example.com";
 const TEST_PASSWORD = "metapass";
 const TEST_OPS = 2;
@@ -347,7 +355,8 @@ const buildMetaMock = async (): Promise<MetaMockState> => {
     };
 };
 
-const buildMetaFetch = (m: MetaMockState) => {
+// `failMLDataFor`: answer 500 to every ML data request that asks for this file.
+const buildMetaFetch = (m: MetaMockState, failMLDataFor?: number) => {
     let srpServer: SrpServer;
     return (async (
         input: RequestInfo | URL,
@@ -403,6 +412,8 @@ const buildMetaFetch = (m: MetaMockState) => {
         }
         if (path === "/files/data/fetch") {
             const body = JSON.parse(init?.body as string);
+            if ((body.fileIDs as number[]).includes(failMLDataFor!))
+                return new Response("server error", { status: 500 });
             const data = (body.fileIDs as number[])
                 .filter((id: number) => m.encryptedMLData[id])
                 .map((id: number) => ({
@@ -636,3 +647,63 @@ describe("quak backup-metadata", () => {
         expect(failedMeta.imageMetadataError).toEqual(expect.any(String));
     });
 });
+
+describe("quak backup-metadata when an ML data request fails", () => {
+    // Run the CLI command against the mock and return its exit code, stderr
+    // and output directory.
+    const runCommand = async (failMLDataFor?: number) => {
+        const client = await Client.login({
+            email: TEST_EMAIL,
+            password: TEST_PASSWORD,
+            apiOptions: {
+                fetch: buildMetaFetch(mock, failMLDataFor),
+                retry: { sleep: async () => {} },
+            },
+        });
+        const outDir = mkdtempSync(join(testDir, "ml-fail-"));
+        let stderr = "";
+        const code = await backupMetadataCommand(
+            {
+                stdout: { write: () => true },
+                stderr: { write: (text: string) => (stderr += text) },
+                sessionDir: testDir,
+                cacheDir: mkdtempSync(join(testDir, "cache-")),
+                loadSession: () => client,
+            },
+            outDir,
+            {},
+        );
+        return { code, stderr, outDir };
+    };
+
+    it("writes every file, marks the failed batch's files, and exits 1", async () => {
+        const { code, stderr, outDir } = await runCommand(200);
+
+        expect(code).toBe(1);
+        expect(stderr).toContain("ML data request for 1 file(s) failed");
+
+        const ok = JSON.parse(
+            readFileSync(
+                join(outDir, "collections", "10-Vacation", "100.json"),
+                "utf-8",
+            ),
+        );
+        expect(ok.mlData.clip.embedding).toEqual([0.5, 0.6, 0.7]);
+        expect(ok.mlDataError).toBeUndefined();
+
+        const failed = JSON.parse(
+            readFileSync(
+                join(outDir, "collections", "20-__Work", "200.json"),
+                "utf-8",
+            ),
+        );
+        expect(failed.metadata.title).toBe("diagram.png");
+        expect(failed.mlData).toBeUndefined();
+        expect(failed.mlDataError).toContain("500");
+    });
+
+    it("exits 0 when every ML data request succeeds", async () => {
+        const { code } = await runCommand();
+        expect(code).toBe(0);
+    });
+});

test/cli/run.test.ts

@@ -0,0 +1,58 @@
+/**
+ * Tests for `run` in `src/cli-run.ts`, which every CLI command goes through.
+ */
+
+import { PassThrough } from "node:stream";
+import { describe, it, expect } from "vitest";
+
+import { run } from "../../src/cli-run.js";
+
+// A stream whose written text is kept in `text`; writes finish at once, so
+// nothing is left waiting to drain.
+const collector = (): { stream: PassThrough; text: () => string } => {
+    const stream = new PassThrough();
+    const chunks: string[] = [];
+    stream.on("data", (chunk: Buffer) => chunks.push(chunk.toString()));
+    return { stream, text: () => chunks.join("") };
+};
+
+const runToExit = async (
+    command: Promise<number>,
+): Promise<{ code: number; stdout: string; stderr: string }> => {
+    const stdout = collector();
+    const stderr = collector();
+    const code = await new Promise<number>((resolve) => {
+        void run(command, stdout.stream, stderr.stream, resolve);
+    });
+    return { code, stdout: stdout.text(), stderr: stderr.text() };
+};
+
+describe("run", () => {
+    it("exits with the code the command returns", async () => {
+        const result = await runToExit(Promise.resolve(3));
+        expect(result).toEqual({ code: 3, stdout: "", stderr: "" });
+    });
+
+    it("prints a thrown error as one line without a stack trace and exits 1", async () => {
+        const failing = async (): Promise<number> => {
+            throw new Error(
+                "ENOTDIR: not a directory, mkdir '/dev/null/x/originals'",
+            );
+        };
+        const result = await runToExit(failing());
+        expect(result.code).toBe(1);
+        expect(result.stdout).toBe("");
+        expect(result.stderr).toBe(
+            "quak: ENOTDIR: not a directory, mkdir '/dev/null/x/originals'\n",
+        );
+    });
+
+    it("prints a thrown value that is not an Error", async () => {
+        const result = await runToExit(Promise.reject("offline"));
+        expect(result).toEqual({
+            code: 1,
+            stdout: "",
+            stderr: "quak: offline\n",
+        });
+    });
+});

