Author SHA1 Message Date
JakeBreath 7ca84f4fea Point desktop updates at the Gitea release feed
CI / Backend tests (push) Successful in 2m41s
CI / Frontend build & lint (push) Successful in 25s
The updater now resolves the newest non-draft desktop-v* release through the
Gitea API at check time (J621_UPDATE_REPO, lowercase because the API path is
case-sensitive), picks the platform's latest*.yml asset and uses that release
as a generic electron-updater feed; J621_UPDATE_URL still overrides
everything. Verified against the live API: release picked, yml fetched,
artifact HEAD 200.

electron-builder's publish.url is now metadata only (still needed so the
build emits latest*.yml). Docs updated: CD release assets are the feed, the
website /desktop/ feed only matters for installs before 0.1.2.
2026-09-23 22:00:49 -05:00
JakeBreath 5f237aa2e3 Widen the header and collapse blacklist listings by default
- the header row is no longer capped at 1600px, so the nav hugs the left
  edge and the status/account controls the right; main content and footer
  keep their width
- OnlinePage's blacklist chips and the Followed page's blacklisted-tag
  cloud start collapsed behind a count toggle (N tags / N entries)
2026-09-23 21:57:48 -05:00
JakeBreath 90ba2ecff7 Bump the desktop app to 0.1.2 2026-09-23 21:40:43 -05:00
JakeBreath 73bf4f9e38 Rework the metadata modal: fullscreen, inline sections, bigger previews
- near-fullscreen panel (up to 1400px / 92vh) with two independently
  scrolling columns; the left preview uses self-start so its border hugs the
  image instead of stretching to the modal height
- J-ID matches render as a larger tile grid (was 48px rows) and IQDB
  candidates get bigger tiles too
- Link to e621 post and Custom metadata are shown inline instead of behind
  tabs, each with its own heading
- new ThumbImage component: spinner while loading and a broken-image icon on
  error, with alt text removed so a pending tile never reads as 'J-7786'
2026-09-23 21:39:48 -05:00
JakeBreath 7c2569522f Make thumbnail generation atomic and warm it on import
- write thumbnails to a .part file and os.replace() them, so concurrent
  requests never read a half-written JPEG
- a stale thumbnail plus a vanished source no longer raises through the
  request (getmtime on a missing file returned 500); it falls back cleanly
- ensure_thumbnail(item) warms the preview when a file is indexed, keeping
  image decoding out of the request path
2026-09-23 21:35:09 -05:00
JakeBreath a62195ffce Re-sign visual-match thumbnails on every detail fetch
Match rows stored a signed URL minted when the scan ran, so it aged out (or
used the pre-stable signing scheme) and the modal showed broken tiles even
after legacy signatures were fixed. Rows now carry item_id/j_id and the
detail serializer mints a fresh thumbnail URL per request; matches whose
item no longer exists are dropped.
2026-09-23 21:33:20 -05:00
JakeBreath c73a81a5f4 Fix 500 on legacy signed URLs
A TimestampSigner value is an HMAC over 'payload:timestamp', so a plain
Signer's HMAC check accepts it and the embedded timestamp then reached the
JSON decoder, raising JSONDecodeError (not BadSignature) and surfacing as a
500. That broke every stored visual-match thumbnail URL minted before the
stable scheme, so the J-ID match tiles never loaded on prod.

Detect the legacy shape by its extra separator and verify it with
TimestampSigner; malformed input returns None instead of raising.
2026-09-23 21:31:34 -05:00
JakeBreath af378e7d71 Update the roadmap for cacheable media and session-only completions
CI / Backend tests (push) Successful in 2m29s
CI / Frontend build & lint (push) Successful in 25s
2026-09-23 18:29:49 -05:00
JakeBreath 59f397a96c Show indexed uploads as session notifications, add a Failed column
The board rendered every persisted completed row with a dismiss button and a
dismiss-all that sent the whole column to the 1000-id bulk endpoint. Now that
the backend deletes completed rows and reports them through the status feed:

- the Auto-uploaded & Indexed column renders the live feed only; a reload or
  leaving the page forgets them, and there is nothing to dismiss (just a
  client-side clear)
- duplicates complete during staging and get their card immediately from the
  upload response
- failures get their own persisted column with per-card retry and discard
- bulk discard chunks requests at 500 ids, so backlogs over the server's
  1000-id cap are still removable in one action
2026-09-23 18:28:35 -05:00
JakeBreath 9ababb8b48 Stop storing completed uploads; announce them through a live feed
Every auto-matched, duplicate or manually resolved upload left a completed
TempUpload row on the board until it was dismissed by hand, so the rows
accumulated without bound and the bulk dismiss (capped at 1000 ids) failed
once there were more. The original app never stored these: they are
notifications, not records.

- complete_temp_upload now appends {filename, J-ID, resolution, post} to a
  bounded recent_completions feed on UploadRun and deletes the staged row
- staging duplicates never create a board record either; the create response
  carries the J-ID and preview so the SPA can show the card immediately
- status_payload returns the feed (newest first, signed thumbnails) for the
  live board; finalize_round counts deleted matches in processed
- resolve/link-bulk return synthetic completion payloads
- migration 0012 adds the field and purges the existing completed backlog
  (and any stray staged files) on deploy
2026-09-23 18:24:38 -05:00
JakeBreath 37085c5dac Make media URLs stable and cacheable, add real image thumbnails
Signed media URLs embedded the current second (TimestampSigner), so every
API response re-minted every raw/thumbnail/staged URL and the browser
re-downloaded each file on every poll or navigation. Responses also carried
no cache headers at all.

- sign with a plain Signer plus a bucket-quantized exp (7d TTL, 24h bucket),
  so a URL is byte-identical across responses and rotates once a day; legacy
  TimestampSigner URLs stay accepted for one release
- add a v=<md5> version parameter to library media URLs so replacing a file
  under the same J-ID (the optimize flow) busts caches exactly when needed
- serve_file now sends ETag/Last-Modified and a private Cache-Control and
  answers conditional requests with 304; library media gets max-age 6d +
  immutable, staged/similarity files 1h
- build cached 480px JPEG thumbnails for images (Pillow, keyed by MD5 under
  MEDIA_ROOT/thumbs) instead of serving full-size originals through the
  thumbnail endpoint; the library grid uses thumbnail_url for images too
2026-09-23 18:13:52 -05:00
JakeBreath ecac4cb8b4 Fix the release asset upload for names with spaces
CI / Backend tests (push) Successful in 2m4s
CI / Frontend build & lint (push) Successful in 23s
curl exit 3 (malformed URL) on 'J621 Setup 0.1.1.exe': percent-encode the
asset name in the query string.

Also replace assets instead of skipping them on re-runs: NSIS builds are
not bit-reproducible, so latest.yml/latest-linux.yml must reference the
installers produced by the same run. Existing assets are deleted by id
before the fresh upload.
2026-09-23 07:14:19 -05:00
JakeBreath 2eb7af0d41 Fix the CD wine build and add per-part toggles
CI / Backend tests (push) Successful in 2m13s
CI / Frontend build & lint (push) Successful in 20s
The Windows NSIS step failed under wine for two reasons: no X display
(nodrv_CreateWindow) and missing 32-bit libraries (failed to load
syswow64\ntdll.dll). The job now installs xvfb + wine32:i386 and runs the
desktop build under xvfb-run, with Gecko/Mono lookups disabled.

Also add `images` and `desktop` dispatch inputs so either half of the CD
can be skipped (e.g. desktop-only or images-only releases).
2026-09-23 00:06:55 -05:00
JakeBreath 7cecfeabc6 Use a minimal registry PAT and the job token for releases
CI / Backend tests (push) Successful in 2m23s
CI / Frontend build & lint (push) Successful in 22s
Gitea's container registry rejects the automatic job token
(go-gitea/gitea#23642 is still open), so the image push keeps a PAT with
only the write:package scope; a preflight step fails clearly when the
REGISTRY_USER/REGISTRY_TOKEN secrets are missing. Release creation needs
no PAT: the desktop job asks for contents: write on the job token.
2026-09-22 23:35:55 -05:00
JakeBreath ed6178d12e Make Actions token permissions explicit
The automatic job token creates the desktop release, so the CD desktop job
asks for contents: write; the images job keeps contents: read and asks for
packages: write so the job token can stand in for the scoped registry PAT.
CI stays read-only. The registry token itself remains a write:package-only
PAT (verified login + pull).
2026-09-22 23:30:37 -05:00
JakeBreath 8f9656ac0e Add the manual CD release workflow
One dispatch builds and pushes both images and builds the desktop packages
into a Gitea release (desktop-v<version>, installers + latest*.yml attached,
idempotent on re-run). The live update feed stays a deploy-host operation:
CI has no SSH key for jakerasp, so push_desktop.sh --no-build remains the
way to publish it.

Repo secrets REGISTRY_USER/REGISTRY_TOKEN are set, so the image push uses
the Gitea registry credentials directly.
2026-09-22 23:23:05 -05:00
JakeBreath 72fc42217f Fix CI for the user-scoped runners
- ci.yml: connect to the test MariaDB as root so Django creates the test
  database itself (no client install/grant step), and drop actions/cache
  (cache: pip/npm): Gitea's cache service hangs the job on restore/save.
- publish.yml: prefer the REGISTRY_USER/REGISTRY_TOKEN secrets (as on other
  repos) and fall back to the automatic Actions token.
- AGENTS.md: note the CI layout, the runner labels and the cache caveat.
2026-09-22 23:12:56 -05:00
JakeBreath d9c1e9e521 Add CI and manual image publishing workflows
CI / Backend tests (push) Successful in 12m54s
CI / Frontend build & lint (push) Failing after 4m59s
- .gitea/workflows/ci.yml: on every push/PR, run Django checks + the full
  backend suite against MariaDB/Redis services and the frontend
  lint/type-check/build. Runs on the nitro-ci runner (ubuntu-latest).
- .gitea/workflows/publish.yml: manual dispatch; multi-arch build+push of
  both images as :latest and :<short-sha> with GIT_HASH baked in.
- push_*.sh: non-interactive registry login for CI (REGISTRY_USER/
  REGISTRY_TOKEN) and a PLATFORMS override.
2026-09-22 22:32:37 -05:00
JakeBreath e2697c0a78 Stop throttling signed media and ease the browser's e621 queue
Signed media URLs are fetched by <img>/<video> tags without an
Authorization header, so they were charged to the anonymous 120/min
bucket: past that, galleries and the fish-greeting download got 429 JSON
instead of image bytes. The raw/thumbnail/staged-file/similarity-file
actions are now exempt, and THROTTLE_ENABLED=false removes the general
anon+user limits for private/tailnet deployments (login/register/proxy
guards stay).

