docker build . stamps the git tag or short commit, not dev (closes #154)

script/build writes the version script/version prints into
dist/package.json, which quak --version reports: the VERSION environment
variable or build arg when one is given, otherwise git describe --tags
--always, otherwise package.json's version. A checkout with .git whose
version comes out empty, dev or unknown fails the build.

.dockerignore no longer leaves out .git, and ARG VERSION has no default,
so a plain docker build . of a clone stamps its commit. script/docker and
script/cibuild still pass the host's version. make build-bin bundles the
built dist/, so the single binary reports the same version.

Model: opus-5-5

.dockerignore

@@ -1,9 +1,9 @@
 # Mirrors .gitignore, with one deliberate exception: .gitignore itself stays
 # in the build context, because prettier 3 reads it as a default ignore file
 # and dropping it would change what the lint phase's prettier check sees.
-
+#
-# VCS
+# .git is deliberately NOT excluded: the build derives the version it stamps
-.git
+# from it (script/version).
 
 # OS
 .DS_Store

Dockerfile

@@ -61,9 +61,12 @@ RUN script/bootstrap
 
 COPY . .
 
-# The version is computed on the host and passed in, because
+# Version stamped into the build: the VERSION build arg when one is given,
-# .dockerignore excludes .git.
+# otherwise what script/version derives from the .git the build context
-ARG VERSION=dev
+# carries (script/bootstrap installed git), so any `docker build .` of a
+# clone stamps its commit. The label can only carry the build arg, and is
+# empty without one.
+ARG VERSION
 LABEL org.opencontainers.image.version="${VERSION}"
 
 RUN make build

Makefile

@@ -26,8 +26,10 @@ check:
 build:
 	@script/build
 
-build-bin:
+# Bundles the built dist/, so the binary reports the version script/build
-	nix-shell -p bun --run "bun build bin/quak.ts --compile --outfile bin/quak"
+# stamped.
+build-bin: build
+	nix-shell -p bun --run "bun build dist/bin/quak.js --compile --outfile bin/quak"
 
 install: build-bin
 	mkdir -p ~/bin

README.md

@@ -8,9 +8,9 @@ and downloads individual images while decrypting them on the way to disk.
 quak also includes a resilient backup command that downloads every file in the
 account into a deduplicated local directory tree, skipping files that already
 exist on disk and continuing past individual download failures instead of
-crashing. It decrypts and persists all three metadata layers (basic, private
+crashing. For each file it persists the basic metadata fields quak keeps (title,
-magic, public magic) per file, including camera info, GPS coordinates, captions,
+file type, creation and modification time, latitude, longitude, content hash),
-and any face/keyword labels the Ente clients have added. A helper subcommand can
+and the private and public magic metadata in full. A helper subcommand can
 detect and regenerate missing thumbnails, encrypting and uploading them back to
 the server.
 
@@ -51,7 +51,8 @@ const client = await Client.login({
 
 // Open a cache-backed library. On an empty cache this awaits one server
 // refresh; on an existing cache it returns immediately and refreshes in the
-// background every `refreshIntervalSeconds` (default 3).
+// background. Later refreshes start `refreshIntervalSeconds` (default 3) after
+// the previous one ends.
 const lib = await Library.open({ client });
 
 // Default reads answer synchronously from the local cache — no network.
@@ -62,7 +63,7 @@ for (const album of lib.albums.list()) {
     }
 }
 
-// Fresh reads await a server round-trip and answer with current state.
+// Fresh reads await a server round-trip, joining one already running.
 const { albums } = await lib.fresh();
 console.log(`${albums.list().length} albums as of now`);
 
@@ -79,12 +80,42 @@ await lib.close();
 The lower-level `Client` (login, session serialization, and the raw
 enumeration/download calls) is exported too and documented under Design below.
 
+## Examples
+
+`examples/download-albums.ts` downloads every album's photos and their metadata
+into a directory, `photos` in the working directory unless you name another. The
+build compiles it; run it after `yarn install`:
+
+```bash
+yarn build
+QUAK_EMAIL=… QUAK_PASSWORD=… node dist/examples/download-albums.js [dir]
+```
+
+It opens the library with `precacheThumbnails` and `precacheOriginals` off, as
+`quak backup` does, so the only file content it fetches is the originals it
+saves. It asks on the terminal for a two-factor or email code when the account
+requires one, and writes:
+
+- each photo's original at its save path under `dir`, as `photo.download()`
+  writes it: `YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID>.<ext>`, and for a live
+  photo its image, its video and the `.livephoto.json` file naming them
+- beside each original, a JSON file named after it with `.json` added, for
+  example `2026-03-01.12345.jpg.json`: the photo's record (`photo.record()`)
+  without its cache paths, and its EXIF fields (`photo.exif()`) under `exif`
+- `albums/<collectionID>.json` for each album: its `collectionID`, its `name`,
+  and under `savePaths` the save paths of its photos relative to `dir`, newest
+  first
+
+A photo in several albums is downloaded once. A second run downloads nothing and
+rewrites only the JSON files whose content changed. A failed download stops the
+run; running it again carries on, since every photo already saved is skipped.
+
 ## Entrypoints
 
 This repository adheres to the
 [Scripts to Rule Them All](https://github.com/github/scripts-to-rule-them-all)
 standard: normalized scripts in `script/` are the entrypoints for the
-development workflow, and the Makefile targets are thin shims that call them.
+development workflow, and most Makefile targets are thin shims that call them.
 The scripts are POSIX sh (not bash) so they run in minimal containers such as
 alpine. We provide:
 
@@ -94,9 +125,12 @@ alpine. We provide:
   `script/bootstrap`, then `script/install-precommit`
 - `script/projectname` — output the project name (our own extension); used by
   `script/docker` for the image tag
-- `script/build` — compile the TypeScript sources into `dist/`, then verify that
+- `script/build` — compile the TypeScript sources into `dist/`, stamp the
-  the entrypoints `package.json` declares (`main`, `types`, `bin`) are among the
+  version into `dist/package.json`, then verify that the entrypoints
-  files the compiler wrote, and make the CLI executable (our own extension)
+  `package.json` declares (`main`, `types`, `bin`) are among the files the
+  compiler wrote, and make the CLI executable (our own extension)
+- `script/version` — print the version `script/build` stamps (our own
+  extension); see Version below
 - `script/test` — run the test suite, by building the `test` phase of the
   `Dockerfile` (vitest, 90s timeout, verbose rerun on failure); requires docker
 - `script/lint` — run eslint and a prettier check, by building the `lint` phase
@@ -122,9 +156,9 @@ alpine. We provide:
 Linting and testing are phases of the `Dockerfile`. The `lint` phase copies the
 repo into a digest-pinned node image and runs eslint and `prettier --check .`;
 the `test` phase does the same with the suite. `script/lint` and `script/test`
-each build one phase with `docker build --no-cache --target <phase>`. There is
+each build one phase with `docker build --no-cache --target <phase>`. Neither
-no host lint or test path: docker is required, and that also works where the
+script runs the tools on the host: docker is required, and that also works where
-docker daemon is remote and bind mounts are impossible.
+the docker daemon is remote and bind mounts are impossible.
 
 The last stage of the `Dockerfile` compiles the package, and it copies a file
 from each phase, so it cannot be built unless lint and the tests pass. That is
@@ -144,6 +178,31 @@ an exact version, installed from `yarn.lock` under `--frozen-lockfile` in both
 places, and reads `.gitignore` as its default ignore file — which is why
 `.dockerignore` keeps `.gitignore` in the build context.
 
+### Version
+
+`quak --version` reports the `version` of `dist/package.json`, which
+`script/build` writes after compiling; the repo's own `package.json` keeps
+`0.0.0`, and that is what the tests, which run from source, report.
+`script/version` decides what is written:
+
+- the `VERSION` environment variable, or the `Dockerfile`'s `VERSION` build arg
+  (`--build-arg VERSION=...`), when one is given and not empty;
+- otherwise, in a checkout with `.git`, `git describe --tags --always`: the tag
+  on a tagged commit, otherwise the short commit;
+- otherwise, as in a source tarball, the version `package.json` declares.
+
+The build fails if the checkout has `.git` and the version still comes out
+empty, `dev` or `unknown`: such a build could not be traced back to its commit.
+
+`.dockerignore` therefore does not leave out `.git`, so any `docker build .` of
+a clone stamps the commit it was built from; a shallow clone of one branch has
+no tags and stamps the short commit. `script/docker` (and so `make docker`) and
+`script/cibuild` pass the version they resolve on the host, with `--dirty`, as
+the build arg, which takes precedence. The image's
+`org.opencontainers.image.version` label carries that build arg only, so a build
+given none leaves it empty. `make build-bin` bundles the built `dist/`, so the
+single binary reports the stamped version too.
+
 ## Rationale
 
 Ente is one of very few photo services with a credible end-to-end encryption
@@ -170,16 +229,19 @@ local cache is reliable, and the UI is responsive on a five-year-old laptop.
 
 All work on quak is test-driven. No exceptions.
 
-1. Every change starts on a feature branch off `main`.
+1. Every change starts on a feature branch off `next`, and its pull request
+   targets `next`.
 2. The first commit on the branch is the test suite for what is being added or
    changed. Those tests must fail at that commit; the branch is red until the
    implementation lands.
 3. Subsequent commits add the implementation and any refactors needed to make
    the tests pass.
-4. A feature branch can only be merged into `main` when `make check` is green.
+4. A pull request can only be merged into `next` when `make check` is green.
-   `main` is always green. CI runs `script/cibuild`, which builds the
+   Once it has passed review, the repository manager squash-merges it into
-   `Dockerfile`: its `lint` and `test` phases, then the compile, so neither a
+   `next`. Only sneak merges `next` into `main`. `main` and `next` are always
-   red branch nor one that does not compile can pass CI.
+   green. CI runs `script/cibuild`, which builds the `Dockerfile`: its `lint`
+   and `test` phases, then the compile, so neither a red branch nor one that
+   does not compile can pass CI.
 5. Tests are the canonical API documentation for this library. Every test file
    is commented thoroughly enough that a reader who has never seen quak can
    learn how to use it from the tests alone. Comments explain why a behavior
@@ -197,7 +259,7 @@ All work on quak is test-driven. No exceptions.
    not the tests, and so not the full `make check`. This is deliberate so the
    TDD red-phase commit (failing tests, no implementation yet) can land. The
    `test` phase is part of the image build, which is what CI executes via
-   `script/cibuild`, so a red branch still cannot reach `main`.
+   `script/cibuild`, so a red branch still cannot reach `next`.
 
 ## Design
 
@@ -209,7 +271,7 @@ the CLI is for humans.
 ```
 quak/
     src/
-        crypto/         libsodium primitives (boxes, secretstreams, KDF, SRP)
+        crypto/         libsodium primitives (boxes, secretstreams, KDF, hash)
         api/            HTTP client (ApiClient class)
         auth/           login flow (SRP + email OTP + TOTP), key unwrap
         model/          decrypted Collection, File, Metadata types + decrypt fns
@@ -219,7 +281,8 @@ quak/
                         and search, request pools
         backup.ts       resilient full-account backup with dedup
         metadata-backup.ts
-                        backup-metadata: all decrypted metadata as JSON
+                        backup-metadata: the metadata quak keeps, as JSON
+        exif.ts         EXIF read from an image's bytes with exifreader
         mldata-fetch.ts fetch + decrypt per-file ML data
         filename.ts     safe file names from server metadata
         errors.ts       error types shared across layers
@@ -234,6 +297,9 @@ quak/
         index.ts        public library exports
     bin/
         quak.ts         CLI entrypoint (commander.js)
+    examples/
+        download-albums.ts
+                        download every album's photos and metadata
     test/               unit + integration tests (vitest)
     Makefile
     Dockerfile          lint phase, test phase, compile
@@ -242,15 +308,18 @@ quak/
 ```
 
 `make build` compiles that tree into `dist/`, preserving its shape: the library
-lands in `dist/src/` and the CLI in `dist/bin/quak.js`, which is what
+lands in `dist/src/`, the examples in `dist/examples/`, and the CLI in
-`package.json` points `main`, `types` and `bin` at. The compiler's `rootDir` is
+`dist/bin/quak.js`, which is what `package.json` points `main`, `types` and
-the repository root rather than `src/`, because `bin/` is compiled too and
+`bin` at. The compiler's `rootDir` is the repository root rather than `src/`,
-`rootDir` has to contain everything that is compiled.
+because `bin/` is compiled too and `rootDir` has to contain everything that is
+compiled.
 
 ### Cryptography
 
 All cryptography is done by `libsodium-wrappers-sumo` (the "sumo" build is
-required for `crypto_pwhash` / Argon2id). No hand-rolled crypto.
+required for `crypto_pwhash` / Argon2id), except the SRP handshake, which uses
+`fast-srp-hap`, and the MD5 checksum sent with a thumbnail upload, which uses
+Node's built-in `node:crypto`. No hand-rolled crypto.
 
 The key hierarchy, derived during login, is:
 
@@ -258,19 +327,23 @@ The key hierarchy, derived during login, is:
 2. Argon2id (`crypto_pwhash`) over the password and a server-issued `kekSalt`,
    with server-issued `memLimit` and `opsLimit`, produces a 32-byte Key
    Encryption Key (KEK).
-3. SRP login: a 16-byte SRP login subkey is derived from the KEK using
+3. SRP login: `crypto_kdf_derive_from_key` (BLAKE2b) derives a 32-byte subkey
-   `crypto_kdf_derive_from_key` (BLAKE2b) with subkey id 1 and context
+   from the KEK with subkey id 1 and context `loginctx`. Its first 16 bytes are
-   `loginctx`. That 16-byte value is the SRP password.
+   the SRP password.
-4. After SRP completes (or after email-OTP fallback), the server returns a blob
+4. When SRP completes, the server returns a blob of "key attributes" plus an
-   of "key attributes" plus an encrypted auth token.
+   encrypted auth token, or first asks for a second factor. quak answers a TOTP
+   request with the code (`POST /users/two-factor/verify`), after which the
+   server returns them, and cannot answer a passkey request. When the account
+   has email MFA on (`isEmailMFAEnabled`), an email OTP replaces SRP and the
+   server returns them after it.
 5. `crypto_secretbox_open_easy` over the encrypted master key with the KEK
    yields the 32-byte master key.
 6. `crypto_secretbox_open_easy` over the encrypted secret key with the master
    key yields the user's X25519 private key. The matching public key is
    delivered in cleartext.
 7. `crypto_box_seal_open` over the encrypted token with the user's keypair
-   yields the URL-safe base64 auth token used in `X-Auth-Token` for all
+   yields the auth token's bytes. Encoded as URL-safe base64 with padding, they
-   subsequent calls.
+   are the `X-Auth-Token` value for all subsequent calls.
 
 Per-collection keys are decrypted with `crypto_secretbox_open_easy` using the
 master key (for owned collections). Per-file keys are decrypted with
@@ -306,7 +379,8 @@ Endpoints used:
 - `POST /users/srp/create-session`: begin SRP handshake.
 - `POST /users/srp/verify-session`: complete SRP, receive 2FA challenge or the
   encrypted token plus key attributes.
-- `POST /users/ott` and `POST /users/verify-email`: email OTP fallback path.
+- `POST /users/ott` and `POST /users/verify-email`: email OTP, used instead of
+  SRP when the SRP attributes have `isEmailMFAEnabled` set.
 - `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
@@ -328,8 +402,9 @@ request is repeated only when repeating it could produce a different answer:
 - Every other 4xx: not retried. A 404 in particular is an answer, and
   `listMissingThumbnails` depends on getting it promptly and once.
 - Transport failures — a `fetch` rejection, `ECONNRESET`, `ETIMEDOUT`, a DNS or
-  TLS failure — and deadline aborts: retried. The errno is looked for in the
+  TLS failure — and deadline aborts: retried. Node's `fetch` rejects with a
-  error's `cause` chain, because that is where Node's `fetch` puts it.
+  plain `TypeError`, so every `TypeError` is retried. The errno is looked for in
+  the error's `cause` chain, because that is where Node's `fetch` puts it.
 - A truncated download: retried.
 - Anything else, including a secretstream authentication failure that is not
   truncation: not retried. The default answer is no. For a backup tool, retrying
@@ -395,10 +470,13 @@ because a socket reset after the response headers have arrived surfaces in the
 download layer rather than in `ApiClient`, and that is the common failure for
 multi-megabyte photos over a CDN. The secretstream pull state is not resumable
 and these endpoints have no Range support, so a retry starts the file over. The
-atomic write stays outside the retry, so a download that needed three attempts
+atomic write is part of the retried unit: each attempt writes its own temporary
-still performs exactly one write and one rename. `runBackup` and
+files, one for most files and two for a live photo (its image and its video),
-`runMetadataBackup` are unchanged: the retry sits below them, and a file that
+and removes them if it fails. Only the attempt that completes renames anything
-fails after exhausting it is still logged, counted, and stepped over.
+into place. The retry sits below `runBackup` and `runMetadataBackup`.
+`runBackup` logs, counts and steps over a file that still fails after its
+retries; `runMetadataBackup`, which downloads only with `--exif`, records the
+error in that file's JSON as `imageMetadataError` and goes on.
 
 One imprecision is deliberate and worth knowing about. When a body ends part-way
 through a secretstream chunk, Poly1305 fails and carries no framing signal, so a
@@ -422,10 +500,11 @@ base64-encoded keys) that the consumer can write to disk, a database, or
 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
+`client.logout()` clears the token and zeroes the key buffers in place; after
-later call on that client throws. It does not contact the server, so the token
+it, every other method on that client throws. It does not contact the server, so
-stays valid there and in any saved snapshot; `await client.logoutOnServer()`
+the token stays valid there and in any saved snapshot;
-first ends the session on the server (`POST /users/logout`).
+`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,
@@ -433,41 +512,43 @@ The CLI stores the snapshot at the platform-appropriate data directory via
 `0600`. The key material is stored in cleartext in the JSON; treat this file as
 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.
+field. Both exit with status 1, except that `quak logout` with no file says
+there is no session and exits 0.
 
 `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 does not delete the cache. When it knows the cache directory, from
-it still holds decrypted data (file keys in `metadata.json`, cached originals
+`--cache-dir` or from a session file it could read, it prints it and says it
-and thumbnails), for the user to delete if they want it gone.
+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 [--cache-dir <path>] <command>                       global: local metadata/content cache location
-quak login                                      interactive or QUAK_EMAIL/QUAK_PASSWORD
+quak login                                                interactive or QUAK_EMAIL/QUAK_PASSWORD
-quak whoami                                     print logged-in account as JSON
+quak whoami                                               print logged-in account as JSON
-quak logout                                     end the session, delete it
+quak logout                                               end the session, delete it
-quak collections [--json]                       list all collections
+quak collections [--json]                                 list all collections
-quak files --collection <id> [--json]           list files in a collection
+quak files --collection <id> [--json]                     list files in a collection
-quak get <fileID> [--out path] [--collection]   download and decrypt a file
+quak get <fileID> [--out path] [--collection]             download and decrypt a file
-quak get-thumb <fileID> [--out] [--collection]  download and decrypt a thumbnail
+quak get-thumb <fileID> [--out] [--collection]            download and decrypt a thumbnail
-quak backup <dir> [--json]                      full incremental backup
+quak backup <dir> [--json]                                full incremental backup
-quak backup-metadata <dir> [--exif]             dump all decrypted metadata as JSON
+quak backup-metadata <dir> [--exif]                       dump the metadata quak keeps as JSON
-quak helper list-missing-thumbnails [--json]    find files with missing thumbnails
+quak helper list-missing-thumbnails [--json]              find files with missing thumbnails
-quak helper fix-missing-thumbnails [--file ids] generate + upload missing thumbnails
+quak helper fix-missing-thumbnails [--file ids] [--json]  generate + upload missing thumbnails
 ```
 
-Every command runs on the same cache-backed library. The read commands —
+Every command except `login`, `whoami` and `logout` runs on the cache-backed
-`collections`, `files`, `get`, `get-thumb`, `backup-metadata`,
+library. The read commands — `collections`, `files`, `get`, `get-thumb`,
-`helper list-missing-thumbnails` and `helper fix-missing-thumbnails` — force a
+`backup-metadata`, `helper list-missing-thumbnails` and
-fresh server round-trip before they answer, so they report current account state
+`helper fix-missing-thumbnails` — force a fresh server round-trip before they
-rather than whatever the cache last held. If that round-trip fails, the command
+answer, so they report current account state rather than whatever the cache last
-prints the error on one line and exits 1. `--cache-dir` overrides where the
+held. If that round-trip fails, the command prints the error on one line and
-cache lives; without it each account gets its own directory under the per-user
+exits 1. `--cache-dir` overrides where the cache lives; without it each account
-cache path.
+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. For a live photo, `get` writes
@@ -475,18 +556,25 @@ its image and its video, each named after the title with its own extension, as
 Ente's clients name them (`IMG_0001.heic` and `IMG_0001.mov`). With
 `--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
+refused. `backup-metadata --exif` (alias `--all`) additionally fetches each
-file to extract full EXIF/IPTC/XMP metadata. The listing and backup commands
+file's original through the cache and records, from it or a live photo's image,
-support `--json` for machine-readable output.
+its XMP metadata, its EXIF metadata and, for a JPEG, its dimensions. EXIF is
+read with [exifreader](https://github.com/mattiasw/ExifReader) from any image
+format it reads, JPEG, HEIC/HEIF, AVIF, PNG and WebP among them. The record's
+`exif` field is exifreader's EXIF tag output: each tag by name, with its
+`value`, `description` and `computed` value. An EXIF block exifreader finds but
+reads no tag from is recorded, base64, as `exifRaw`, with the reason in
+`exifError`. `collections`, `files`, `backup`, `helper list-missing-thumbnails`
+and `helper fix-missing-thumbnails` take `--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
+fails, the error is logged, each of its files is written with the reason in an
-with the reason in an `mlDataError` field instead of `mlData`, and the dump goes
+`mlDataError` field instead of `mlData`, and the dump goes on. The exit code is
-on. The exit code is non-zero if any ML data request failed.
+non-zero if any ML data request failed.
 
-`helper fix-missing-thumbnails` regenerates thumbnails for baseline JPEG images
+`helper fix-missing-thumbnails` regenerates thumbnails for JPEG images only,
-only, because the bundled decoder (`jpeg-js`) decodes only JPEG. A non-JPEG
+because the bundled decoder (`jpeg-js`) decodes only JPEG. A non-JPEG image
-image (PNG, HEIC) or a video is reported as `skipped` (unsupported format), kept
+(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. 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
@@ -502,23 +590,36 @@ the smallest does not.
 