test/crypto/hash.test.ts

@@ -0,0 +1,33 @@
+import { beforeAll, describe, expect, it } from "vitest";
+import {
+    chunkHashFinal,
+    chunkHashInit,
+    chunkHashUpdate,
+    init,
+} from "../../src/crypto/index.js";
+
+beforeAll(async () => {
+    await init();
+});
+
+describe("content hash", () => {
+    // RFC 7693 Appendix A: BLAKE2b-512 of "abc".
+    const abc = Buffer.from(
+        "ba80a53f981c4d0d6a2797b69f12f6e94c212f14685ac4b74b12bb6fdbffa2d1" +
+            "7d87c5392aab792dc252d5de4533cc9518d38aa8dbf1925ab92386edd4009923",
+        "hex",
+    ).toString("base64");
+
+    it("is unkeyed BLAKE2b-512 in standard base64", () => {
+        const state = chunkHashInit();
+        chunkHashUpdate(state, new TextEncoder().encode("abc"));
+        expect(chunkHashFinal(state)).toBe(abc);
+    });
+
+    it("gives the same hash when the input arrives in chunks", () => {
+        const state = chunkHashInit();
+        chunkHashUpdate(state, new TextEncoder().encode("a"));
+        chunkHashUpdate(state, new TextEncoder().encode("bc"));
+        expect(chunkHashFinal(state)).toBe(abc);
+    });
+});

test/crypto/kdf.test.ts

@@ -29,10 +29,10 @@ describe("crypto.deriveKEK (Argon2id)", () => {
     });
 
     /**
-     * Cheap parameters used so the test suite stays under the 30-second
+     * Cheap parameters used so the test suite stays under the 90-second
-     * budget. The real production parameters Ente uses are larger
+     * `timeout` in the `test` phase of the `Dockerfile`. The real production
-     * (memLimit up to 1 GiB, opsLimit 3-16). The algorithm is the same
+     * parameters Ente uses are larger (memLimit up to 1 GiB, opsLimit 3-16).
-     * regardless of parameters.
+     * The algorithm is the same regardless of parameters.
      */
     const TEST_OPS = 2;
     const TEST_MEM = 64 * 1024 * 1024; // 64 MiB

test/download/download.test.ts

@@ -55,12 +55,14 @@ import {
     readFileSync,
     rmSync,
     mkdtempSync,
+    statSync,
     writeFileSync,
 } from "node:fs";
 import { dirname, join } from "node:path";
 import { tmpdir } from "node:os";
 import { createHash } from "node:crypto";
 import sodium from "libsodium-wrappers-sumo";
+import { zipSync } from "fflate";
 import {
     beforeAll,
     beforeEach,
@@ -80,6 +82,7 @@ import {
     writeAtomic,
 } from "../../src/download/index.js";
 import type { EnteFile, FileMetadata } from "../../src/model/types.js";
+import { IMAGE, livePhotoHash, livePhotoZip, VIDEO } from "../live-photo.js";
 
 // ---------------------------------------------------------------------------
 // Test helpers
@@ -105,6 +108,8 @@ import type { EnteFile, FileMetadata } from "../../src/model/types.js";
 const renameHook = vi.hoisted(() => ({
     calls: [] as { from: string; to: string; sourceExisted: boolean }[],
     failWith: null as Error | null,
+    // When set, only a rename to this path fails.
+    failTo: null as string | null,
 }));
 
 /**
@@ -180,7 +185,10 @@ vi.mock("node:fs/promises", async (importOriginal) => {
                 sourceExisted: sourceExists(from),
             });
             durabilityHook.events.push(`rename:${to}`);
-            if (renameHook.failWith !== null) {
+            if (
+                renameHook.failWith !== null &&
+                (renameHook.failTo === null || renameHook.failTo === to)
+            ) {
                 throw renameHook.failWith;
             }
             await actual.rename(from, to);
@@ -188,9 +196,34 @@ vi.mock("node:fs/promises", async (importOriginal) => {
     };
 });
 
+/**
+ * `chunkHashUpdate` is wrapped to record the length of every piece hashed, so
+ * a test can show that a live photo entry reaches the hash in pieces far
+ * smaller than the entry, rather than decompressed whole first.
+ */
+const hashHook = vi.hoisted(() => ({
+    lengths: [] as number[],
+}));
+
+vi.mock("../../src/crypto/index.js", async (importOriginal) => {
+    const actual =
+        await importOriginal<typeof import("../../src/crypto/index.js")>();
+    return {
+        ...actual,
+        chunkHashUpdate: (
+            ...args: Parameters<typeof actual.chunkHashUpdate>
+        ): void => {
+            hashHook.lengths.push(args[1].length);
+            actual.chunkHashUpdate(...args);
+        },
+    };
+});
+
 beforeEach(() => {
+    hashHook.lengths.length = 0;
     renameHook.calls.length = 0;
     renameHook.failWith = null;
+    renameHook.failTo = null;
     durabilityHook.events.length = 0;
     writeHook.writes.length = 0;
 });
@@ -223,7 +256,8 @@ afterAll(() => {
  * `sodium.randombytes_buf` goes through the wasm wrapper a byte at a time and
  * costs roughly 20 seconds for the 4 MiB chunk below — about two hundred
  * times what it costs to encrypt the same buffer, and on its own enough to
- * push `make test` past the 30-second cap in `script/test`. This loop fills
+ * push `make test` past the 90-second `timeout` in the `test` phase of the
+ * `Dockerfile`. This loop fills
  * 4 MiB in a few milliseconds.
  */
 const patternBytes = (length: number, seed: number): Uint8Array => {
@@ -1009,13 +1043,10 @@ describe.each(entryPoints)(
             expect(readdirSync(dir)).toEqual([]);
         });
 
-        // Root ignores directory permissions, so this cannot fail as root
+        // Root ignores directory permissions, so this fails when run as root.
-        // (the Docker test image runs as root).
+        // The `test` phase of the `Dockerfile` runs as the `node` user.
-        it.skipIf(process.getuid?.() === 0)(
+        it("fails without creating anything when the destination directory is not writable", async () => {
-            "fails without creating anything when the destination directory is not writable",
+            const key = sodium.crypto_secretstream_xchacha20poly1305_keygen();
-            async () => {
-                const key =
-                    sodium.crypto_secretstream_xchacha20poly1305_keygen();
             const { header, ciphertext } = encryptFileBody(
                 patternBytes(64, 34),
                 key,
@@ -1033,8 +1064,7 @@ describe.each(entryPoints)(
             }
 
             expect(readdirSync(dir)).toEqual([]);
-            },
+        });
-        );
     },
 );
 