The SPA's e621 client also stops self-throttling so hard: 1s gap between
browsing calls (2.5s for the stricter IQDB endpoint) and a 15s cooldown
instead of 60s when e621 answers 429.
2026-09-22 22:32:37 -05:00
JakeBreath 474403ffe2 Upload updates 2026-09-21 09:01:01 -05:00
JakeBreath 98025e9e6d Prune old installers from the remote feed on push
rsync without --delete left every previous version on the server (the
screenshot showed 0.1.0 and 0.1.1 side by side). The push now removes
non-current installers over ssh first, so the remote feed mirrors the local
one whether the transfer uses rsync or tar.
2026-09-20 21:39:43 -05:00
JakeBreath 3183a3bece Prune old desktop builds from release/ on every build
build_desktop.sh now reads the version first and removes anything in
desktop/release/ that is not that version (plus the regenerated unpacked
trees), so a version bump never leaves old installers lying around — the
same rule push_desktop.sh applies to the feed.
2026-09-20 21:16:56 -05:00
JakeBreath 041a9d4471 Keep desktop builds and the feed to one version
Both scripts now read the version from desktop/package.json: the build
report and checksums only cover the current version's artifacts, the feed
copy ignores older files, and publishing prunes previous installers from
the feed (latest*.yml only ever points at the current one).
2026-09-20 21:11:19 -05:00
JakeBreath d84fdd0e98 Bump the desktop app to 0.1.1
The icon set, e621 referrer fix and setup-screen corrections shipped after
0.1.0, and the updater compares versions, so installed 0.1.0 builds would
never have seen them.
2026-09-20 21:09:42 -05:00
JakeBreath c992a63b8f Fix e621 images and the setup screen in the desktop shell
e621's CDN answers cross-site image loads that carry no Referer with a 403
(Chromium sends none from a custom-scheme page, then blocks the response as
ORB), so images never appeared in the desktop app. The main process now
attaches an e621 referrer to requests for its hosts.

The shell also answers /api, /admin, /static and /health with a 404 JSON
instead of the SPA fallback — that fallback made the setup screen's empty-URL
connection test report "Connected" against the shell itself. The setup screen
is now desktop-aware (no same-origin option, no "Use this server", clearer
copy), and the smoke test runs against a throwaway profile and covers both
regressions.
2026-09-20 21:08:56 -05:00
JakeBreath bd2417aff8 Keep a broken Redis from 500ing the whole API
Redis backs the DRF throttles, and the stock RedisCache raises inside the
throttle check when Redis is unreachable or refusing writes (a failed RDB
snapshot disables writes by default) — turning a cache problem into a
blanket 500, which is exactly what took prod down. ResilientRedisCache
treats backend failures as cache misses, logs the first one per worker, and
lets rate limits degrade until Redis is back.
2026-09-20 21:08:38 -05:00
JakeBreath 70d7a4f606 Default push_desktop.sh to the jakerasp deploy checkout
The remote feed copy now happens by default (overridable with
J621_DESKTOP_FEED_HOST or --host, skippable with --local), so a release is
one command on the build machine.
2026-09-20 20:25:37 -05:00
JakeBreath f4b534aa09 Let push_desktop.sh copy the feed to a remote deploy checkout
--host user@server:/path syncs deploy/data/desktop to the same path on the
server with rsync, falling back to tar over ssh when the server has no
rsync. A failed transfer now exits non-zero instead of claiming the feed is
live.
2026-09-20 20:22:59 -05:00
JakeBreath 1c6735cde8 Include the desktop origin in generated CORS settings
The backend only allows app://j621 through CORS_ALLOWED_ORIGINS, so
gen_env.sh now always writes that origin (plus the split-deploy frontend
when one is given) and --update keeps hand-added origins instead of
overwriting the list.
2026-09-20 19:44:56 -05:00
JakeBreath 7f64c6b635 Ship a full icon set so KDE resolves the launcher icon
The Linux packages only installed a single 1024x1024 hicolor PNG, and
Plasma's icon lookup returns nothing for a lone oversized icon — confirmed
with kiconfinder6 under Papirus-Dark. Generate 16-1024 px PNGs from the
favicon and point electron-builder at the directory so every standard
hicolor size is installed.
2026-09-20 19:39:19 -05:00
JakeBreath 69f324ead7 Add a desktop build script for manual releases
deploy/build_desktop.sh builds the Arch, Debian and Windows packages into
desktop/release/ without publishing anything, lists what it produced with
sizes and SHA-256 sums for release notes, and prints the suggested Gitea
tag. Installing locally, handing the files out and attaching them to a
Gitea release all stay manual; push_desktop.sh remains the update-feed
publisher.
2026-09-20 18:08:19 -05:00
JakeBreath e839c84bf0 Make the desktop smoke test mode-aware
The dev path (J621_DEV_SERVER) has the Vite origin and no /setup screen, so
the expectations now follow the mode instead of reporting false failures.
2026-09-20 17:34:01 -05:00
JakeBreath 7c39383655 Wire desktop updates through a generic feed
electron-builder now publishes latest-linux.yml / latest.yml and embeds
app-update.yml plus package-type, so electron-updater runs pacman -U or
dpkg -i through pkexec for packages and updates the per-user NSIS install
without elevation. The app checks only when asked (menu item), asks before
downloading and before installing, and J621_UPDATE_URL overrides the feed
for tests or forks.

deploy/push_desktop.sh builds and publishes the artifacts to
deploy/data/desktop, which the frontend nginx mounts read-only at
/desktop/. Verified detection and up-to-date handling against a local feed
with the packaged Arch build.
2026-09-20 17:32:41 -05:00
JakeBreath 84ec431441 Package the desktop app for Arch, Debian and Windows
electron-builder produces j621-desktop_*_amd64.deb,
j621-desktop-*.pkg.tar.zst (zstd, lean Arch dependencies instead of the
Electron 2 era default set) and a per-user NSIS installer cross-built with
wine. The launcher entry and Electron's desktopName now agree on
io.j621.desktop so window association works, and package metadata points
at the repository homepage and the non-commercial licence.
2026-09-20 17:18:53 -05:00
JakeBreath cf129714be Add an Electron desktop shell for the SPA
desktop/ serves the normal frontend build over a privileged app://j621
scheme, so localStorage, OPFS, Web Workers, WebCodecs and history routing
behave exactly like Chrome. Development points at the Vite dev server;
`npm run smoke` runs headless Electron and checks the bundled app.

External links open in the system browser, and download navigations
(?download=1 or media URLs) are rerouted through webContents.downloadURL,
since preventing them cancels the download. The setup screen names the
shell's origin when the connection test fails, and the deploy docs list
app://j621 for CORS_ALLOWED_ORIGINS.
2026-09-20 17:07:22 -05:00
JakeBreath 3a07481dfc Wait for all uploads, then batch MD5 -> visual -> IQDB with bulk links
Batching must not start while files are still being uploaded, and the MD5
phase must move a whole chunk at once instead of one resolve per file:

- the upload queue drains completely first (failed uploads included) before
  any matching starts;
- phase 1 asks e621 for every md5 (75 per posts.json request), builds the
  md5 -> post map from the response, and sends the matches to the new
  POST /api/uploads/link-bulk/ action, so a whole 75-file chunk moves into
  Indexed in a single board update;
- link-bulk indexes the staged file directly when the post's MD5 matches
  (identical bytes), so there is no per-file download round trip;
- phase 2 runs local visual similarity for whatever stayed pending, phase 3
  the IQDB queue.

Verified end to end with real e621 files: one md5 query for the batch, one
link-bulk call, both matching files flipped to Indexed together, then the
visual and IQDB phases. 23 library tests green (link-bulk, visual phase,
deferred visual matching).
2026-09-19 11:10:24 -05:00
JakeBreath e2bf1c457f Make upload processing phase-based: MD5 -> visual -> IQDB over the batch
Uploads were doing md5 + local visual matching inside the upload request
(backend create) while the frontend later ran its own e621 MD5 pass, so the
pipeline looked interleaved per file. Now every step is a phase applied to
the whole batch in order:

1. upload (fast: md5 + exact-duplicate check only),
2. e621 MD5 lookup, 75 md5: metatags per posts.json request,
3. local visual similarity, one file at a time via the new
   POST /api/uploads/<id>/visual-match/ action,
4. IQDB through the existing serial queue.

The board shows the active phase with its own progress bar (e621 MD5 in
peach, visual in lavender, IQDB in teal) and every step updates the staged
list as it lands. Verified from a headless run: one batched posts.json
request for 10 files, then 10 visual-match calls, then IQDB.
2026-09-19 10:25:57 -05:00
JakeBreath e62d7af42f Live-updating upload board, self-clearing tiles, original MD5 batch size
- MD5 auto-match now sends 75 md5: metatags per posts.json query, the same
  batch size the original J621-Django app used (was 20); the limit cap no
  longer truncates batches.
- every settled upload is upserted into the staged list right away, so the
  Pending / Visual Similarity / Auto-uploaded columns move as files land
  instead of waiting for the whole batch (auto-matched resolves and IQDB
  results use the same path).
- finished upload tiles fade out and remove themselves ~2s after completing;
  failures stay until cleared. Batch counters are tracked separately from the
  visible tiles so the header and progress bar stay accurate as tiles vanish.
- AGENTS.md now points at the original Django app for reference behavior.

Verified live with a headless upload run: the columns showed the new files
immediately and the 8 tiles were gone ~3s after finishing.
2026-09-19 01:18:50 -05:00
JakeBreath 9641862515 Back off from e621 rate limits and pace requests more conservatively
e621 intermittently answers 429 to the IQDB endpoint; browsers hide that
status behind CORS ('Access-Control-Allow-Origin missing'), so the SPA
cannot read it. Treat every network-level failure as a possible rate limit
and pause all e621 traffic for a minute. The cooldown is shared through
localStorage so extra tabs respect it, requests are spaced 1.5s apart
instead of 1s, user-cancelled requests do not trigger a cooldown, and the
upload queue waits the cooldown out with a countdown instead of looking
stuck.