 ```
 <dir>/
-    originals/
+    YYYY/YYYY-MM/YYYY-MM-DD/
-        <fileID>.<ext>          actual file content (one per unique file,
+        YYYY-MM-DD.<fileID>.<ext>   actual file content, at its save path (one
-                                two for a live photo: see below)
+                                    per unique file, two for a live photo: see
-        <fileID>.json           all decrypted metadata for that file
+                                    below)
-        <fileID>.livephoto.json which of a live photo's two files is which
+        YYYY-MM-DD.<fileID>.json    the file's basic metadata fields quak
+                                    keeps, and its private and public magic
+                                    metadata
+        YYYY-MM-DD.<fileID>.livephoto.json
+                                    which of a live photo's two files is which
     collections/
         <name>/
-            <title> -> ../../originals/<fileID>.<ext>   (symlink)
+            <title> -> ../../YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID>.<ext>
-        <name>.json             collection metadata + file list
+                                    (symlink)
-    failures.json               files that failed and have not yet succeeded
+        <name>.json                 collection metadata + file list
+    failures.json                   files that failed and have not yet succeeded
 ```
 
+Each original is saved at its save path, the same path `photo.savePath` gives
+and `photo.download()` writes (see Read surface below). The date is the photo's
+`takenAt` (the date set in Ente if it was edited, else its creation time) in the
+time zone of the machine running quak. The extension is the one in the file's
+name as uploaded, case kept, or `.bin` when it has none or it holds anything but
+letters and digits. When the date or the time zone changes, the next run saves
+the original at its new path and leaves the old copy where it is.
+
 `failures.json` records each failed file with the kind of failure, how many
 times it has been tried and when it was last tried. A file leaves it once it
 succeeds, or once it is no longer in the library or in the backup's scope. The
 library's `lib.backup({ includeThumbnails: true })` also writes
-`thumbnails/<fileID>.jpg` beside `originals/`; `quak backup` does not.
+`thumbnails/<fileID>.jpg` beside `collections/`; `quak backup` does not.
 
 A collection's directory and JSON are named after the collection, and a symlink
 after the file's title, both with unsafe characters replaced. When two
@@ -530,45 +631,47 @@ 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
+`YYYY-MM-DD.<fileID>.<ext>` with the extension it has inside the ZIP (for
-`12345.heic` and `12345.mov`), and `<fileID>.livephoto.json` names the two. The
+example `2026-03-01.12345.heic` and `2026-03-01.12345.mov`), and
-live photo counts as stored only when both files are present and not empty. Its
+`YYYY-MM-DD.<fileID>.livephoto.json` names the two. The live photo counts as
-album folder links both, each named after the title with that file's extension
+stored only when both files are present and not empty. Its album folder links
-(`IMG_0001.heic` and `IMG_0001.mov`). A live photo that an earlier version of
+both, each named after the title with that file's extension (`IMG_0001.heic` and
-quak stored as the ZIP, under the image's name, is replaced by its two files on
+`IMG_0001.mov`).
-the next run, and the ZIP and its link are removed.
 
-Each run removes the symlinks into `originals/` that no longer belong in their
+Each run removes the symlinks into the date folders that no longer belong in
-collection's directory, and the directories (and JSON) of collections that were
+their collection's directory, and the directories (and JSON) of collections that
-deleted or renamed. Nothing else in `collections/` is touched: a file or a
+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
+appears in, and written once: straight to its save path, 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.
+originals precache off, so the only file content it fetches is the originals the
+backup stores. The library's ML data fetch still runs and fills the cache's
+`mldata/`.
 
 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
+`.quak-backup-YYYY-MM-DD.<fileID>.<ext>-<pid>-<random>.tmp`. A run that is
-one of these temporary files behind; the next backup deletes those whose process
+killed can leave one of these temporary files behind; the next backup deletes
-is no longer running. The content cache uses the same scheme, and opening a
+those whose process is no longer running. The content cache uses the same
-library deletes the temporary files in the cache whose process is no longer
+scheme, and opening a library deletes the temporary files in the cache whose
-running, so a download another process has in progress in the same cache is left
+process is no longer running, so a download another process has in progress in
-alone. The rename replaces whatever was at the destination rather than writing
+the same cache is left alone. The rename replaces whatever was at the
-through it: a symlink there is replaced, not followed, and the new file has the
+destination rather than writing through it: a symlink there is replaced, not
-temporary file's permissions, not those of the file it replaced.
+followed, and the new file has the temporary file's permissions, not those of
+the file it replaced.
 
 ## TODO
 
-- [x] Retry policy: no retry on 4xx, exponential backoff on 5xx and network
+- [x] Retry policy: no retry on 4xx (except `408` and `429`), exponential
-      errors
+      backoff on 5xx and network errors
 - [x] Update the API reference section below to match the current implementation
 - [x] `make docker` green
 - [x] Store live photos in a form a photo viewer can open
@@ -591,7 +694,8 @@ Future (desktop client, separate repo):
 The library's primary surface is the cache-backed `Library`; the lower-level
 `Client` sits underneath it and is covered by the Design sections above. The
 test suite is the canonical, executable documentation — `test/library/` and
-`test/client/usage.test.ts` walk every operation, and `yarn test` verifies them.
+`test/client/usage.test.ts` walk most operations, `test/cli/backup.test.ts`
+walks `lib.backup()`, and `yarn test` verifies them.
 
 ### Opening a library
 
@@ -604,21 +708,24 @@ background, so an unreachable server does not block opening.
 
 `LibraryOptions`:
 
-| Option                   | Default                     | Meaning                                                               |
+| Option                   | Default                                            | Meaning                                                              |
-| ------------------------ | --------------------------- | --------------------------------------------------------------------- |
+| ------------------------ | -------------------------------------------------- | -------------------------------------------------------------------- |
-| `client`                 | required                    | the account client (a `Client`, or any `LibraryClient`)               |
+| `client`                 | required                                           | the account client (a `Client`, or any `LibraryClient`)              |
-| `cacheDirectory`         | `<XDG cache>/quak/<userID>` | where `metadata.json` and the content cache live                      |
+| `cacheDirectory`         | `quak/<userID>` under the per-user cache directory | where `metadata.json` and the content cache live                     |
-| `downloadDirectory`      | none                        | backup destination; an original already stored there counts as cached |
+| `downloadDirectory`      | `photos` in the working directory                  | root of the save paths; an original stored there counts as cached    |
-| `refreshIntervalSeconds` | `3`                         | background refresh cadence                                            |
+| `refreshIntervalSeconds` | `3`                                                | background refresh cadence                                           |
-| `precacheThumbnails`     | `true`                      | prefetch every thumbnail, newest first                                |
+| `precacheThumbnails`     | `true`                                             | prefetch every thumbnail, newest first                               |
-| `precacheOriginals`      | `true`                      | prefetch the favorites album and the latest-window originals          |
+| `precacheOriginals`      | `true`                                             | prefetch the favorites album and the latest-window originals         |
-| `precacheOriginalsDays`  | `7`                         | length in days of that latest window                                  |
+| `precacheOriginalsDays`  | `7`                                                | length in days of that latest window                                 |
-| `cacheOriginalsMaxBytes` | 100 GiB                     | hard ceiling on the originals cache                                   |
+| `cacheOriginalsMaxBytes` | 100 GiB                                            | size limit on the originals cache; pinned originals can exceed it    |
-| `freeBelowBytes`         | 50 GiB                      | free space to protect on the volume; the effective limit adapts down  |
+| `freeBelowBytes`         | 50 GiB                                             | free space to protect on the volume; the effective limit adapts down |
-| `isOriginalPinned`       | none                        | extra predicate for originals that must never be evicted              |
+| `isOriginalPinned`       | none                                               | extra predicate for originals that must never be evicted             |
-| `pools`                  | fresh `RequestPools`        | the bounded request pools (sets concurrency)                          |
+| `pools`                  | fresh `RequestPools`                               | the bounded request pools (sets concurrency)                         |
-| `onProgress`             | none                        | refresh/ML/precache progress callback (`RefreshEvent`)                |
+| `onProgress`             | none                                               | refresh/ML/precache progress callback (`RefreshEvent`)               |
-| `contentSource`          | the client's own            | override the byte source (mainly for tests)                           |
+| `contentSource`          | the client's own                                   | override the byte source (mainly for tests)                          |
+
+The default `downloadDirectory` is resolved against the working directory once,
+when the library opens; `lib.downloadDirectory` holds the result.
 
 Concurrency is set through `pools`: construct
 `new RequestPools({ metadataConcurrency, contentConcurrency, thumbnailConcurrency })`
@@ -635,17 +742,21 @@ can then be removed.
 ### Default reads vs. fresh reads
 
 Default reads — `lib.albums`, `lib.photos`, `lib.timeline` — answer
-synchronously from the last refreshed copy held in RAM and never touch the
+synchronously from the copy held in RAM and never touch the network. A refresh
-network. The background timer refreshes that copy every
+changes that copy only once all its server requests have succeeded, and the
-`refreshIntervalSeconds`, so a default read is immediate but may be up to one
+background timer starts the next refresh `refreshIntervalSeconds` after the
-interval stale.
+previous one ends. So a default read is immediate, but only as current as the
+last refresh whose requests all succeeded; right after opening an existing
+cache, it is the copy on disk.
 
-`await lib.fresh()` forces a refresh, waits for it to complete and persist, and
+`await lib.fresh()` waits for a refresh to complete and persist, and returns the
-returns the same `{ albums, photos, timeline }` namespaces — now guaranteed to
+same `{ albums, photos, timeline }` namespaces, which then reflect a completed
-reflect a completed server round-trip. Concurrent `fresh()` calls coalesce onto
+server round-trip. When a refresh is already running, background or not,
-one refresh, and a refresh that fails rejects the caller (default reads stay
+`fresh()` waits for that one, so its answer can come from requests made before
-silent and keep serving the last good copy). The CLI's read commands use fresh
+the call; only when none is running does it start one. A refresh that fails
-reads (issue https://git.eeqj.de/sneak/quak/issues/75).
+rejects the caller (default reads stay silent and keep serving the last good
+copy). The CLI's read commands use fresh reads (issue
+https://git.eeqj.de/sneak/quak/issues/75).
 
 ### Read surface
 
@@ -662,17 +773,57 @@ reads (issue https://git.eeqj.de/sneak/quak/issues/75).
   `includeArchived`; hidden photos are always excluded.
 
 An `Album` exposes its record fields and `album.photos.list()` → `Photo[]`
-(newest first). A `Photo` exposes its record fields, `photo.record()` →
+(newest first). A `Photo` exposes its record fields other than `thumbnailPath`
-`PhotoRecord`, and two content methods:
+and `originalPath`, and `photo.year`, the local-time year of `takenAt`, all as
+synchronous getters read from RAM; `photo.record()` → `PhotoRecord`. Two more
+synchronous getters look at the disk and never touch the network:
+
+- `photo.savePath` → `string` — where `photo.download()` and `lib.backup()` put
+  the original, whether or not it is there yet:
+  `YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID>.<ext>` under the library's
+  `downloadDirectory` (see Backup layout above for the date and the extension).
+  For a live photo already stored, its image. For a live photo not yet stored,
+  it carries the title's extension, and the image may be stored under a
+  different one, found inside the live photo. It needs no content source.
+- `photo.isLocal` → `boolean` — whether the whole original is at `savePath`. A
+  copy only in the cache does not count.
+
+These async methods may download:
 
 - `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.download()` → `{ path, bytes, videoPath? }`, as `original()` —
+  puts the original at `savePath`, creating its folders. When it is already
+  there, nothing is written. When the cache holds it, it is copied from the
+  cache; otherwise it is fetched straight to `savePath`, with no copy left in
+  the cache. Afterwards `isLocal` is true.
 - `await photo.thumbnail(opts?)` → `{ path, bytes }`.
+- `await photo.content(opts?)` → `Uint8Array` — the original's bytes, read
+  through `original()`; for a live photo, its image's.
+- `await photo.exif(opts?)` → `PhotoExif` — `make`, `model`, `lensModel`,
+  `dateTimeOriginal`, `offsetTimeOriginal`, `exposureTime`, `fNumber`, `iso`,
+  `focalLength`, `orientation`, `gpsLatitude`, `gpsLongitude` and `gpsAltitude`,
+  each absent when the file lacks it. GPS values are signed decimal degrees and
+  metres. `dateTimeOriginal` is the camera's clock reading held in the `Date`'s
+  UTC fields; `offsetTimeOriginal`, when present, is that clock's offset from
+  UTC. EXIF is read from any image format exifreader reads (such as JPEG,
+  HEIC/HEIF, AVIF, PNG, WebP and TIFF), a live photo's image included. Any other
+  original gives `{}`, and a video gives `{}` without being downloaded.
+- `await photo.make(opts?)`, and likewise `model()`, `lensModel()`,
+  `dateTimeOriginal()`, `offsetTimeOriginal()`, `exposureTime()`, `fNumber()`,
+  `iso()`, `focalLength()`, `orientation()`, `gpsLatitude()`, `gpsLongitude()`
+  and `gpsAltitude()` → one field of `exif()` each, typed as in `PhotoExif`, or
+  `undefined` when the file lacks it. Each calls `exif()` with its `opts`, so
+  each call reads the original again.
 
-Both serve from the on-disk content cache when the bytes are present and
+They serve from the on-disk content cache when the bytes are present and
-otherwise fetch through the pools; `opts.onProgress` reports per-file progress.
+otherwise fetch through the pools; `original()`, `content()` and `exif()` also
-They throw when the library was opened without a content source.
+serve an original already stored at its save path. `opts.onProgress` reports
+per-file progress. They throw when the library was opened without a content
+source. An original that `original()`, `content()` or `exif()` downloads lands
+in the cache, which does not make `isLocal` true; only `download()` and
+`lib.backup()` do.
 
 Lower-level accessors that return decrypted model objects (which hold key
 material) are also available: `listCollections()`, `getCollection(id)`,
@@ -684,10 +835,12 @@ material) are also available: `listCollections()`, `getCollection(id)`,
 The GUI-facing records hold no key material and no binary, so they survive
 `structuredClone`/JSON across the Electron IPC boundary:
 
-- `PhotoRecord`: `fileID`, `albumIDs`, `title`, `takenAt` (milliseconds),
+- `PhotoRecord`: `fileID`, `albumIDs`, `title`, `takenAt` and `modifiedAt`
-  `fileType`, optional `caption` / `width` / `height` / `latitude` /
+  (milliseconds), `fileType`, optional `caption` / `width` / `height` /
-  `longitude`, `isArchived`, `isHidden`, and `thumbnailPath` / `originalPath`
+  `latitude` / `longitude`, optional `hash` (the content hash recorded at
-  once the bytes are cached (for a live photo, `originalPath` is its image).
+  upload; very old files have none), `isArchived`, `isHidden`, and
+  `thumbnailPath` / `originalPath` 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 }`.
@@ -712,16 +865,16 @@ photos newest first). `lib.subscribe({ onChange })` delivers a `LibraryChange`
   `SimilarResult[]` (`{ fileID, score }`, cosine similarity, most similar first,
   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
+- `await lib.backup(opts?)` → `BackupResult`. It waits for a refresh as
-  in-scope original not already in the backup (and, with `includeThumbnails`,
+  `fresh()` does, puts every in-scope original not already at its save path
+  there as `photo.download()` does (and, with `includeThumbnails`, fetches
   thumbnails) through the content cache, and rebuilds the on-disk backup tree
-  with a durable failure ledger. A fetched original is written straight into the
+  with a durable failure ledger. A fetched original is written straight to its
-  backup's `originals/` and not into the cache, which then counts it as present;
+  save path and not into the cache, which then counts it as present; one the
-  one the cache already held is copied from there. `BackupOptions`:
+  cache already held is copied from there. `BackupOptions`: `downloadDirectory`
-  `downloadDirectory` (falls back to the one `open()` was given),
+  (falls back to the library's), `includeOriginals` (default `true`),
-  `includeOriginals` (default `true`), `includeThumbnails` (default `false`),
+  `includeThumbnails` (default `false`), `onlyAlbumNames`, and `onProgress`. See
-  `onlyAlbumNames`, and `onProgress`. See Backup layout above for the tree it
+  Backup layout above for the tree it writes.
-  writes.
 
 ### Request pools
 
@@ -752,7 +905,7 @@ When `metadata.json` belongs to a different account than the client's,
 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,
+A live photo's original is cached as at its save path: 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: the library
@@ -776,12 +929,13 @@ from a very old client, is stored unchecked.
 - `src/library/index.ts`: `Library`, `LibraryOptions`, `LibraryStatus`,
   `LibraryClient`, `RefreshEvent`
 - `src/library/read.ts`: `Album`, `Photo`, `AlbumsAPI`, `PhotosAPI`,
-  `TimelineAPI`, `PhotoFilter`, `TimelineGroup`, `GroupBy`
+  `TimelineAPI`, `PhotoFilter`, `TimelineGroup`, `GroupBy`, `SavePathLookup`
 - `src/library/content.ts`: `ContentResult`, `ContentOptions`, `ThumbnailsAPI`,
   `EnsureOptions`, `EnsureResult`, `ContentSource`
 - `src/library/records.ts`: `PhotoRecord`, `AlbumRecord`, `LibrarySnapshot`,
   `LibraryChange`
 - `src/library/mlsearch.ts`: `MLDataAPI`, `SimilarResult`
+- `src/exif.ts`: `PhotoExif`
 - `src/library/pools.ts`: `RequestPools`, `RequestPoolsOptions`, `BoundedPool`
 - `src/backup.ts`: `BackupOptions`, `BackupResult`, `BackupError`
 - `src/client.ts`: `Client`, `LoginOptions`, `ClientSnapshot`
@@ -813,14 +967,15 @@ documents:
   `yarn.lock`. Never `git add -A`. Never force-push to main.
 
 - **The "Development workflow" section above.** All changes go on feature
-  branches. Tests are written first and committed in a failing state before the
+  branches off `next`, and every pull request targets `next`; only sneak merges
-  implementation. Tests are the canonical API documentation and must be
+  `next` into `main`. Tests are written first and committed in a failing state
-  commented thoroughly. `main` is always green.
+  before the implementation. Tests are the canonical API documentation and must
+  be commented thoroughly. `main` and `next` are always green.
 
 - **Required checks before every commit:** `make lint` must pass — that is
   eslint plus the prettier check, and it builds the `lint` phase of the
   `Dockerfile`, so it needs docker. The pre-commit hook enforces exactly that.
-  `make check` (which also runs the tests) must pass before merging to `main`.
+  `make check` (which also runs the tests) must pass before merging into `next`.
   `make fmt-check` is available for a host-side formatting check on its own, but
   it is not a separate requirement: `make lint` already covers it, and running
   both would check formatting twice. Never invoke eslint or prettier directly;

TODO.md

@@ -1,12 +1,15 @@
 # Workflow
 
-- branch (from `main`)
+- branch from `next`
 - do the work in Next Step
 - move Next Step to the top of Completed Steps
 - move the top item of Future Steps into Next Step
 - commit (`TODO.md` changes in the same commit as the work)
-- merge to `main` if the branch is not protected, otherwise open a PR
 - push
+- open a pull request that targets `next`
+- once the pull request has passed review, the repository manager squash-merges
+  it into `next`
+- only sneak merges `next` into `main`
 
 # Status
 
@@ -14,13 +17,107 @@ pre-1.0
 
 # Next Step
 
-None: every issue still open is done on `next` and waits for it to reach `main`.
+None: no implementation work is open. The cache design,
+https://git.eeqj.de/sneak/quak/issues/36, waits on sneak's review.
 
 Tagging and releases are decided by sneak alone, and happen only when he
 declares one.
 
 # Completed Steps
 
+- 2026-10-02: `docker build .` stamps the commit's tag or short commit, not
+  `dev` (issue 154). `script/build` writes the version `script/version` prints
+  into `dist/package.json`: the `VERSION` environment variable or build arg when
+  one is given, otherwise `git describe --tags --always`, otherwise the version
+  `package.json` declares. `.dockerignore` sends `.git`, and a checkout with
+  `.git` whose version comes out empty, `dev` or `unknown` fails the build.
+  `make build-bin` bundles the built `dist/`, so the single binary reports the
+  same version.
+
+- 2026-10-01: `examples/download-albums.ts` logs in, opens the library, and for
+  every album downloads each photo to its save path, writes the photo's record
+  and EXIF fields to a JSON file beside it, and writes the album's photos to
+  `albums/<collectionID>.json` (issue 144). The build compiles it to
+  `dist/examples/`; the README's "Examples" section says how to run it.
+
+- 2026-10-01: A `Photo` has one async method for each field of `exif()`, named
+  and typed as in `PhotoExif`: `make()`, `model()`, `lensModel()`,
+  `dateTimeOriginal()`, `offsetTimeOriginal()`, `exposureTime()`, `fNumber()`,
+  `iso()`, `focalLength()`, `orientation()`, `gpsLatitude()`, `gpsLongitude()`
+  and `gpsAltitude()` (issue 148). Each calls `exif()` and returns its one
+  field, or undefined when the file lacks it. `Photo` implements a type with one
+  method per `PhotoExif` field, so the build's type check fails when a field has
+  no method. A test checks, on the JPEG and the HEIC, that each method gives the
+  same value as `exif()`.
+
+- 2026-10-01: Each original's save path is
+  `YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID>.<ext>` under the library's
+  download directory, which defaults to `photos` in the working directory (issue
+  143). The date is the photo's `takenAt` in the machine's time zone.
+  `photo.savePath` is always a string, with or without a content cache, and
+  `isLocal` is true only when the original is at its save path.
+  `photo.download()` puts the original there, copied from the cache when the
+  cache holds it and fetched otherwise. `lib.backup()` does the same for each
+  file, writes each file's JSON beside its original, and links `collections/` to
+  the save paths.
+
+- 2026-10-01: `photo.exif()` and `backup-metadata --exif` read EXIF from
+  HEIC/HEIF originals, a live photo's HEIC image included, as well as JPEG and
+  the other image formats `exifreader` reads (issue 145). `exifreader` replaces
+  `exif-reader` and the JPEG segment scan; `PhotoExif` is unchanged. The `exif`
+  field of `backup-metadata --exif` is now exifreader's tag output, and
+  `exifRaw` holds the whole EXIF block it could not read. The tests use a real
+  HEIC, `test/exif.heic`.
+
+- 2026-10-01: A `Photo` has `savePath`, `isLocal`, `content()`, `exif()`,
+  `modifiedAt`, `hash` and `year` (issue 141). `savePath` is where
+  `lib.backup()` writes the original under the library's download directory; for
+  a live photo not yet stored, it carries the title's extension, and the backup
+  may store the image under a different one. `isLocal` says whether the whole
+  original is there. Both look only at the disk. `content()` returns the
+  original's bytes and `exif()` the common EXIF fields of a JPEG; both may
+  download the original, and `exif()` downloads no video. `PhotoRecord` gains
+  `modifiedAt` and `hash`, and the JPEG EXIF scan moved to `src/exif.ts`.
+
+- 2026-09-29: The "Workflow" list at the top of this file now says to branch
+  from `next` and open a pull request that targets `next`, that the repository
+  manager squash-merges a reviewed pull request into `next`, and that only sneak
+  merges `next` into `main`, as the README does (issue 137).
+
+- 2026-09-29: The README's "Development workflow" and "For LLMs" sections now
+  say work branches from `next`, every pull request targets `next`, the
+  repository manager squash-merges reviewed pull requests into `next`, and only
+  sneak merges `next` into `main` (issue 135).
+
+- 2026-09-29: Brought the README and this file in line with `next` after the
+  milestone merge (issue 132). The Next Step says no implementation work is open
+  and the cache design (issue 36) waits on sneak's review. README corrections:
+  the Getting Started comments say when the library refreshes; most, not all,
+  Makefile targets call a script; `script/lint` and `script/test` have no host
+  path, though `yarn test` does; the SRP handshake uses `fast-srp-hap`, outside
+  `crypto/`, and a thumbnail upload's MD5 uses `node:crypto`; the SRP password
+  is the first 16 bytes of a 32-byte subkey; the key attributes and token come
+  after SRP, after the TOTP code SRP may ask for, or after the email OTP that
+  replaces SRP when the account has email MFA on, and quak cannot answer a
+  passkey; the auth token is sent as URL-safe base64 with padding; every
+  `TypeError` is retried; each download attempt writes its own temporary files,
+  two for a live photo, and only the attempt that completes renames them into
+  place; `runMetadataBackup` records a failed download in the file's JSON; a
+  second `client.logout()` does not throw; `quak logout` with no session exits
+  0, and names the cache directory only when it knows it; `login`, `whoami` and
+  `logout` open no library; `--exif` records XMP and, for a JPEG, EXIF, and no
+  IPTC; which commands take `--json`; a failed ML data request may not have been
+  retried; the thumbnail fixer is not limited to baseline JPEG; `quak backup`
+  still fetches ML data; a backup's JSON holds the basic metadata fields quak
+  keeps and the private and public magic metadata, not every decrypted field,
+  and the README no longer lists what the magic metadata holds; the default
+  cache directory is the per-user one, not an XDG path on macOS; pinned
+  originals can exceed `cacheOriginalsMaxBytes`; which tests cover which
+  operations; a default read is only as current as the last refresh whose
+  requests all succeeded; `fresh()` and `lib.backup()` join a refresh already
+  running; a `Photo` has no `thumbnailPath` or `originalPath`; and `408` and