@@ -1538,7 +1568,13 @@ describe("writeAtomic", () => {
         // directory so that new entry is on disk too. Do the directory fsync
         // before the rename, or skip it, and a crash can lose the rename.
         expect(durabilityHook.events).toHaveLength(3);
-        expect(durabilityHook.events[0]).toMatch(/^sync:w:.*\.tmp$/);
+        // The temp name carries this process's ID, so a library opening the
+        // same cache can tell a write in progress from a leftover.
+        const tempSync = durabilityHook.events[0]!;
+        expect(tempSync.startsWith(`sync:w:${dir}/`)).toBe(true);
+        expect(tempSync.slice(`sync:w:${dir}/`.length)).toMatch(
+            new RegExp(`^\\.quak-${process.pid}-[0-9a-f]{32}\\.tmp$`),
+        );
         expect(durabilityHook.events[1]).toBe(`rename:${dest}`);
         expect(durabilityHook.events[2]).toBe(`sync:r:${dir}`);
     });
@@ -1613,3 +1649,250 @@ describe.each(entryPoints)("$name progress", ({ name, download }) => {
         expectSameBytes(readFileSync(outPath), plaintext);
     });
 });
+
+// Node's own BLAKE2b-512 is the reference, so the tests below do not depend on
+// the code under test to compute what they expect.
+const blake2b = (bytes: Uint8Array): string =>
+    createHash("blake2b512").update(bytes).digest("base64");
+
+// Serve `plaintext` encrypted as file 999 with the given metadata, to be
+// written to `f.bin` in a fresh directory. Four responses are scripted so a
+// retried failure would show in `requests`.
+const setup = (plaintext: Uint8Array, metadata: Partial<FileMetadata>) => {
+    const key = sodium.crypto_secretstream_xchacha20poly1305_keygen();
+    const { header, ciphertext } = encryptFileBody(plaintext, key);
+    const file = buildMockEnteFile(key, header, header);
+    file.metadata = { ...file.metadata, ...metadata };
+    const body = { kind: "body", bytes: ciphertext } as const;
+    const { fetch, requests } = scriptedCdnFetch(body, body, body, body);
+    const api = new ApiClient({ fetch, retry: { ...noWait, attempts: 4 } });
+    const dir = mkdtempSync(join(testDir, "hash-"));
+    const outPath = join(dir, "f.bin");
+    return {
+        run: () => downloadFile(api, file, outPath),
+        api,
+        file,
+        dir,
+        outPath,
+        requests,
+    };
+};
+
+describe("downloadFile content hash", () => {
+    it("stores a file whose hash matches", async () => {
+        const plaintext = patternBytes(700, 80);
+        const t = setup(plaintext, { hash: blake2b(plaintext) });
+
+        await t.run();
+
+        expectSameBytes(readFileSync(t.outPath), plaintext);
+    });
+
+    it("rejects a mismatch, stores nothing, names the file and does not retry", async () => {
+        const t = setup(patternBytes(700, 80), {
+            hash: blake2b(patternBytes(700, 79)),
+        });
+
+        await expect(t.run()).rejects.toThrow(
+            /file 999: content hash .* does not match/,
+        );
+
+        expect(readdirSync(t.dir)).toEqual([]);
+        expect(t.requests()).toBe(1);
+    });
+
+    it("stores a file with no recorded hash unchecked", async () => {
+        const plaintext = patternBytes(700, 80);
+        const t = setup(plaintext, { hash: undefined });
+
+        await t.run();
+
+        expectSameBytes(readFileSync(t.outPath), plaintext);
+    });
+
+    it("stores a live photo whose image and video hashes match", async () => {
+        const t = setup(livePhotoZip(), {
+            fileType: "livePhoto",
+            hash: livePhotoHash(),
+        });
+
+        await t.run();
+
+        expect(readFileSync(join(t.dir, "f.heic"))).toEqual(Buffer.from(IMAGE));
+        expect(readFileSync(join(t.dir, "f.mov"))).toEqual(Buffer.from(VIDEO));
+    });
+
+    it("writes a large live photo entry as it decompresses, never whole", async () => {
+        // 64 MiB of zeros deflates to a few kilobytes, the shape of a ZIP
+        // that would exhaust memory if expanded whole.
+        const image = new Uint8Array(64 * 1024 * 1024);
+        const video = patternBytes(900, 83);
+        const zip = zipSync({ "image.heic": image, "video.mov": video });
+        const t = setup(zip, {
+            fileType: "livePhoto",
+            hash: livePhotoHash(image, video),
+        });
+
+        await t.run();
+
+        expect(statSync(join(t.dir, "f.heic")).size).toBe(image.length);
+        expectSameBytes(readFileSync(join(t.dir, "f.mov")), video);
+        const hashed = hashHook.lengths.reduce((a, b) => a + b, 0);
+        expect(hashed).toBe(image.length + video.length);
+        expect(Math.max(...hashHook.lengths)).toBeLessThanOrEqual(
+            2 * STREAM_CHUNK_SIZE,
+        );
+        const written = writeHook.writes.filter((w) => w.path.endsWith(".tmp"));
+        expect(Math.max(...written.map((w) => w.length))).toBeLessThanOrEqual(
+            2 * STREAM_CHUNK_SIZE,
+        );
+    });
+
+    it("rejects a live photo whose hash does not match, keeping what was there", async () => {
+        // The whole ZIP's hash is not the recorded one: each part is hashed.
+        const zip = livePhotoZip();
+        const t = setup(zip, { fileType: "livePhoto", hash: blake2b(zip) });
+        writeFileSync(t.outPath, "an earlier download");
+
+        await expect(t.run()).rejects.toThrow(