Server side: the per-process e621 gap goes from 0.5s to 1s so two gunicorn
workers cannot together exceed e621's 2/s hard limit.
2026-09-19 00:37:32 -05:00
JakeBreath b96c311235 Fix cross-origin staging URLs, e621 UA format and IQDB queue stalls
The dev Vite proxy rewrites the request Host to 127.0.0.1:8000, so the
backend's absolute signed file URLs pointed at a different origin than the
SPA (localhost:5173). Images tolerated it, but the auth'd fetch that reads
the staging blob for IQDB was blocked ('Cross-Origin Request Blocked') and
every similarity check died before reaching e621.

apiUrl() now keeps API-built absolute URLs on the page's origin whenever the
SPA is in same-origin mode (dev proxy, deploy nginx) and leaves them
absolute when an explicit backend URL is configured. All consumers use it:
staging previews and the bulk modal, library cards, optimizer (range sniff +
worker), IQDB card, delete page, similar page.

e621 identification now follows the documented 'App/version (developer)'
form: server-side requests send 'J621/<hash> (JakeBreath)' and the browser
_client gets the same string, with the hash baked into the frontend image
(GIT_HASH build arg; guarded at runtime so the dev server still works).

IQDB stalls: requests now time out after 20s (a hung fetch used to block the
serialized e621 queue forever), and all checks run through one serial drain
so repeated 'Check similarity' clicks can no longer start overlapping runs
that re-download the same staging blobs. Auth/rate-limit/timeout/network
failures stop the queue with the reason and a retry button instead of
grinding through the rest.

Verified live: staged file URL is same-origin through the proxy and fetches
200 through it.
2026-09-19 00:30:42 -05:00
JakeBreath 2f613b7027 Stop Firefox from squashing upload tiles and show IQDB queue progress
The upload grid was its own scroll container (max-height + overflow on the
same element). Firefox sizes auto grid rows to min-content in that setup,
so every tile collapsed to its footer height and the image was clipped to a
wide strip; Chromium sizes them to max-content, which is why this only
showed up in the user's browser. auto-rows-max pins rows to max-content in
both; verified with headless Firefox screenshots and Chromium measurements
(60 tiles render 152x194 each, 1845px of content in a 639px scroller).
The box is now 70vh so it behaves as a proper fixed gallery area with its
own scrollbar instead of shrinking with the item count.

IQDB progress: the page header now shows a live 'Checking IQDB — x/y'
counter with a bar while background checks run, so the queue is visible
without opening the per-file modal (which keeps its spinner, candidates
and per-file errors).
2026-09-19 00:06:05 -05:00
JakeBreath bc7494e7be Show IQDB checks in the metadata modal and run them for visual matches
The modal held a snapshot of the staged upload, so IQDB results that landed
from the background check queue never appeared until it was closed and
reopened — the only hint a check was running was the e621 request history.
It now follows the live uploads query, so candidates, progress and errors
show up in place.

Related gaps fixed along the way:
- files flagged by the local visual-similarity check were skipped by the
  IQDB pass entirely (only 'pending' files were checked), so their modal
  could only ever show 'already in your library'; unresolved files of both
  statuses are now checked, and the check button shows on visual-match
  cards too;
- a check with no candidates posted nothing, leaving 'never checked' and
  'checked, no match' indistinguishable; results are stored even when
  empty and the modal now says which one it is;
- per-file failures surface in the modal instead of being swallowed, the
  modal shows a spinner while the query runs and a check now/re-check
  button, and auto-runs skip files already checked (and videos, since IQDB
  is image-only).

Backend production code unchanged; tests pin the empty-result recording
(18 library tests, full suite 56 green).
2026-09-18 23:42:45 -05:00
JakeBreath b076903ecd Show live progress while the bulk rating tool moves files
A 200-file bulk move is one long server-side copy+index chain per file, so
the old single request sat on a spinner the whole time (and got uncomfortably
close to the 120s proxy timeout). The modal now resolves the selection in
chunks of 8:

- footer switches to a progress bar with 'n moved / done / total / percent'
  while running, and the header explains that files are being indexed;
- the rating pills, selection actions, grid and close button are locked
  while it runs so progress can't be lost by accident;
- failures are collected with their filenames, the modal stays open for a
  summary, and 'Retry failed' re-selects only the files that are still
  pending; a clean run still auto-closes with a toast.
2026-09-18 23:09:42 -05:00
JakeBreath a761def65e Scrollable upload grid and bulk rating for the pending backlog
Upload board:
- the tile grid no longer re-sorts itself as files finish (that reshuffled
  the list under the cursor); it keeps insertion order, uses auto-fill tiles
  of ~150px so they hold a readable size, scrolls inside a 60vh area and no
  longer chains the page scroll (overscroll-contain);
- the files currently in flight are pinned in a small live strip above the
  grid (name, percent, bar) so progress stays visible while the grid is
  scrolled with hundreds of tiles.

Bulk rating: a 'bulk rate' button in the Pending & Unmatched header opens a
large modal with Safe/Questionable/Explicit pills, a tickable thumbnail grid
(Select all / Clear) and one confirm that moves every selected upload into
the library with that rating. Backed by POST /api/uploads/resolve-bulk/
(temp_ids + rating, own rows only): each staged file is resolved as a custom
entry (keeps its staged tags/notes), and already-completed or foreign ids are
reported per entry instead of failing the whole batch. Built for the
358-file backlog.

Tests: 4 bulk-resolve tests (resolution with the rating, input validation,
foreign ids untouched, mixed completed+pending) — full backend suite 53
green. Verified live end to end: staged a file, bulk-resolved it as 'q', saw
J-96 created with that rating, then removed the item, temp row and test
token.
2026-09-18 21:50:22 -05:00
JakeBreath 39307cb141 Unpaginate staged uploads and make big upload batches visible
The upload board partitions /api/uploads/ into Pending / Visual similarity /
Auto-uploaded, but the endpoint was paginated at 48 — a 69-file batch
silently lost 21 entries, and the similarity sweep (which reads the same
list back after uploading) only ever saw the first page. The staged-upload
list is now unpaginated: it is a transient per-user set, still limited to
the caller's rows and the uploader role. The page takes a plain array.

Watching progress with dozens of files was also poor:
- the queue uploads three files at a time instead of strictly one at a time;
- the Uploads section now shows a batch bar and 'n/m uploaded · x%' next to
  the count, so the overall progress never scrolls out of sight;
- entries are ordered active-first (uploading, queued, failed, done) so the
  file being uploaded is always at the top of the grid;
- tiles are larger (4 columns at lg instead of 5);
- the header reads 'Uploading n/m…' and 'Checking n file(s) against IQDB…'
  instead of a bare spinner.

Tests: staged-upload list unpaginated past 48, per-user, uploader-only
(3 new; full suite 49 green). Live-checked the bare-array response.
2026-09-18 19:58:17 -05:00
JakeBreath 1adb761c8d Fix following past 48 entries and make the e621 page size configurable
Follow lists were paginated at the API default of 48, but the SPA treats
them as complete sets: the tag/pool toggles read their state from page one
(so the 49th follow looked unfollowed and its spinner waited for a page that
could never contain it) and the Followed page rendered only 48 cards while
showing that as the count. Both follow endpoints are now unpaginated — they
are per-user sets and still restricted to the caller's rows — and the three
consumers take plain arrays.

Post visibility: the old J621-Django online view fetched limit=320 (e621's
maximum) while ours hard-coded 48, and fetchPostsByIds capped id batches at
100. The Online browser now has a 'Posts per page' setting (48/100/200/320)
in its sidebar, mirrored in Account -> Browsing preferences, stored per user
as e621_per_page and also used for pool loading; the id-batch cap is raised
to 320.

Tests: follow list shape/isolation (4) and preference validation/merge (3)
added; the full backend suite is 46 green. Live-checked the array response
shape and the preference bounds (200 accepted, 500 rejected).
2026-09-18 19:06:18 -05:00
JakeBreath d16a77907a Keep the online detail's download state on its own post
Two stale-state bugs came from react-router reusing the detail component
between posts (parent/child links hit the same /detail/<id> route):

- the previously viewed post's download panel kept rendering, so a freshly
  opened post could claim 'Downloaded to the library J-xx'. The task view is
  now gated on the task's post_id as well, and DetailPage keys the detail
  views per item — component state (download panel, delete confirmation,
  optimizer modal) cannot survive a post change any more.
- 'Your last download for this post finished' appeared on every revisit. It
  now only shows when this visit actually saw the download running (derived
  state, set during render), which still reports a re-attached download
  finishing while staying quiet on later visits; the 'In library' button
  remains the persistent indicator.

Library detail gets the same per-item key, so its delete confirmation and
optimizer modal reset between items too.

tsc, oxlint and the build are clean (the derived-state pattern was chosen
over a ref read in render / setState-in-effect, both flagged by the linter).
2026-09-18 17:52:50 -05:00
JakeBreath a17dd5a4ef Keep secrets and runtime data out of the Docker images
The build context is the repository root and there was no .dockerignore, so
'COPY backend/ ./' swept backend/venv (327 MB), backend/media (the actual
library, 224 MB), backend/logs, backend/staticfiles and backend/.env
(SECRET_KEY plus the database password) into the backend image: 1.43 GB per
architecture, including secrets headed for the registry. The frontend build
stage also copied the host's node_modules over the fresh install.

- Root .dockerignore excludes .git, virtualenvs, __pycache__, db.sqlite3,
  logs/staticfiles, .env files, media/, node_modules, dist and the deploy
  runtime state (data/, tailscale-state/).
- The backend Dockerfile now asserts the context is clean (.env, venv,
  media/library, db.sqlite3 all absent) before collectstatic, so a missing
  ignore file fails the build instead of leaking.
- Rebuilt: backend 1.43 GB -> 876 MB ('COPY backend/' is now 268 kB),
  frontend stays at 65 MB. Verified by booting the compose stack with the
  new image: migrations applied, /health ok, no .env or venv inside, and
  /app/media is the mounted (empty) volume; the scheduler runs too.
- Removed the stale local images that still contained the library and the
  dev .env.
2026-09-18 16:04:55 -05:00
JakeBreath 1170e6e9c1 Updates 2026-09-18 14:11:10 -05:00
JakeBreath b3ceac52ed fish greeting: win over distro configs that define fish_greeting inline
On CachyOS (and OMF-style setups) config.fish sources a distro file that
defines fish_greeting while the shell starts. A function defined that way
beats autoloading from functions/, so the installed greeting never ran.

install.fish now appends a guarded block to ~/.config/fish/config.fish that
sources the greeting after everything else (idempotent via a marker, skipped
with --no-config), and the README documents the symptom, the check
(functions --details fish_greeting) and the manual one-liner.