+  `429` are retried.
+
 - 2026-09-28: Tested the live-photo writer's fsyncs (issue 130). A test checks
   that the image's and the video's temp files are fsynced before either is
   renamed into place, and the directory after both renames, as the `writeAtomic`

examples/download-albums.ts

@@ -0,0 +1,129 @@
+// Download every album's photos to a directory, with each photo's metadata
+// beside it, using only quak's public API. The README's "Examples" section
+// describes the files it writes.
+//
+//     yarn build
+//     QUAK_EMAIL=… QUAK_PASSWORD=… node dist/examples/download-albums.js [dir]
+
+import { realpathSync } from "node:fs";
+import { mkdir, readFile, writeFile } from "node:fs/promises";
+import { join, relative } from "node:path";
+import { stdin, stdout } from "node:process";
+import { createInterface } from "node:readline/promises";
+import { pathToFileURL } from "node:url";
+
+import { Client, Library } from "../src/index.js";
+
+// Write `text` to `path` unless the file already holds exactly that, so a
+// second run rewrites nothing.
+async function writeIfChanged(path: string, text: string): Promise<void> {
+    const current = await readFile(path, "utf-8").catch(() => undefined);
+    if (current !== text) await writeFile(path, text);
+}
+
+const pretty = (value: unknown): string =>
+    JSON.stringify(value, null, 2) + "\n";
+
+// For every album in `lib`, download each photo to its save path, write the
+// photo's metadata to `{savePath}.json`, and write the album's photos to
+// `{dir}/albums/{collectionID}.json`. Returns how many photos it downloaded
+// and how many were already at their save paths.
+export async function downloadAlbums(
+    lib: Library,
+    dir: string,
+): Promise<{ downloaded: number; alreadyLocal: number }> {
+    let downloaded = 0;
+    let alreadyLocal = 0;
+    // A photo in several albums is handled once.
+    const done = new Set<number>();
+    // fresh() waits for a refresh from the server and throws if it fails, so
+    // albums and photos added since the cache was last written are included.
+    const { albums } = await lib.fresh();
+    for (const album of albums.list()) {
+        const savePaths: string[] = [];
+        for (const photo of album.photos.list()) {
+            if (!done.has(photo.fileID)) {
+                done.add(photo.fileID);
+                if (photo.isLocal) alreadyLocal++;
+                else downloaded++;
+                await photo.download();
+                // The cache paths say where quak's cache keeps copies, not
+                // anything about the photo.
+                const record = { ...photo.record() };
+                delete record.thumbnailPath;
+                delete record.originalPath;
+                const exif = await photo.exif();
+                // savePath is read after download(): a live photo's names its
+                // image only once the image is stored.
+                await writeIfChanged(
+                    `${photo.savePath}.json`,
+                    pretty({ ...record, exif }),
+                );
+            }
+            savePaths.push(relative(dir, photo.savePath));
+        }
+        await mkdir(join(dir, "albums"), { recursive: true });
+        await writeIfChanged(
+            join(dir, "albums", `${album.collectionID}.json`),
+            pretty({
+                collectionID: album.collectionID,
+                name: album.name,
+                savePaths,
+            }),
+        );
+    }
+    return { downloaded, alreadyLocal };
+}
+
+// Ask for a login code on the terminal. Client.login calls this only when the
+// account requires a code.
+async function ask(question: string): Promise<string> {
+    const terminal = createInterface({ input: stdin, output: stdout });
+    try {
+        return await terminal.question(question);
+    } finally {
+        terminal.close();
+    }
+}
+
+async function main(): Promise<void> {
+    const email = process.env.QUAK_EMAIL;
+    const password = process.env.QUAK_PASSWORD;
+    if (!email || !password) {
+        console.error(
+            "Set QUAK_EMAIL and QUAK_PASSWORD to the account's email and password.",
+        );
+        process.exit(1);
+    }
+    const dir = process.argv[2] ?? "photos";
+
+    const client = await Client.login({
+        email,
+        password,
+        totp: () => ask("Two-factor code: "),
+        emailOTP: () => ask("Code sent to your email: "),
+    });
+    const lib = await Library.open({
+        client,
+        downloadDirectory: dir,
+        // As in `quak backup`: fetch only the originals this script saves,
+        // not every thumbnail and the recent originals into the cache too.
+        precacheThumbnails: false,
+        precacheOriginals: false,
+    });
+    try {
+        const { downloaded, alreadyLocal } = await downloadAlbums(lib, dir);
+        console.log(
+            `${downloaded} photos downloaded, ${alreadyLocal} already local, in ${dir}`,
+        );
+    } finally {
+        await lib.close();
+    }
+}
+
+// Run main() when node runs this file, not when a test imports it. argv[1] is
+// the path as given, and import.meta.url has symlinks resolved.
+const script = process.argv[1];
+if (script && pathToFileURL(realpathSync(script)).href === import.meta.url) {
+    await main();
+}

package.json

@@ -46,7 +46,7 @@
         "@inquirer/prompts": "8.5.2",
         "commander": "14.0.3",
         "env-paths": "4.0.0",
-        "exif-reader": "2.0.3",
+        "exifreader": "4.46.0",
         "fast-srp-hap": "2.0.4",
         "fflate": "0.8.3",
         "jpeg-js": "0.4.4",

script/build

@@ -1,7 +1,7 @@
 #!/bin/sh
-# script/build: compile the TypeScript sources into dist/, then verify that
+# script/build: compile the TypeScript sources into dist/, stamp the version
-# the artifacts package.json advertises are among the files the compiler
+# script/version prints into it, then verify that the artifacts package.json
-# actually wrote. tsc reports success by exit status alone and knows nothing
+# advertises are among the files the compiler actually wrote. tsc reports success by exit status alone and knows nothing
 # about the manifest, so without this step a green build can still ship a
 # package whose main, types or bin resolve to nothing. Our own extension to
 # scripts-to-rule-them-all.
@@ -46,13 +46,24 @@ for (const bin of bins) {
 }
 
 # src/index.ts imports ../package.json for the version, which tsc copies to
-# dist/package.json. Running the built CLI proves that import resolves from
+# dist/package.json. The version script/version prints is written into that
-# dist/ and reports the version package.json declares.
+# copy only; the repo's own package.json is left as it is.
+stamp_version() {
+    node -e '
+const { readFileSync, writeFileSync } = require("node:fs");
+
+const pkg = JSON.parse(readFileSync("dist/package.json", "utf-8"));
+pkg.version = process.argv[1];
+writeFileSync("dist/package.json", JSON.stringify(pkg, null, 4) + "\n");
+' "$1"
+}
+
+# Running the built CLI proves the import resolves from dist/ and reports
+# the stamped version.
 verify_version() {
     built="$(node dist/bin/quak.js --version)"
-    declared="$(node -p 'require("./package.json").version')"
+    if [ "$built" != "$1" ]; then
-    if [ "$built" != "$declared" ]; then
+        echo "build: dist/bin/quak.js reports $built, the build stamped $1" >&2
-        echo "build: dist/bin/quak.js reports $built, package.json declares $declared" >&2
         exit 1
     fi
     echo "build: dist/bin/quak.js reports version $built"
@@ -60,9 +71,12 @@ verify_version() {
 
 main() {
     cd "$ROOT"
+    # Own line, so that a failing script/version stops the build.
+    version="$("$ROOT/script/version")"
     yarn run tsc
+    stamp_version "$version"
     verify_entrypoints
-    verify_version
+    verify_version "$version"
 }
 
 main "$@"

script/cibuild

@@ -15,9 +15,9 @@ main() {
     cd "$ROOT"
     # Own line: a failing command substitution inside an argument does
     # not trip `set -e`, so the inline form degrades silently to an
-    # empty constant. VERSION is computed here because .dockerignore
+    # empty constant. The version resolved here goes in as the VERSION
-    # excludes .git, so `git describe` in a build stage yields an empty
+    # build arg, which takes precedence over what the build would derive
-    # version without failing.
+    # from the .git in its context.
     version="$(git describe --tags --always --dirty 2>/dev/null || true)"
     [ -n "$version" ] || version="unknown"
     docker build --no-cache \

script/docker

@@ -12,9 +12,9 @@ main() {
     cd "$ROOT"
     # Own line: a failing command substitution inside an argument does
     # not trip `set -e`, so the inline form degrades silently to an
-    # empty constant. VERSION is computed here because .dockerignore
+    # empty constant. The version resolved here goes in as the VERSION
-    # excludes .git, so `git describe` in a build stage yields an empty
+    # build arg, which takes precedence over what the build would derive
-    # version without failing.
+    # from the .git in its context.
     version="$(git describe --tags --always --dirty 2>/dev/null || true)"
     [ -n "$version" ] || version="unknown"
     docker build --no-cache \

script/version

@@ -0,0 +1,39 @@
+#!/bin/sh
+# script/version: print the version script/build stamps into the built
+# package. Our own extension to scripts-to-rule-them-all.
+#
+# Order of precedence:
+#
+#   1. $VERSION, if set and not empty: an explicit value, such as the
+#      Dockerfile's VERSION build arg.
+#   2. If this checkout has .git, `git describe --tags --always`: the tag
+#      on a tagged commit, otherwise the short commit.
+#   3. Otherwise, as in a source tarball, the version package.json declares.
+#
+# A checkout with .git whose version still comes out empty, dev or unknown
+# fails: git is missing or could not read the checkout, and the build could
+# not be traced back to its commit.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    version="${VERSION:-}"
+    if [ -e .git ]; then
+        if [ -z "$version" ]; then
+            version="$(git describe --tags --always || true)"
+        fi
+        case "$version" in
+            "" | dev | unknown)
+                echo "version: $ROOT has .git, but the version came out '$version'" >&2
+                exit 1
+                ;;
+        esac
+    elif [ -z "$version" ]; then
+        version="$(node -p 'require("./package.json").version')"
+    fi
+    echo "$version"
+}
+
+main "$@"

src/backup.ts

@@ -2,29 +2,30 @@
 //
 // `lib.backup()` waits for a completed refresh of the library (a failed one
 // fails the backup before any file is touched), then, for every file in scope,
-// gets its original bytes onto disk under `downloadDirectory` and rebuilds the
+// puts its original at its save path under `downloadDirectory`, as
-// derived views (per-file sidecars, per-collection symlink trees,
+// `Photo.download()` does, and rebuilds the derived views (per-file sidecars,
-// per-collection JSON) from the model. The on-disk layout is the historical
+// per-collection symlink trees, per-collection JSON) from the model. The
-// one:
+// on-disk layout:
 //
 //     <downloadDirectory>/
-//         originals/<fileID>.<ext>     the decrypted bytes
+//         YYYY/YYYY-MM/YYYY-MM-DD/
-//         originals/<fileID>.json      per-file metadata sidecar
+//             YYYY-MM-DD.<fileID>.<ext>    the decrypted bytes (the save path)
-//         collections/<name>/<title>   symlink into ../../originals
+//             YYYY-MM-DD.<fileID>.json     per-file metadata sidecar
-//         collections/<name>.json      per-collection metadata
+//         collections/<name>/<title>       symlink to the original
-//         failures.json                durable ledger of unresolved failures
+//         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
+// A live photo's original is its image and its video, each with its own
-// with its own extension, and `originals/<fileID>.livephoto.json` naming them;
+// extension, beside `YYYY-MM-DD.<fileID>.livephoto.json` naming them; its album
-// its album folders link both.
+// 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
+// original appears at its save path 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
+// interrupted run resumes by looking at the save paths. The derived views hold
-// unique state, so they are rebuilt every run; that repairs stale sidecars and
+// no unique state, so they are rebuilt every run; that repairs stale sidecars
-// missing or broken symlinks left by an earlier crash. A rebuild also removes
+// and missing or broken symlinks left by an earlier crash. A rebuild also
-// the symlinks into originals/ that no longer belong to an album, and the
+// removes the symlinks to 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
@@ -48,24 +49,25 @@ import {
     symlinkSync,
     writeFileSync,
 } from "node:fs";
-import { copyFile, rename, rm } from "node:fs/promises";
+import { dirname, extname, join, relative, resolve } from "node:path";
-import { basename, dirname, extname, join, relative } from "node:path";
 
-import { fsyncPath, removeLeftoverTempFiles } from "./download/index.js";
+import { removeLeftoverTempFiles } from "./download/index.js";
 import { sanitizeFileName, withExtension } from "./filename.js";
 import {
-    nameInOriginals,
+    copyAtomic,
-    storedOriginal,
+    placeOriginal,
-    writeLivePhotoJSON,
+    savePath,
+    storedAtSavePath,
 } from "./library/content.js";
+import { representative } from "./library/records.js";
 import type { Collection, EnteFile } from "./model/types.js";
 
 export type ProgressCallback = (message: string) => void;
 
 export interface BackupOptions {
-    // Where the backup tree lives. Required: with none, `backup()` throws
+    // Where the backup tree lives. `lib.backup()` defaults it to the library's
-    // before any network traffic. A library opened with a `downloadDirectory`
+    // download directory; `runBackup` with none throws before any network
-    // supplies the default.
+    // traffic.
     downloadDirectory?: string;
     // Fetch and store full-resolution originals. Default true.
     includeOriginals?: boolean;
@@ -89,7 +91,7 @@ export interface BackupResult {
     totalFiles: number;
     // Originals fetched (or copied from the cache) this run.
     downloaded: number;
-    // Originals already present and left untouched.
+    // Originals already at their save path and left untouched.
     skipped: number;
     // Files with an unresolved failure after this run (the ledger size); the
     // CLI exits non-zero while this is above zero. A file can be both
@@ -107,8 +109,8 @@ export interface BackupLibrary {
     listFiles(collectionID: number): EnteFile[];
     // Get an original's bytes onto disk through the content cache/pools,
     // returning where they landed: `destination` when they were fetched now,
-    // otherwise wherever they already were (the cache, or a prior backup). A
+    // otherwise wherever they already were (the cache, or the library's save
-    // live photo lands as its image and its video, fetched now beside
+    // path). A live photo lands as its image and its video, fetched now beside
     // `destination`.
     original(
         fileID: number,
@@ -168,58 +170,6 @@ const classify = (err: unknown): FailureClass => {
 const errorMessage = (err: unknown): string =>
     err instanceof Error ? err.message : String(err);
 
-// Copy bytes into `dest` via a temp file in the same directory plus rename, so
-// `dest` appears only once it is whole ("present means complete"). As in the
-// download writer, the temp file is fsynced before the rename and the directory
-// after it, so a power cut cannot leave a correctly named but short original.
-// The temp name carries this process's ID so a later run can tell a leftover
-// from a copy still in progress (see `removeLeftoverTempFiles`).
-const copyAtomic = async (src: string, dest: string): Promise<void> => {
-    if (src === dest) return;
-    const tmp = join(
-        dirname(dest),
-        `.quak-backup-${basename(dest)}-${process.pid}-${Math.random()
-            .toString(36)
-            .slice(2)}.tmp`,
-    );
-    try {
-        await copyFile(src, tmp);
-        await fsyncPath(tmp);
-        // `rename` replaces the destination's directory entry: an existing
-        // symlink at `dest` is replaced, not followed, and the new file has
-        // the temp file's permissions (copied from `src`).
-        await rename(tmp, dest);
-        await fsyncPath(dirname(dest));
-    } finally {
-        await rm(tmp, { force: true });
-    }
-};
-
-// Put an original the library returned at `dest` in originals/, where a fresh
-// fetch already wrote it. A live photo's image and video go beside `dest`: when
-// they came from the cache they are copied, after removing whatever was at
-// `dest` (an earlier version's ZIP of the two). Then the JSON file naming them
-// is written, which is what makes the live photo count as stored.
-const placeOriginal = async (
-    file: EnteFile,
-    dest: string,
-    got: { path: string; videoPath?: string },
-): Promise<void> => {
-    if (got.videoPath === undefined) {
-        await copyAtomic(got.path, dest);
-        return;
-    }
-    const originalsDir = dirname(dest);
-    const path = join(originalsDir, basename(got.path));
-    const videoPath = join(originalsDir, basename(got.videoPath));
-    if (got.path !== path) {
-        await rm(dest, { force: true });
-        await copyAtomic(got.path, path);
-        await copyAtomic(got.videoPath, videoPath);
-    }
-    await writeLivePhotoJSON(originalsDir, file.id, { path, videoPath });
-};
-
 // Ensure `linkPath` is a symlink to `target`, rebuilding a missing, wrong, or
 // non-symlink entry. Throws on failure (a directory in the way, no permission)
 // so the caller records it and moves on rather than aborting the run.
@@ -291,36 +241,55 @@ const linksFor = (
     }));
 };
 
-// Remove the symlinks in the album directory `dir` that point into
+// Every date folder (`YYYY/YYYY-MM/YYYY-MM-DD/`) under `root`, whether or not a
-// `originalsDir` and are not named in `keep`. Nothing else in the directory
+// file in this backup is saved there. A folder that cannot be read is skipped.
-// is touched: anything else there was put there by the user.
+const dateFolders = (root: string): string[] => {
+    const subfolders = (dir: string, name: RegExp): string[] => {
+        try {
+            return readdirSync(dir, { withFileTypes: true })
+                .filter((e) => e.isDirectory() && name.test(e.name))
+                .map((e) => join(dir, e.name));
+        } catch {
+            return [];
+        }
+    };
+    return subfolders(root, /^\d{4}$/)
+        .flatMap((year) => subfolders(year, /^\d{4}-\d\d$/))
+        .flatMap((month) => subfolders(month, /^\d{4}-\d\d-\d\d$/));
+};
+
+// Whether the entry at `path` is a symlink a backup to `root` made: one to an
+// original in a `YYYY/YYYY-MM/YYYY-MM-DD/` folder of `root`.
+const linksToOriginal = (path: string, root: string): boolean => {
+    if (!lstatSync(path).isSymbolicLink()) return false;
+    const target = relative(root, resolve(dirname(path), readlinkSync(path)));
+    return /^\d{4}\/\d{4}-\d\d\/\d{4}-\d\d-\d\d\/[^/]+$/.test(target);
+};
+
+// Remove the symlinks in the album directory `dir` that point to an original
+// in `root` 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,
+    root: string,
 ): void => {
-    const target = relative(dir, originalsDir);
     for (const name of readdirSync(dir)) {
         if (keep.has(name)) continue;
         const path = join(dir, name);
-        if (
+        if (linksToOriginal(path, root)) rmSync(path);
-            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
+// symlinks to originals in `root` are removed; if that leaves it empty, it and
-// JSON are deleted, otherwise both stay for what the user put there.
+// its JSON are deleted, otherwise both stay for what the user put there.
 const removeStaleAlbumDirs = (
     collectionsDir: string,
     current: Set<string>,
-    originalsDir: string,
+    root: string,
 ): void => {
     for (const entry of readdirSync(collectionsDir, { withFileTypes: true })) {
         if (!entry.isDirectory() || current.has(entry.name)) continue;
@@ -334,7 +303,7 @@ const removeStaleAlbumDirs = (
             continue;
         }
         const dir = join(collectionsDir, entry.name);
-        removeStaleLinks(dir, new Set(), originalsDir);
+        removeStaleLinks(dir, new Set(), root);
         if (readdirSync(dir).length > 0) continue;
         rmdirSync(dir);
         rmSync(jsonPath);
@@ -402,34 +371,48 @@ export const runBackup = async (
     log("Refreshing library...");
     await lib.refresh();
 
-    const originalsDir = join(downloadDirectory, "originals");
     const collectionsDir = join(downloadDirectory, "collections");
     const thumbnailsDir = join(downloadDirectory, "thumbnails");
-    mkdirSync(originalsDir, { recursive: true });
     mkdirSync(collectionsDir, { recursive: true });
     if (includeThumbnails) mkdirSync(thumbnailsDir, { recursive: true });
-    removeLeftoverTempFiles(originalsDir);
     removeLeftoverTempFiles(thumbnailsDir);
+    for (const dir of dateFolders(downloadDirectory)) {
+        removeLeftoverTempFiles(dir);
+    }
 
     const ledgerPath = join(downloadDirectory, "failures.json");
     const ledger = loadLedger(ledgerPath);
     const now = Date.now();
 
     // Collections in scope, and the distinct files across them (a file shared
-    // by two albums is one original).
+    // by two albums is one original). Each file is the membership
+    // `representative` picks from all of its albums, in scope or not, so it is
+    // saved at the path `photo.savePath` names.
     const allCollections = lib.listCollections();
     const collections = allCollections.filter((c) =>
         only ? only.has(c.name) : true,
     );
     const collectionName = new Map<number, string>();
-    for (const c of collections) collectionName.set(c.id, c.name);
+    for (const c of allCollections) collectionName.set(c.id, c.name);
 
-    const distinct = new Map<number, EnteFile>();
+    const memberships = new Map<number, EnteFile[]>();
     const filesByCollection = new Map<number, EnteFile[]>();
-    for (const c of collections) {
+    for (const c of allCollections) {
         const files = lib.listFiles(c.id);
         filesByCollection.set(c.id, files);
-        for (const f of files) if (!distinct.has(f.id)) distinct.set(f.id, f);
+        for (const f of files) {
+            const arr = memberships.get(f.id);
+            if (arr) arr.push(f);
+            else memberships.set(f.id, [f]);
+        }
+    }
+    const distinct = new Map<number, EnteFile>();
+    for (const c of collections) {
+        for (const f of filesByCollection.get(c.id)!) {
+            if (!distinct.has(f.id)) {
+                distinct.set(f.id, representative(memberships.get(f.id)!));
+            }
+        }
     }
 
     const errors: BackupError[] = [];
@@ -465,25 +448,22 @@ export const runBackup = async (
         failedThisRun.add(file.id);
     };
 
-    // Phase 1: get the bytes. Fetch each pending original (and optional
+    // Phase 1: get the bytes. Put each pending original at its save path
-    // thumbnail) through the content cache/pools and place it under the backup
+    // through the content cache/pools, as `Photo.download()` does, and fetch
-    // tree; a present file is left as is.
+    // the optional thumbnails; a present file is left as is.
     if (includeOriginals) {
         for (const [fileID, file] of distinct) {
-            if (storedOriginal(originalsDir, file) !== undefined) {
+            if (storedAtSavePath(downloadDirectory, file) !== undefined) {
                 skipped++;
                 continue;
             }
-            const dest = join(originalsDir, nameInOriginals(file));
             try {
                 log(`Fetching original ${file.metadata.title} (${fileID})...`);
-                // A fetched original is written straight to `dest` (a live
+                // A fetched original is written straight to its save path (a
-                // photo beside it); only one that was already cached elsewhere
+                // live photo beside it); only one that was already cached
-                // is copied.
+                // elsewhere is copied.
-                await placeOriginal(
+                await placeOriginal(downloadDirectory, file, (dest) =>
-                    file,
+                    lib.original(fileID, dest),
-                    dest,
-                    await lib.original(fileID, dest),
                 );
                 downloaded++;
             } catch (err) {
@@ -519,9 +499,10 @@ export const runBackup = async (
     // Phase 2: rebuild the derived views from the model. Sidecars first, for
     // every present original (this repairs stale ones).
     if (includeOriginals) {
-        for (const [fileID, file] of distinct) {
+        for (const file of distinct.values()) {
-            if (storedOriginal(originalsDir, file) !== undefined) {
+            if (storedAtSavePath(downloadDirectory, file) !== undefined) {
-                writeSidecar(join(originalsDir, `${fileID}.json`), file);
+                const path = savePath(downloadDirectory, file);
+                writeSidecar(withExtension(path, ".json"), file);
             }
         }
     }
@@ -543,7 +524,11 @@ export const runBackup = async (
         allCollections.map((c, i) => [c.id, dirNames[i]!]),
     );
     try {
-        removeStaleAlbumDirs(collectionsDir, new Set(dirNames), originalsDir);
+        removeStaleAlbumDirs(
+            collectionsDir,
+            new Set(dirNames),
+            downloadDirectory,
+        );
     } catch (err) {
         log(`FAILED removing old album directories: ${errorMessage(err)}`);
     }
@@ -553,13 +538,18 @@ export const runBackup = async (
         const colDir = join(collectionsDir, colDirName);
         mkdirSync(colDir, { recursive: true });
 
+        // Every album links the one original, saved from the file's entry in
+        // `distinct`.
         const files = filesByCollection.get(c.id) ?? [];
         const links = files.flatMap((f) =>
-            linksFor(f, storedOriginal(originalsDir, f)),
+            linksFor(
+                f,
+                storedAtSavePath(downloadDirectory, distinct.get(f.id)!),
+            ),
         );
         const linkNames = uniqueNames(links, true);
         try {
-            removeStaleLinks(colDir, new Set(linkNames), originalsDir);
+            removeStaleLinks(colDir, new Set(linkNames), downloadDirectory);
         } catch (err) {
             log(`FAILED removing old links in ${c.name}: ${errorMessage(err)}`);
         }

src/exif.ts

@@ -0,0 +1,115 @@
+// EXIF in an original's bytes, read with exifreader, which reads it from JPEG,
+// HEIC/HEIF, AVIF, PNG, WebP and the other image formats it supports.
+// `backup-metadata --exif` records every EXIF tag it finds except the
+// thumbnail's; `Photo.exif()` returns the common fields picked from them here.
+
+import ExifReader, { type ExpandedTags } from "exifreader";
+
+// The EXIF tags in `bytes` (`exif`), the GPS position exifreader computes from
+// them (`gps`), and where the EXIF block lies in `bytes` (`metadataRange`).
+// Undefined when exifreader cannot read the file at all, such as a video. An
+// EXIF block it finds but reads no tag from comes back as an empty `exif`.
+// `exif` holds every tag except the thumbnail's; a tag exifreader has no name
+// for is keyed `undefined-<tag number>`. Each tag's `computed` holds its value
+// as a string or number, or as an array of them for a tag with several values,
+// such as `GPSLatitude`'s `[40, 26, 46]`. A fraction with a zero denominator
+// computes to null.
+export const readExifTags = (bytes: Uint8Array): ExpandedTags | undefined => {
+    try {
+        return ExifReader.loadView(
+            new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength),
+            {
+                expanded: true,
+                computed: true,
+                includeOffsets: true,
+                includeUnknown: true,
+                includeTags: { exif: true, gps: true },
+            },
+        );
+    } catch {
+        return undefined;
+    }
+};
+
+// The common EXIF fields of an original. Each is absent when the file lacks it.
+export interface PhotoExif {
+    make?: string;
+    model?: string;
+    lensModel?: string;
+    // When the photo was taken, by the camera's clock. EXIF writes this as text