+            /file 999: content hash .* does not match/,
+        );
+
+        expect(readdirSync(t.dir)).toEqual(["f.bin"]);
+        expect(readFileSync(t.outPath, "utf-8")).toBe("an earlier download");
+    });
+
+    it("rejects a live photo that is not a readable ZIP and does not retry", async () => {
+        // Bytes 8-9 of a ZIP entry's local header name its compression
+        // method; 99 is one no reader knows, so the entry cannot be read.
+        const zip = livePhotoZip();
+        zip[8] = 99;
+        zip[9] = 0;
+        const t = setup(zip, { fileType: "livePhoto", hash: livePhotoHash() });
+
+        await expect(t.run()).rejects.toThrow(
+            /file 999: live photo is not a readable ZIP/,
+        );
+
+        expect(readdirSync(t.dir)).toEqual([]);
+        expect(t.requests()).toBe(1);
+    });
+
+    it("rejects a live photo ZIP with no image entry", async () => {
+        const zip = livePhotoZip({ "video.mov": VIDEO });
+        const t = setup(zip, { fileType: "livePhoto", hash: livePhotoHash() });
+
+        await expect(t.run()).rejects.toThrow(
+            /file 999: live photo ZIP does not hold both an image and a video/,
+        );
+
+        expect(readdirSync(t.dir)).toEqual([]);
+    });
+
+    it("rejects a live photo ZIP with no video entry", async () => {
+        const zip = livePhotoZip({ "image.heic": IMAGE });
+        const t = setup(zip, { fileType: "livePhoto", hash: livePhotoHash() });
+
+        await expect(t.run()).rejects.toThrow(
+            /file 999: live photo ZIP does not hold both an image and a video/,
+        );
+
+        expect(readdirSync(t.dir)).toEqual([]);
+    });
+});
+
+// ---------------------------------------------------------------------------
+// Live photos
+//
+// A live photo arrives as a ZIP of its image and its video. It is written as
+// those two files, which a photo viewer can open, each named after the
+// destination with its own extension from the ZIP, the way Ente's clients name
+// them when they save one.
+// ---------------------------------------------------------------------------
+
+describe("downloadFile live photos", () => {
+    const livePhoto = { fileType: "livePhoto", hash: livePhotoHash() } as const;
+
+    it("names the image and the video after the title when no outPath is given", async () => {
+        const t = setup(livePhotoZip(), {
+            ...livePhoto,
+            title: "IMG_1234.HEIC",
+        });
+
+        const result = await inDirectory(t.dir, () =>
+            downloadFile(t.api, t.file),
+        );
+
+        // `bytesWritten` is the length of the decrypted ZIP.
+        expect(result).toEqual({
+            path: "IMG_1234.heic",
+            videoPath: "IMG_1234.mov",
+            bytesWritten: livePhotoZip().length,
+        });
+        expect(readdirSync(t.dir).sort()).toEqual([
+            "IMG_1234.heic",
+            "IMG_1234.mov",
+        ]);
+    });
+
+    it("gives each part its own extension from the ZIP, letters and digits only", async () => {
+        const zip = livePhotoZip({ "image.JPG": IMAGE, "video.m-4v": VIDEO });
+        const t = setup(zip, livePhoto);
+
+        const result = await downloadFile(t.api, t.file, join(t.dir, "f.HEIC"));
+
+        expect(result.path).toBe(join(t.dir, "f.JPG"));
+        expect(result.videoPath).toBe(join(t.dir, "f.bin"));
+        expect(readdirSync(t.dir).sort()).toEqual(["f.JPG", "f.bin"]);
+    });
+
+    it("replaces what was at the destination, such as an earlier ZIP of the two", async () => {
+        const t = setup(livePhotoZip(), livePhoto);
+        writeFileSync(t.outPath, livePhotoZip());
+
+        await t.run();
+
+        expect(readdirSync(t.dir).sort()).toEqual(["f.heic", "f.mov"]);
+    });
+
+    it("renames the image and then the video into place, each from its own temp file", async () => {
+        const t = setup(livePhotoZip(), livePhoto);
+
+        await t.run();
+
+        expect(
+            renameHook.calls.map((c) => [
+                dirname(c.from),
+                c.to,
+                c.sourceExisted,
+            ]),
+        ).toEqual([
+            [t.dir, join(t.dir, "f.heic"), true],
+            [t.dir, join(t.dir, "f.mov"), true],
+        ]);
+    });
+
+    it("stores neither part when the video cannot be renamed into place", async () => {
+        const t = setup(livePhotoZip(), livePhoto);
+        renameHook.failWith = new Error("simulated rename failure");
+        renameHook.failTo = join(t.dir, "f.mov");
+
+        await expect(t.run()).rejects.toThrow("simulated rename failure");
+
+        expect(readdirSync(t.dir)).toEqual([]);
+    });
+
+    it("refuses an image and a video with the same extension, storing nothing", async () => {
+        // On a file system that ignores case, the two would be one file.
+        const zip = livePhotoZip({ "image.mov": IMAGE, "video.MOV": VIDEO });
+        const t = setup(zip, livePhoto);
+        writeFileSync(t.outPath, "an earlier download");
+
+        await expect(t.run()).rejects.toThrow(
+            /file 999: live photo's image and video have the same extension/,
+        );
+
+        expect(readdirSync(t.dir)).toEqual(["f.bin"]);
+        expect(t.requests()).toBe(1);
+    });
+});