Verified in a sandbox HOME that reproduces the CachyOS setup: before the
install fish resolves the distro file and prints its message, after it
resolves ~/.config/fish/functions/fish_greeting.fish and runs ours; a second
install leaves a single block. Applied to this machine's real config as
well, where fish_greeting now resolves to the user function and the
__j621_fetch_random helper is loaded.
2026-09-18 13:48:02 -05:00
JakeBreath bcff184a64 Make extras/fish_greeting/install.fish executable
It has a '#!/usr/bin/env fish' shebang but was committed as mode 644, so
./install.fish answered 'Permission denied' on a fresh clone. Mode is now
100755 like the deploy scripts; 'fish install.fish' worked either way.
2026-09-18 13:43:12 -05:00
JakeBreath 770b1e5ee6 Scoped API tokens for the random endpoint, with a management page
Backend: a GreetingToken model stores only a SHA-256 hash of a j621r_…
key (shown once at creation) plus label, prefix, created/last-used. A
dedicated GreetingTokenAuthentication understands the usual
'Authorization: Token …' header but is registered only on RandomItemView
(alongside the normal token auth), so a greeting token authenticates
/api/random/ and is rejected with 401 everywhere else — exactly the scope
shell greetings need. Endpoints: GET/POST /api/auth/greeting-tokens/ and
DELETE /api/auth/greeting-tokens/{id}/ (own tokens only; the list never
returns keys or hashes).

Frontend: /tokens page (Account → Shell tokens card, command palette entry)
lists tokens with label, prefix, created/last-used and revoke (shared
confirm dialog). Creating one shows the key with Copy and 'Copy for fish'
buttons plus a pointer to extras/fish_greeting.

Tests: apps/accounts/tests/test_greeting_tokens.py — 9 tests covering
create-once semantics and hashing, hidden keys in listings, the scope
guarantee (random 200 with a signed URL; 401 on files, storage, me, tags
cloud, delete and the token list itself), unknown/revoked keys, cross-user
revocation, last-used tracking and label limits.

Verified live: created a token, rolled /random (signed URL), got 401 from
four other endpoints, saw the list omit secrets, revoked it (204) and the
same key then 401'd on /random. Full suite: 39 tests green.
2026-09-18 13:37:29 -05:00
JakeBreath 2d9493d9fe fish_greeting installer asks for the API origin and probes the backend
install.fish now:
- asks for the API origin (default https://j621.rainbow-herring.ts.net) and
  an optional token when run interactively, or takes --url/--token;
- writes ~/.config/j621Greeting/config.fish with mode 600 (it may hold a
  token), keeps an existing config unless --force/--url is given, and
  --no-prompt runs fully unattended;
- verifies the setup: /health must answer 'ok' and a fastfetch random roll is
  attempted, reporting the API's own message when it finds nothing; exits
  non-zero when the backend is unreachable so scripts notice.

README documents the prompts/flags, the chmod 600 config, and that the
greeting is designed to run without a token (guest mode: unsigned links,
guest-visible items only) with a token adding signed links and hidden items.

Tested with fish 4.9.3 in throwaway HOME dirs: flag-driven install with a
token, interactive install with a piped origin and no token, re-run keeping
the existing config, and an unreachable backend exiting 1.
2026-09-18 13:22:55 -05:00
JakeBreath aee29de34a Port the fish_greeting shell greeting to the new API
The original script (J621-Django/extras/fish_greeting system) downloaded
image bytes from /random/?rating=X and read the X-File-* headers. The new
API answers with JSON and a signed link, so the greeting now:

- asks /api/random/?fastfetch=1&rating=<mode> for JSON;
- parses url/j_id/filename/md5 with jq, python3, or a grep/sed fallback;
- downloads the signed link and keeps the original display path (fastfetch
  kitty/kitty-icat logos, gifsicle preprocessing for GIFs, recursion guard,
  logging with rotation, greeting_mode 0/1/2 = NSFW/SFW/Questionable);
- is configured through ~/.config/j621Greeting/config.fish or universal
  variables (J621_BASE, J621_WEB, J621_TOKEN, J621_FASTFETCH_CONFIG) instead
  of a hardcoded host, and prints the /detail/<J-ID> link on J621_WEB;
- reports the API's own 404 message when a rating has no images, and keeps
  curl quiet so failures do not spill into the greeting.

extras/fish_greeting/ contains the function, an install.fish (copies it into
the fish functions dir, creates the config once, checks dependencies) and a
README with the old-vs-new table and troubleshooting. The existing
gm-switch helper and .desktop launchers keep working (same mode file).

Tested with fish 4.9.3: syntax check on every file, guest and token runs
against the dev API (unsigned vs signed URLs), the empty-rating and
unreachable-API paths, and the grep/sed fallback with jq and python3
unavailable.
2026-09-18 13:14:41 -05:00
JakeBreath f8667c1037 Add a Random image endpoint and SPA page (with fastfetch mode)
Backend: GET /api/random/ (aliases /random and /random/) returns a random
library image with:
- rating=s,q,e filtering (comma separated, default any);
- fastfetch mode (?fastfetch=1 or any User-Agent containing "fastfetch")
  that only considers png/jpg/gif - what terminal viewers can show;
- JSON with j_id, filename, extension, rating, size, e621 id plus absolute
  url/download_url/thumbnail_url. Authenticated callers get signed URLs so
  fastfetch and image viewers can load them without headers; guests get
  unsigned URLs and never receive hidden_from_guests items.

Tests: apps/library/tests/test_random.py (8 tests) covering the response
contract, guest signatures, image-only default, the fastfetch format
restriction (flag and User-Agent), rating filters, guest visibility and the
short alias.

Frontend: /random page with rating pills, R to roll, Open/Download and a
library link, plus navigation and command palette entries; needs a backend,
hidden in local mode.

nginx: /random negotiates on Accept so browsers keep getting the SPA while
scripts get the JSON (verified with the proxy and frontend containers).

Also fixes a regression from the SSRF change: the guest download proxy
still referenced the removed 'parsed' variable on its success path, so
every proxied download would have 500'd. Redirect hops are now covered by
tests with a mocked requests.get.
2026-09-18 13:06:36 -05:00
JakeBreath f25526782c Run the periodic commands in a scheduler service (no host cron)
Adds deploy/scheduler-entrypoint.sh to the backend image (entrypoint
j621-scheduler) and a scheduler service to the four composes that have a
backend. It waits until the database and migrations are ready, runs every
job once, then keeps to the intervals:

  sync_followed_tags + sync_followed_pools   every 30 min (J621_SYNC_EVERY)
  cleanup_similarity                         hourly      (J621_CLEAN_EVERY)
  refresh_guest_blacklist                    daily       (J621_BLACKLIST_EVERY)

It shares the backend image, media volume and env file, so commands see
the same library and database; failures are logged and retried next
interval. Output goes to docker compose logs scheduler; the frontend-only
composes have no backend and therefore no scheduler.

Verified against a real stack: the scheduler waited for migrations, ran all
four commands on start (follow syncs as anonymous, similarity cleanup, and
a guest blacklist refresh that pulled the real 14-tag list from e621), and
kept looping. All six composes still validate.
2026-09-18 12:51:17 -05:00
JakeBreath 93cce6b9fd deploy/gen_env.sh: generate .env with openssl secrets
Builds deploy/.env from .env.example, generating SECRET_KEY and both
database passwords with openssl (base64/hex only, so nothing needs quoting
in the env file or the compose parser). Derives TS_HOSTNAME and
ALLOWED_HOSTS from the tailnet hostname, optionally sets
CORS_ALLOWED_ORIGINS/CSRF_TRUSTED_ORIGINS for split deployments, forces
DEBUG=False and writes the file with mode 600.

Modes: --no-prompt (defaults only), --update (refresh hostnames/auth key
while keeping the existing secrets, reading the stored FQDN from
ALLOWED_HOSTS), --force (rotate everything, with the SECRET_KEY warning in
the docs). Refuses to overwrite an existing file otherwise.

Verified: all modes, updated FQDN preservation, mode 600, and
docker compose config accepting the generated file.
2026-09-18 12:46:56 -05:00
JakeBreath 30b1a1a4b0 Tailnet-only (Serve, no Funnel) compose variants
Three more composes — compose.tailnet.yml, compose.tailnet.frontend.yml,
compose.tailnet.backend.yml — mirror the funnel set exactly but mount
serve.default/frontend/backend.tailnet.json, which drop AllowFunnel. The
sidecar still registers and serves HTTPS with a tailnet certificate, but
nothing is exposed publicly; Serve also needs no ACL change.

Project names carry a -tailnet suffix so both sets can coexist, and the
README explains that each set needs its own data directory (or host), plus
how to switch a host between funnel and tailnet by starting the other file
with the same .env.