+    // with no time zone, and it is read as if it were UTC: the Date's UTC
+    // fields are the clock reading, which is the moment it was taken only when
+    // the clock was set to UTC.
+    dateTimeOriginal?: Date;
+    // The camera clock's offset from UTC, such as "+02:00".
+    offsetTimeOriginal?: string;
+    // Seconds.
+    exposureTime?: number;
+    fNumber?: number;
+    iso?: number;
+    // Millimetres.
+    focalLength?: number;
+    // The EXIF orientation code, 1 to 8.
+    orientation?: number;
+    // Decimal degrees, negative south of the equator and west of Greenwich.
+    gpsLatitude?: number;
+    gpsLongitude?: number;
+    // Metres, negative below sea level.
+    gpsAltitude?: number;
+}
+
+// exifreader gives "<faulty value>" for a tag whose value lies outside the
+// file; that tag is left out like one the file lacks.
+const asString = (v: unknown): string | undefined =>
+    typeof v === "string" && v.length > 0 && v !== "<faulty value>"
+        ? v
+        : undefined;
+
+const asNumber = (v: unknown): number | undefined =>
+    typeof v === "number" && Number.isFinite(v) ? v : undefined;
+
+// EXIF writes a date and time as "2021:07:15 14:30:00". This is that reading
+// in a Date's UTC fields.
+const asDate = (v: unknown): Date | undefined => {
+    const m =
+        typeof v === "string"
+            ? /^(\d{4}):(\d{2}):(\d{2}) (\d{2}:\d{2}:\d{2})$/.exec(v)
+            : null;
+    if (!m) return undefined;
+    const date = new Date(`${m[1]}-${m[2]}-${m[3]}T${m[4]}Z`);
+    return Number.isNaN(date.getTime()) ? undefined : date;
+};
+
+// The common fields of an original's EXIF: `{}` when the file has no EXIF,
+// exifreader cannot read its EXIF, or it is not an image exifreader reads.
+export const readPhotoExif = (bytes: Uint8Array): PhotoExif => {
+    const tags = readExifTags(bytes);
+    const exif = tags?.exif;
+    const gps = tags?.gps;
+    const altitude = asNumber(exif?.GPSAltitude?.computed);
+    const fields: PhotoExif = {
+        make: asString(exif?.Make?.computed),
+        model: asString(exif?.Model?.computed),
+        lensModel: asString(exif?.LensModel?.computed),
+        dateTimeOriginal: asDate(exif?.DateTimeOriginal?.computed),
+        offsetTimeOriginal: asString(exif?.OffsetTimeOriginal?.computed),
+        exposureTime: asNumber(exif?.ExposureTime?.computed),
+        fNumber: asNumber(exif?.FNumber?.computed),
+        // Only when the tag holds a single number, as most cameras write it.
+        iso: asNumber(exif?.ISOSpeedRatings?.computed),
+        focalLength: asNumber(exif?.FocalLength?.computed),
+        orientation: asNumber(exif?.Orientation?.computed),
+        gpsLatitude: asNumber(gps?.Latitude),
+        gpsLongitude: asNumber(gps?.Longitude),
+        // A GPSAltitudeRef of 1 means the altitude is below sea level.
+        gpsAltitude:
+            altitude !== undefined && exif?.GPSAltitudeRef?.value === 1
+                ? -altitude
+                : altitude,
+    };
+    // Leave out what the file lacks, so a missing field is absent rather than
+    // present and undefined.
+    return Object.fromEntries(
+        Object.entries(fields).filter(([, v]) => v !== undefined),
+    ) as PhotoExif;
+};

src/index.ts

@@ -53,6 +53,7 @@ export {
     type PhotoFilter,
     type TimelineGroup,
     type GroupBy,
+    type SavePathLookup,
     type ContentSource,
     type ContentResult,
     type ContentEvent,
@@ -84,6 +85,7 @@ export type {
     LibrarySnapshot,
     LibraryChange,
 } from "./library/records.js";
+export type { PhotoExif } from "./exif.js";
 export { decryptCollection, decryptFile } from "./model/index.js";
 export { downloadFile, downloadThumbnail } from "./download/index.js";
 export type {

src/library/content.ts

@@ -23,6 +23,11 @@
 // original with no recorded hash is stored unchecked, as the upstream client
 // does; thumbnails have none. On top of that this module refuses to record a
 // stored file that came out empty.
+//
+// It also names each original's save path under the download directory
+// (`savePath`), where `Photo.download()` and `lib.backup()` put it. A copy in
+// the cache does not count as saved there, but is copied there rather than
+// fetched again.
 
 import {
     closeSync,
@@ -34,8 +39,10 @@ import {
 } from "node:fs";
 import {
     chmod,
+    copyFile,
     mkdir,
     readdir,
+    rename,
     rm,
     stat,
     statfs,
@@ -47,13 +54,15 @@ import type { ApiClient } from "../api/client.js";
 import {
     downloadFile,
     downloadThumbnail,
+    fsyncPath,
     type ProgressCallback,
     removeLeftoverTempFiles,
     writeAtomic,
 } from "../download/index.js";
-import { safeExtension } from "../filename.js";
+import { safeExtension, withExtension } from "../filename.js";
 import type { EnteFile } from "../model/types.js";
 import type { Priority, RequestPools } from "./pools.js";
+import { takenAtOf } from "./records.js";
 
 const DIR_MODE = 0o700;
 const FILE_MODE = 0o600;
@@ -108,6 +117,9 @@ export interface ContentOptions {
 export interface PhotoContent {
     original(fileID: number, opts?: ContentOptions): Promise<ContentResult>;
     thumbnail(fileID: number, opts?: ContentOptions): Promise<ContentResult>;
+    // Put the original at the save path of `file`, the copy the `Photo` holds,
+    // and return it there.
+    download(file: EnteFile): Promise<ContentResult>;
 }
 
 export interface EnsureResult {
@@ -194,10 +206,10 @@ export interface ContentCacheOptions {
     pools: RequestPools;
     source: ContentSource;
     cacheDirectory: string;
-    // The backup destination (issue-level `downloadDirectory`). An original
+    // The root of the save paths. An original already stored at its save path
-    // already stored there by a backup counts as present, so the cache serves
+    // counts as present, so the cache serves it rather than fetching a second
-    // it rather than fetching a second copy.
+    // copy.
-    downloadDirectory?: string;
+    downloadDirectory: string;
     // Resolve any membership of a file; every membership shares the underlying
     // content key, so any one decrypts the same bytes.
     getFile: (fileID: number) => EnteFile | undefined;
@@ -224,11 +236,27 @@ class AbortDrop extends Error {
     }
 }
 
-// The name of a file's original in originals/: `<fileID><ext>`, the extension
+// The name of a file's original in the cache's originals/: `<fileID><ext>`, the
-// taken from the title (or `.bin`). A backup names its originals the same way.
+// extension taken from the title (or `.bin`).
 export const nameInOriginals = (file: EnteFile): string =>
     `${file.id}${safeExtension(file.metadata.title)}`;
 
+const pad = (n: number): string => String(n).padStart(2, "0");
+
+// Where the original of `file` is saved under `root`:
+// `YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID><ext>`, dated by the photo's
+// `takenAt` in this machine's time zone, with the extension taken from the
+// title (or `.bin`). A live photo is stored as its image and its video beside
+// this path, each with the extension found inside the live photo.
+export const savePath = (root: string, file: EnteFile): string => {
+    const taken = new Date(takenAtOf(file));
+    const year = String(taken.getFullYear());
+    const month = `${year}-${pad(taken.getMonth() + 1)}`;
+    const day = `${month}-${pad(taken.getDate())}`;
+    const ext = safeExtension(file.metadata.title);
+    return join(root, year, month, day, `${day}.${file.id}${ext}`);
+};
+
 // The fileID a cache filename encodes, or undefined when the name is not one
 // the cache writes (`<digits><ext>`).
 const fileIDFromName = (name: string): number | undefined => {
@@ -274,28 +302,29 @@ const isZip = (path: string): boolean => {
 
 // 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, a JSON file of this name beside them names both.
+// save path stores one, a JSON file of this name beside them names both.
-const livePhotoJSONName = (fileID: number): string =>
+// `name` is the original's name without its extension: `<fileID>` in the
-    `${fileID}.livephoto.json`;
+// cache, `YYYY-MM-DD.<fileID>` at the save path.
+const livePhotoJSONName = (name: string): string => `${name}.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
+// undefined when there is none. Only `name` with an extension is taken, so the
-// taken, so the file cannot point outside `dir`.
+// file cannot point outside `dir`.
 const readLivePhotoJSON = (
     dir: string,
-    fileID: number,
+    name: string,
 ): { path: string; videoPath: string } | undefined => {
-    const valid = (name: unknown): name is string =>
+    const valid = (part: unknown): part is string =>
-        typeof name === "string" && name === `${fileID}${safeExtension(name)}`;
+        typeof part === "string" && part === `${name}${safeExtension(part)}`;
     try {
         const { image, video } = JSON.parse(
-            readFileSync(join(dir, livePhotoJSONName(fileID)), "utf-8"),
+            readFileSync(join(dir, livePhotoJSONName(name)), "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.
+        // No such file, or not one quak wrote.
     }
     return undefined;
 };
@@ -303,11 +332,11 @@ const readLivePhotoJSON = (
 // Write the JSON file naming a live photo's image and video, both in `dir`.
 export const writeLivePhotoJSON = (
     dir: string,
-    fileID: number,
+    name: string,
     stored: { path: string; videoPath: string },
 ): Promise<void> =>
     writeAtomic(
-        join(dir, livePhotoJSONName(fileID)),
+        join(dir, livePhotoJSONName(name)),
         new TextEncoder().encode(
             JSON.stringify({
                 image: basename(stored.path),
@@ -316,17 +345,19 @@ export const writeLivePhotoJSON = (
         ),
     );
 
-// The original of `file` as the cache or a backup stored it in `dir`, when all
+// The original of `file` as stored in `dir` under `name` (without its
-// of it is there: `<fileID><ext>`, or a live photo's image and video.
+// extension), when all of it is there: `<name><ext>`, or a live photo's image
+// and video.
 export const storedOriginal = (
     dir: string,
+    name: string,
     file: EnteFile,
 ): { path: string; videoPath?: string } | undefined => {
     if (file.metadata.fileType !== "livePhoto") {
-        const path = join(dir, nameInOriginals(file));
+        const path = join(dir, `${name}${safeExtension(file.metadata.title)}`);
         return hasContent(path) ? { path } : undefined;
     }
-    const stored = readLivePhotoJSON(dir, file.id);
+    const stored = readLivePhotoJSON(dir, name);
     return stored !== undefined &&
         hasContent(stored.path) &&
         hasContent(stored.videoPath)
@@ -334,10 +365,76 @@ export const storedOriginal = (
         : undefined;
 };
 
+// The original of `file` as stored at its save path under `root`, when all of
+// it is there.
+export const storedAtSavePath = (
+    root: string,
+    file: EnteFile,
+): { path: string; videoPath?: string } | undefined => {
+    const path = savePath(root, file);
+    return storedOriginal(dirname(path), basename(path, extname(path)), file);
+};
+
+// Copy bytes into `dest` via a temp file in the same directory plus rename, so
+// `dest` appears only once it is whole ("present means complete"). As in the
+// download writer, the temp file is fsynced before the rename and the directory
+// after it, so a power cut cannot leave a correctly named but short original.
+// The temp name carries this process's ID so a later run can tell a leftover
+// from a copy still in progress (see `removeLeftoverTempFiles`).
+export const copyAtomic = async (src: string, dest: string): Promise<void> => {
+    if (src === dest) return;
+    const tmp = join(
+        dirname(dest),
+        `.quak-backup-${basename(dest)}-${process.pid}-${Math.random()
+            .toString(36)
+            .slice(2)}.tmp`,
+    );
+    try {
+        await copyFile(src, tmp);
+        await fsyncPath(tmp);
+        // `rename` replaces the destination's directory entry: an existing
+        // symlink at `dest` is replaced, not followed, and the new file has
+        // the temp file's permissions (copied from `src`).
+        await rename(tmp, dest);
+        await fsyncPath(dirname(dest));
+    } finally {
+        await rm(tmp, { force: true });
+    }
+};
+
+// Put the original of `file` at its save path under `root`, creating its
+// folders. `get` is given the save path and returns where the original is: a
+// fetch writes it there, and a copy the cache holds is copied there. A live
+// photo's image and video go beside the save path, each with its own
+// extension: when they came from the cache they are copied. Then the JSON file
+// naming them is written, which is what makes the live photo count as stored.
+export const placeOriginal = async (
+    root: string,
+    file: EnteFile,
+    get: (dest: string) => Promise<{ path: string; videoPath?: string }>,
+): Promise<{ path: string; videoPath?: string }> => {
+    const dest = savePath(root, file);
+    await mkdir(dirname(dest), { recursive: true });
+    const got = await get(dest);
+    if (got.videoPath === undefined) {
+        await copyAtomic(got.path, dest);
+        return { path: dest };
+    }
+    const path = withExtension(dest, extname(got.path));
+    const videoPath = withExtension(dest, extname(got.videoPath));
+    if (got.path !== path) {
+        await copyAtomic(got.path, path);
+        await copyAtomic(got.videoPath, videoPath);
+    }
+    const name = basename(dest, extname(dest));
+    await writeLivePhotoJSON(dirname(dest), name, { path, videoPath });
+    return { path, videoPath };
+};
+
 export class ContentCache implements PhotoContent, ThumbnailsAPI {
     private readonly pools: RequestPools;
     private readonly source: ContentSource;
-    private readonly downloadDirectory?: string;
+    private readonly downloadDirectory: string;
     private readonly getFile: (fileID: number) => EnteFile | undefined;
     private readonly originalsDir: string;
     private readonly thumbnailsDir: string;
@@ -435,7 +532,23 @@ 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
+    // Put the original at the save path of `file` under the download directory
+    // and return it there. `file` is the copy the `Photo` holds, so the path is
+    // the one its `savePath` names, even after a refresh changed the date. One
+    // already stored there is returned as it is; one the cache holds is copied
+    // from it; any other is fetched straight to the save path, with no copy
+    // left in the cache.
+    async download(file: EnteFile): Promise<ContentResult> {
+        const root = this.downloadDirectory;
+        const saved =
+            storedAtSavePath(root, file) ??
+            (await placeOriginal(root, file, (dest) =>
+                this.backupOriginal(file.id, dest),
+            ));
+        return { ...saved, bytes: fileSize(saved.path) ?? 0 };
+    }
+
+    // Get an original for a save path. 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(
@@ -605,12 +718,9 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
             known.delete(fileID);
         }
 
-        // An original a backup already stored counts as present.
+        // An original already stored at its save path counts as present.
-        if (kind === "original" && this.downloadDirectory !== undefined) {
+        if (kind === "original") {
-            const stored = storedOriginal(
+            const stored = storedAtSavePath(this.downloadDirectory, file);
-                join(this.downloadDirectory, "originals"),
-                file,
-            );
             if (stored !== undefined) {
                 this.originals.set(fileID, stored);
                 return {
@@ -646,7 +756,7 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
                         ? this.beginOriginalWrite(fileID)
                         : null;
                 try {
-                    const stored = await this.download(
+                    const stored = await this.fetchInto(
                         file,
                         dest,
                         kind,
@@ -661,12 +771,12 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
                             );
                         }
                     }
-                    // A backup records its own live photos.
+                    // `placeOriginal` records a live photo it saves.
                     if (
                         stored.videoPath !== undefined &&
                         opts?.destination === undefined
                     ) {
-                        await writeLivePhotoJSON(dir, fileID, {
+                        await writeLivePhotoJSON(dir, String(fileID), {
                             path: stored.path,
                             videoPath: stored.videoPath,
                         });
@@ -689,7 +799,7 @@ 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(
+    private async fetchInto(
         file: EnteFile,
         destination: string,
         kind: Kind,
@@ -714,10 +824,10 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
         await utimes(path, now, now).catch(() => undefined);
     }
 
-    // Every stored original that lives under `originalsDir` (a backup-directory
+    // Every stored original that lives under `originalsDir` (a save-path hit
-    // hit recorded in the map is excluded), with its size and mtime; a live
+    // recorded in the map is excluded), with its size and mtime; a live
     // photo's size includes its video. Entries whose file has vanished are
-    // dropped from the map. Backups and thumbnails are never counted.
+    // dropped from the map. Save paths and thumbnails are never counted.
     private async measureOriginals(): Promise<{
         entries: {
             fileID: number;
@@ -823,7 +933,7 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
                         await rm(
                             join(
                                 this.originalsDir,
-                                livePhotoJSONName(e.fileID),
+                                livePhotoJSONName(String(e.fileID)),
                             ),
                             { force: true },
                         );
@@ -877,8 +987,8 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
             // earlier version stored under the image's name, and is removed.
             // Any other is left alone: another process may have just stored
             // it and not yet written the JSON file.
-            const livePhoto = names.has(livePhotoJSONName(id))
+            const livePhoto = names.has(livePhotoJSONName(String(id)))
-                ? readLivePhotoJSON(dir, id)
+                ? readLivePhotoJSON(dir, String(id))
                 : undefined;
             if (livePhoto !== undefined) {
                 into.set(id, livePhoto);

src/library/index.ts

@@ -27,7 +27,7 @@
 // never masked by a subsequent empty refresh.
 
 import { rm } from "node:fs/promises";
-import { join } from "node:path";
+import { join, resolve } from "node:path";
 import envPaths from "env-paths";
 
 import { MetadataStore } from "./store.js";
@@ -49,9 +49,12 @@ import {
     type PhotosAPI,
     type TimelineAPI,
     type FreshReads,
+    type SavePathLookup,
 } from "./read.js";
 import {
     ContentCache,
+    savePath,
+    storedAtSavePath,
     type ContentSource,
     type ThumbnailsAPI,
     type EnsureOptions,
@@ -70,6 +73,7 @@ export {
     type PhotoFilter,
     type TimelineGroup,
     type GroupBy,
+    type SavePathLookup,
 } from "./read.js";
 export {
     type ContentSource,
@@ -169,8 +173,10 @@ export interface LibraryOptions {
     // Where `metadata.json` lives. Defaults to the env-paths cache directory
     // plus the user id, so each account has its own cache.
     cacheDirectory?: string;
-    // Persistent backup destination. The refresh loop does not use it; the
+    // The root of every photo's save path, where `Photo.download()` and
-    // content cache treats an original already stored there as present.
+    // `lib.backup()` put originals. Defaults to `photos` in the working
+    // directory at open. The content cache treats an original already stored
+    // at its save path as present.
     downloadDirectory?: string;
     refreshIntervalSeconds?: number;
     onProgress?: RefreshProgressCallback;
@@ -234,7 +240,7 @@ export interface LibraryStatus {
 
 export class Library {
     readonly cacheDirectory: string;
-    readonly downloadDirectory?: string;
+    readonly downloadDirectory: string;
 
     // The in-process read surface (issue #44). Each namespace answers
     // synchronously from the live record projection; no read touches the
@@ -296,7 +302,7 @@ export class Library {
         store: MetadataStore;
         userID: number;
         cacheDirectory: string;
-        downloadDirectory?: string;
+        downloadDirectory: string;
         intervalMs: number;
         onProgress?: RefreshProgressCallback;
         pools: RequestPools;
@@ -320,8 +326,14 @@ export class Library {
         // The read namespaces derive fresh from the store on each call, so they
         // always reflect the latest refresh.
         const derive = (): DerivedRecords => this.deriveNow();
-        this.albums = makeAlbumsAPI(derive, this.cache);
+        const root = this.downloadDirectory;
-        this.photos = makePhotosAPI(derive, this.cache);
+        const saves: SavePathLookup = {
+            savePath: (file) =>
+                storedAtSavePath(root, file)?.path ?? savePath(root, file),
+            isLocal: (file) => storedAtSavePath(root, file) !== undefined,
+        };
+        this.albums = makeAlbumsAPI(derive, saves, this.cache);
+        this.photos = makePhotosAPI(derive, saves, this.cache);
         this.timeline = makeTimelineAPI(derive);
         this.thumbnails = {
             ensure: (opts: EnsureOptions): Promise<EnsureResult[]> => {
@@ -350,6 +362,13 @@ export class Library {
         const { userID } = opts.client.whoami();
         const cacheDirectory =
             opts.cacheDirectory ?? defaultCacheDirectory(userID);
+        if (opts.downloadDirectory === "") {
+            throw new Error(
+                "library: downloadDirectory is empty (leave it out to save " +
+                    "under photos/ in the working directory)",
+            );
+        }
+        const downloadDirectory = opts.downloadDirectory ?? resolve("photos");
         const metadataPath = join(cacheDirectory, "metadata.json");
         let store = await MetadataStore.load(metadataPath);
         // A cache directory given explicitly can hold another account's cache.
@@ -404,7 +423,7 @@ export class Library {
                 pools,
                 source,
                 cacheDirectory,
-                downloadDirectory: opts.downloadDirectory,
+                downloadDirectory,
                 getFile: (fileID) => store.getFileByID(fileID),
                 cacheOriginalsMaxBytes: opts.cacheOriginalsMaxBytes,
                 freeBelowBytes: opts.freeBelowBytes,
@@ -426,7 +445,7 @@ export class Library {
             store,
             userID,
             cacheDirectory,
-            downloadDirectory: opts.downloadDirectory,
+            downloadDirectory,
             intervalMs,
             onProgress: opts.onProgress,
             pools,
@@ -470,9 +489,10 @@ export class Library {
         return this.store.getFile(collectionID, fileID);
     }
 
-    // Any membership of a file, addressed by file id alone. A file's own
+    // The membership of a file its record is read from, addressed by file id
-    // metadata (title, creationTime) is identical across the collections it
+    // alone. A file's own metadata (title, creationTime) is identical across
-    // belongs to, so this serves the point commands that hold only a fileID.
+    // the collections it belongs to, so this serves the point commands that
+    // hold only a fileID.
     getFileByID(fileID: number): EnteFile | undefined {
         return this.store.getFileByID(fileID);
     }
@@ -543,27 +563,20 @@ export class Library {
         };
     }
 
-    // Back up every in-scope file to `downloadDirectory` in the historical
+    // Back up every in-scope file to `opts.downloadDirectory`, or else the
-    // on-disk layout, with a durable failure ledger (issue #51). Waits for a
+    // library's, each original at its save path, with a durable failure
-    // completed refresh first, as `fresh()` does, joining one already running,
+    // ledger (issue #51). Waits for a completed refresh first, as `fresh()`
-    // and rejects before touching any file when it fails. Then fetches pending
+    // does, joining one already running, and rejects before touching any file
-    // originals (and optional thumbnails) through the content cache and pools,
+    // when it fails. Then puts pending originals at their save paths as
-    // and rebuilds the derived symlink/JSON views from the model. Throws before
+    // `Photo.download()` does (and optional thumbnails) through the content
-    // any network work when no download directory is available or no content
+    // cache and pools, and rebuilds the derived symlink/JSON views from the
-    // cache backs the originals it must fetch.
+    // model. Throws before any network work when no content cache backs the
+    // originals it must fetch.
     backup(opts?: BackupOptions): Promise<BackupResult> {
         const downloadDirectory =
             opts?.downloadDirectory ?? this.downloadDirectory;
         const includeOriginals = opts?.includeOriginals ?? true;
         const includeThumbnails = opts?.includeThumbnails ?? false;
-        if (!downloadDirectory) {
-            return Promise.reject(
-                new Error(
-                    "backup requires a downloadDirectory (pass one to " +
-                        "backup() or open the library with one)",
-                ),
-            );
-        }
         if ((includeOriginals || includeThumbnails) && !this.cache) {
             return Promise.reject(
                 new Error(

src/library/read.ts

@@ -11,15 +11,27 @@
 // access and, for an album, its photos. They are not sent across IPC — the
 // plain records are the serializable surface, and `record()` returns one.
 //
-// A `Photo` also fetches its own bytes: `original()` and `thumbnail()` go
+// A `Photo` also fetches its own bytes: `original()`, `thumbnail()`,
-// through the on-disk content cache (issue #46), the one place in this module
+// `download()`, `content()`, `exif()` and the methods that each return one
-// that is not synchronous and RAM-only. A library opened without a content
+// field of `exif()` go through the on-disk content cache (issue #46), and are
-// source leaves that cache absent, and those two methods then throw.
+// the one place in this module that may touch the network. A library opened
+// without a content source leaves that cache absent, and those methods then
+// throw. `savePath` and `isLocal` look only at the disk and need no cache.
 
-import type { CollectionType, FileType } from "../model/types.js";
+import { readFile } from "node:fs/promises";
+
+import { readPhotoExif, type PhotoExif } from "../exif.js";
+import type { CollectionType, EnteFile, FileType } from "../model/types.js";
 import type { ContentOptions, ContentResult, PhotoContent } from "./content.js";
 import type { AlbumRecord, PhotoRecord, DerivedRecords } from "./records.js";
 
+// Where a photo's original is saved, and whether all of it is there. The
+// library answers both from the disk, with or without a content cache.
+export interface SavePathLookup {
+    savePath(file: EnteFile): string;
+    isLocal(file: EnteFile): boolean;
+}
+
 // Newest first, with fileID as a stable tiebreak so equal-timed files order
 // deterministically — the same order the record projection uses.
 const byNewest = (a: PhotoRecord, b: PhotoRecord): number =>
@@ -30,12 +42,23 @@ const byNewest = (a: PhotoRecord, b: PhotoRecord): number =>
 const byNewestAlbum = (a: AlbumRecord, b: AlbumRecord): number =>
     b.updationTime - a.updationTime || b.collectionID - a.collectionID;
 
+// A method for each `PhotoExif` field, named after it, taking the options
+// `exif()` takes and giving that field. `Photo` implements it, so the build's
+// type check fails when `PhotoExif` has a field `Photo` has no method for.
+type PhotoExifMethods = {
+    [K in keyof PhotoExif]-?: (opts?: ContentOptions) => Promise<PhotoExif[K]>;
+};
+
 // A single photo. Field access mirrors `PhotoRecord`; `record()` returns the
-// underlying plain record for callers that need the IPC-safe value.
+// underlying plain record for callers that need the IPC-safe value. `file` is
-export class Photo {