test/library/content.test.ts

@@ -31,8 +31,11 @@ import {
     existsSync,
     writeFileSync,
     mkdirSync,
+    readdirSync,
+    readFileSync,
     statSync,
 } from "node:fs";
+import { spawnSync } from "node:child_process";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
 
@@ -43,6 +46,13 @@ import {
 } from "../../src/library/content.js";
 import { RequestPools } from "../../src/library/pools.js";
 import type { EnteFile } from "../../src/model/types.js";
+import {
+    asLivePhoto,
+    cdnSource,
+    IMAGE,
+    livePhotoZip,
+    VIDEO,
+} from "../live-photo.js";
 
 const file = (id: number, title = `file-${id}.jpg`): EnteFile => ({
     id,
@@ -161,15 +171,28 @@ describe("ContentCache.open", () => {
         expect(statSync(thumbnails).mode & 0o777).toBe(0o700);
     });
 
-    it("reaps orphan temp files but keeps complete content", async () => {
+    it("removes temp files of an exited process, keeping those of a running one and complete content", async () => {
         const originals = join(cacheDir, "originals");
         const thumbnails = join(cacheDir, "thumbnails");
         mkdirSync(originals, { recursive: true });
         mkdirSync(thumbnails, { recursive: true });
-        const orphan = join(originals, ".quak-abc123.tmp");
+        // A child that has already exited: its process ID is not running.
+        const exitedPID = spawnSync(process.execPath, ["-e", ""]).pid;
+        const orphan = join(originals, `.quak-${exitedPID}-abc123.tmp`);
+        const orphanThumb = join(thumbnails, `.quak-${exitedPID}-abc456.tmp`);
+        // This test's own process stands in for another process still
+        // downloading into the same cache.
+        const inProgress = join(originals, `.quak-${process.pid}-def123.tmp`);
+        const inProgressThumb = join(
+            thumbnails,
+            `.quak-${process.pid}-def456.tmp`,
+        );
         const complete = join(originals, "1.jpg");
         const thumb = join(thumbnails, "2.jpg");
         writeFileSync(orphan, "half-written");
+        writeFileSync(orphanThumb, "half-written");
+        writeFileSync(inProgress, "half-written");
+        writeFileSync(inProgressThumb, "half-written");
         writeFileSync(complete, "whole");
         writeFileSync(thumb, "whole-thumb");
 
@@ -177,8 +200,12 @@ describe("ContentCache.open", () => {
         await cache.open();
 
         expect(existsSync(orphan)).toBe(false);
+        expect(existsSync(orphanThumb)).toBe(false);
+        expect(existsSync(inProgress)).toBe(true);
+        expect(existsSync(inProgressThumb)).toBe(true);
         expect(existsSync(complete)).toBe(true);
         expect(existsSync(thumb)).toBe(true);
+        expect(cache.pathsFor(1)).toEqual({ originalPath: complete });
     });
 
     it("records already-cached files so their paths appear in pathsFor", async () => {
@@ -440,3 +467,125 @@ describe("ContentCache.ensureThumbnails", () => {
         expect(results[1]?.error).toMatch(/unknown file/i);
     });
 });
+
+// A live photo's original is two files, its image and its video, which a
+// photo viewer can open, and a JSON file naming them: the two are named with
+// the extensions from inside the ZIP, so the names alone do not say which is
+// which. These tests download a live photo ZIP through the real download
+// layer (test/live-photo.ts).
+describe("ContentCache live photos", () => {
+    const originals = (): string => join(cacheDir, "originals");
+
+    // A cache over the stand-in server, which holds `bodies` by file ID.
+    const cacheOf = (
+        files: EnteFile[],
+        bodies: Map<number, Uint8Array>,
+    ): ContentCache => buildCache({ files, source: cdnSource(bodies) }).cache;
+
+    it("stores a live photo as its image and its video and a JSON file naming them", async () => {
+        const { file: live, body } = await asLivePhoto(file(5, "IMG_5.HEIC"));
+        const cache = cacheOf([live], new Map([[5, body]]));
+        await cache.open();
+
+        const result = await cache.original(5);
+
+        expect(result).toEqual({
+            path: join(originals(), "5.heic"),
+            videoPath: join(originals(), "5.mov"),
+            bytes: IMAGE.length,
+        });
+        expect(readFileSync(result.path)).toEqual(Buffer.from(IMAGE));
+        expect(readFileSync(result.videoPath!)).toEqual(Buffer.from(VIDEO));
+        expect(statSync(result.videoPath!).mode & 0o777).toBe(0o600);
+        expect(
+            JSON.parse(
+                readFileSync(join(originals(), "5.livephoto.json"), "utf-8"),
+            ),
+        ).toEqual({ image: "5.heic", video: "5.mov" });
+        expect(cache.pathsFor(5)).toEqual({ originalPath: result.path });
+    });
+
+    it("serves a stored live photo from disk after the cache is opened again", async () => {
+        const { file: live, body } = await asLivePhoto(file(5, "IMG_5.HEIC"));
+        const first = cacheOf([live], new Map([[5, body]]));
+        await first.open();
+        const stored = await first.original(5);
+
+        // This server has nothing, so a fetch would fail.
+        const second = cacheOf([live], new Map());
+        await second.open();
+        const events: string[] = [];
+        const served = await second.original(5, {
+            onProgress: (e) => events.push(e.status),
+        });
+
+        expect(served).toEqual(stored);
+        expect(events).toEqual(["skipped"]);
+    });
+
+    it("replaces a live photo an earlier version stored as a ZIP under the image's name", async () => {
+        const { file: live, body } = await asLivePhoto(file(5, "IMG_5.HEIC"));
+        mkdirSync(originals(), { recursive: true });
+        writeFileSync(join(originals(), "5.HEIC"), livePhotoZip());
+        const cache = cacheOf([live], new Map([[5, body]]));
+        await cache.open();
+
+        const result = await cache.original(5);
+
+        expect(result.videoPath).toBe(join(originals(), "5.mov"));
+        expect(readdirSync(originals()).sort()).toEqual([
+            "5.heic",
+            "5.livephoto.json",
+            "5.mov",
+        ]);
+    });
+
+    it("evicts a live photo's image, video and JSON file together", async () => {
+        const a = await asLivePhoto(file(5, "a.HEIC"));
+        const b = await asLivePhoto(file(6, "b.HEIC"));
+        const size = IMAGE.length + VIDEO.length;
+        const cache = new ContentCache({
+            pools: new RequestPools(),
+            source: cdnSource(
+                new Map([
+                    [5, a.body],
+                    [6, b.body],
+                ]),
+            ),
+            cacheDirectory: cacheDir,
+            getFile: (id) => [a.file, b.file].find((f) => f.id === id),
+            // Room for one live photo, on a disk with plenty free.
+            cacheOriginalsMaxBytes: size,
+            freeBelowBytes: 0,
+            statfs: async () => ({ bsize: 1, bavail: 1e12 }),
+        });
+        await cache.open();
+
+        await cache.original(5);
+        await cache.original(6);
+
+        expect(readdirSync(originals()).sort()).toEqual([
+            "6.heic",
+            "6.livephoto.json",
+            "6.mov",
+        ]);
+        expect(cache.originalsStatus().usedBytes).toBe(size);
+    });
+
+    it("stores nothing when a live photo does not match its recorded hash", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(5, "IMG_5.HEIC"),
+            livePhotoZip(),
+            "not:the recorded hash",
+        );
+        const cache = cacheOf([live], new Map([[5, body]]));
+        await cache.open();
+
+        await expect(cache.original(5)).rejects.toThrow(
+            /file 5: content hash .* does not match/,
+        );
+
+        expect(readdirSync(originals())).toEqual([]);
+        expect(cache.pathsFor(5)).toEqual({});
+    });
+});