Verified: all six composes validate with docker compose config, and the
six serve configs split cleanly into funnel (AllowFunnel present) and
tailnet-only (absent).
2026-09-18 11:21:48 -05:00
119 changed files with 13198 additions and 816 deletions
+35
View File
@@ -0,0 +1,35 @@
# Build context is the repository root (see deploy/J621-Backend and
# deploy/J621-Frontend). Keep the context small and, more importantly, keep
# secrets and runtime data out of the images.
# Version control / tooling
.git
.gitignore
.dockerignore
# Python: the dev virtualenv, caches, and anything generated at runtime
backend/venv/
**/__pycache__/
**/*.py[cod]
backend/db.sqlite3
backend/logs/
backend/staticfiles/
# Secrets and media: the .env holds SECRET_KEY and DB passwords, media/ is
# the actual library. The deploy composes mount these at runtime instead.
backend/.env
backend/.env.*
backend/media/
# Frontend build artefacts (npm ci installs fresh from the lockfile)
frontend/node_modules/
frontend/dist/
# Deploy runtime state and env
deploy/.env
deploy/.env.*
deploy/data/
deploy/tailscale-state/
# Misc
*.log
+197
View File
@@ -0,0 +1,197 @@
# J621 CD — manual release workflow (Actions tab -> "Run workflow").
#
# One dispatch does everything; each half can be skipped with the `images`
# and `desktop` inputs:
# * builds and pushes the backend + frontend images (multi-arch, :latest
# and :<short-sha>, GIT_HASH baked in for the version pill),
# * builds the desktop packages and attaches them (plus the update
# metadata) to the Gitea release tagged `desktop-v<package.json version>`.
#
# The desktop build also attaches the update metadata (latest*.yml) to the
# release; that is the desktop updater's feed, resolved through the Gitea API
# at check time (see desktop/README.md). The older website feed
# (deploy/data/desktop, served at /desktop/) is runtime state on the deploy
# host and only needed for installs before 0.1.2; it is refreshed with
# `deploy/push_desktop.sh` from a machine that can reach the deploy host.
#
# Registry login uses a repo PAT with the minimal write:package scope (the
# Gitea registry rejects the automatic job token, go-gitea/gitea#23642);
# release creation uses the automatic job token. Jobs run on the user-scoped
# nitro-ci runner (ubuntu-latest).
name: CD
on:
workflow_dispatch:
inputs:
images:
description: Build and push the Docker images
required: false
default: "true"
desktop:
description: Build the desktop release
required: false
default: "true"
platforms:
description: Image platforms (comma separated)
required: false
default: linux/amd64,linux/arm64
windows:
description: Also cross-build the Windows installer (needs wine, slow)
required: false
default: "false"
concurrency:
group: cd
cancel-in-progress: false
jobs:
images:
name: Build & push images
if: ${{ inputs.images != 'false' }}
runs-on: ubuntu-latest
# The Gitea container registry does not accept the automatic job token
# (go-gitea/gitea#23642 is still open), so the push uses a repo PAT with
# the minimal write:package scope. Releases use the job token instead.
permissions:
contents: read
env:
REGISTRY_USER: ${{ secrets.REGISTRY_USER }}
REGISTRY_TOKEN: ${{ secrets.REGISTRY_TOKEN }}
PLATFORMS: ${{ inputs.platforms || 'linux/amd64,linux/arm64' }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Check the registry credentials
run: |
if [ -z "$REGISTRY_USER" ] || [ -z "$REGISTRY_TOKEN" ]; then
echo "Set the REGISTRY_USER and REGISTRY_TOKEN repo secrets" >&2
echo "(a PAT with the write:package scope)." >&2
exit 1
fi
- name: Register binfmt (multi-arch builds)
run: docker run --privileged --rm tonistiigi/binfmt --install all
- name: Build & push both images
run: |
set -euo pipefail
SHA="$(git rev-parse --short HEAD)"
echo "Publishing $SHA for $PLATFORMS"
PLATFORMS="$PLATFORMS" ./deploy/push_frontend.sh "$SHA"
PLATFORMS="$PLATFORMS" ./deploy/push_backend.sh "$SHA"
desktop:
name: Desktop release
if: ${{ inputs.desktop != 'false' }}
runs-on: ubuntu-latest
# Creating the release and uploading its assets uses the automatic job
# token, so it needs write access to the repository's releases.
permissions:
contents: write
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: "22"
- name: Install packaging tools
run: |
sudo apt-get update
sudo apt-get install -y --no-install-recommends fakeroot libarchive-tools
if [ "${{ inputs.windows }}" = "true" ]; then
# electron-builder runs the 32-bit NSIS installer under wine to
# build the uninstaller: that needs a virtual display (Xvfb) and
# 32-bit wine libraries.
sudo dpkg --add-architecture i386
sudo apt-get update
sudo apt-get install -y --no-install-recommends xvfb wine wine32:i386
fi
- name: Install frontend + desktop dependencies
run: |
npm --prefix frontend ci --no-audit --no-fund
npm --prefix desktop ci --no-audit --no-fund
- name: Build desktop packages
env:
# Keep wine from trying to fetch Gecko/Mono on first run.
WINEDLLOVERRIDES: mscoree,mshtml=
run: |
if [ "${{ inputs.windows }}" = "true" ]; then
xvfb-run -a ./deploy/build_desktop.sh --all
else
./deploy/build_desktop.sh --linux
fi
- name: Add the Gitea release
env:
GITEA_TOKEN: ${{ secrets.GITEA_TOKEN || secrets.GITHUB_TOKEN }}
run: |
set -euo pipefail
VERSION="$(node -p "require('./desktop/package.json').version")"
TAG="desktop-v$VERSION"
API="${{ github.server_url }}/api/v1/repos/${{ github.repository }}"
AUTH="Authorization: token $GITEA_TOKEN"
NOTES="$(printf 'J621 desktop %s\n\n' "$VERSION"
cd desktop/release
sha256sum ./*.deb ./*.pkg.tar.zst ./*.exe 2>/dev/null || true)"
RELEASE_ID="$(curl -sf -H "$AUTH" "$API/releases/tags/$TAG" \
| python3 -c 'import json,sys; print(json.load(sys.stdin).get("id",""))' \
2>/dev/null || true)"
if [ -z "$RELEASE_ID" ]; then
echo "Creating release $TAG"
PAYLOAD="$(python3 - "$TAG" "${{ github.sha }}" "$NOTES" <<'PY'
import json, sys
print(json.dumps({
"tag_name": sys.argv[1],
"name": sys.argv[1],
"body": sys.argv[3],
"target_commitish": sys.argv[2],
}))
PY
)"
RELEASE_ID="$(curl -sf -X POST -H "$AUTH" \
-H "Content-Type: application/json" -d "$PAYLOAD" "$API/releases" \
| python3 -c 'import json,sys; print(json.load(sys.stdin)["id"])')"
else
echo "Release $TAG already exists (id $RELEASE_ID); attaching missing files."
fi
EXISTING="$(curl -sf -H "$AUTH" "$API/releases/$RELEASE_ID/assets" \
|| echo '[]')"
for FILE in desktop/release/*"$VERSION"*.deb \
desktop/release/*"$VERSION"*.pkg.tar.zst \
desktop/release/latest-linux.yml \
desktop/release/latest.yml \
desktop/release/*"$VERSION"*.exe \
desktop/release/*"$VERSION"*.exe.blockmap; do
[ -e "$FILE" ] || continue
NAME="$(basename "$FILE")"
# Replace the asset when it is already there: latest*.yml must
# reference the installers built by *this* run (NSIS builds are
# not bit-reproducible), so old copies are deleted first.
ASSET_ID="$(printf '%s' "$EXISTING" | python3 -c '
import json, sys
name = sys.argv[1]
print(next((str(a["id"]) for a in json.load(sys.stdin) if a["name"] == name), ""))
' "$NAME")"
if [ -n "$ASSET_ID" ]; then
echo " replacing $NAME"
curl -sf -X DELETE -H "$AUTH" \
"$API/releases/$RELEASE_ID/assets/$ASSET_ID" >/dev/null
else
echo " attaching $NAME"
fi
# Names like "J621 Setup 0.1.1.exe" contain spaces: encode them
# or curl refuses the URL (exit 3).
ENCODED="$(python3 -c 'import sys, urllib.parse; print(urllib.parse.quote(sys.argv[1]))' "$NAME")"
curl -sf -X POST -H "$AUTH" -H "Content-Type: application/octet-stream" \
--data-binary @"$FILE" "$API/releases/$RELEASE_ID/assets?name=$ENCODED" >/dev/null
done
echo "Release: ${{ github.server_url }}/${{ github.repository }}/releases/tag/$TAG"
+101
View File
@@ -0,0 +1,101 @@
# J621 CI — runs on every push (and pull request): Django checks + the full
# backend test suite against MariaDB/Redis, and the frontend type-check,
# lint and production build.
#
# Runner: the "nitro-ci" act_runner with the custom `ubuntu-latest` label.
name: CI
# Tests and builds only need to read the repository; the automatic job token
# stays read-only.
permissions:
contents: read
on:
push:
branches: ["**"]
tags-ignore: ["**"]
pull_request:
workflow_dispatch:
concurrency:
group: ci-${{ github.ref }}
cancel-in-progress: true
jobs:
backend:
name: Backend tests
runs-on: ubuntu-latest
services:
mariadb:
image: mariadb:11.4
env:
MARIADB_ROOT_PASSWORD: root
MARIADB_DATABASE: j621
MARIADB_USER: j621
MARIADB_PASSWORD: j621
options: >-
--health-cmd="healthcheck.sh --connect --innodb_initialized"
--health-interval=5s
--health-timeout=5s
--health-retries=12
redis:
image: redis:7-alpine
options: >-
--health-cmd="redis-cli ping"
--health-interval=5s
--health-timeout=5s
--health-retries=12
env:
# Connect as root so Django can create the test database itself;
# everything else mirrors the development defaults.
DB_HOST: mariadb
DB_PORT: "3306"
DB_NAME: j621
DB_USER: root
DB_PASSWORD: root
REDIS_URL: redis://redis:6379/1
steps:
- uses: actions/checkout@v4
- uses: actions/setup-python@v5
with:
python-version: "3.14"
- name: Install backend dependencies
run: pip install -r backend/requirements.txt
- name: Django system checks
working-directory: backend
run: python manage.py check
- name: Backend tests
working-directory: backend
run: >-
python manage.py test
apps.core.tests
apps.library.tests
apps.follows.tests
apps.accounts.tests
frontend:
name: Frontend build & lint
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: "22"
- name: Install frontend dependencies
working-directory: frontend
run: npm ci
- name: Lint
working-directory: frontend
run: npm run lint
- name: Type-check & build
working-directory: frontend
run: npm run build
+1
View File
@@ -25,6 +25,7 @@ frontend/dist/
.vscode/
.idea/
.DS_Store
.directory
# Tooling
*.log
+32 -5
View File
@@ -2,6 +2,11 @@ Use the venv on backend/venv/
DO NOT mess with the system's python
The original J621 (Django) app is kept for reference at
/mnt/Disco/Proyects/Python3.13/J621-Django/
Consult it when matching original behavior (batch sizes for e621 lookups,
per-page sets, upload flow) instead of guessing.
For Frontend work:
read both Frontend Design related Markdown files on the frontend/ folder
@@ -34,10 +39,27 @@ Project constraints (do not regress):