+// the membership the record is read from, so the save path carries the date of
+// `takenAt` and stays known after a refresh removes the file from the library.
+export class Photo implements PhotoExifMethods {
     constructor(
         private readonly rec: PhotoRecord,
-        private readonly content?: PhotoContent,
+        private readonly file: EnteFile,
+        private readonly saves: SavePathLookup,
+        private readonly cache?: PhotoContent,
     ) {}
 
     get fileID(): number {
@@ -50,6 +73,13 @@ export class Photo {
     get takenAt(): number {
         return this.rec.takenAt;
     }
+    get modifiedAt(): number {
+        return this.rec.modifiedAt;
+    }
+    // The local-time year of `takenAt`.
+    get year(): number {
+        return new Date(this.rec.takenAt).getFullYear();
+    }
     get fileType(): FileType {
         return this.rec.fileType;
     }
@@ -68,6 +98,9 @@ export class Photo {
     get longitude(): number | undefined {
         return this.rec.longitude;
     }
+    get hash(): string | undefined {
+        return this.rec.hash;
+    }
     get isArchived(): boolean {
         return this.rec.isArchived;
     }
@@ -75,30 +108,131 @@ export class Photo {
         return this.rec.isHidden;
     }
 
+    // Where `download()` and `lib.backup()` put the original under the
+    // library's download directory, whether or not it is there yet:
+    // `YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID><ext>`. For a live photo
+    // already stored, its image. For a live photo not yet stored, it carries
+    // the title's extension, and the image may be stored under a different
+    // one.
+    get savePath(): string {
+        return this.saves.savePath(this.file);
+    }
+
+    // Whether the whole original is at `savePath`. A copy only in the cache
+    // does not count.
+    get isLocal(): boolean {
+        return this.saves.isLocal(this.file);
+    }
+
     record(): PhotoRecord {
         return this.rec;
     }
 
     // Fetch and cache the full-resolution original, returning its on-disk path
     // 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)
+    // `videoPath`. Served from the cache (or the save path) when already
-    // when already present, otherwise fetched through the content pool.
+    // present, otherwise fetched through the content pool.
     async original(opts?: ContentOptions): Promise<ContentResult> {
-        return this.contentOrThrow().original(this.rec.fileID, opts);
+        return this.cacheOrThrow().original(this.rec.fileID, opts);
+    }
+
+    // Put the original at `savePath` and return it there, as `original()`
+    // does. When it is already there, nothing is written. When the cache holds
+    // it, it is copied from there; otherwise it is fetched straight to
+    // `savePath`.
+    async download(): Promise<ContentResult> {
+        return this.cacheOrThrow().download(this.file);
     }
 
     // As `original`, for the thumbnail, through the thumbnail pool.
     async thumbnail(opts?: ContentOptions): Promise<ContentResult> {
-        return this.contentOrThrow().thumbnail(this.rec.fileID, opts);
+        return this.cacheOrThrow().thumbnail(this.rec.fileID, opts);
     }
 
-    private contentOrThrow(): PhotoContent {
+    // The original's bytes, read from where `original()` puts it. For a live
-        if (!this.content) {
+    // photo, its image's.
+    async content(opts?: ContentOptions): Promise<Uint8Array> {
+        const { path } = await this.original(opts);
+        return readFile(path);
+    }
+
+    // The common EXIF fields of the original, read from `content()`, so this
+    // may download it. EXIF is read from any image format exifreader reads,
+    // JPEG and HEIC/HEIF among them; any other file gives `{}`, and a video
+    // gives it without fetching anything. Like the other content methods, it
+    // throws when there is no content cache, video or not.
+    async exif(opts?: ContentOptions): Promise<PhotoExif> {
+        this.cacheOrThrow();
+        if (this.rec.fileType === "video") return {};
+        return readPhotoExif(await this.content(opts));
+    }
+
+    // One field of `exif()` each, named and typed as in `PhotoExif`, and
+    // undefined when the file lacks it. Each call runs `exif()`, which reads
+    // the original again.
+    async make(opts?: ContentOptions): Promise<PhotoExif["make"]> {
+        return (await this.exif(opts)).make;
+    }
+    async model(opts?: ContentOptions): Promise<PhotoExif["model"]> {
+        return (await this.exif(opts)).model;
+    }
+    async lensModel(opts?: ContentOptions): Promise<PhotoExif["lensModel"]> {
+        return (await this.exif(opts)).lensModel;
+    }
+    async dateTimeOriginal(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["dateTimeOriginal"]> {
+        return (await this.exif(opts)).dateTimeOriginal;
+    }
+    async offsetTimeOriginal(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["offsetTimeOriginal"]> {
+        return (await this.exif(opts)).offsetTimeOriginal;
+    }
+    async exposureTime(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["exposureTime"]> {
+        return (await this.exif(opts)).exposureTime;
+    }
+    async fNumber(opts?: ContentOptions): Promise<PhotoExif["fNumber"]> {
+        return (await this.exif(opts)).fNumber;
+    }
+    async iso(opts?: ContentOptions): Promise<PhotoExif["iso"]> {
+        return (await this.exif(opts)).iso;
+    }
+    async focalLength(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["focalLength"]> {
+        return (await this.exif(opts)).focalLength;
+    }
+    async orientation(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["orientation"]> {
+        return (await this.exif(opts)).orientation;
+    }
+    async gpsLatitude(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["gpsLatitude"]> {
+        return (await this.exif(opts)).gpsLatitude;
+    }
+    async gpsLongitude(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["gpsLongitude"]> {
+        return (await this.exif(opts)).gpsLongitude;
+    }
+    async gpsAltitude(
+        opts?: ContentOptions,
+    ): Promise<PhotoExif["gpsAltitude"]> {
+        return (await this.exif(opts)).gpsAltitude;
+    }
+
+    private cacheOrThrow(): PhotoContent {
+        if (!this.cache) {
             throw new Error(
                 "Photo content requires a library opened with a content cache",
             );
         }
-        return this.content;
+        return this.cache;
     }
 }
 
@@ -108,6 +242,7 @@ export class Album {
     constructor(
         private readonly rec: AlbumRecord,
         private readonly records: DerivedRecords,
+        private readonly saves: SavePathLookup,
         private readonly content?: PhotoContent,
     ) {}
 
@@ -142,7 +277,10 @@ export class Album {
         const out: Photo[] = [];
         for (const id of this.rec.fileIDs) {
             const p = this.records.photos.get(id);
-            if (p) out.push(new Photo(p, this.content));
+            const file = this.records.files.get(id);
+            if (p && file) {
+                out.push(new Photo(p, file, this.saves, this.content));
+            }
         }
         return out;
     }
@@ -205,18 +343,19 @@ export interface FreshReads {
 
 export const makeAlbumsAPI = (
     derive: () => DerivedRecords,
+    saves: SavePathLookup,
     content?: PhotoContent,
 ): AlbumsAPI => ({
     list: (): Album[] => {
         const records = derive();
         return [...records.albums.values()]
             .sort(byNewestAlbum)
-            .map((rec) => new Album(rec, records, content));
+            .map((rec) => new Album(rec, records, saves, content));
     },
     byID: ({ collectionID }): Album | undefined => {
         const records = derive();
         const rec = records.albums.get(collectionID);
-        return rec ? new Album(rec, records, content) : undefined;
+        return rec ? new Album(rec, records, saves, content) : undefined;
     },
     byName: ({ albumName }): Album | undefined => {
         const records = derive();
@@ -225,17 +364,20 @@ export const makeAlbumsAPI = (
         const match = [...records.albums.values()]
             .sort(byNewestAlbum)
             .find((rec) => rec.name === albumName);
-        return match ? new Album(match, records, content) : undefined;
+        return match ? new Album(match, records, saves, content) : undefined;
     },
 });
 
 export const makePhotosAPI = (
     derive: () => DerivedRecords,
+    saves: SavePathLookup,
     content?: PhotoContent,
 ): PhotosAPI => ({
     byID: ({ fileID }): Photo | undefined => {
-        const rec = derive().photos.get(fileID);
+        const records = derive();
-        return rec ? new Photo(rec, content) : undefined;
+        const rec = records.photos.get(fileID);
+        const file = records.files.get(fileID);
+        return rec && file ? new Photo(rec, file, saves, content) : undefined;
     },
     records: ({ fileIDs }): PhotoRecord[] => {
         const { photos } = derive();

src/library/records.ts

@@ -5,10 +5,11 @@
 // owner ruling 5). The decrypted `Collection`/`EnteFile` objects stay in RAM in
 // the main process; the window only ever sees these records.
 //
-// Ente holds edited/basic times in microseconds; records expose `takenAt` in
+// Ente holds edited/basic times in microseconds; records expose `takenAt` and
-// milliseconds. The magic-metadata field names below are the ones the Ente
+// `modifiedAt` in milliseconds. The magic-metadata field names below are the
-// clients write, confirmed against the repo's own fixtures: `w`/`h` in
+// ones the Ente clients write, confirmed against the repo's own fixtures:
-// test/cli/metadata-backup.test.ts, `visibility` in test/library/store.test.ts.
+// `w`/`h` in test/cli/metadata-backup.test.ts, `visibility` in
+// test/library/store.test.ts.
 
 import type {
     Collection,
@@ -33,12 +34,17 @@ export interface PhotoRecord {
     // Milliseconds. `pubMagicMetadata.editedTime` when the user edited the
     // date, else basic-metadata `creationTime`.
     takenAt: number;
+    // Milliseconds. Basic-metadata `modificationTime`.
+    modifiedAt: number;
     fileType: FileType;
     caption?: string;
     width?: number;
     height?: number;
     latitude?: number;
     longitude?: number;
+    // The content hash the uploader recorded (`FileMetadata.hash`); files from
+    // very old clients have none.
+    hash?: string;
     isArchived: boolean;
     isHidden: boolean;
     // Local cache paths, set once a later phase caches the bytes; unset here.
@@ -80,6 +86,10 @@ export interface LibraryChange {
 export interface DerivedRecords {
     albums: Map<number, AlbumRecord>;
     photos: Map<number, PhotoRecord>;
+    // The membership each photo's record is read from, for its `Photo`'s save
+    // path. It holds the file's key, so it stays in this process: no snapshot
+    // or change carries it.
+    files: Map<number, EnteFile>;
 }
 
 const asString = (v: unknown): string | undefined =>
@@ -95,18 +105,19 @@ const microsToMillis = (micros: number): number => Math.floor(micros / 1000);
 const byNewestPhoto = (a: PhotoRecord, b: PhotoRecord): number =>
     b.takenAt - a.takenAt || b.fileID - a.fileID;
 
-// Build one PhotoRecord from every membership of a file. The memberships share
+// A photo's `takenAt` in milliseconds: `pubMagicMetadata.editedTime` when the
-// the same underlying file, so metadata is read from a single representative
+// user edited the date, else basic-metadata `creationTime`.
-// (the most recently synced, lowest collection id to break ties); `albumIDs`
+export const takenAtOf = (file: EnteFile): number =>
-// gathers them all.
+    microsToMillis(
-const toPhotoRecord = (
+        asNumber(file.pubMagicMetadata?.editedTime) ??
-    fileID: number,
+            file.metadata.creationTime,
-    memberships: EnteFile[],
+    );
-): PhotoRecord => {
+
-    const albumIDs = memberships
+// The membership a file's record is read from: the most recently synced, lowest
-        .map((m) => m.collectionID)
+// collection id to break ties. Whatever dates a file's save path takes this
-        .sort((a, b) => a - b);
+// membership too, so the path always carries the record's `takenAt`.
-    const rep = memberships.reduce((best, m) =>
+export const representative = (memberships: EnteFile[]): EnteFile =>
+    memberships.reduce((best, m) =>
         m.updationTime > best.updationTime ||
         (m.updationTime === best.updationTime &&
             m.collectionID < best.collectionID)
@@ -114,17 +125,29 @@ const toPhotoRecord = (
             : best,
     );
 
+// Build one PhotoRecord from every membership of a file. The memberships share
+// the same underlying file, so metadata is read from a single representative;
+// `albumIDs` gathers them all.
+const toPhotoRecord = (
+    fileID: number,
+    memberships: EnteFile[],
+): PhotoRecord => {
+    const albumIDs = memberships
+        .map((m) => m.collectionID)
+        .sort((a, b) => a - b);
+    const rep = representative(memberships);
+
     const pub = rep.pubMagicMetadata ?? {};
     const priv = rep.magicMetadata ?? {};
 
-    const takenAtMicros = asNumber(pub.editedTime) ?? rep.metadata.creationTime;
     const visibility = asNumber(priv.visibility);
 
     const record: PhotoRecord = {
         fileID,
         albumIDs,
         title: asString(pub.editedName) ?? rep.metadata.title,
-        takenAt: microsToMillis(takenAtMicros),
+        takenAt: takenAtOf(rep),
+        modifiedAt: microsToMillis(rep.metadata.modificationTime),
         fileType: rep.metadata.fileType,
         isArchived: visibility === VISIBILITY_ARCHIVED,
         isHidden: visibility === VISIBILITY_HIDDEN,
@@ -140,6 +163,7 @@ const toPhotoRecord = (
         record.latitude = rep.metadata.latitude;
     if (rep.metadata.longitude !== undefined)
         record.longitude = rep.metadata.longitude;
+    if (rep.metadata.hash !== undefined) record.hash = rep.metadata.hash;
 
     return record;
 };
@@ -190,6 +214,7 @@ export const deriveRecords = (
     }
 
     const photos = new Map<number, PhotoRecord>();
+    const photoFiles = new Map<number, EnteFile>();
     const takenAtByFile = new Map<number, number>();
     for (const [fileID, memberships] of byFileID) {
         const record = toPhotoRecord(fileID, memberships);
@@ -201,6 +226,7 @@ export const deriveRecords = (
                 record.thumbnailPath = paths.thumbnailPath;
         }
         photos.set(fileID, record);
+        photoFiles.set(fileID, representative(memberships));
         takenAtByFile.set(fileID, record.takenAt);
     }
 
@@ -209,7 +235,7 @@ export const deriveRecords = (
         albums.set(c.id, toAlbumRecord(c, files, takenAtByFile));
     }
 
-    return { albums, photos };
+    return { albums, photos, files: photoFiles };
 };
 
 // Sorted, GUI-ready arrays: albums newest updated first, photos newest first.

src/library/store.ts

@@ -16,6 +16,7 @@ import { dirname } from "node:path";
 
 import { writeAtomic } from "../download/index.js";
 import type { Collection, EnteFile, Microseconds } from "../model/types.js";
+import { representative } from "./records.js";
 
 // Bumped only when the on-disk shape changes incompatibly. A file written
 // under a different version is discarded on load (see `load`): re-fetching
@@ -179,14 +180,16 @@ export class MetadataStore {
         return this.files.get(fileKey(collectionID, fileID));
     }
 
-    // Any membership of a file, or undefined. Every membership re-wraps the
+    // The membership of a file its record is read from (`representative`), or
-    // same underlying content key, so any one is enough to fetch the bytes;
+    // undefined. Any membership could fetch the bytes, but the content cache
-    // the content cache resolves a fileID to a file this way.
+    // resolves a fileID to a file this way so that it dates the save path
+    // from the same membership as `photo.savePath`.
     getFileByID(fileID: number): EnteFile | undefined {
+        const memberships: EnteFile[] = [];
         for (const file of this.files.values()) {
-            if (file.id === fileID) return file;
+            if (file.id === fileID) memberships.push(file);
         }
-        return undefined;
+        return memberships.length > 0 ? representative(memberships) : undefined;
     }
 
     listFiles(collectionID: number): EnteFile[] {

src/metadata-backup.ts

@@ -1,8 +1,8 @@
 import { mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { join } from "node:path";
 import * as jpeg from "jpeg-js";
-import exifReader from "exif-reader";
 import type { Client } from "./client.js";
+import { readExifTags } from "./exif.js";
 import type { Library, Photo } from "./library/index.js";
 import { sanitizeFileName } from "./filename.js";
 import {
@@ -19,63 +19,10 @@ export interface MetadataBackupOptions {
     onProgress?: ProgressCallback;
 }
 
-// Find the raw EXIF APP1 segment in JPEG bytes. Returns `exif` (the segment
+// Extract dimensions, EXIF and XMP from a file's bytes. `exif` is the EXIF tags
-// data, starting at the "Exif\0\0" header) when there is one, nothing when the
+// exifreader returns, from any image format it reads. When it finds an EXIF
-// bytes are not a JPEG or carry no EXIF, and `error` when the segment layout is
+// block but reads no tag from it, the record keeps the block's bytes, base64,
-// malformed. Each segment length is checked against the bytes that remain and
+// in `exifRaw`, with the reason in `exifError`.
-// each step moves forward by at least 4 bytes, so the scan ends on any input.
-export const extractExifFromJpeg = (
-    buf: Uint8Array,
-): { exif?: Buffer; error?: string } => {
-    if (buf[0] !== 0xff || buf[1] !== 0xd8) return {};
-    let offset = 2;
-    while (offset < buf.length) {
-        if (offset + 2 > buf.length)
-            return { error: `truncated segment marker at byte ${offset}` };
-        if (buf[offset] !== 0xff)
-            return { error: `no segment marker at byte ${offset}` };
-        const marker = buf[offset + 1]!;
-        if (marker === 0xda) return {}; // start of scan, no more markers
-        if (offset + 4 > buf.length)
-            return { error: `truncated segment length at byte ${offset}` };
-        const len = (buf[offset + 2]! << 8) | buf[offset + 3]!;
-        // The length counts its own two bytes, so anything under 2 is invalid.
-        if (len < 2)
-            return {
-                error: `segment length ${len} at byte ${offset} is too small`,
-            };
-        if (offset + 2 + len > buf.length)
-            return {
-                error: `segment length ${len} at byte ${offset} runs past the end of the file`,
-            };
-        if (marker === 0xe1) {
-            // APP1 — check for "Exif\0\0" header. A length under 8 cannot hold
-            // the six-byte header, so the segment is not EXIF; below 6 the
-            // bytes compared would also lie past the segment.
-            if (
-                len >= 8 &&
-                buf[offset + 4] === 0x45 &&
-                buf[offset + 5] === 0x78 &&
-                buf[offset + 6] === 0x69 &&
-                buf[offset + 7] === 0x66
-            ) {
-                return {
-                    exif: Buffer.from(
-                        buf.buffer,
-                        buf.byteOffset + offset + 4,
-                        len - 2,
-                    ),
-                };
-            }
-        }
-        offset += 2 + len;
-    }
-    return { error: "file ends before the image data" };
-};
-
-// Extract dimensions, EXIF and XMP from a file's bytes. When the EXIF segment
-// is malformed or cannot be parsed, the record carries the reason in
-// `exifError`.
 export const extractImageMetadata = (
     fileBytes: Uint8Array,
 ): Record<string, unknown> | undefined => {
@@ -92,19 +39,21 @@ export const extractImageMetadata = (
         result.height = decoded.height;
     } catch {
         // Not every original is a JPEG (PNG, HEIC, video), so a failed decode
-        // is expected and only means no dimensions; a malformed JPEG is still
+        // is expected and only means no dimensions; unreadable EXIF is still
         // reported below through `exifError`.
     }
 
-    const { exif, error } = extractExifFromJpeg(fileBytes);
+    const tags = readExifTags(fileBytes);
-    if (error) result.exifError = error;
+    if (tags?.exif && Object.keys(tags.exif).length > 0) {
-    if (exif) {
+        result.exif = tags.exif;
-        try {
+    } else if (tags?.exif) {
-            result.exif = exifReader(exif);
+        const block = tags.metadataRange?.blocks.find((b) => b.type === "exif");
-        } catch (err) {
+        if (block) {
-            result.exifRaw = exif.toString("base64");
+            result.exifRaw = Buffer.from(
-            result.exifError = err instanceof Error ? err.message : String(err);
+                fileBytes.subarray(block.start, block.end),
+            ).toString("base64");
         }
+        result.exifError = "no tag could be read from the EXIF block";
     }
 
     // Extract XMP (look for "http://ns.adobe.com/xap" in the bytes)

test/cli/backup.test.ts

@@ -1,16 +1,16 @@
 /**
  * Tests for the `quak backup` logic, now built on the library API (issue #51).
  *
- * `lib.backup({ downloadDirectory })` refreshes the library, fetches each
+ * `lib.backup({ downloadDirectory })` refreshes the library, puts each pending
- * pending file's original through the content cache/pools, and materialises the
+ * file's original at its save path through the content cache/pools, and
- * unchanged on-disk layout:
+ * materialises the on-disk layout:
  *
  *     <downloadDirectory>/
- *         originals/
+ *         YYYY/YYYY-MM/YYYY-MM-DD/
- *             <fileID>.<ext>          the decrypted bytes ("present means complete")
+ *             YYYY-MM-DD.<fileID>.<ext>   the decrypted bytes ("present means complete")
- *             <fileID>.json           per-file metadata sidecar (rebuilt each run)
+ *             YYYY-MM-DD.<fileID>.json    per-file metadata sidecar (rebuilt each run)
  *         collections/
- *             <name>/<title>          symlink into ../originals (rebuilt each run)
+ *             <name>/<title>          symlink to the original (rebuilt each run)
  *             <name>.json             per-collection metadata (rebuilt each run)
  *         failures.json               durable ledger of unresolved failures
  *
@@ -94,6 +94,18 @@ const USER_ID = 42;
 // Decrypted-byte length each stub original writes, keyed by fileID.
 const SIZE_BY_ID: Record<number, number> = { 100: 3000, 101: 2000, 200: 1500 };
 
+// Every file is taken at noon local time on 2026-03-01, in microseconds as Ente
+// stores times, so the machine's time zone cannot move it to another day. Its
+// original is saved in the folder `DAY`, named `2026-03-01.<fileID>.<ext>`.
+const TAKEN = new Date(2026, 2, 1, 12).getTime() * 1000;
+const DAY = join("2026", "2026-03", "2026-03-01");
+
+// Where the backup in `outDir` saves `name` (`<fileID>.<ext>`), and the target
+// of an album folder's symlink to it.
+const saved = (outDir: string, name: string): string =>
+    join(outDir, DAY, `2026-03-01.${name}`);
+const linkTo = (name: string): string => `../../${DAY}/2026-03-01.${name}`;
+
 const collection = (id: number, name: string): Collection => ({
     id,
     ownerID: USER_ID,
@@ -112,7 +124,7 @@ const file = (id: number, collectionID: number, title: string): EnteFile => ({
     metadata: {
         title,
         fileType: "image",
-        creationTime: 1,
+        creationTime: TAKEN,
         modificationTime: 1,
     },
     file: { decryptionHeader: "aGVhZGVy" },
@@ -253,7 +265,11 @@ describe("lib.backup", () => {
     it("throws before any network when no downloadDirectory is given", async () => {
         const source = stubSource();
         const lib = await openLibrary(source);
-        await expect(lib.backup()).rejects.toThrow(/downloadDirectory/i);
+        // The library always has a download directory, so only an empty one
+        // given to backup() reaches runBackup as none.
+        await expect(lib.backup({ downloadDirectory: "" })).rejects.toThrow(
+            /downloadDirectory/i,
+        );
         expect(source.originalCalls).toBe(0);
         lib.close();
     });
@@ -271,28 +287,22 @@ describe("lib.backup", () => {
         expect(result.failed).toBe(0);
         expect(result.errors).toEqual([]);
 
-        // Originals under originals/<fileID>.<ext>.
+        // Originals at YYYY/YYYY-MM/YYYY-MM-DD/YYYY-MM-DD.<fileID>.<ext>.
-        expect(readFileSync(join(outDir, "originals", "100.jpg")).length).toBe(
+        expect(readFileSync(saved(outDir, "100.jpg")).length).toBe(3000);
-            3000,
+        expect(readFileSync(saved(outDir, "101.jpg")).length).toBe(2000);
-        );
+        expect(readFileSync(saved(outDir, "200.png")).length).toBe(1500);
-        expect(readFileSync(join(outDir, "originals", "101.jpg")).length).toBe(
-            2000,
-        );
-        expect(readFileSync(join(outDir, "originals", "200.png")).length).toBe(
-            1500,
-        );
 
         // Per-file metadata sidecar.
         const sidecar = JSON.parse(
-            readFileSync(join(outDir, "originals", "100.json"), "utf-8"),
+            readFileSync(saved(outDir, "100.json"), "utf-8"),
         );
         expect(sidecar.id).toBe(100);
         expect(sidecar.metadata.title).toBe("beach.jpg");
 
-        // Collection dirs contain symlinks into ../originals.
+        // Collection dirs contain symlinks into the date folders.
         const beach = join(outDir, "collections", "Vacation", "beach.jpg");
         expect(lstatSync(beach).isSymbolicLink()).toBe(true);
-        expect(readlinkSync(beach)).toContain("originals");
+        expect(readlinkSync(beach)).toContain(DAY);
         expect(readFileSync(beach).length).toBe(3000);
 
         // Per-collection metadata JSON.
@@ -316,7 +326,7 @@ describe("lib.backup", () => {
 
         expect(result.failed).toBe(0);
         // The title has no usable extension, so the original is `.bin`.
-        expect(existsSync(join(outDir, "originals", "300.bin"))).toBe(true);
+        expect(existsSync(saved(outDir, "300.bin"))).toBe(true);
         const link = join(
             outDir,
             "collections",
@@ -366,9 +376,9 @@ describe("lib.backup", () => {
         expect(result.errors[0]!.title).toBe("sunset.jpg");
 
         // The two good files are on disk; the failed one is not.
-        expect(existsSync(join(outDir, "originals", "100.jpg"))).toBe(true);
+        expect(existsSync(saved(outDir, "100.jpg"))).toBe(true);
-        expect(existsSync(join(outDir, "originals", "200.png"))).toBe(true);
+        expect(existsSync(saved(outDir, "200.png"))).toBe(true);
-        expect(existsSync(join(outDir, "originals", "101.jpg"))).toBe(false);
+        expect(existsSync(saved(outDir, "101.jpg"))).toBe(false);
         expect(
             existsSync(join(outDir, "collections", "Vacation", "sunset.jpg")),
         ).toBe(false);
@@ -403,7 +413,7 @@ describe("lib.backup", () => {
         const r3 = await lib.backup({ downloadDirectory: outDir });
         expect(r3.failed).toBe(0);
         expect(r3.skipped).toBe(2);
-        expect(existsSync(join(outDir, "originals", "101.jpg"))).toBe(true);
+        expect(existsSync(saved(outDir, "101.jpg"))).toBe(true);
         // A ledger with no remaining failures is removed.
         expect(existsSync(join(outDir, "failures.json"))).toBe(false);
         lib.close();
@@ -423,8 +433,8 @@ describe("lib.backup", () => {
         const result = await lib.backup({ downloadDirectory: outDir });
 
         // Every original still downloads despite the symlink failure.
-        expect(existsSync(join(outDir, "originals", "100.jpg"))).toBe(true);
+        expect(existsSync(saved(outDir, "100.jpg"))).toBe(true);
-        expect(existsSync(join(outDir, "originals", "200.png"))).toBe(true);
+        expect(existsSync(saved(outDir, "200.png"))).toBe(true);
         // The other symlinks are still built.
         expect(
             lstatSync(
@@ -454,7 +464,7 @@ describe("lib.backup", () => {
         await lib.backup({ downloadDirectory: outDir });
 
         // Corrupt a sidecar and delete a symlink between runs.
-        writeFileSync(join(outDir, "originals", "100.json"), "not json");
+        writeFileSync(saved(outDir, "100.json"), "not json");
         rmSync(join(outDir, "collections", "Vacation", "beach.jpg"));
 
         const result = await lib.backup({ downloadDirectory: outDir });
@@ -462,7 +472,7 @@ describe("lib.backup", () => {
 
         // The derived views are repaired from the model.
         const sidecar = JSON.parse(