test/library/mldata.test.ts

@@ -13,6 +13,8 @@
  *   2. `Library` wiring: after each refresh the library fetches ML data through
  *      the metadata pool for every known file not yet cached, is incremental on
  *      later refreshes, and refetches a file whose `updationTime` advanced.
+ *      Opening a cache directory written by another account starts empty,
+ *      its ML data included (issue #104).
  *
  * Embedding values are chosen to be exactly representable as float32 so the
  * round-trip through `clip.f32` compares equal.
@@ -498,4 +500,63 @@ describe("Library ML-data fetch on refresh", () => {
             await lib.close();
         }
     });
+
+    it("starts empty when the cache directory holds another account's cache", async () => {
+        // Account A fills the cache directory: metadata and ML data.
+        const clientA = new MLMockClient();
+        clientA.collectionsQueue.push({
+            collections: [collection(1, 100)],
+            deleted: [],
+            cursor: 100,
+        });
+        clientA.filesFor(1, {
+            files: [file(1001, 1, 90)],
+            deleted: [],
+            cursor: 90,
+        });
+        clientA.mlByFile.set(1001, payload([0.5, 0.25, 0.75]));
+        const libA = await Library.open({
+            client: clientA,
+            cacheDirectory,
+            refreshIntervalSeconds: 3600,
+        });
+        try {
+            await vi.waitFor(
+                () => expect(libA.status().lastMLFetchAt).toBeGreaterThan(0),
+                { timeout: 2000, interval: 5 },
+            );
+        } finally {
+            await libA.close();
+        }
+
+        // Account B opens the same directory.
+        const clientB = new MLMockClient();
+        clientB.userID = USER_ID + 1;
+        const sinceTimes: number[] = [];
+        const realCollectionsSince = clientB.collectionsSince.bind(clientB);
+        clientB.collectionsSince = async (args) => {
+            sinceTimes.push(args.sinceTime);
+            return realCollectionsSince(args);
+        };
+        const libB = await Library.open({
+            client: clientB,
+            cacheDirectory,
+            refreshIntervalSeconds: 3600,
+        });
+        try {
+            expect(sinceTimes[0]).toBe(0);
+            expect(libB.status().userID).toBe(USER_ID + 1);
+            expect(libB.listCollections()).toEqual([]);
+            expect(libB.getFile(1, 1001)).toBeUndefined();
+            expect(await libB.mldata.forFile({ fileID: 1001 })).toBeUndefined();
+            expect(
+                libB.mldata.searchByEmbedding({ embedding: [0.5, 0.25, 0.75] }),
+            ).toEqual([]);
+            expect(
+                existsSync(join(cacheDirectory, "mldata", "1001.json")),
+            ).toBe(false);
+        } finally {
+            await libB.close();
+        }
+    });
 });