- deploy/ is the deployment source of truth: J621-Frontend (SPA on static
nginx) and J621-Backend (gunicorn + whitenoise, ffmpeg, migrations on
start), three compose variants (both / frontend-only / backend-only) behind
a shared nginx proxy service, and a Tailscale sidecar per compose whose
serve configs only use funnel ports 443 / 8443 / 10000. Images are pushed
to the Gitea registry with deploy/push_*.sh (multi-arch, :latest + :sha,
GIT_HASH baked in for the version pill).
a shared nginx proxy service, and a Tailscale sidecar per compose.
serve.*.json are the public Funnel variants (only ports 443 / 8443 / 10000);
serve.*.tailnet.json + compose.tailnet*.yml are the same stacks without
AllowFunnel (tailnet-only, no public exposure). Images are pushed to the
Gitea registry with deploy/push_*.sh (multi-arch, :latest + :sha, GIT_HASH
baked in for the version pill); deploy/gen_env.sh generates .env with
openssl secrets (--update keeps SECRET_KEY, --force rotates it).
- Periodic commands (follow syncs, similarity cleanup, guest blacklist
refresh) run in the composes' `scheduler` service — the backend image with
the j621-scheduler entrypoint, intervals via J621_*_EVERY. No host cron.
- CI/CD lives in .gitea/workflows: ci.yml runs on every push/PR (Django checks
+ the full backend suite against MariaDB/Redis service containers, frontend
lint/type-check/build); cd.yml is manual and builds/pushes both images
multi-arch plus the desktop packages (attached to the Gitea release
`desktop-v<version>`). Jobs run on the user-scoped runners: `ubuntu-latest`
on nitro-ci, `desktop` on msi-mortar-ci. Do not add actions/cache
(`cache: pip`/`npm`) to these workflows: Gitea's cache service hangs the job
on restore/save. Desktop updates read the release assets (latest*.yml) from
the Gitea API at check time; the older website feed (deploy/data/desktop)
only matters for installs before 0.1.2 and is refreshed with
`deploy/push_desktop.sh` from a machine with SSH to the deploy host.
- Security/permission tests live in backend/apps/core/tests and need a
one-time grant: GRANT ALL ON `test_j621`.* TO 'j621'@'%';
@@ -74,7 +96,12 @@ Security hardening (do not weaken):
SECRET_KEY (apps/accounts/crypto.py); rotating SECRET_KEY invalidates them
(and all signed media URLs), so users must re-enter the key.
- API throttles live in REST_FRAMEWORK (env-overridable): anon 120/min,
user 600/min, login 5/min, register 20/hour, e621_proxy 60/hour.
user 600/min, login 5/min, register 20/hour, e621_proxy 60/hour. Signed
media URLs (raw/thumbnail/staged-file/similarity-file actions) are exempt
on purpose: <img>/<video> tags fetch them without an Authorization header,
so a gallery would otherwise drain the anonymous bucket and get 429 JSON
instead of images. THROTTLE_ENABLED=false removes the anon+user limits for
private/tailnet deployments (the login/register/proxy guards stay).
- Only admins (superusers) may grant/revoke the staff role or delete
staff/admin accounts; staff manage regular/uploader accounts only.
- Storage, duplicates, delete, temp-clear, uploads and downloads require
+50 -4
View File
@@ -5,8 +5,10 @@ Self-hosted media library and e621 archive manager, rebuilt as a **React SPA + D
## Structure
```
backend/ Django 6 + DRF API (SQLite in dev, MariaDB in production)
backend/ Django 6 + DRF API (MariaDB + Redis via docker compose)
frontend/ Vite + React + TypeScript SPA
deploy/ Docker images, compose variants and Tailscale serve configs
extras/ shell integrations (fish_greeting with fastfetch)
```
## Development
@@ -33,9 +35,53 @@ npm run dev # http://localhost:5173, proxies /api to
### Production
Not wired up yet — planned: Nginx serving the SPA build, `/media` and `/library` media
directly, and proxying `/api` to Waitress/Django. See `frontend/` design docs for the UI
specification.
Docker: see [`deploy/`](deploy/README.md) for the two images (SPA on static
nginx, API on gunicorn), the three compose variants (both / frontend-only /
backend-only) behind a shared nginx service and a Tailscale sidecar, and the
public-funnel or tailnet-only serve configs. `deploy/push_*.sh` builds and
pushes the multi-arch images to the Gitea registry.
## Random image endpoint
Used by the SPA's Random page and by shell greetings (fish_greeting +
fastfetch):
```bash
curl -H "Authorization: Token <token>" \
"https://j621.example.ts.net/api/random/?rating=s,q&fastfetch=1"
```
```json
{
"j_id": "J-59",
"filename": "J-59.jpg",
"extension": "jpg",
"rating": "e",
"url": "https://j621.example.ts.net/api/files/J-59/raw/?sig=…",
"download_url": "https://j621.example.ts.net/api/files/J-59/raw/?sig=…&download=1",
"thumbnail_url": "https://j621.example.ts.net/api/files/J-59/thumbnail/?sig=…",
"fastfetch": true
}
```
- `rating` — comma separated subset of `s`, `q`, `e` (default: any).
- `fastfetch=1`, or any request whose User-Agent contains `fastfetch`, limits
the roll to `png`/`jpg`/`gif` so terminals can display it. Images are the
only candidates in both modes.
- `url` is absolute, and signed for authenticated callers, so fastfetch can
load it without headers. Guests get an unsigned URL and only see
guest-visible items.
- `/random` and `/random/` are aliases of `/api/random/` for scripts. Behind
the bundled nginx those aliases negotiate on `Accept`: browsers get the SPA
page, requesters like curl/wget/fastfetch get the JSON. `/api/random/` is
the unambiguous path for scripts; 404 when nothing matches the filters.
- A ready-made shell greeting that uses this endpoint lives in
[`extras/fish_greeting/`](extras/fish_greeting/README.md).
- **Scoped tokens for scripts**: the Account page's *Shell tokens* section
(also `/tokens`) issues `j621r_…` tokens that only authenticate
`/api/random/` — the rest of the API rejects them. They are stored as
hashes, shown once, and revocable any time
(`/api/auth/greeting-tokens/`).
## Licence
+42 -15
View File
@@ -20,7 +20,23 @@ they land.
- [x] Tag cloud in the sidebar (click to search, hidden from guests for
blacklisted items)
- [x] Status filter (matched / not_found / deleted / custom / unknown)
- [x] **Cacheable media serving**
- [x] Stable signed URLs (7 d TTL, 24 h rotation) plus a `v=<md5>` cache
buster, so replacing a file under the same J-ID invalidates it
- [x] ETag/Last-Modified and a private `Cache-Control` (immutable for
versioned media), conditional 304s
- [x] Real 480 px image thumbnails (cached by MD5) instead of serving
full-size originals through the thumbnail endpoint
- [x] **Random image** endpoint and page: `GET /api/random/` (aliases
`/random`, `/random/`) with `rating=s,q,e` filters; fastfetch mode
(`?fastfetch=1` or a Fastfetch User-Agent) only returns png/jpg/gif as
JSON with a signed absolute link, for the fish_greeting scripts; the
`/random` SPA page rolls with rating pills and the `R` key
- **Scoped `j621r_…` greeting tokens** (Account → Shell tokens, `/tokens`):
only `/api/random/` accepts them, they are stored hashed, shown once and
revocable; `install.fish` in `extras/fish_greeting` fills them into the
shell greeting config
- [x] **Ephemeral similarity check** (`/similar`)
- [x] Drop a file: exact MD5 match, perceptual matches against the library,
and e621 IQDB candidates (auto-run for images)
@@ -86,18 +102,25 @@ Files now stage first and are resolved before entering the library.
- [x] `cleanup_temp_uploads` command for old staged files
- [x] **Auto-upload / auto-match**
- [x] MD5 computed on staging; exact duplicates resolve immediately
- [x] MD5 batch-checked against e621; matches auto-complete with post
metadata stored and rating seeded
- [x] **IQDB similarity on upload** (SPA-driven)
- [x] Automatic + manual IQDB checks with candidate posts
- [x] "Visual Similarity Detected" state with candidate picker
- [x] Perceptual-hash comparison against the library (staged uploads are
flagged with their library matches as soon as they land)
- [x] MD5 batch-checked against e621 in chunks of 75; matches auto-complete
with post metadata stored and rating seeded
- [x] **Background pipeline** (server-side)
- [x] Daemon-thread worker runs MD5 → visual similarity → IQDB for every
staged upload, so the work continues after the page or tab is closed
- [x] Durable progress (`UploadRun` + per-file phase flags) polled by the
shell indicator; rate-limited runs retry with backoff
- [x] Perceptual-hash comparison against the library, loaded once per batch
- [x] IQDB candidates stored with one batched enrichment request
- [x] Indexed uploads are announced through a bounded per-run feed and the
staged row is deleted; nothing persists on the board to dismiss
- [x] **Upload UI**
- [x] Three-column board: Pending & Unmatched / Visual Similarity Detected /
Auto-uploaded & Indexed
- [x] Four-column board: Pending & Unmatched / Visual Similarity Detected /
Auto-uploaded & Indexed (session feed) / Failed, private per user
(staff included)
- [x] Metadata modal (link to e621 post, IQDB candidates, custom metadata)
- [x] Per-file progress plus batch processing indicator
- [x] Per-file progress plus background pipeline status
- [x] Bulk actions: bulk rate and discard all (chunked past the API's
1000-id cap)
## 4. Staff tools
@@ -136,7 +159,9 @@ Files now stage first and are resolved before entering the library.
- [x] Profile pictures: staff Users page and a self-service Account picker
(searchable library grid, remove supported) set avatars from library J-IDs
- [x] Profile extras: per-user landing page, default rating filter/sort, items
per page and thumbnail size (synced to the account, applied on load)
per page and thumbnail size (synced to the account, applied on load),
plus e621 posts per page (48/100/200/320) for the Online browser and
pool loading
- [x] Backend-less local mode: with no backend connected the SPA runs on the
e621-facing pages only (Online, Pools) using credentials stored in the
browser, and the shell offers a "Setup Backend" button instead
@@ -145,10 +170,12 @@ Files now stage first and are resolved before entering the library.
## 6. Infrastructure
- [ ] Guest blacklist refresh on a timer (in-container scheduler)