-            readFileSync(join(outDir, "originals", "100.json"), "utf-8"),
+            readFileSync(saved(outDir, "100.json"), "utf-8"),
         );
         expect(sidecar.metadata.title).toBe("beach.jpg");
         expect(
@@ -485,8 +495,8 @@ describe("lib.backup", () => {
 
         expect(result.totalFiles).toBe(1);
         expect(result.downloaded).toBe(1);
-        expect(existsSync(join(outDir, "originals", "200.png"))).toBe(true);
+        expect(existsSync(saved(outDir, "200.png"))).toBe(true);
-        expect(existsSync(join(outDir, "originals", "100.jpg"))).toBe(false);
+        expect(existsSync(saved(outDir, "100.jpg"))).toBe(false);
         expect(existsSync(join(outDir, "collections", "Work.json"))).toBe(true);
         expect(existsSync(join(outDir, "collections", "Vacation.json"))).toBe(
             false,
@@ -530,7 +540,7 @@ describe("lib.backup", () => {
 
         expect(result.totalFiles).toBe(1);
         expect(result.failed).toBe(0);
-        expect(existsSync(join(outDir, "originals", "200.png"))).toBe(true);
+        expect(existsSync(saved(outDir, "200.png"))).toBe(true);
         expect(existsSync(join(outDir, "failures.json"))).toBe(false);
         lib.close();
     });
@@ -572,7 +582,7 @@ describe("lib.backup", () => {
     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.
+        // to its save path, and the cache records it there.
         const source = stubSource();
         const lib = await openLibrary(source);
         const outDir = join(root, "backup");
@@ -582,23 +592,78 @@ describe("lib.backup", () => {
         expect(result.downloaded).toBe(3);
         expect(source.originalCalls).toBe(3);
         expect(readdirSync(join(root, "cache", "originals"))).toEqual([]);
-        const stored = readdirSync(join(outDir, "originals")).filter(
+        const stored = readdirSync(join(outDir, DAY)).filter(
             (name) => !name.endsWith(".json"),
         );
-        expect(stored.sort()).toEqual(["100.jpg", "101.jpg", "200.png"]);
+        expect(stored.sort()).toEqual([
+            "2026-03-01.100.jpg",
+            "2026-03-01.101.jpg",
+            "2026-03-01.200.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(read.path).toBe(saved(outDir, "100.jpg"));
         expect(source.originalCalls).toBe(3);
         await lib.close();
     });
 
+    it("saves a file in two albums at its photo's save path and links it from both", async () => {
+        // The date edited in Ente has reached Work's copy, synced later, but
+        // Vacation's copy still has an earlier edit.
+        const older = file(100, 1, "beach.jpg");
+        older.pubMagicMetadata = {
+            editedTime: new Date(2026, 1, 1, 12).getTime() * 1000,
+        };
+        const newer = file(100, 2, "beach.jpg");
+        newer.updationTime = 2;
+        newer.pubMagicMetadata = {
+            editedTime: new Date(2026, 3, 15, 12).getTime() * 1000,
+        };
+        class SharedFileClient extends MockClient {
+            override async filesSince(args: {
+                collectionID: number;
+            }): Promise<FilesPage> {
+                const files = [args.collectionID === 1 ? older : newer];
+                return { files, deleted: [], cursor: 1 };
+            }
+        }
+        const outDir = join(root, "backup");
+        const lib = await Library.open({
+            client: new SharedFileClient(),
+            cacheDirectory: join(root, "cache"),
+            downloadDirectory: outDir,
+            contentSource: stubSource(),
+            refreshIntervalSeconds: 3600,
+            precacheThumbnails: false,
+            precacheOriginals: false,
+        });
+        const photo = lib.photos.byID({ fileID: 100 })!;
+
+        const result = await lib.backup();
+
+        expect(result.totalFiles).toBe(1);
+        expect(result.downloaded).toBe(1);
+        expect(result.failed).toBe(0);
+        expect(photo.savePath).toBe(
+            join(outDir, "2026", "2026-04", "2026-04-15", "2026-04-15.100.jpg"),
+        );
+        expect(photo.isLocal).toBe(true);
+        expect(existsSync(join(outDir, "2026", "2026-02"))).toBe(false);
+        const target = "../../2026/2026-04/2026-04-15/2026-04-15.100.jpg";
+        for (const album of ["Vacation", "Work"]) {
+            expect(
+                readlinkSync(join(outDir, "collections", album, "beach.jpg")),
+            ).toBe(target);
+        }
+        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 day = join(outDir, DAY);
-        const dest = join(originals, "100.jpg");
+        const dest = saved(outDir, "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();
@@ -609,34 +674,50 @@ describe("lib.backup", () => {
         const at = fsEvents.indexOf(`rename:${dest}`);
         expect(at).toBeGreaterThan(0);
         expect(fsEvents[at - 1]).toMatch(
-            /^sync:.*\/\.quak-backup-100\.jpg-\d+-[0-9a-z]*\.tmp$/,
+            /^sync:.*\/\.quak-backup-2026-03-01\.100\.jpg-\d+-[0-9a-z]*\.tmp$/,
         );
-        expect(fsEvents[at + 1]).toBe(`sync:${originals}`);
+        expect(fsEvents[at + 1]).toBe(`sync:${day}`);
         lib.close();
     });
 
     it("removes temp files left by a killed backup but not those of one still running", async () => {
         const outDir = join(root, "backup");
-        const originals = join(outDir, "originals");
+        const day = join(outDir, DAY);
-        mkdirSync(originals, { recursive: true });
+        mkdirSync(day, { recursive: true });
         // A child that has already exited: its process ID is not running.
         const exitedPID = spawnSync(process.execPath, ["-e", ""]).pid;
-        const leftover = `.quak-backup-100.jpg-${exitedPID}-abc123.tmp`;
+        const leftover = `.quak-backup-2026-03-01.100.jpg-${exitedPID}-abc123.tmp`;
         // This test's own process stands in for a backup running at the same
         // time.
-        const inProgress = `.quak-backup-101.jpg-${process.pid}-def456.tmp`;
+        const inProgress = `.quak-backup-2026-03-01.101.jpg-${process.pid}-def456.tmp`;
-        writeFileSync(join(originals, leftover), "partial");
+        writeFileSync(join(day, leftover), "partial");
-        writeFileSync(join(originals, inProgress), "partial");
+        writeFileSync(join(day, inProgress), "partial");
         const lib = await openLibrary(stubSource());
 
         await lib.backup({ downloadDirectory: outDir });
 
-        const names = readdirSync(originals);
+        const names = readdirSync(day);
         expect(names).not.toContain(leftover);
         expect(names).toContain(inProgress);
         lib.close();
     });
 
+    it("removes temp files left in a date folder no file in the backup is saved in", async () => {
+        const outDir = join(root, "backup");
+        // As for a file since deleted, or given another date, after a run was
+        // killed while writing it.
+        const otherDay = join(outDir, "2025", "2025-01", "2025-01-02");
+        mkdirSync(otherDay, { recursive: true });
+        const exitedPID = spawnSync(process.execPath, ["-e", ""]).pid;
+        writeFileSync(join(otherDay, `.quak-${exitedPID}-abc123.tmp`), "x");
+        const lib = await openLibrary(stubSource());
+
+        await lib.backup({ downloadDirectory: outDir });
+
+        expect(readdirSync(otherDay)).toEqual([]);
+        lib.close();
+    });
+
     it("removes leftover temp files in thumbnails/ but not those of a backup still running", async () => {
         const outDir = join(root, "backup");
         const thumbnails = join(outDir, "thumbnails");
@@ -709,7 +790,7 @@ describe("the refresh before a backup", () => {
         const result = await backup;
 
         expect(result.totalFiles).toBe(4);
-        expect(existsSync(join(outDir, "originals", "300.jpg"))).toBe(true);
+        expect(existsSync(saved(outDir, "300.jpg"))).toBe(true);
         await lib.close();
     });
 
@@ -728,7 +809,7 @@ describe("the refresh before a backup", () => {
 
         expect(source.originalCalls).toBe(0);
         expect(readFileSync(ledgerPath, "utf-8")).toBe(ledgerBefore);
-        expect(existsSync(join(outDir, "originals"))).toBe(false);
+        expect(existsSync(join(outDir, DAY))).toBe(false);
         await lib.close();
     });
 
@@ -819,13 +900,13 @@ describe("backup album folders", () => {
         expect(result.failed).toBe(0);
         expect(tree(outDir)).toEqual([
             "Trip (10)/",
-            "Trip (10)/IMG_0001 (1).JPG -> ../../originals/1.JPG",
+            `Trip (10)/IMG_0001 (1).JPG -> ${linkTo("1.JPG")}`,
-            "Trip (10)/IMG_0001 (2).JPG -> ../../originals/2.JPG",
+            `Trip (10)/IMG_0001 (2).JPG -> ${linkTo("2.JPG")}`,
-            "Trip (10)/img_0001 (4).jpg -> ../../originals/4.jpg",
+            `Trip (10)/img_0001 (4).jpg -> ${linkTo("4.jpg")}`,
-            "Trip (10)/other.jpg -> ../../originals/3.jpg",
+            `Trip (10)/other.jpg -> ${linkTo("3.jpg")}`,
             "Trip (10).json",
             "Trip (11)/",
-            "Trip (11)/other.jpg -> ../../originals/3.jpg",
+            `Trip (11)/other.jpg -> ${linkTo("3.jpg")}`,
             "Trip (11).json",
         ]);
         expect(albumID(outDir, "Trip (10).json")).toBe(10);
@@ -858,14 +939,14 @@ describe("backup album folders", () => {
         expect(result.failed).toBe(0);
         expect(tree(outDir)).toEqual([
             "Trip (10)/",
-            "Trip (10)/IMG (6) (5).JPG -> ../../originals/5.JPG",
+            `Trip (10)/IMG (6) (5).JPG -> ${linkTo("5.JPG")}`,
-            "Trip (10)/IMG (6).JPG -> ../../originals/6.JPG",
+            `Trip (10)/IMG (6).JPG -> ${linkTo("6.JPG")}`,
-            "Trip (10)/IMG (7).JPG -> ../../originals/7.JPG",
+            `Trip (10)/IMG (7).JPG -> ${linkTo("7.JPG")}`,
             "Trip (10).json",
             "Trip (11)/",
-            "Trip (11)/a.jpg -> ../../originals/8.jpg",
+            `Trip (11)/a.jpg -> ${linkTo("8.jpg")}`,
             "Trip (11) (12)/",
-            "Trip (11) (12)/b.jpg -> ../../originals/9.jpg",
+            `Trip (11) (12)/b.jpg -> ${linkTo("9.jpg")}`,
             "Trip (11) (12).json",
             "Trip (11).json",
         ]);
@@ -931,13 +1012,13 @@ describe("backup album folders", () => {
         expect(scoped.failed).toBe(0);
         expect(before).toEqual([
             "Trip (10)/",
-            "Trip (10)/a.jpg -> ../../originals/1.jpg",
+            `Trip (10)/a.jpg -> ${linkTo("1.jpg")}`,
             "Trip (10).json",
             "Work/",
-            "Work/c.jpg -> ../../originals/3.jpg",
+            `Work/c.jpg -> ${linkTo("3.jpg")}`,
             "Work.json",
             "trip (11)/",
-            "trip (11)/b.jpg -> ../../originals/2.jpg",
+            `trip (11)/b.jpg -> ${linkTo("2.jpg")}`,
             "trip (11).json",
         ]);
         expect(tree(outDir)).toEqual(before);
@@ -990,13 +1071,13 @@ describe("backup album folders", () => {
             "Mine/",
             "Mine/keep.txt",
             "Office/",
-            "Office/a.jpg -> ../../originals/5.jpg",
+            `Office/a.jpg -> ${linkTo("5.jpg")}`,
             "Office.json",
             "Trip/",
-            "Trip/IMG_0001.JPG -> ../../originals/1.JPG",
+            `Trip/IMG_0001.JPG -> ${linkTo("1.JPG")}`,
             "Trip/mine -> ../elsewhere",
             "Trip/notes.txt",
-            "Trip/other.jpg -> ../../originals/3.jpg",
+            `Trip/other.jpg -> ${linkTo("3.jpg")}`,
             "Trip.json",
             "Work/",
             "Work/keep.txt",
@@ -1005,7 +1086,7 @@ describe("backup album folders", () => {
     });
 
     // 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.
+    // symlink to an original, with `json` (if given) as its sibling JSON.
     const backupWithUserFolder = async (
         json: string | undefined,
     ): Promise<{ outDir: string; failed: number }> => {
@@ -1019,10 +1100,7 @@ describe("backup album folders", () => {
         await runBackup(lib, { downloadDirectory: outDir });
         const collectionsDir = join(outDir, "collections");
         mkdirSync(join(collectionsDir, "Mine"));
-        symlinkSync(
+        symlinkSync(linkTo("1.jpg"), join(collectionsDir, "Mine", "a.jpg"));
-            "../../originals/1.jpg",
-            join(collectionsDir, "Mine", "a.jpg"),
-        );
         if (json !== undefined) {
             writeFileSync(join(collectionsDir, "Mine.json"), json);
         }
@@ -1036,9 +1114,9 @@ describe("backup album folders", () => {
         expect(failed).toBe(0);
         expect(tree(outDir)).toEqual([
             "Mine/",
-            "Mine/a.jpg -> ../../originals/1.jpg",
+            `Mine/a.jpg -> ${linkTo("1.jpg")}`,
             "Trip/",
-            "Trip/a.jpg -> ../../originals/1.jpg",
+            `Trip/a.jpg -> ${linkTo("1.jpg")}`,
             "Trip.json",
         ]);
     });
@@ -1050,10 +1128,10 @@ describe("backup album folders", () => {
         expect(failed).toBe(0);
         expect(tree(outDir)).toEqual([
             "Mine/",
-            "Mine/a.jpg -> ../../originals/1.jpg",
+            `Mine/a.jpg -> ${linkTo("1.jpg")}`,
             "Mine.json",
             "Trip/",
-            "Trip/a.jpg -> ../../originals/1.jpg",
+            `Trip/a.jpg -> ${linkTo("1.jpg")}`,
             "Trip.json",
         ]);
         expect(
@@ -1087,23 +1165,16 @@ describe("backup of live photos", () => {
     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
+    const stored = [
-    // image's name, and its link.
+        "2026-03-01.500.heic",
-    const earlierZIP = (outDir: string): void => {
+        "2026-03-01.500.json",
-        mkdirSync(join(outDir, "originals"), { recursive: true });
+        "2026-03-01.500.livephoto.json",
-        mkdirSync(join(outDir, "collections", "Trip"), { recursive: true });
+        "2026-03-01.500.mov",
-        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.heic -> ${linkTo("500.heic")}`,
-        "Trip/IMG_0500.mov -> ../../originals/500.mov",
+        `Trip/IMG_0500.mov -> ${linkTo("500.mov")}`,
         "Trip.json",
     ];
 
@@ -1113,16 +1184,15 @@ describe("backup of live photos", () => {
         );
         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(readdirSync(join(outDir, DAY)).sort()).toEqual(stored);
-        expect(readFileSync(join(originals, "500.heic"))).toEqual(
+        expect(readFileSync(saved(outDir, "500.heic"))).toEqual(
             Buffer.from(IMAGE),
         );
-        expect(readFileSync(join(originals, "500.mov"))).toEqual(
+        expect(readFileSync(saved(outDir, "500.mov"))).toEqual(
             Buffer.from(VIDEO),
         );
         expect(tree(outDir)).toEqual(linked);
@@ -1149,39 +1219,22 @@ describe("backup of live photos", () => {
 
         expect(tree(outDir)).toEqual([
             "Trip/",
-            "Trip/IMG_0001 (500).heic -> ../../originals/500.heic",
+            `Trip/IMG_0001 (500).heic -> ${linkTo("500.heic")}`,
-            "Trip/IMG_0001 (500).mov -> ../../originals/500.mov",
+            `Trip/IMG_0001 (500).mov -> ${linkTo("500.mov")}`,
-            "Trip/IMG_0001 (501).heic -> ../../originals/501.heic",
+            `Trip/IMG_0001 (501).heic -> ${linkTo("501.heic")}`,
-            "Trip/IMG_0001 (501).mov -> ../../originals/501.mov",
+            `Trip/IMG_0001 (501).mov -> ${linkTo("501.mov")}`,
             "Trip.json",
         ]);
         await lib.close();
     });
 
-    it("replaces the ZIP an earlier version stored, and its link", async () => {
+    it("stores nothing for a live photo that fails its hash", 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 });
@@ -1189,12 +1242,8 @@ describe("backup of live photos", () => {
         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(readdirSync(join(outDir, DAY))).toEqual([]);
-        expect(tree(outDir)).toEqual([
+        expect(tree(outDir)).toEqual(["Trip/", "Trip.json"]);
-            "Trip/",
-            "Trip/IMG_0500.HEIC -> ../../originals/500.HEIC",
-            "Trip.json",
-        ]);
         await lib.close();
     });
 
@@ -1209,8 +1258,8 @@ describe("backup of live photos", () => {
         const result = await lib.backup({ downloadDirectory: outDir });
 
         expect(result).toMatchObject({ downloaded: 1, failed: 0 });
-        expect(readdirSync(join(outDir, "originals")).sort()).toEqual(stored);
+        expect(readdirSync(join(outDir, DAY)).sort()).toEqual(stored);
-        expect(readFileSync(join(outDir, "originals", "500.mov"))).toEqual(
+        expect(readFileSync(saved(outDir, "500.mov"))).toEqual(
             Buffer.from(VIDEO),
         );
         expect(tree(outDir)).toEqual(linked);
@@ -1219,23 +1268,6 @@ describe("backup of live photos", () => {
         await lib.close();
     });
 
-    it("replaces an earlier ZIP and its link with the image and video the cache holds", async () => {
-        const { file: live, body } = await asLivePhoto(
-            file(500, 10, "IMG_0500.HEIC"),
-        );
-        const lib = await open([live], new Map([[500, body]]));
-        await lib.photos.byID({ fileID: 500 })!.original();
-        const outDir = join(root, "backup");
-        earlierZIP(outDir);
-
-        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.each(["missing", "empty"])(
         "fetches a live photo again when the video its JSON file names is %s",
         async (state) => {
@@ -1245,7 +1277,7 @@ describe("backup of live photos", () => {
             const lib = await open([live], new Map([[500, body]]));
             const outDir = join(root, "backup");
             await lib.backup({ downloadDirectory: outDir });
-            const video = join(outDir, "originals", "500.mov");
+            const video = saved(outDir, "500.mov");
             if (state === "missing") rmSync(video);
             else writeFileSync(video, "");
 
@@ -1271,7 +1303,7 @@ describe("backup of live photos", () => {
             const lib = await open([live], new Map([[500, body]]));
             const outDir = join(root, "backup");
             await lib.backup({ downloadDirectory: outDir });
-            const image = join(outDir, "originals", "500.heic");
+            const image = saved(outDir, "500.heic");
             if (state === "missing") rmSync(image);
             else writeFileSync(image, "");
 
@@ -1310,8 +1342,8 @@ describe("backup of live photos", () => {
         const read = await reader.photos.byID({ fileID: 500 })!.original();
 
         expect(read).toEqual({
-            path: join(outDir, "originals", "500.heic"),
+            path: saved(outDir, "500.heic"),
-            videoPath: join(outDir, "originals", "500.mov"),
+            videoPath: saved(outDir, "500.mov"),
             bytes: IMAGE.length,
         });
         await reader.close();

test/cli/commands.test.ts

@@ -56,6 +56,7 @@ import type { ContentSource } from "../../src/library/content.js";
 import type { Collection, EnteFile } from "../../src/model/types.js";
 import { init, toBase64 } from "../../src/crypto/index.js";
 import { defaultCacheDirectory } from "../../src/library/index.js";
+import { HEIC_WITH_EXIF } from "../exif-heic.js";
 import {
     asLivePhoto,
     cdnSource,
@@ -653,6 +654,31 @@ describe("a live photo", () => {
             height: 4,
         });
     });
+
+    it("backup-metadata --exif records the EXIF of a HEIC image", async () => {
+        const dir = join(root, "dump");
+
+        expect(
+            await backupMetadataCommand(
+                context(await livePhotoClient(HEIC_WITH_EXIF)),
+                dir,
+                { exif: true },
+            ),
+        ).toBe(0);
+
+        const record = JSON.parse(
+            readFileSync(
+                join(dir, "collections", "1-Vacation", "300.json"),
+                "utf-8",
+            ),
+        );
+        expect(record.imageMetadata.exifError).toBeUndefined();
+        expect(record.imageMetadata.exif).toMatchObject({
+            Make: { value: ["Canon"] },
+            Model: { value: ["EOS R5"] },
+            DateTimeOriginal: { value: ["2021:07:15 14:30:00"] },
+        });
+    });
 });
 
 describe("backup", () => {
@@ -731,7 +757,7 @@ describe("backup", () => {
         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);
+        expect(existsSync(dir)).toBe(false);
     });
 });
 

test/cli/metadata-exif.test.ts

@@ -1,21 +1,24 @@
 /**
- * Tests for the JPEG EXIF scan behind `quak backup-metadata --exif`.
+ * Tests for reading EXIF (`src/exif.ts`) and the image metadata
+ * `quak backup-metadata --exif` records.
  *
- * The originals come from users' libraries, so a truncated or corrupt JPEG
+ * The originals come from users' libraries, so a truncated or corrupt file
- * must neither hang the scan nor throw out of it, and a malformed file must be
+ * must neither hang the read nor throw out of it: `readPhotoExif` gives `{}`,
- * told apart from one that simply has no EXIF: the record carries the reason in
+ * and `backup-metadata` tells an EXIF block it cannot read apart from a file
- * `exifError`. Each input below is a short hand-built byte array.
+ * that simply has no EXIF, carrying the reason in `exifError`. Each JPEG below
+ * is a short hand-built byte array; the HEIC is a real file.
  */
 
 import { describe, expect, it } from "vitest";
-import {
+import { readPhotoExif } from "../../src/exif.js";
-    extractExifFromJpeg,
+import { extractImageMetadata } from "../../src/metadata-backup.js";
-    extractImageMetadata,
+import { HEIC_WITH_EXIF } from "../exif-heic.js";
-} from "../../src/metadata-backup.js";
 
 const SOI = [0xff, 0xd8]; // start of image
-const SOS = [0xff, 0xda, 0x00, 0x02]; // start of scan, where the scan stops
+const SOS = [0xff, 0xda, 0x00, 0x02]; // start of scan
 const EXIF_HEADER = [0x45, 0x78, 0x69, 0x66, 0x00, 0x00]; // "Exif\0\0"
+const APP0 = [0xff, 0xe0, 0x00, 0x04, 0x00, 0x00];
+const ZERO_LENGTH_APP0 = [0xff, 0xe0, 0x00, 0x00];
 
 // A big-endian TIFF block with one IFD entry: Orientation (0x0112), SHORT, 6.
 const TIFF_ORIENTATION_6 = [
@@ -24,6 +27,66 @@ const TIFF_ORIENTATION_6 = [
     0x00, 0x00,
 ];
 
+// A big-endian TIFF block holding Orientation 6 and DateTimeOriginal
+// "0000:00:00 00:00:00", which a camera with an unset clock writes.
+const TIFF_UNSET_DATE = [
+    ...[0x4d, 0x4d, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x08], // the first IFD at 8
+    // The first IFD, at 8: two entries, then no next IFD.
+    ...[0x00, 0x02],
+    // Orientation (0x0112), SHORT, 6.
+    ...[0x01, 0x12, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x06, 0x00, 0x00],
+    // The Exif IFD's offset (0x8769), LONG, 38.
+    ...[0x87, 0x69, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x26],
+    ...[0x00, 0x00, 0x00, 0x00],
+    // The Exif IFD, at 38: one entry, then no next IFD.
+    ...[0x00, 0x01],
+    // DateTimeOriginal (0x9003), 20 ASCII bytes at 56.
+    ...[0x90, 0x03, 0x00, 0x02, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x38],
+    ...[0x00, 0x00, 0x00, 0x00],
+    ...new TextEncoder().encode("0000:00:00 00:00:00\0"), // at 56
+];
+
+// A big-endian TIFF block holding a GPSAltitude of 12.5 m and no
+// GPSAltitudeRef.
+const TIFF_ALTITUDE_WITHOUT_REF = [
+    ...[0x4d, 0x4d, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x08], // the first IFD at 8
+    // The first IFD, at 8: one entry, then no next IFD.
+    ...[0x00, 0x01],
+    // The GPS IFD's offset (0x8825), LONG, 26.
+    ...[0x88, 0x25, 0x00, 0x04, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x1a],
+    ...[0x00, 0x00, 0x00, 0x00],
+    // The GPS IFD, at 26: one entry, then no next IFD.
+    ...[0x00, 0x01],
+    // GPSAltitude (0x0006), one RATIONAL at 44.
+    ...[0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2c],
+    ...[0x00, 0x00, 0x00, 0x00],
+    ...[0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x02], // 25/2, at 44
+];
+
+// A big-endian TIFF block holding Orientation 6 and a Make whose value lies
+// past the end of the file.
+const TIFF_MAKE_PAST_END = [
+    ...[0x4d, 0x4d, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x08], // the first IFD at 8
+    // The first IFD, at 8: two entries, then no next IFD.
+    ...[0x00, 0x02],
+    // Make (0x010f), 6 ASCII bytes at 4096.
+    ...[0x01, 0x0f, 0x00, 0x02, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x10, 0x00],
+    // Orientation (0x0112), SHORT, 6.
+    ...[0x01, 0x12, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x06, 0x00, 0x00],
+    ...[0x00, 0x00, 0x00, 0x00],
+];
+
+// A big-endian TIFF block with one IFD entry that exifreader has no name for:
+// tag 0xc000 (49152), SHORT, 7.
+const TIFF_UNNAMED_TAG = [
+    ...[0x4d, 0x4d, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x08], // the first IFD at 8
+    // The first IFD, at 8: one entry, then no next IFD.
+    ...[0x00, 0x01],
+    // Tag 0xc000, SHORT, 7.
+    ...[0xc0, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x00],
+    ...[0x00, 0x00, 0x00, 0x00],
+];
+
 // An APP1 segment whose length field matches its data.
 const app1 = (data: number[]): number[] => {
     const len = data.length + 2;
@@ -33,74 +96,87 @@ const app1 = (data: number[]): number[] => {
 const bytes = (...parts: number[][]): Uint8Array =>
     new Uint8Array(parts.flat());
 
-describe("extractExifFromJpeg", () => {
+describe("readPhotoExif", () => {
-    it("returns the EXIF segment of a valid JPEG", () => {
+    it("reads the common fields of a valid JPEG", () => {
         const data = [...EXIF_HEADER, ...TIFF_ORIENTATION_6];
-        const scan = extractExifFromJpeg(bytes(SOI, app1(data), SOS));
+        expect(readPhotoExif(bytes(SOI, app1(data), SOS))).toStrictEqual({
-        expect(scan.error).toBeUndefined();
+            orientation: 6,
-        expect([...scan.exif!]).toEqual(data);
+        });
     });
 
-    it("returns nothing for a file that is not a JPEG", () => {
+    it("gives no dateTimeOriginal for a DateTimeOriginal of 0000:00:00 00:00:00", () => {
-        const png = bytes([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a]);
+        const data = [...EXIF_HEADER, ...TIFF_UNSET_DATE];
-        expect(extractExifFromJpeg(png)).toEqual({});
+        expect(readPhotoExif(bytes(SOI, app1(data), SOS))).toStrictEqual({
+            orientation: 6,
+        });
     });
 
-    it("returns nothing for a JPEG without EXIF", () => {
+    it("reads a GPSAltitude without GPSAltitudeRef as above sea level", () => {
-        const app0 = [0xff, 0xe0, 0x00, 0x04, 0x00, 0x00];
+        const data = [...EXIF_HEADER, ...TIFF_ALTITUDE_WITHOUT_REF];
-        expect(extractExifFromJpeg(bytes(SOI, app0, SOS))).toEqual({});
+        expect(readPhotoExif(bytes(SOI, app1(data), SOS))).toStrictEqual({
+            gpsAltitude: 12.5,
+        });
     });
 
-    it("ignores an APP1 segment too short to hold the Exif header", () => {
+    it("gives no make for a Make whose value lies past the end of the file", () => {
-        // A length under 8 cannot hold the six-byte "Exif\0\0" header, so the
+        const data = [...EXIF_HEADER, ...TIFF_MAKE_PAST_END];
-        // segment is not EXIF. This one has length 7 and holds only "Exif\0",
+        expect(readPhotoExif(bytes(SOI, app1(data), SOS))).toStrictEqual({
-        // which the old code, lacking the length check, returned as EXIF.
+            orientation: 6,
-        const short = app1(EXIF_HEADER.slice(0, 5));
+        });
-        expect(extractExifFromJpeg(bytes(SOI, short, SOS))).toEqual({});
     });
 
-    it("accepts an APP1 segment of length 8 holding just the Exif header", () => {
+    it.each([
-        const scan = extractExifFromJpeg(bytes(SOI, app1(EXIF_HEADER), SOS));
+        [
-        expect(scan.error).toBeUndefined();
+            "a file that is not an image",
-        expect([...scan.exif!]).toEqual(EXIF_HEADER);
+            new TextEncoder().encode("just some text, not an image"),
-    });
+        ],
-
+        [
-    it("reports a JPEG truncated inside a segment header", () => {
+            "a PNG without EXIF",
-        const scan = extractExifFromJpeg(bytes(SOI, [0xff, 0xe1, 0x00]));
+            bytes([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a]),
-        expect(scan.exif).toBeUndefined();
+        ],
-        expect(scan.error).toMatch(/truncated segment length/);
+        ["a JPEG without EXIF", bytes(SOI, APP0, SOS)],
-    });
+        // A length under 8 cannot hold the six-byte "Exif\0\0" header. This one
-
+        // has length 7 and holds only "Exif\0", so a read past its end would
-    it("reports a JPEG that ends before the image data", () => {
+        // take the next segment's bytes as EXIF.
-        const app0 = [0xff, 0xe0, 0x00, 0x04, 0x00, 0x00];
+        [
-        const scan = extractExifFromJpeg(bytes(SOI, app0));
+            "an APP1 segment too short to hold the Exif header",