test/live-photo.ts

@@ -0,0 +1,77 @@
+/**
+ * Live photo fixtures for the download, content cache, backup and CLI tests.
+ *
+ * Ente stores a live photo as one ZIP holding its image and its video, which
+ * Ente's clients name `image.<ext>` and `video.<ext>`. The ZIP is built here
+ * with fflate from small fixed bytes and encrypted the way the server serves a
+ * file under 4 MiB, as one secretstream chunk. `cdnSource` serves it to the
+ * real download layer, so a test checks what quak stores for a real one.
+ */
+
+import { createHash } from "node:crypto";
+import { zipSync } from "fflate";
+
+import { ApiClient } from "../src/api/client.js";
+import { encryptBlob, init, toBase64 } from "../src/crypto/index.js";
+import {
+    makeDownloadContentSource,
+    type ContentSource,
+} from "../src/library/content.js";
+import type { EnteFile } from "../src/model/types.js";
+
+export const IMAGE = new TextEncoder().encode("the still image");
+export const VIDEO = new TextEncoder().encode("the few seconds of video");
+
+// A live photo's ZIP; by default the entries an iPhone's live photo gets.
+export const livePhotoZip = (
+    entries: Record<string, Uint8Array> = {
+        "image.heic": IMAGE,
+        "video.mov": VIDEO,
+    },
+): Uint8Array => zipSync(entries);
+
+const blake2b = (bytes: Uint8Array): string =>
+    createHash("blake2b512").update(bytes).digest("base64");
+
+// The hash Ente's clients record for a live photo: the unkeyed BLAKE2b-512 of
+// the image and of the video, each in standard base64, joined by a colon.
+export const livePhotoHash = (image = IMAGE, video = VIDEO): string =>
+    `${blake2b(image)}:${blake2b(video)}`;
+
+// `file` as a live photo whose original is `zip` and whose recorded hash is
+// `hash`, and `body`, what the server serves for it: `zip` encrypted under the
+// file's key and header.
+export const asLivePhoto = async (
+    file: EnteFile,
+    zip = livePhotoZip(),
+    hash = livePhotoHash(),
+): Promise<{ file: EnteFile; body: Uint8Array }> => {
+    await init();
+    const key = new Uint8Array(32).fill(file.id & 0xff);
+    const { header, ciphertext } = encryptBlob(zip, key);
+    return {
+        file: {
+            ...file,
+            key,
+            metadata: { ...file.metadata, fileType: "livePhoto", hash },
+            file: { decryptionHeader: toBase64(header) },
+        },
+        body: ciphertext,
+    };
+};
+
+// A content source that downloads through the real download layer from a
+// stand-in server, which serves `bodies` by file ID and a 404 for any other.
+export const cdnSource = (bodies: Map<number, Uint8Array>): ContentSource =>
+    makeDownloadContentSource(
+        new ApiClient({
+            fetch: (async (url: string | URL) => {
+                const fileID = new URL(String(url)).searchParams.get("fileID");
+                const body = bodies.get(Number(fileID));
+                return body === undefined
+                    ? new Response("not found", { status: 404 })
+                    : new Response(body);
+            }) as typeof globalThis.fetch,
+            retry: { attempts: 1 },
+        }),
+    );

test/model/decrypt.test.ts

@@ -351,6 +351,46 @@ describe("model.decryptFile", () => {
         }
     });
 
+    it("reads the recorded content hash, joining an older live photo's two parts", () => {
+        const masterKey = sodium.crypto_secretbox_keygen();
+        const { collectionKey } = buildRawCollection(masterKey);
+        const hashOf = (metadata: Record<string, unknown>) =>
+            decryptFile(
+                buildRawFile(collectionKey, {
+                    metadata: { title: "x", ...metadata },
+                }),
+                collectionKey,
+            ).metadata.hash;
+
+        expect(hashOf({ fileType: 0, hash: "H" })).toBe("H");
+        expect(
+            hashOf({ fileType: 2, hash: "H", imageHash: "I", videoHash: "V" }),
+        ).toBe("H");
+        expect(hashOf({ fileType: 2, imageHash: "I", videoHash: "V" })).toBe(
+            "I:V",
+        );
+        expect(hashOf({ fileType: 2, imageHash: "I" })).toBeUndefined();
+        expect(
+            hashOf({ fileType: 0, imageHash: "I", videoHash: "V" }),
+        ).toBeUndefined();
+        expect(hashOf({ fileType: 0 })).toBeUndefined();
+        expect(hashOf({ fileType: 0, hash: 42 })).toBeUndefined();
+        expect(
+            hashOf({ fileType: 2, imageHash: "I", videoHash: 7 }),
+        ).toBeUndefined();
+        // An empty string counts as absent, not as a hash to match.
+        expect(hashOf({ fileType: 0, hash: "" })).toBeUndefined();
+        expect(
+            hashOf({ fileType: 2, hash: "", imageHash: "I", videoHash: "V" }),
+        ).toBe("I:V");
+        expect(
+            hashOf({ fileType: 2, imageHash: "", videoHash: "V" }),
+        ).toBeUndefined();
+        expect(
+            hashOf({ fileType: 2, imageHash: "I", videoHash: "" }),
+        ).toBeUndefined();
+    });
+
     it("maps fileType numbers to FileType strings", () => {
         // Ente uses: 0=image, 1=video, 2=livePhoto
         const masterKey = sodium.crypto_secretbox_keygen();

test/thumbnails/thumbnails.test.ts

@@ -215,6 +215,9 @@ const buildThumbMock = async (opts?: {
             thumbnail: {
                 decryptionHeader: toBase64(sodium.randombytes_buf(24)),
             },
+            // The encrypted size of the thumbnail the server records; large
+            // enough here that the default encoding fits.
+            info: { thumbSize: 1_000_000 },
             updationTime: TEST_TIME,
         };
     };
@@ -446,6 +449,32 @@ const openLib = (client: Client): Promise<Library> =>
         precacheOriginals: false,
     });
 