- [ ] Follow sync on a timer (in-container scheduler, e.g. every 30 minutes)
- [ ] Similarity temp cleanup on a timer (in-container scheduler; the TTL
also cleans lazily when new checks are created)
- [x] Guest blacklist refresh on a timer (the composes' `scheduler` service;
`J621_BLACKLIST_EVERY`, default daily)
- [x] Follow sync on a timer (same scheduler: `sync_followed_tags` +
`sync_followed_pools`, default every 30 minutes)
- [x] Similarity temp cleanup on a timer (same scheduler, default hourly;
the TTL also cleans lazily when new checks are created)
- [x] Production setup: two Docker images (SPA on static nginx, API on
gunicorn + whitenoise with ffmpeg) and three compose variants (both /
frontend-only / backend-only) behind a shared nginx proxy service, each
+42
View File
@@ -0,0 +1,42 @@
"""Authentication for scope-limited bearer tokens.
`GreetingTokenAuthentication` understands the same header a normal API token
uses (``Authorization: Token <key>``) but only resolves tokens issued for the
random-image endpoint. It is registered per-view (currently only
`RandomItemView`), so a greeting token is rejected everywhere else by the
regular DRF token authentication.
"""
from rest_framework import authentication, exceptions
from .models import GreetingToken
class GreetingTokenAuthentication(authentication.BaseAuthentication):
keyword = b"token"
def authenticate_header(self, request):
# DRF answers 401 (instead of 403) for AuthenticationFailed only when
# the first authenticator can name the scheme.
return "Token"
def authenticate(self, request):
header = authentication.get_authorization_header(request).split()
if not header or header[0].lower() != self.keyword:
return None
if len(header) != 2:
raise exceptions.AuthenticationFailed("Invalid token header.")
try:
key = header[1].decode()
except UnicodeError:
raise exceptions.AuthenticationFailed("Invalid token header.")
# Not one of ours: let the regular token authentication handle it.
if not key.startswith(GreetingToken.PREFIX):
return None
token = GreetingToken.resolve(key)
if token is None:
raise exceptions.AuthenticationFailed("Invalid token.")
token.touch()
return (token.user, token)
@@ -0,0 +1,30 @@
# Generated by Django 6.1.1 on 2026-09-18 18:25
import django.db.models.deletion
from django.conf import settings
from django.db import migrations, models
class Migration(migrations.Migration):
dependencies = [
('accounts', '0006_encrypt_e621_api_keys'),
]
operations = [
migrations.CreateModel(
name='GreetingToken',
fields=[
('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
('key_hash', models.CharField(max_length=64, unique=True)),
('prefix', models.CharField(max_length=16)),
('label', models.CharField(blank=True, default='', max_length=100)),
('created_at', models.DateTimeField(auto_now_add=True)),
('last_used_at', models.DateTimeField(blank=True, null=True)),
('user', models.ForeignKey(on_delete=django.db.models.deletion.CASCADE, related_name='greeting_tokens', to=settings.AUTH_USER_MODEL)),
],
options={
'ordering': ['-created_at'],
},
),
]
+64
View File
@@ -1,5 +1,9 @@
import secrets
from django.conf import settings
from django.contrib.auth.models import AbstractUser
from django.db import models
from django.utils import timezone
class User(AbstractUser):
@@ -50,3 +54,63 @@ class User(AbstractUser):
return bool(
self.is_superuser or self.is_staff or self.role == self.ROLE_STAFF
)
def hash_bearer_token(value):
"""SHA-256 of a high-entropy bearer token (no salt needed)."""
import hashlib
return hashlib.sha256(value.encode()).hexdigest()
class GreetingToken(models.Model):
"""Long-lived token that only authenticates the random-image endpoint.
Meant for shell greetings and similar scripts, so it is safe to keep in a
config file: it cannot read the library, upload, or touch an account. Only
the SHA-256 hash is stored; the plaintext is returned once at creation.
"""
PREFIX = "j621r_"
user = models.ForeignKey(
settings.AUTH_USER_MODEL,
on_delete=models.CASCADE,
related_name="greeting_tokens",
)
key_hash = models.CharField(max_length=64, unique=True)
prefix = models.CharField(max_length=16)
label = models.CharField(max_length=100, blank=True, default="")
created_at = models.DateTimeField(auto_now_add=True)
last_used_at = models.DateTimeField(null=True, blank=True)
class Meta:
ordering = ["-created_at"]
def __str__(self):
return f"{self.prefix}… ({self.user})"
@classmethod
def issue(cls, user, label=""):
"""Create a token and return ``(token, plaintext_key)``."""
key = cls.PREFIX + secrets.token_hex(20)
token = cls.objects.create(
user=user,
key_hash=hash_bearer_token(key),
prefix=key[:12],
label=label.strip()[:100],
)
return token, key
@classmethod
def resolve(cls, key):
if not key.startswith(cls.PREFIX):
return None
return (
cls.objects.select_related("user")
.filter(key_hash=hash_bearer_token(key))
.first()
)
def touch(self):
GreetingToken.objects.filter(pk=self.pk).update(last_used_at=timezone.now())
+10 -1
View File
@@ -6,7 +6,7 @@ from rest_framework import serializers
from apps.library.services import VIDEO_EXTENSIONS
from apps.library.services import signed_media_url as signed_library_url
from .models import User
from .models import User, GreetingToken
def signed_media_url(request, item):
@@ -92,6 +92,13 @@ class UserUpdateSerializer(serializers.Serializer):
role = serializers.ChoiceField(choices=User.ROLE_CHOICES, required=False)
class GreetingTokenSerializer(serializers.ModelSerializer):
class Meta:
model = GreetingToken
fields = ["id", "prefix", "label", "created_at", "last_used_at"]
read_only_fields = fields
class RegisterSerializer(serializers.ModelSerializer):
password = serializers.CharField(write_only=True, validators=[validate_password])
@@ -127,5 +134,7 @@ class PreferencesSerializer(serializers.Serializer):
ordering = serializers.CharField(max_length=30, required=False)
per_page = serializers.IntegerField(min_value=12, max_value=200, required=False)
zoom = serializers.IntegerField(min_value=120, max_value=400, required=False)
# e621 renders at most 320 posts per request.
e621_per_page = serializers.IntegerField(min_value=12, max_value=320, required=False)
# Open Online with the order:hot metatag when no search is given.
online_hot_default = serializers.BooleanField(required=False)
@@ -0,0 +1,166 @@
"""Scope-limited greeting tokens (`j621r_…`).
They exist so shell greetings and scripts can hold a credential that only
authenticates `/api/random/` — everything else must reject them — and they
are stored hashed, shown once.
"""
import hashlib
import json
import shutil
import tempfile
import time
from pathlib import Path
from django.contrib.auth import get_user_model
from django.test import Client, TestCase, override_settings
from rest_framework.authtoken.models import Token
from apps.accounts.models import GreetingToken, hash_bearer_token
from apps.library.models import MediaItem, MediaLocation
User = get_user_model()
def jpost(client, path, body=None):
return client.post(path, data=json.dumps(body or {}), content_type="application/json")
class GreetingTokenTests(TestCase):
@classmethod
def setUpClass(cls):
super().setUpClass()
cls._tmp = tempfile.mkdtemp(prefix="j621-tokens-")
cls._watched = Path(cls._tmp) / "library"
cls._watched.mkdir(parents=True, exist_ok=True)
cls._settings = override_settings(
MEDIA_ROOT=cls._tmp, WATCHED_FOLDER=str(cls._watched)
)
cls._settings.enable()
@classmethod
def tearDownClass(cls):
cls._settings.disable()
shutil.rmtree(cls._tmp, ignore_errors=True)
super().tearDownClass()
def setUp(self):
self.user = User.objects.create_user(
username="token-user", password="token-pass-123456"
)
self.other = User.objects.create_user(
username="token-other", password="token-pass-123456"
)
self.client = self.api_client(self.user)
self.other_client = self.api_client(self.other)
# One image so the random endpoint can answer.
path = self._watched / "token-test.png"
path.write_bytes(b"token-test")
self.item = MediaItem.objects.create(
md5=hashlib.md5(b"token-test").hexdigest(),
size=path.stat().st_size,
rating="s",
uploaded_by=self.user,
)
MediaLocation.objects.create(
item=self.item, path=str(path), rel_path=path.name, mtime=time.time()
)
def api_client(self, user):
client = Client()
client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=user).key}"
)
return client
def token_client(self, key):
client = Client()
client.defaults["HTTP_AUTHORIZATION"] = f"Token {key}"
return client
def issue(self, client=None, label=""):
response = jpost(client or self.client, "/api/auth/greeting-tokens/", {"label": label})
self.assertEqual(response.status_code, 201)
return response.json()
def test_create_returns_the_key_once_and_stores_only_a_hash(self):
created = self.issue(self.client, "shell")
key = created["key"]
self.assertTrue(key.startswith("j621r_"))
self.assertEqual(created["label"], "shell")
token = GreetingToken.objects.get(pk=created["id"])
self.assertEqual(token.key_hash, hash_bearer_token(key))
self.assertNotIn(key, token.key_hash)
self.assertEqual(token.prefix, key[:12])
self.assertIsNone(token.last_used_at)
def test_list_hides_keys_and_hashes(self):
self.issue(self.client, "one")
self.issue(self.client, "two")
rows = self.client.get("/api/auth/greeting-tokens/").json()
self.assertEqual([row["label"] for row in rows], ["two", "one"])
for row in rows:
self.assertNotIn("key", row)
self.assertNotIn("key_hash", row)
self.assertTrue(row["prefix"].startswith("j621r_"))
def test_token_authenticates_random_and_nothing_else(self):
key = self.issue(self.other_client, "shell")["key"]
client = self.token_client(key)
response = client.get("/api/random/")
self.assertEqual(response.status_code, 200)
self.assertIn("sig=", response.json()["url"])
for method, path, body in (
("get", "/api/files/", None),
("get", "/api/storage/", None),
("get", "/api/auth/me/", None),
("get", "/api/tags/cloud/", None),
("post", "/api/delete/", {"j_ids": []}),
("get", "/api/auth/greeting-tokens/", None),
):
call = getattr(client, method)
if body is None:
self.assertEqual(call(path).status_code, 401, path)
else:
self.assertEqual(jpost(client, path, body).status_code, 401, path)
def test_normal_api_token_still_authenticates_random(self):