-        expect(scan.error).toMatch(/ends before the image data/);
+            bytes(SOI, app1(EXIF_HEADER.slice(0, 5)), SOS),
-    });
+        ],
-
+        [
-    it("stops on a zero-length segment instead of looping", () => {
+            "an APP1 segment holding just the Exif header",
-        // A length of 0 would otherwise step the scan by 2 bytes at a time
+            bytes(SOI, app1(EXIF_HEADER), SOS),
-        // through the rest of the file, reading garbage as markers.
+        ],
-        const zero = [0xff, 0xe0, 0x00, 0x00];
+        [
-        const scan = extractExifFromJpeg(
+            "a JPEG truncated inside a segment header",
-            bytes(SOI, zero, zero, zero, zero, SOS),
+            bytes(SOI, [0xff, 0xe1, 0x00]),
-        );
+        ],
-        expect(scan.error).toMatch(/segment length 0 at byte 2 is too small/);
+        ["a JPEG that ends before the image data", bytes(SOI, APP0)],
-    });
+        // A length of 0 would step a scan by 2 bytes at a time through the
-
+        // rest of the file, reading garbage as markers.
-    it("stops on a segment length of 1", () => {
+        [
-        const scan = extractExifFromJpeg(
+            "a zero-length segment",
-            bytes(SOI, [0xff, 0xe0, 0x00, 0x01], SOS),
+            bytes(
-        );
+                SOI,
-        expect(scan.error).toMatch(/segment length 1 at byte 2 is too small/);
+                ZERO_LENGTH_APP0,
-    });
+                ZERO_LENGTH_APP0,
-
+                ZERO_LENGTH_APP0,
-    it("reports a segment length that runs past the end of the file", () => {
+                ZERO_LENGTH_APP0,
+                SOS,
+            ),
+        ],
+        ["a segment length of 1", bytes(SOI, [0xff, 0xe0, 0x00, 0x01], SOS)],
         // APP1 claims 0x4000 bytes but only the "Exif\0\0" header follows.
-        const scan = extractExifFromJpeg(
+        [
+            "a segment length that runs past the end of the file",
             bytes(SOI, [0xff, 0xe1, 0x40, 0x00], EXIF_HEADER),
-        );
+        ],
-        expect(scan.exif).toBeUndefined();
+        // "XX" where the TIFF byte order belongs.
-        expect(scan.error).toMatch(/runs past the end of the file/);
+        [
+            "an EXIF block that cannot be parsed",
+            bytes(SOI, app1([...EXIF_HEADER, 0x58, 0x58]), SOS),
+        ],
+    ])("returns no fields for %s", (_, input) => {
+        expect(readPhotoExif(input)).toStrictEqual({});
     });
 });
 
@@ -110,10 +186,30 @@ describe("extractImageMetadata", () => {
             bytes(SOI, app1([...EXIF_HEADER, ...TIFF_ORIENTATION_6]), SOS),
         );
         expect(meta?.exifError).toBeUndefined();
-        expect(meta?.exif).toMatchObject({ Image: { Orientation: 6 } });
+        expect(meta?.exif).toMatchObject({ Orientation: { value: 6 } });
     });
 
-    it("returns nothing for a file that is not a JPEG", () => {
+    it("parses EXIF from a HEIC", () => {
+        const meta = extractImageMetadata(HEIC_WITH_EXIF);
+        expect(meta?.exifError).toBeUndefined();
+        expect(meta?.exif).toMatchObject({
+            Make: { value: ["Canon"] },
+            Model: { value: ["EOS R5"] },
+            DateTimeOriginal: { value: ["2021:07:15 14:30:00"] },
+            Orientation: { value: 6 },
+            GPSLatitudeRef: { value: ["N"] },
+        });
+    });
+
+    it("keys a tag exifreader has no name for by its number", () => {
+        const meta = extractImageMetadata(
+            bytes(SOI, app1([...EXIF_HEADER, ...TIFF_UNNAMED_TAG]), SOS),
+        );
+        expect(meta?.exifError).toBeUndefined();
+        expect(meta?.exif).toMatchObject({ "undefined-49152": { value: 7 } });
+    });
+
+    it("returns nothing for a file that is not an image", () => {
         const text = new TextEncoder().encode("just some text, not an image");
         expect(extractImageMetadata(text)).toBeUndefined();
     });
@@ -123,14 +219,20 @@ describe("extractImageMetadata", () => {
             bytes(SOI, [0xff, 0xe1, 0x40, 0x00], EXIF_HEADER),
         );
         expect(meta?.exif).toBeUndefined();
-        expect(meta?.exifError).toMatch(/runs past the end of the file/);
+        expect(meta?.exifError).toBe(
+            "no tag could be read from the EXIF block",
+        );
     });
 
     it("keeps the raw bytes and the reason when EXIF cannot be parsed", () => {
-        const data = [...EXIF_HEADER, 0x58, 0x58];
+        // The raw bytes are the whole EXIF block as exifreader finds it: for a
-        const meta = extractImageMetadata(bytes(SOI, app1(data), SOS));
+        // JPEG, the APP1 segment, marker and length included.
+        const segment = app1([...EXIF_HEADER, 0x58, 0x58]);
+        const meta = extractImageMetadata(bytes(SOI, segment, SOS));
         expect(meta?.exif).toBeUndefined();
-        expect(meta?.exifRaw).toBe(Buffer.from(data).toString("base64"));
+        expect(meta?.exifRaw).toBe(Buffer.from(segment).toString("base64"));
-        expect(meta?.exifError).toEqual(expect.any(String));
+        expect(meta?.exifError).toBe(
+            "no tag could be read from the EXIF block",
+        );
     });
 });

test/examples/download-albums.test.ts

@@ -0,0 +1,317 @@
+/**
+ * The example script `examples/download-albums.ts` (issue #144), run twice
+ * against a stand-in account, as a user would run it twice.
+ *
+ * The account has two albums sharing one photo, and one of its photos is a
+ * live photo whose image is a HEIC with EXIF. The first run puts every
+ * original at its save path, writes each photo's metadata beside it and each
+ * album's photos under `albums/`. Before the second run the account gains an
+ * album holding a new photo. The second run downloads that photo and writes
+ * its files and the new album's, and fetches nothing else and changes no
+ * other file.
+ */
+
+import { describe, it, expect, beforeEach, afterEach } from "vitest";
+import {
+    existsSync,
+    mkdtempSync,
+    readdirSync,
+    readFileSync,
+    rmSync,
+    statSync,
+    utimesSync,
+    writeFileSync,
+} from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+
+import { downloadAlbums } from "../../examples/download-albums.js";
+import { Library, type ContentSource } from "../../src/index.js";
+import type { CollectionsPage, FilesPage } from "../../src/client.js";
+import type { Collection, EnteFile } from "../../src/model/types.js";
+import { HEIC_WITH_EXIF } from "../exif-heic.js";
+import {
+    asLivePhoto,
+    cdnSource,
+    livePhotoHash,
+    livePhotoZip,
+    VIDEO,
+} from "../live-photo.js";
+
+const USER_ID = 7;
+
+// Every photo is taken at noon local time on 2026-03-01, so the machine's time
+// zone cannot move it to another day; it is saved in the folder `DAY`. Ente
+// stores times in microseconds.
+const TAKEN_MS = new Date(2026, 2, 1, 12).getTime();
+const DAY = join("2026", "2026-03", "2026-03-01");
+
+const collection = (id: number, name: string): Collection => ({
+    id,
+    ownerID: USER_ID,
+    key: new Uint8Array([id]),
+    name,
+    type: "album",
+    updationTime: 1,
+    isShared: false,
+});
+
+const file = (id: number, collectionID: number): EnteFile => ({
+    id,
+    collectionID,
+    ownerID: USER_ID,
+    key: new Uint8Array([id]),
+    metadata: {
+        title: `file-${id}.jpg`,
+        fileType: "image",
+        creationTime: TAKEN_MS * 1000,
+        modificationTime: TAKEN_MS * 1000,
+    },
+    file: { decryptionHeader: "aGVhZGVy" },
+    thumbnail: { decryptionHeader: "dGh1bWI=" },
+    updationTime: 1,
+});
+
+let root: string;
+
+beforeEach(() => {
+    root = mkdtempSync(join(tmpdir(), "quak-download-albums-"));
+});
+
+afterEach(() => {
+    if (root && existsSync(root))
+        rmSync(root, { recursive: true, force: true });
+});
+
+// Every file and directory under `dir`, by path.
+const entries = (dir: string): string[] =>
+    readdirSync(dir, { recursive: true, encoding: "utf-8" });
+
+// Set the modification time of everything under `dir` to the epoch, so that
+// anything written there afterwards has a later one, however soon it comes.
+const backdate = (dir: string): void => {
+    for (const name of entries(dir)) utimesSync(join(dir, name), 0, 0);
+};
+
+// The modification time of everything under `dir`, by path.
+const mtimes = (dir: string): Map<string, number> =>
+    new Map(
+        entries(dir).map((name) => [name, statSync(join(dir, name)).mtimeMs]),
+    );
+
+const readJSON = (path: string): unknown =>
+    JSON.parse(readFileSync(path, "utf-8"));
+
+describe("examples/download-albums.ts", () => {
+    it("downloads every album's photos with their metadata, and on a second run only what the account gained", async () => {
+        // Album 1, "Trip", holds photos 1 and 2. Album 2, "Family", holds
+        // photo 2 and photo 3, a live photo.
+        const live = await asLivePhoto(
+            file(3, 2),
+            livePhotoZip({ "image.heic": HEIC_WITH_EXIF, "video.mov": VIDEO }),
+            livePhotoHash(HEIC_WITH_EXIF, VIDEO),
+        );
+        const collections = [collection(1, "Trip"), collection(2, "Family")];
+        const filesByAlbum = new Map<number, EnteFile[]>([
+            [1, [file(1, 1), file(2, 1)]],
+            [2, [file(2, 2), live.file]],
+        ]);
+        const client = {
+            whoami: () => ({ email: "u@example.com", userID: USER_ID }),
+            collectionsSince: async (): Promise<CollectionsPage> => ({
+                collections: [...collections],
+                deleted: [],
+                cursor: 1,
+            }),
+            filesSince: async (args: {
+                collectionID: number;
+            }): Promise<FilesPage> => ({
+                files: filesByAlbum.get(args.collectionID) ?? [],
+                deleted: [],
+                cursor: 1,
+            }),
+        };
+
+        // Photo 3 comes from a stand-in server, encrypted as Ente serves a
+        // live photo. Any other original is a few bytes naming its photo.
+        // `calls` counts every fetch, thumbnails included.
+        const server = cdnSource(new Map([[3, live.body]]));
+        let calls = 0;
+        const source: ContentSource = {
+            original: async (args) => {
+                calls++;
+                if (args.file.id === 3) return server.original(args);
+                const bytes = `original-${args.file.id}`;
+                writeFileSync(args.destination, bytes);
+                return { bytesWritten: bytes.length };
+            },
+            thumbnail: async (args) => {
+                calls++;
+                return server.thumbnail(args);
+            },
+        };
+
+        const dir = join(root, "photos");
+        const open = (): Promise<Library> =>
+            Library.open({
+                client,
+                cacheDirectory: join(root, "cache"),
+                downloadDirectory: dir,
+                contentSource: source,
+                refreshIntervalSeconds: 3600,
+                precacheThumbnails: false,
+                precacheOriginals: false,
+            });
+
+        const first = await open();
+        // The cache already holds photo 1's original, so its record names a
+        // cache path, which the metadata leaves out. download() copies it
+        // from the cache rather than fetching it again.
+        await first.photos.byID({ fileID: 1 })!.original();
+        expect(await downloadAlbums(first, dir)).toEqual({
+            downloaded: 3,
+            alreadyLocal: 0,
+        });
+        await first.close();
+        expect(calls).toBe(3);
+
+        // Each original at its save path, a live photo as its image, its video
+        // and the file naming them, and each photo's metadata beside it.
+        const day = join(dir, DAY);
+        expect(readdirSync(day).sort()).toEqual([
+            "2026-03-01.1.jpg",
+            "2026-03-01.1.jpg.json",
+            "2026-03-01.2.jpg",
+            "2026-03-01.2.jpg.json",
+            "2026-03-01.3.heic",
+            "2026-03-01.3.heic.json",
+            "2026-03-01.3.livephoto.json",
+            "2026-03-01.3.mov",
+        ]);
+        expect(readFileSync(join(day, "2026-03-01.1.jpg"), "utf-8")).toBe(
+            "original-1",
+        );
+        expect(readFileSync(join(day, "2026-03-01.2.jpg"), "utf-8")).toBe(
+            "original-2",
+        );
+        expect(readFileSync(join(day, "2026-03-01.3.heic"))).toEqual(
+            Buffer.from(HEIC_WITH_EXIF),
+        );
+        expect(readFileSync(join(day, "2026-03-01.3.mov"))).toEqual(
+            Buffer.from(VIDEO),
+        );
+
+        // The metadata is the photo's record and its EXIF fields. The
+        // originals of photos 1 and 2 are not image data, so they have no
+        // EXIF fields. Photo 3's image holds a camera, an exposure and a
+        // position, and the date it was taken is written as an ISO 8601
+        // string. `record` holds the fields the three records share.
+        const record = {
+            takenAt: TAKEN_MS,
+            modifiedAt: TAKEN_MS,
+            fileType: "image",
+            isArchived: false,
+            isHidden: false,
+        };
+        expect(readJSON(join(day, "2026-03-01.1.jpg.json"))).toEqual({
+            ...record,
+            fileID: 1,
+            albumIDs: [1],
+            title: "file-1.jpg",
+            exif: {},
+        });
+        expect(readJSON(join(day, "2026-03-01.2.jpg.json"))).toEqual({
+            ...record,
+            fileID: 2,
+            albumIDs: [1, 2],
+            title: "file-2.jpg",
+            exif: {},
+        });
+        expect(readJSON(join(day, "2026-03-01.3.heic.json"))).toEqual({
+            ...record,
+            fileID: 3,
+            albumIDs: [2],
+            title: "file-3.jpg",
+            fileType: "livePhoto",
+            hash: livePhotoHash(HEIC_WITH_EXIF, VIDEO),
+            exif: {
+                make: "Canon",
+                model: "EOS R5",
+                lensModel: "RF50mm F1.8 STM",
+                dateTimeOriginal: "2021-07-15T14:30:00.000Z",
+                offsetTimeOriginal: "+02:00",
+                exposureTime: 1 / 250,
+                fNumber: 2.8,
+                iso: 400,
+                focalLength: 50,
+                orientation: 6,
+                gpsLatitude: 40 + 26 / 60 + 46 / 3600,
+                gpsLongitude: -(79 + 58 / 60 + 56 / 3600),
+                gpsAltitude: -12.5,
+            },
+        });
+
+        // Each album's photos, newest first, by save path relative to `dir`.
+        // Photo 2 is in both.
+        expect(readdirSync(join(dir, "albums")).sort()).toEqual([
+            "1.json",
+            "2.json",
+        ]);
+        expect(readJSON(join(dir, "albums", "1.json"))).toEqual({
+            collectionID: 1,
+            name: "Trip",
+            savePaths: [
+                join(DAY, "2026-03-01.2.jpg"),
+                join(DAY, "2026-03-01.1.jpg"),
+            ],
+        });
+        expect(readJSON(join(dir, "albums", "2.json"))).toEqual({
+            collectionID: 2,
+            name: "Family",
+            savePaths: [
+                join(DAY, "2026-03-01.3.heic"),
+                join(DAY, "2026-03-01.2.jpg"),
+            ],
+        });
+
+        // Before the second run the account gains album 3, "Garden", holding
+        // a new photo 4. The second run, with a newly opened library, opens
+        // the cache written by the first and still downloads photo 4. It finds
+        // the other photos already local and fetches nothing else.
+        collections.push(collection(3, "Garden"));
+        filesByAlbum.set(3, [file(4, 3)]);
+        backdate(dir);
+        const before = mtimes(dir);
+        const second = await open();
+        expect(await downloadAlbums(second, dir)).toEqual({
+            downloaded: 1,
+            alreadyLocal: 3,
+        });
+        await second.close();
+        expect(calls).toBe(4);
+        expect(readFileSync(join(day, "2026-03-01.4.jpg"), "utf-8")).toBe(
+            "original-4",
+        );
+        expect(readJSON(join(dir, "albums", "3.json"))).toEqual({
+            collectionID: 3,
+            name: "Garden",
+            savePaths: [join(DAY, "2026-03-01.4.jpg")],
+        });
+
+        // The second run adds only photo 4's files and album 3's, and
+        // rewrites, renames or removes no file from the first run. The two
+        // directories that gain a file are the only other changes.
+        const after = mtimes(dir);
+        expect(
+            [...after.keys()].filter((name) => !before.has(name)).sort(),
+        ).toEqual([
+            join(DAY, "2026-03-01.4.jpg"),
+            join(DAY, "2026-03-01.4.jpg.json"),
+            join("albums", "3.json"),
+        ]);
+        for (const [name, mtime] of before) {
+            if (name === DAY || name === "albums") continue;
+            expect(after.get(name), name).toBe(mtime);
+        }
+    });
+});

test/exif-heic.ts

@@ -0,0 +1,27 @@
+/**
+ * `exif.heic`, beside this file: a real 64x64 HEIC whose EXIF holds the same
+ * values as the hand-built JPEG in `library/content-library.test.ts`, for the
+ * tests of `exif()` and `backup-metadata --exif`.
+ *
+ * It was made once, in a throwaway node:22-alpine container (Alpine 3.23.3),
+ * with libheif 1.23.0 and exiftool 13.55:
+ *
+ *     apk add libheif-tools exiftool imagemagick
+ *     magick -size 64x64 gradient:red-blue -depth 8 in.png
+ *     heif-enc -q 30 -o exif.heic in.png
+ *     exiftool -overwrite_original \
+ *         -Make=Canon -Model="EOS R5" -LensModel="RF50mm F1.8 STM" \
+ *         -DateTimeOriginal="2021:07:15 14:30:00" -OffsetTimeOriginal="+02:00" \
+ *         -ExposureTime=1/250 -FNumber=2.8 -ISO=400 -FocalLength=50 \
+ *         -Orientation#=6 \
+ *         -GPSLatitude="40 26 46" -GPSLatitudeRef=N \
+ *         -GPSLongitude="79 58 56" -GPSLongitudeRef=W \
+ *         -GPSAltitude=12.5 -GPSAltitudeRef#=1 \
+ *         exif.heic
+ */
+
+import { readFileSync } from "node:fs";
+
+export const HEIC_WITH_EXIF = new Uint8Array(
+    readFileSync(new URL("exif.heic", import.meta.url)),
+);

test/library/content-eviction.test.ts

@@ -142,6 +142,7 @@ const buildCache = (args: {
         pools: new RequestPools(),
         source,
         cacheDirectory: cacheDir,
+        downloadDirectory: join(root, "photos"),
         getFile: (id) => byID.get(id),
         statfs: args.statfs,
         cacheOriginalsMaxBytes: args.cacheOriginalsMaxBytes,

test/library/content-library.test.ts

@@ -6,15 +6,18 @@
  * `Photo` objects that fetch through it, `lib.thumbnails.ensure` drives it, and
  * a cached path shows up on the projected record. A library opened without a
  * content source leaves those methods throwing rather than silently doing
- * nothing.
+ * nothing. It also covers a `Photo`'s `savePath`, `isLocal`, `download()`,
+ * `content()`, `exif()` and the methods that each return one field of `exif()`.
  */
 
-import { describe, it, expect, beforeEach, afterEach } from "vitest";
+import { describe, it, expect, beforeEach, afterEach, vi } from "vitest";
 import {
     mkdtempSync,
     readdirSync,
+    readFileSync,
     rmSync,
     existsSync,
+    statSync,
     writeFileSync,
 } from "node:fs";
 import { tmpdir } from "node:os";
@@ -24,10 +27,25 @@ import { Library, type LibraryOptions } 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, livePhotoZip } from "../live-photo.js";
+import type { PhotoExif } from "../../src/exif.js";
+import { HEIC_WITH_EXIF } from "../exif-heic.js";
+import {
+    asLivePhoto,
+    cdnSource,
+    IMAGE,
+    livePhotoHash,
+    livePhotoZip,
+    VIDEO,
+} from "../live-photo.js";
 
 const USER_ID = 7;
 
+// Every file is taken at noon local time on 2026-03-01, in microseconds as Ente
+// stores times, so the machine's time zone cannot move it to another day; it
+// is saved in the folder `DAY`.
+const TAKEN = new Date(2026, 2, 1, 12).getTime() * 1000;
+const DAY = join("2026", "2026-03", "2026-03-01");
+
 const collection = (id: number): Collection => ({
     id,
     ownerID: USER_ID,
@@ -46,7 +64,7 @@ const file = (id: number, collectionID: number): EnteFile => ({
     metadata: {
         title: `file-${id}.jpg`,
         fileType: "image",
-        creationTime: 1,
+        creationTime: TAKEN,
         modificationTime: 1,
     },
     file: { decryptionHeader: "aGVhZGVy" },
@@ -70,14 +88,23 @@ class MockClient {
     }
 }
 
-// A content source that writes a marker file and counts thumbnail fetches.
+// A content source that writes `original` as every original and a marker file
-const stubSource = (): ContentSource & { thumbCalls: () => number } => {
+// as every thumbnail, and counts the fetches of each.
+const stubSource = (
+    original: string | Uint8Array = "orig-bytes",
+): ContentSource & {
+    originalCalls: () => number;
+    thumbCalls: () => number;
+} => {
+    let originalCalls = 0;
     let thumbCalls = 0;
     return {
+        originalCalls: () => originalCalls,
         thumbCalls: () => thumbCalls,
         original: async ({ destination }) => {
-            writeFileSync(destination, "orig-bytes");
+            originalCalls++;
-            return { bytesWritten: 10 };
+            writeFileSync(destination, original);
+            return { bytesWritten: original.length };
         },
         thumbnail: async ({ destination }) => {
             thumbCalls++;
@@ -205,9 +232,523 @@ describe("Library content wiring", () => {
         await expect(
             lib.photos.byID({ fileID: 1 })!.thumbnail(),
         ).rejects.toThrow(/content cache/i);
+        await expect(
+            lib.photos.byID({ fileID: 1 })!.download(),
+        ).rejects.toThrow(/content cache/i);
         await expect(
             lib.thumbnails.ensure({ fileIDs: [1], priority: "visible" }),
         ).rejects.toThrow(/content cache/i);
         await lib.close();
     });
 });
+
+// Big-endian bytes for the hand-built JPEG below.
+const u16 = (n: number): number[] => [n >> 8, n & 0xff];
+const u32 = (n: number): number[] => [...u16(n >>> 16), ...u16(n & 0xffff)];
+const ascii = (s: string): number[] => [...new TextEncoder().encode(s), 0];
+const rational = (num: number, den: number): number[] => [
+    ...u32(num),
+    ...u32(den),
+];
+// One IFD entry: tag, type (1 BYTE, 2 ASCII, 3 SHORT, 4 LONG, 5 RATIONAL),
+// count, then the value when it fits in 4 bytes, else its offset.
+const entry = (
+    tag: number,
+    type: number,
+    count: number,
+    value: number[],
+): number[] => [...u16(tag), ...u16(type), ...u32(count), ...value];
+
+// The TIFF block of a JPEG's EXIF segment, holding every field `exif()` picks:
+// the camera in the first IFD, the exposure in the Exif IFD, and a GPS position
+// of 40°26'46" N, 79°58'56" W, 12.5 m below sea level. Offsets count from the
+// start of this block.
+const TIFF = [
+    ...[0x4d, 0x4d, 0x00, 0x2a], // big-endian TIFF
+    ...u32(8), // the first IFD's offset
+    // The first IFD, at 8: five entries, then no next IFD.
+    ...u16(5),
+    ...entry(0x010f, 2, 6, u32(74)), // Make
+    ...entry(0x0110, 2, 7, u32(80)), // Model
+    ...entry(0x0112, 3, 1, [...u16(6), 0, 0]), // Orientation
+    ...entry(0x8769, 4, 1, u32(88)), // the Exif IFD's offset
+    ...entry(0x8825, 4, 1, u32(246)), // the GPS IFD's offset
+    ...u32(0),
+    ...ascii("Canon"), // at 74
+    ...ascii("EOS R5"), // at 80
+    0, // a pad byte
+    // The Exif IFD, at 88: seven entries, then no next IFD.
+    ...u16(7),
+    ...entry(0x829a, 5, 1, u32(178)), // ExposureTime
+    ...entry(0x829d, 5, 1, u32(186)), // FNumber
+    ...entry(0x8827, 3, 1, [...u16(400), 0, 0]), // ISOSpeedRatings
+    ...entry(0x9003, 2, 20, u32(194)), // DateTimeOriginal
+    ...entry(0x9011, 2, 7, u32(214)), // OffsetTimeOriginal
+    ...entry(0x920a, 5, 1, u32(222)), // FocalLength
+    ...entry(0xa434, 2, 16, u32(230)), // LensModel
+    ...u32(0),
+    ...rational(1, 250), // at 178
+    ...rational(28, 10), // at 186
+    ...ascii("2021:07:15 14:30:00"), // at 194
+    ...ascii("+02:00"), // at 214
+    0, // a pad byte
+    ...rational(50, 1), // at 222
+    ...ascii("RF50mm F1.8 STM"), // at 230
+    // The GPS IFD, at 246: six entries, then no next IFD.
+    ...u16(6),
+    ...entry(0x0001, 2, 2, [...ascii("N"), 0, 0]), // GPSLatitudeRef
+    ...entry(0x0002, 5, 3, u32(324)), // GPSLatitude
+    ...entry(0x0003, 2, 2, [...ascii("W"), 0, 0]), // GPSLongitudeRef
+    ...entry(0x0004, 5, 3, u32(348)), // GPSLongitude
+    ...entry(0x0005, 1, 1, [1, 0, 0, 0]), // GPSAltitudeRef: below sea level
+    ...entry(0x0006, 5, 1, u32(372)), // GPSAltitude
+    ...u32(0),
+    ...[...rational(40, 1), ...rational(26, 1), ...rational(46, 1)], // at 324
+    ...[...rational(79, 1), ...rational(58, 1), ...rational(56, 1)], // at 348
+    ...rational(25, 2), // at 372
+];
+
+const JPEG_WITH_EXIF = new Uint8Array([
+    ...[0xff, 0xd8], // start of image
+    ...[0xff, 0xe1, ...u16(2 + 6 + TIFF.length)], // APP1 and its length
+    ...[...ascii("Exif"), 0], // "Exif\0\0"
+    ...TIFF,
+    ...[0xff, 0xda, 0x00, 0x02], // start of scan
+]);
+
+// A JPEG whose EXIF segment is laid out correctly but holds "XX" where the TIFF
+// byte order belongs, so exifreader cannot parse it.
+const JPEG_WITH_BAD_EXIF = new Uint8Array([
+    ...[0xff, 0xd8], // start of image
+    ...[0xff, 0xe1, ...u16(2 + 6 + 2)], // APP1 and its length
+    ...[...ascii("Exif"), 0], // "Exif\0\0"
+    ...[0x58, 0x58], // "XX"
+    ...[0xff, 0xda, 0x00, 0x02], // start of scan
+]);
+
+describe("Photo save path, local copy, content and EXIF", () => {
+    // The same account, with `files` in its album instead.
+    class FilesClient extends MockClient {
+        constructor(private readonly files: EnteFile[]) {
+            super();
+        }
+        override async filesSince(): Promise<FilesPage> {
+            return { files: this.files, deleted: [], cursor: 1 };
+        }
+    }
+
+    const open = (opts: Partial<LibraryOptions> = {}): Promise<Library> =>
+        Library.open({
+            client: new MockClient(),
+            cacheDirectory: join(root, "cache"),
+            downloadDirectory: join(root, "backup"),
+            contentSource: stubSource(),
+            refreshIntervalSeconds: 3600,
+            precacheThumbnails: false,
+            precacheOriginals: false,
+            ...opts,
+        });
+
+    it("names where a backup writes the original, which is local once the backup has written it", async () => {
+        const lib = await open();
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        const savePath = join(root, "backup", DAY, "2026-03-01.1.jpg");
+        expect(photo.savePath).toBe(savePath);
+        expect(photo.isLocal).toBe(false);
+
+        await lib.backup();
+        expect(photo.savePath).toBe(savePath);
+        expect(existsSync(savePath)).toBe(true);
+        expect(photo.isLocal).toBe(true);
+        await lib.close();
+    });
+
+    it("returns the original's bytes, and a copy only in the cache is not local", async () => {
+        const lib = await open();
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        expect(await photo.content()).toEqual(Buffer.from("orig-bytes"));
+        expect(existsSync(join(root, "cache", "originals", "1.jpg"))).toBe(
+            true,
+        );
+        expect(existsSync(photo.savePath)).toBe(false);
+        expect(photo.isLocal).toBe(false);
+        await lib.close();
+    });
+
+    it("saves under photos/ in the working directory at open without a download directory", async () => {
+        const cwd = vi.spyOn(process, "cwd").mockReturnValue(root);
+        const lib = await open({ downloadDirectory: undefined });
+        cwd.mockRestore();
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        expect(lib.downloadDirectory).toBe(join(root, "photos"));
+        expect(photo.savePath).toBe(
+            join(root, "photos", DAY, "2026-03-01.1.jpg"),
+        );
+        expect(photo.isLocal).toBe(false);
+        await lib.close();
+    });
+
+    it("refuses an empty download directory", async () => {
+        await expect(open({ downloadDirectory: "" })).rejects.toThrow(
+            /downloadDirectory is empty/,
+        );
+    });
+
+    it("keeps a photo's save path after a refresh removes its file", async () => {
+        // The same account, whose album is deleted on the second refresh.
+        class AlbumDeletedClient extends MockClient {
+            override async collectionsSince(): Promise<CollectionsPage> {
+                if (!this.served) return super.collectionsSince();
+                return { collections: [], deleted: [1], cursor: 2 };
+            }
+        }
+        const lib = await open({ client: new AlbumDeletedClient() });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        await photo.download();