+/** The mock's raw record for one file, for a test to change before login. */
+const rawFile = (m: ThumbMockState, fileID: number): Record<string, unknown> =>
+    m.filesByCollection[1]!.find((f) => f.id === fileID)!;
+
+/** Replace the original the mock serves for one file. */
+const replaceOriginal = (
+    m: ThumbMockState,
+    fileID: number,
+    body: Uint8Array,
+): void => {
+    const push = sodium.crypto_secretstream_xchacha20poly1305_init_push(
+        m.fileKeys[fileID]!,
+    );
+    m.fileCiphertexts[fileID] =
+        sodium.crypto_secretstream_xchacha20poly1305_push(
+            push.state,
+            body,
+            null,
+            sodium.crypto_secretstream_xchacha20poly1305_TAG_FINAL,
+        );
+    rawFile(m, fileID).file = { decryptionHeader: toBase64(push.header) };
+};
+
+const isOriginalDownload = (url: string): boolean =>
+    url.includes("files.ente.io") || url.includes("/files/download/");
+
 const login = (fetch: typeof globalThis.fetch, retry?: RetryOptions) =>
     Client.login({
         email: TEST_EMAIL,
@@ -581,6 +610,33 @@ describe("listMissingThumbnails", () => {
         // Should still be 2, not 4 (each file checked only once)
         expect(missing.length).toBe(2);
     });
+
+    it("skips a file another account owns without fetching its thumbnail", async () => {
+        const otherMock = await buildThumbMock();
+        rawFile(otherMock, 102).ownerID = 7;
+        const logs: string[] = [];
+        const counted = countingFetch(
+            buildThumbFetch(otherMock),
+            (url) => url.includes("thumbnails.ente.io") && url.includes("102"),
+        );
+        const client = await login(counted.fetch);
+        const lib = await openLib(client);
+
+        const missing = await listMissingThumbnails(lib, client, (msg) =>
+            logs.push(msg),
+        );
+        lib.close();
+
+        expect(missing.map((m) => m.fileID)).toEqual([101]);
+        expect(counted.matched()).toBe(0);
+        expect(
+            logs.some(
+                (l) =>
+                    l.includes("Skipping file-102.jpg") &&
+                    l.includes("another account"),
+            ),
+        ).toBe(true);
+    });
 });
 
 describe("fixMissingThumbnails", () => {
@@ -688,6 +744,93 @@ describe("fixMissingThumbnails", () => {
         expect(fixMock.uploadedThumbnails.length).toBe(1);
         expect(fixMock.uploadedThumbnails[0]!.fileID).toBe(101);
     });
+
+    it("skips a file another account owns without downloading it", async () => {
+        // The server accepts a thumbnail only from the file's owner.
+        const fixMock = await buildThumbMock();
+        rawFile(fixMock, 101).ownerID = 7;
+        const counted = countingFetch(
+            buildThumbFetch(fixMock),
+            isOriginalDownload,
+        );
+        const client = await login(counted.fetch);
+        const lib = await openLib(client);
+
+        const results = await fixMissingThumbnails(lib, client, [101]);
+        lib.close();
+
+        expect(results[0]!.status).toBe("skipped");
+        expect(results[0]!.reason).toContain("another account");
+        expect(counted.matched()).toBe(0);
+        expect(fixMock.uploadedThumbnails.length).toBe(0);
+    });
+
+    it("skips a file whose recorded thumbnail size is 0 without downloading it", async () => {
+        // The server refuses a thumbnail larger than the one it records, and
+        // no thumbnail is 0 bytes.
+        const fixMock = await buildThumbMock();
+        rawFile(fixMock, 101).info = { thumbSize: 0 };
+        const counted = countingFetch(
+            buildThumbFetch(fixMock),
+            isOriginalDownload,
+        );
+        const client = await login(counted.fetch);
+        const lib = await openLib(client);
+
+        const results = await fixMissingThumbnails(lib, client, [101]);
+        lib.close();
+
+        expect(results[0]!.status).toBe("skipped");
+        expect(results[0]!.reason).toContain("recorded thumbnail size is 0");
+        expect(counted.matched()).toBe(0);
+        expect(fixMock.uploadedThumbnails.length).toBe(0);
+    });
+
+    it("re-encodes smaller until the thumbnail fits the recorded size", async () => {
+        // A noisy 400x300 JPEG, which the default encoding (quality 50, not
+        // resized because it is under 720 px) cannot compress below the size
+        // recorded here: one byte less than that encoding's ciphertext.
+        const fixMock = await buildThumbMock();
+        const w = 400;
+        const h = 300;
+        const noisy = new Uint8Array(
+            jpegJs.encode(
+                {
+                    data: sodium.randombytes_buf(w * h * 4),
+                    width: w,
+                    height: h,
+                },
+                90,
+            ).data,
+        );
+        replaceOriginal(fixMock, 101, noisy);
+        const decoded = jpegJs.decode(noisy, {
+            useTArray: true,
+            formatAsRGBA: true,
+        });
+        const defaultSize =
+            jpegJs.encode(decoded, 50).data.length +
+            sodium.crypto_secretstream_xchacha20poly1305_ABYTES;
+        const recordedSize = defaultSize - 1;
+        rawFile(fixMock, 101).info = { thumbSize: recordedSize };
+
+        const client = await login(buildThumbFetch(fixMock));
+        const lib = await openLib(client);
+
+        const results = await fixMissingThumbnails(lib, client, [101]);
+        lib.close();
+
+        expect(results[0]!.status).toBe("fixed");
+        const upload = fixMock.uploadedThumbnails[0]!;
+        expect(upload.ciphertext.length).toBeLessThanOrEqual(recordedSize);
+        const decrypted = decryptBlob(
+            upload.ciphertext,
+            fromBase64(upload.decryptionHeader),
+            fixMock.fileKeys[101]!,
+        );
+        expect(decrypted[0]).toBe(0xff);
+        expect(decrypted[1]).toBe(0xd8);
+    });
 });
 
 describe("Client.getApiClient", () => {

yarn.lock

@@ -1069,6 +1069,11 @@ fastq@^1.6.0:
   dependencies:
     reusify "^1.0.4"
 
+fflate@0.8.3:
+  version "0.8.3"
+  resolved "https://registry.yarnpkg.com/fflate/-/fflate-0.8.3.tgz#bc27d8eb30343d4d512abb03480202ce65d825fc"
+  integrity sha512-tbZNuJrLwGUp3zshBtdy4W+ORxZuIh8a5ilyIEQDC5rY1f3U20JMry0Ll3WBzU58EZKsEuJFXhb5gwv8CsPvgA==
+
 file-entry-cache@^8.0.0:
   version "8.0.0"
   resolved "https://registry.yarnpkg.com/file-entry-cache/-/file-entry-cache-8.0.0.tgz#7787bddcf1131bffb92636c69457bbc0edd6d81f"