+
+        await lib.fresh();
+
+        expect(lib.photos.byID({ fileID: 1 })).toBeUndefined();
+        expect(photo.savePath).toBe(
+            join(root, "backup", DAY, "2026-03-01.1.jpg"),
+        );
+        expect(photo.isLocal).toBe(true);
+        await lib.close();
+    });
+
+    it("dates the save path and download() by the membership its takenAt comes from", async () => {
+        // One file in two albums. The date edited in Ente has reached album 2's
+        // copy, synced later, but album 1 still has an earlier edit.
+        const older = file(1, 1);
+        older.pubMagicMetadata = {
+            editedTime: new Date(2026, 1, 1, 12).getTime() * 1000,
+        };
+        const newer = file(1, 2);
+        newer.updationTime = 2;
+        newer.pubMagicMetadata = {
+            editedTime: new Date(2026, 3, 15, 12).getTime() * 1000,
+        };
+        const client = {
+            whoami: () => ({ email: "u@example.com", userID: USER_ID }),
+            collectionsSince: async (): Promise<CollectionsPage> => ({
+                collections: [collection(1), collection(2)],
+                deleted: [],
+                cursor: 1,
+            }),
+            filesSince: async (args: {
+                collectionID: number;
+            }): Promise<FilesPage> => ({
+                files: [args.collectionID === 1 ? older : newer],
+                deleted: [],
+                cursor: 1,
+            }),
+        };
+        const lib = await open({ client });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+
+        expect(photo.takenAt).toBe(new Date(2026, 3, 15, 12).getTime());
+        expect(photo.savePath).toBe(
+            join(
+                root,
+                "backup",
+                "2026",
+                "2026-04",
+                "2026-04-15",
+                "2026-04-15.1.jpg",
+            ),
+        );
+
+        // download() writes to that same path, so the photo is then local.
+        const saved = await photo.download();
+        expect(saved.path).toBe(photo.savePath);
+        expect(photo.isLocal).toBe(true);
+        expect(existsSync(join(root, "backup", "2026", "2026-02"))).toBe(false);
+        await lib.close();
+    });
+
+    it("downloads a photo held across a refresh that edits its date to the save path it names", async () => {
+        // The same account, whose second refresh brings a date edited in Ente.
+        const edited = file(1, 1);
+        edited.updationTime = 2;
+        edited.pubMagicMetadata = {
+            editedTime: new Date(2026, 3, 15, 12).getTime() * 1000,
+        };
+        class DateEditedClient extends MockClient {
+            refreshes = 0;
+            override async collectionsSince(): Promise<CollectionsPage> {
+                this.refreshes++;
+                return {
+                    collections: [
+                        { ...collection(1), updationTime: this.refreshes },
+                    ],
+                    deleted: [],
+                    cursor: this.refreshes,
+                };
+            }
+            override async filesSince(): Promise<FilesPage> {
+                return {
+                    files: [this.refreshes === 1 ? file(1, 1) : edited],
+                    deleted: [],
+                    cursor: this.refreshes,
+                };
+            }
+        }
+        const lib = await open({ client: new DateEditedClient() });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+
+        await lib.fresh();
+        expect(lib.photos.byID({ fileID: 1 })!.takenAt).toBe(
+            new Date(2026, 3, 15, 12).getTime(),
+        );
+
+        const saved = await photo.download();
+        expect(saved.path).toBe(photo.savePath);
+        expect(saved.path).toBe(join(root, "backup", DAY, "2026-03-01.1.jpg"));
+        expect(photo.isLocal).toBe(true);
+        await lib.close();
+    });
+
+    it("has a save path, and is not local, without a content source", async () => {
+        const lib = await open({ contentSource: undefined });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        expect(photo.savePath).toBe(
+            join(root, "backup", DAY, "2026-03-01.1.jpg"),
+        );
+        expect(photo.isLocal).toBe(false);
+        await lib.close();
+    });
+
+    it("gives a live photo's image as its save path once a backup has stored it", async () => {
+        const { file: live, body } = await asLivePhoto(file(1, 1));
+        const lib = await open({
+            client: new FilesClient([live]),
+            contentSource: cdnSource(new Map([[1, body]])),
+        });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        const day = join(root, "backup", DAY);
+        // Until then the name comes from the title, file-1.jpg; the backup
+        // stores the image with the extension found inside the live photo.
+        expect(photo.savePath).toBe(join(day, "2026-03-01.1.jpg"));
+
+        await lib.backup();
+        expect(photo.savePath).toBe(join(day, "2026-03-01.1.heic"));
+        expect(photo.isLocal).toBe(true);
+        expect(await photo.content()).toEqual(Buffer.from(IMAGE));
+        await lib.close();
+    });
+
+    it("downloads an original the cache holds by copying it, without fetching it again", async () => {
+        const source = stubSource();
+        const lib = await open({ contentSource: source });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        await photo.original();
+        expect(source.originalCalls()).toBe(1);
+        expect(photo.isLocal).toBe(false);
+
+        const saved = await photo.download();
+
+        expect(source.originalCalls()).toBe(1);
+        expect(saved).toEqual({
+            path: photo.savePath,
+            bytes: "orig-bytes".length,
+        });
+        expect(readFileSync(photo.savePath, "utf-8")).toBe("orig-bytes");
+        expect(photo.isLocal).toBe(true);
+        // The cache keeps its own copy.
+        expect(existsSync(join(root, "cache", "originals", "1.jpg"))).toBe(
+            true,
+        );
+        await lib.close();
+    });
+
+    it("downloads an original the cache does not hold straight to its save path", async () => {
+        const source = stubSource();
+        const lib = await open({ contentSource: source });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+
+        const saved = await photo.download();
+
+        expect(source.originalCalls()).toBe(1);
+        expect(saved.path).toBe(join(root, "backup", DAY, "2026-03-01.1.jpg"));
+        expect(readFileSync(saved.path, "utf-8")).toBe("orig-bytes");
+        expect(photo.isLocal).toBe(true);
+        expect(readdirSync(join(root, "cache", "originals"))).toEqual([]);
+        await lib.close();
+    });
+
+    it("does nothing when download() finds the original already at its save path", async () => {
+        const source = stubSource();
+        const lib = await open({ contentSource: source });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        const first = await photo.download();
+        const written = statSync(first.path).ino;
+
+        const second = await photo.download();
+
+        expect(second).toEqual(first);
+        expect(source.originalCalls()).toBe(1);
+        expect(statSync(second.path).ino).toBe(written);
+        await lib.close();
+    });
+
+    it("downloads a live photo the cache holds as its image, its video and the JSON file naming them", async () => {
+        const { file: live, body } = await asLivePhoto(file(1, 1));
+        const lib = await open({
+            client: new FilesClient([live]),
+            contentSource: cdnSource(new Map([[1, body]])),
+        });
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        await photo.original();
+        const day = join(root, "backup", DAY);
+
+        const saved = await photo.download();
+
+        expect(saved).toEqual({
+            path: join(day, "2026-03-01.1.heic"),
+            videoPath: join(day, "2026-03-01.1.mov"),
+            bytes: IMAGE.length,
+        });
+        expect(readdirSync(day).sort()).toEqual([
+            "2026-03-01.1.heic",
+            "2026-03-01.1.livephoto.json",
+            "2026-03-01.1.mov",
+        ]);
+        expect(
+            JSON.parse(
+                readFileSync(join(day, "2026-03-01.1.livephoto.json"), "utf-8"),
+            ),
+        ).toEqual({ image: "2026-03-01.1.heic", video: "2026-03-01.1.mov" });
+        expect(photo.savePath).toBe(saved.path);
+        expect(photo.isLocal).toBe(true);
+        await lib.close();
+    });
+
+    it("reads the common EXIF fields of a JPEG original", async () => {
+        const lib = await open({ contentSource: stubSource(JPEG_WITH_EXIF) });
+        expect(await lib.photos.byID({ fileID: 1 })!.exif()).toStrictEqual({
+            make: "Canon",
+            model: "EOS R5",
+            lensModel: "RF50mm F1.8 STM",
+            // The camera's clock reading, held in the Date's UTC fields.
+            dateTimeOriginal: new Date(Date.UTC(2021, 6, 15, 14, 30)),
+            offsetTimeOriginal: "+02:00",
+            exposureTime: 1 / 250,
+            fNumber: 2.8,
+            iso: 400,
+            focalLength: 50,
+            orientation: 6,
+            gpsLatitude: 40 + 26 / 60 + 46 / 3600,
+            gpsLongitude: -(79 + 58 / 60 + 56 / 3600),
+            gpsAltitude: -12.5,
+        });
+        await lib.close();
+    });
+
+    // What exif() returns for HEIC_WITH_EXIF, and for JPEG_WITH_EXIF, which
+    // holds the same values.
+    const heicFields: PhotoExif = {
+        make: "Canon",
+        model: "EOS R5",
+        lensModel: "RF50mm F1.8 STM",
+        dateTimeOriginal: new Date(Date.UTC(2021, 6, 15, 14, 30)),
+        offsetTimeOriginal: "+02:00",
+        exposureTime: 1 / 250,
+        fNumber: 2.8,
+        iso: 400,
+        focalLength: 50,
+        orientation: 6,
+        gpsLatitude: 40 + 26 / 60 + 46 / 3600,
+        gpsLongitude: -(79 + 58 / 60 + 56 / 3600),
+        gpsAltitude: -12.5,
+    };
+
+    it("reads the same common EXIF fields from a HEIC original", async () => {
+        const lib = await open({ contentSource: stubSource(HEIC_WITH_EXIF) });
+        expect(await lib.photos.byID({ fileID: 1 })!.exif()).toStrictEqual(
+            heicFields,
+        );
+        await lib.close();
+    });
+
+    it("reads the EXIF of a live photo whose image is a HEIC", async () => {
+        const { file: live, body } = await asLivePhoto(
+            file(1, 1),
+            livePhotoZip({ "image.heic": HEIC_WITH_EXIF, "video.mov": VIDEO }),
+            livePhotoHash(HEIC_WITH_EXIF, VIDEO),
+        );
+        const lib = await open({
+            client: new FilesClient([live]),
+            contentSource: cdnSource(new Map([[1, body]])),
+        });
+        expect(await lib.photos.byID({ fileID: 1 })!.exif()).toStrictEqual(
+            heicFields,
+        );
+        await lib.close();
+    });
+
+    // The build's type check, not this test, makes sure `Photo` has a method
+    // for every `PhotoExif` field, whatever the fixtures hold: `Photo`
+    // implements a type with one method per field. This test checks that each
+    // method gives the same value as exif().
+    it.each([
+        ["JPEG", JPEG_WITH_EXIF],
+        ["HEIC", HEIC_WITH_EXIF],
+    ])(
+        "has a method for each field exif() returns, giving the same value, for a %s",
+        async (_, bytes) => {
+            const lib = await open({ contentSource: stubSource(bytes) });
+            const photo = lib.photos.byID({ fileID: 1 })!;
+            const exif = await photo.exif();
+            // The file holds every field, so every method is checked.
+            expect(exif).toStrictEqual(heicFields);
+            for (const [field, value] of Object.entries(exif)) {
+                expect(await photo[field as keyof PhotoExif]()).toStrictEqual(
+                    value,
+                );
+            }
+            await lib.close();
+        },
+    );
+
+    it("returns no EXIF fields for an original that is not an image", async () => {
+        const lib = await open();
+        const photo = lib.photos.byID({ fileID: 1 })!;
+        expect(await photo.exif()).toStrictEqual({});
+        expect(await photo.dateTimeOriginal()).toBeUndefined();
+        await lib.close();
+    });
+
+    it("returns no EXIF fields for a JPEG whose EXIF cannot be parsed", async () => {
+        const lib = await open({
+            contentSource: stubSource(JPEG_WITH_BAD_EXIF),
+        });
+        expect(await lib.photos.byID({ fileID: 1 })!.exif()).toStrictEqual({});
+        await lib.close();
+    });
+
+    it("throws from exif() on a video without a content source, as the other content methods do", async () => {
+        const video = file(1, 1);
+        video.metadata.fileType = "video";
+        const lib = await open({
+            client: new FilesClient([video]),
+            contentSource: undefined,
+        });
+        await expect(lib.photos.byID({ fileID: 1 })!.exif()).rejects.toThrow(
+            /content cache/i,
+        );
+        await lib.close();
+    });
+
+    it("returns no EXIF fields for a video, without fetching it", async () => {
+        const video = file(1, 1);
+        video.metadata.fileType = "video";
+        const source = stubSource(JPEG_WITH_EXIF);
+        const lib = await open({
+            client: new FilesClient([video]),
+            contentSource: source,
+        });
+        expect(await lib.photos.byID({ fileID: 1 })!.exif()).toStrictEqual({});
+        expect(source.originalCalls()).toBe(0);
+        await lib.close();
+    });
+});

test/library/content.test.ts

@@ -7,8 +7,8 @@
  * 1. **Fetch once, then serve from disk.** The first `original`/`thumbnail`
  *    fetches through the request pool and stores the bytes; the next finds the
  *    file present and returns its path with a single `skipped` event and no
- *    network. A file already sitting in the backup `downloadDirectory` counts
+ *    network. A file already stored at its save path under the
- *    as present too.
+ *    `downloadDirectory` counts as present too.
  * 2. **Present-means-complete.** Content appears only by the streaming atomic
  *    writer's rename, so a file that exists is whole. The directory listing
  *    taken at `open()` is the record of what is cached, and the orphan temp
@@ -41,6 +41,7 @@ import { join } from "node:path";
 
 import {
     ContentCache,
+    savePath,
     type ContentSource,
     type EnsureEvent,
 } from "../../src/library/content.js";
@@ -152,12 +153,48 @@ const buildCache = (
         pools: args.pools ?? new RequestPools(),
         source,
         cacheDirectory: cacheDir,
-        downloadDirectory: args.downloadDirectory,
+        downloadDirectory: args.downloadDirectory ?? join(root, "photos"),
         getFile: (id) => byID.get(id),
     });
     return { cache, source };
 };
 
+// Microseconds, as Ente stores times, for noon local time on a day, so the
+// machine's time zone cannot move the photo to another day.
+const noon = (year: number, month: number, day: number): number =>
+    new Date(year, month - 1, day, 12).getTime() * 1000;
+
+describe("savePath", () => {
+    it("files an original by year, month and day under the root", () => {
+        const f = file(12345, "IMG_0001.HEIC");
+        f.metadata.creationTime = noon(2026, 3, 1);
+
+        expect(savePath("/photos", f)).toBe(
+            "/photos/2026/2026-03/2026-03-01/2026-03-01.12345.HEIC",
+        );
+    });
+
+    it("dates an original by the date the user set, when there is one", () => {
+        const f = file(7, "a.jpg");
+        f.metadata.creationTime = noon(2026, 3, 1);
+        f.pubMagicMetadata = { editedTime: noon(1999, 12, 31) };
+
+        expect(savePath("/photos", f)).toBe(
+            "/photos/1999/1999-12/1999-12-31/1999-12-31.7.jpg",
+        );
+    });
+
+    it("takes the extension from the title it was uploaded with, not a new name", () => {
+        const f = file(8, "upload");
+        f.metadata.creationTime = noon(2026, 3, 1);
+        f.pubMagicMetadata = { editedName: "renamed.png" };
+
+        expect(savePath("/photos", f)).toBe(
+            "/photos/2026/2026-03/2026-03-01/2026-03-01.8.bin",
+        );
+    });
+});
+
 describe("ContentCache.open", () => {
     it("creates the cache directories with 0700 permissions", async () => {
         const { cache } = buildCache();
@@ -274,11 +311,17 @@ describe("ContentCache.original / thumbnail", () => {
 
     it("serves a file already present in the download directory without fetching", async () => {
         const downloadDirectory = join(root, "backup");
-        mkdirSync(join(downloadDirectory, "originals"), { recursive: true });
+        const day = join(downloadDirectory, "2026", "2026-03", "2026-03-01");
-        const backupPath = join(downloadDirectory, "originals", "1.jpg");
+        mkdirSync(day, { recursive: true });
+        const backupPath = join(day, "2026-03-01.1.jpg");
         writeFileSync(backupPath, "from-backup");
+        const f = file(1);
+        f.metadata.creationTime = noon(2026, 3, 1);
 
-        const { cache, source } = buildCache({ downloadDirectory });
+        const { cache, source } = buildCache({
+            downloadDirectory,
+            files: [f],
+        });
         await cache.open();
 
         const events: EnsureEvent["status"][] = [];
@@ -649,6 +692,7 @@ describe("ContentCache live photos", () => {
                 ]),
             ),
             cacheDirectory: cacheDir,
+            downloadDirectory: join(root, "photos"),
             getFile: (id) => [a.file, b.file].find((f) => f.id === id),
             // Room for one live photo, on a disk with plenty free.
             cacheOriginalsMaxBytes: size,

test/library/precache.test.ts

@@ -280,6 +280,7 @@ describe("Precache eviction integration", () => {
             pools: new RequestPools(),
             source,
             cacheDirectory: cacheDir,
+            downloadDirectory: join(root, "photos"),
             getFile: (id) => byID.get(id),
             statfs,
             cacheOriginalsMaxBytes: 25, // holds two 10-byte originals
@@ -348,6 +349,7 @@ describe("Precache preemption", () => {
             pools: new RequestPools({ contentConcurrency: 1 }),
             source,
             cacheDirectory: join(root, "cache"),
+            downloadDirectory: join(root, "photos"),
             getFile: (id) => byID.get(id),
             statfs: async () => ({ bsize: 1, bavail: 1_000_000_000 }),
             freeBelowBytes: 0,

test/library/read.test.ts

@@ -36,6 +36,7 @@ import {
     makeAlbumsAPI,
     makePhotosAPI,
     makeTimelineAPI,
+    type SavePathLookup,
     type TimelineGroup,
 } from "../../src/library/read.js";
 import { Library } from "../../src/library/index.js";
@@ -101,12 +102,19 @@ const file = (
 };
 
 // Build the three API objects over one fixed projection, the way `Library`
-// wires them over its live store.
+// wires them over its live store. Save paths are covered in
+// content-library.test.ts; these tests never ask for one.
 const apis = (records: DerivedRecords) => {
     const derive = () => records;
+    const saves: SavePathLookup = {
+        savePath: () => {
+            throw new Error("no save paths in these tests");
+        },
+        isLocal: () => false,
+    };
     return {
-        albums: makeAlbumsAPI(derive),
+        albums: makeAlbumsAPI(derive, saves),
-        photos: makePhotosAPI(derive),
+        photos: makePhotosAPI(derive, saves),
         timeline: makeTimelineAPI(derive),
     };
 };
@@ -209,6 +217,29 @@ describe("lib.photos", () => {
         expect("key" in photo.record()).toBe(false);
     });
 
+    it("byID exposes modifiedAt, hash and the year taken", () => {
+        // Mid-July, so the year is 2021 in every time zone.
+        const takenAt = Date.UTC(2021, 6, 15, 12);
+        const records = deriveRecords(
+            [collection(1)],
+            [
+                file(1001, 1, {
+                    metadata: {
+                        title: "IMG.jpg",
+                        fileType: "image",
+                        creationTime: micros(takenAt),
+                        modificationTime: micros(1_700_000_123_456),
+                        hash: "aGFzaA==",
+                    },
+                }),
+            ],
+        );
+        const photo = apis(records).photos.byID({ fileID: 1001 })!;
+        expect(photo.modifiedAt).toBe(ms(1_700_000_123_456));
+        expect(photo.hash).toBe("aGFzaA==");
+        expect(photo.year).toBe(2021);
+    });
+
     it("byID returns undefined for an unknown file id", () => {
         const records = deriveRecords([collection(1)], [file(1, 1)]);
         expect(apis(records).photos.byID({ fileID: 999 })).toBeUndefined();

test/library/records.test.ts

@@ -159,6 +159,7 @@ describe("deriveRecords: photo mapping", () => {
         expect("width" in rec).toBe(false);
         expect("height" in rec).toBe(false);
         expect("latitude" in rec).toBe(false);
+        expect("hash" in rec).toBe(false);
     });
 
     it("reads archived and hidden from private magicMetadata.visibility", () => {

test/packaging/build-context.test.ts

@@ -9,8 +9,10 @@
 // Excluding too much: Prettier 3 reads `.gitignore` as a default ignore file,
 // so dropping it from the context silently changes which files the lint
 // phase's prettier check looks at compared to `make fmt-check` on the host.
+// And without `.git`, a `docker build .` given no `VERSION` build arg cannot
+// derive the version (`script/version`) and stamps `package.json`'s instead.
 //
-// Neither shows up as a build failure, so they are asserted here.
+// None of these shows up as a build failure, so they are asserted here.
 import { describe, expect, it } from "vitest";
 import { existsSync, readFileSync } from "node:fs";
 import { fileURLToPath } from "node:url";
@@ -47,6 +49,11 @@ describe(".dockerignore", () => {
         expect(dockerignore).not.toContain(".gitignore");
     });
 
+    it("leaves .git in the build context for the version", () => {
+        expect(dockerignore).not.toContain(".git");
+        expect(dockerignore).not.toContain(".git/");
+    });
+
     // BuildKit lets a `Dockerfile.dockerignore` shadow the root one; such a
     // file would silently give the build a different, unreviewed context —
     // and eslint's flat config does not ignore dot-directories, so a stray

test/packaging/version.test.ts

@@ -0,0 +1,136 @@
+// `script/version` prints the version `script/build` stamps into
+// `dist/package.json`, which is what `quak --version` reports from a build.
+// A `docker build .` of a clone is given no `VERSION` build arg, so the
+// version has to come from the `.git` in its context: the tag on a tagged
+// commit, otherwise the short commit. A checkout with `.git` that still yields
+// no usable version must fail the build, not ship a version nobody can trace
+// back to its commit.
+//
+// Each test copies the script into a fresh directory, which the script then
+// treats as the checkout, and executes it there.
+import { afterEach, describe, expect, it } from "vitest";
+import { execFileSync, spawnSync } from "node:child_process";
+import {
+    chmodSync,
+    copyFileSync,
+    mkdirSync,
+    mkdtempSync,
+    rmSync,
+    writeFileSync,
+} from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { fileURLToPath } from "node:url";
+
+const repoRoot = fileURLToPath(new URL("../../", import.meta.url));
+
+let checkout = "";
+
+afterEach(() => {
+    rmSync(checkout, { recursive: true, force: true });
+});
+
+// A checkout holding the script and a package.json that declares 0.0.0, with
+// no .git yet.
+const makeCheckout = (): void => {
+    checkout = mkdtempSync(join(tmpdir(), "quak-version-"));
+    mkdirSync(join(checkout, "script"));
+    copyFileSync(
+        join(repoRoot, "script/version"),
+        join(checkout, "script/version"),
+    );
+    chmodSync(join(checkout, "script/version"), 0o755);
+    writeFileSync(join(checkout, "package.json"), '{ "version": "0.0.0" }\n');
+};
+
+// git in the checkout, with an identity and no commit signing, whatever the
+// host's own git config says.
+const git = (...args: string[]): string =>
+    execFileSync(
+        "git",
+        [
+            "-c",
+            "user.name=quak",
+            "-c",
+            "user.email=quak@example.invalid",
+            "-c",
+            "commit.gpgsign=false",
+            ...args,
+        ],
+        { cwd: checkout, encoding: "utf-8", stdio: ["ignore", "pipe", "pipe"] },
+    ).trim();
+
+const makeCommittedCheckout = (): void => {
+    makeCheckout();
+    git("init", "-q");
+    git("add", "package.json");
+    git("commit", "-q", "-m", "first");
+};
+
+// Runs the script with nothing in its environment but PATH and, when given,
+// VERSION.
+const runVersion = (version?: string) =>
+    spawnSync(join(checkout, "script/version"), {
+        cwd: checkout,
+        encoding: "utf-8",
+        env: { PATH: process.env.PATH, VERSION: version },
+    });
+
+describe("script/version", () => {
+    it("prints the short commit of an untagged commit", () => {
+        makeCommittedCheckout();
+        expect(runVersion().stdout.trim()).toBe(
+            git("rev-parse", "--short", "HEAD"),
+        );
+    });
+
+    it("prints the tag of a tagged commit", () => {
+        makeCommittedCheckout();
+        git("tag", "v1.2.3");
+        expect(runVersion().stdout.trim()).toBe("v1.2.3");
+    });
+
+    // script/docker and script/cibuild pass the version they resolve on the
+    // host as the VERSION build arg.
+    it("prints the VERSION it is given over what git would derive", () => {
+        makeCommittedCheckout();
+        expect(runVersion("x").stdout.trim()).toBe("x");
+    });
+
+    // `--build-arg VERSION=` must not stamp an empty version.
+    it("treats an empty VERSION as unset", () => {
+        makeCommittedCheckout();
+        expect(runVersion("").stdout.trim()).toBe(
+            git("rev-parse", "--short", "HEAD"),
+        );
+    });
+
+    // A source tarball has no .git: it keeps the version package.json
+    // declares, and must still build.
+    it("prints package.json's version where there is no .git", () => {
+        makeCheckout();
+        const result = runVersion();
+        expect(result.status).toBe(0);
+        expect(result.stdout.trim()).toBe("0.0.0");
+    });
+
+    // A repository with no commits stands in for any .git that git cannot
+    // describe: git missing from the image, or refusing to read the checkout.
+    it("fails where .git yields no version", () => {
+        makeCheckout();
+        git("init", "-q");
+        const result = runVersion();
+        expect(result.status).not.toBe(0);
+        expect(result.stdout).toBe("");
+    });
+
+    it.each(["dev", "unknown"])(
+        "fails where there is .git and the version is %s",
+        (version) => {
+            makeCommittedCheckout();
+            const result = runVersion(version);
+            expect(result.status).not.toBe(0);
+            expect(result.stdout).toBe("");
+        },
+    );
+});

tsconfig.json

@@ -18,5 +18,5 @@
         "sourceMap": true,
         "resolveJsonModule": true
     },
-    "include": ["src/**/*", "bin/**/*"]
+    "include": ["src/**/*", "bin/**/*", "examples/**/*"]
 }

yarn.lock

@@ -702,6 +702,11 @@
     loupe "^3.1.2"
     tinyrainbow "^1.2.0"
 
+"@xmldom/xmldom@^0.9.10":
+  version "0.9.12"
+  resolved "https://registry.yarnpkg.com/@xmldom/xmldom/-/xmldom-0.9.12.tgz#1f84c07cb95ccf28202299f77b5fd7fc257151e8"
+  integrity sha512-5AXjrcMClTryPe9LgZrygpB1lj7s0S9E0+W+AHaVKAVyHanafK86iPSvG5xHVSp/jC+VH1UXu0TAEmY279xH7A==
+
 acorn-jsx@^5.3.2:
   version "5.3.2"
   resolved "https://registry.yarnpkg.com/acorn-jsx/-/acorn-jsx-5.3.2.tgz#7ed5bb55908b3b2f1bc55c6af1653bada7f07937"
@@ -1002,10 +1007,12 @@ esutils@^2.0.2:
   resolved "https://registry.yarnpkg.com/esutils/-/esutils-2.0.3.tgz#74d2eb4de0b8da1293711910d50775b9b710ef64"
   integrity sha512-kVscqXk4OCp68SZ0dkgEKVi6/8ij300KBWTJq32P/dYeWTSwK41WyTxalN1eRmA5Z9UU/LX9D7FWSmV9SAYx6g==
 
-exif-reader@2.0.3:
+exifreader@4.46.0:
-  version "2.0.3"
+  version "4.46.0"
-  resolved "https://registry.yarnpkg.com/exif-reader/-/exif-reader-2.0.3.tgz#259997735080bc6bb959c37b32c60f004ec4391d"
+  resolved "https://registry.yarnpkg.com/exifreader/-/exifreader-4.46.0.tgz#b6216eae512997587c45114f972cc14ca979205f"
-  integrity sha512-zFbQvguwT9JkqyYhR7pjE1Yn8SagwaGLNRU0Oh14xFa1paSf5Gzxn4gxgk0XhnudI0UIqU+HgnBX93+nva592A==
+  integrity sha512-ksHTpjXKWzbckY+bYlGaomG0EobHJkaMWLg5OPbzlTPda04F2dfrDfYSz8iAPp/kXYUSQdINBVI6nCAjQ8PQ/Q==
+  optionalDependencies:
+    "@xmldom/xmldom" "^0.9.10"
 
 expect-type@^1.1.0:
   version "1.3.0"