response = self.client.get("/api/random/")
self.assertEqual(response.status_code, 200)
self.assertIn("sig=", response.json()["url"])
def test_unknown_greeting_key_is_rejected(self):
client = self.token_client("j621r_" + "0" * 40)
self.assertEqual(client.get("/api/random/").status_code, 401)
def test_revoked_token_stops_working(self):
created = self.issue(self.client, "temporary")
client = self.token_client(created["key"])
self.assertEqual(client.get("/api/random/").status_code, 200)
response = self.client.delete(f"/api/auth/greeting-tokens/{created['id']}/")
self.assertEqual(response.status_code, 204)
self.assertEqual(client.get("/api/random/").status_code, 401)
self.assertFalse(GreetingToken.objects.filter(pk=created["id"]).exists())
def test_cannot_revoke_someone_elses_token(self):
created = self.issue(self.other_client, "theirs")
response = self.client.delete(f"/api/auth/greeting-tokens/{created['id']}/")
self.assertEqual(response.status_code, 404)
self.assertEqual(self.token_client(created["key"]).get("/api/random/").status_code, 200)
def test_last_used_is_recorded(self):
created = self.issue(self.client, "used")
self.token_client(created["key"]).get("/api/random/")
token = GreetingToken.objects.get(pk=created["id"])
self.assertIsNotNone(token.last_used_at)
def test_long_labels_are_rejected(self):
response = jpost(
self.client, "/api/auth/greeting-tokens/", {"label": "x" * 101}
)
self.assertEqual(response.status_code, 400)
@@ -0,0 +1,49 @@
"""Per-user browse preferences: validation and merge semantics."""
import json
from django.contrib.auth import get_user_model
from django.test import Client, TestCase
from rest_framework.authtoken.models import Token
User = get_user_model()
class PreferenceTests(TestCase):
def setUp(self):
self.user = User.objects.create_user(
username="prefs-user", password="prefs-pass-123456"
)
self.client = Client()
self.client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=self.user).key}"
)
def post(self, payload):
return self.client.post(
"/api/auth/preferences/",
data=json.dumps(payload),
content_type="application/json",
)
def test_e621_per_page_round_trips_within_range(self):
response = self.post({"e621_per_page": 320})
self.assertEqual(response.status_code, 200)
self.assertEqual(response.json()["e621_per_page"], 320)
self.assertEqual(
self.client.get("/api/auth/me/").json()["preferences"]["e621_per_page"],
320,
)
def test_e621_per_page_rejects_out_of_range(self):
for value in (0, 8, 321, 1000):
self.assertEqual(self.post({"e621_per_page": value}).status_code, 400, value)
def test_merge_keeps_other_keys(self):
self.post({"ratings": ["s"], "e621_per_page": 100})
response = self.post({"per_page": 24})
data = response.json()
self.assertEqual(data["ratings"], ["s"])
self.assertEqual(data["e621_per_page"], 100)
self.assertEqual(data["per_page"], 24)
+12
View File
@@ -3,6 +3,8 @@ from django.urls import path
from .views import (
AvatarView,
E621CredentialsView,
GreetingTokenDetailView,
GreetingTokenListView,
LoginView,
LogoutView,
MeView,
@@ -18,4 +20,14 @@ urlpatterns = [
path("avatar/", AvatarView.as_view(), name="avatar"),
path("preferences/", PreferencesView.as_view(), name="preferences"),
path("e621/", E621CredentialsView.as_view(), name="e621_credentials"),
path(
"greeting-tokens/",
GreetingTokenListView.as_view(),
name="greeting_tokens",
),
path(
"greeting-tokens/<int:pk>/",
GreetingTokenDetailView.as_view(),
name="greeting_token",
),
]
+49 -1
View File
@@ -1,5 +1,6 @@
import logging
from django.http import Http404
from rest_framework import mixins, status, viewsets
from rest_framework.authtoken.models import Token
from rest_framework.authtoken.views import ObtainAuthToken
@@ -13,9 +14,10 @@ from apps.core.permissions import IsAppStaff
from apps.library.models import MediaItem
from .crypto import encrypt_secret
from .models import User
from .models import GreetingToken, User
from .serializers import (
E621CredentialsSerializer,
GreetingTokenSerializer,
PreferencesSerializer,
RegisterSerializer,
UserListSerializer,
@@ -157,6 +159,52 @@ class PreferencesView(APIView):
return Response(preferences)
class GreetingTokenListView(APIView):
"""List and create the caller's random-endpoint tokens.
The plaintext key is returned once on creation; only its hash is stored,
and it only authenticates `/api/random/` (see GreetingToken).
"""
permission_classes = [IsAuthenticated]
def get(self, request):
tokens = GreetingToken.objects.filter(user=request.user)
return Response(GreetingTokenSerializer(tokens, many=True).data)
def post(self, request):
label = str(request.data.get("label") or "").strip()
if len(label) > 100:
return Response(
{"detail": "Label is too long (100 characters max)."},
status=status.HTTP_400_BAD_REQUEST,
)
token, key = GreetingToken.issue(request.user, label)
logger.info(
"Greeting token %s created by %s", token.prefix, request.user.username
)
return Response(
{**GreetingTokenSerializer(token).data, "key": key},
status=status.HTTP_201_CREATED,
)
class GreetingTokenDetailView(APIView):
"""Revoke one of the caller's tokens."""
permission_classes = [IsAuthenticated]
def delete(self, request, pk):
token = GreetingToken.objects.filter(pk=pk, user=request.user).first()
if token is None:
raise Http404
logger.info(
"Greeting token %s revoked by %s", token.prefix, request.user.username
)
token.delete()
return Response(status=status.HTTP_204_NO_CONTENT)
class UserViewSet(
mixins.ListModelMixin,
mixins.RetrieveModelMixin,
+108
View File
@@ -0,0 +1,108 @@
"""Redis cache that degrades instead of taking the whole API down.
Redis backs the DRF throttles and a few caches (storage stats, guest
blacklist, tag clouds). With Django's stock ``RedisCache``, a Redis that is
unreachable — or merely refusing writes because its RDB snapshot failed, the
default ``stop-writes-on-bgsave-error`` behaviour — raises inside the
throttle check on every request, so a cache outage becomes a blanket 500.
This backend treats cache failures as misses: rate limits and cached values
simply stop working until Redis is back, and the first failure per worker is
logged once so the cause is still visible.
"""
import logging
from django.core.cache.backends.redis import RedisCache
logger = logging.getLogger(__name__)
_warned = False
def _degrade(operation: str, error: Exception, default):
global _warned
if not _warned:
_warned = True
logger.warning(
"Cache unavailable (%s failed: %s) — continuing without it.",
operation,
error,
)
return default
class ResilientRedisCache(RedisCache):
"""``RedisCache`` where a broken Redis behaves like an empty cache."""
def add(self, *args, **kwargs):
try:
return super().add(*args, **kwargs)
except Exception as error: # noqa: BLE001 - any backend failure degrades
return _degrade("add", error, False)
def get(self, key, default=None, version=None):
try:
return super().get(key, default, version)
except Exception as error: # noqa: BLE001
return _degrade("get", error, default)
def set(self, *args, **kwargs):
try:
return super().set(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("set", error, None)
def touch(self, *args, **kwargs):
try:
return super().touch(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("touch", error, False)
def delete(self, *args, **kwargs):
try:
return super().delete(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("delete", error, False)
def get_many(self, *args, **kwargs):
try:
return super().get_many(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("get_many", error, {})
def has_key(self, *args, **kwargs):
try:
return super().has_key(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("has_key", error, False)
def incr(self, *args, **kwargs):
try:
return super().incr(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("incr", error, None)
def set_many(self, *args, **kwargs):
try:
return super().set_many(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("set_many", error, [])
def delete_many(self, *args, **kwargs):
try:
return super().delete_many(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("delete_many", error, None)
def clear(self, *args, **kwargs):
try:
return super().clear(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("clear", error, None)
def close(self, *args, **kwargs):
try:
return super().close(*args, **kwargs)
except Exception as error: # noqa: BLE001
return _degrade("close", error, None)
+36
View File
@@ -0,0 +1,36 @@
"""A broken Redis must degrade the cache, not 500 the API.
A Redis that cannot persist (the default ``stop-writes-on-bgsave-error``)
or is simply unreachable used to raise inside the DRF throttle check on
every request; ``ResilientRedisCache`` treats that as a cache miss.
"""
from unittest import mock
from django.core.cache import cache
from django.core.cache.backends.redis import RedisCacheClient
from django.test import SimpleTestCase
def failing(method: str):
return mock.patch.object(
RedisCacheClient,
method,
side_effect=RuntimeError("redis is down"),
)
class ResilientCacheTests(SimpleTestCase):
def test_get_returns_the_default_when_redis_fails(self):
with failing("get"):
self.assertIsNone(cache.get("j621-cache-test"))
self.assertEqual(cache.get("j621-cache-test", "fallback"), "fallback")
def test_writes_report_failure_without_raising(self):
with failing("set"):
self.assertIsNone(cache.set("j621-cache-test", "value"))
def test_bulk_and_delete_operations_degrade(self):
with failing("get_many"), failing("delete"):
self.assertEqual(cache.get_many(["a", "b"]), {})
self.assertFalse(cache.delete("a"))
+94 -21
View File
@@ -17,6 +17,7 @@ import shutil
import tempfile
import time
from pathlib import Path
from unittest import mock
from django.contrib.auth import get_user_model
from django.core import signing
@@ -35,6 +36,7 @@ from apps.library.models import (
TempUpload,
)
from apps.library.services import MEDIA_FILE_SALT
from apps.library.signing_urls import sign_payload
User = get_user_model()
@@ -121,21 +123,15 @@ class SecurityTestCase(TestCase):
return item
def old_signature(self, item, action="raw", age=3 * 86400):
"""A valid signature minted `age` seconds ago."""
real_time = signing.time
class Backdated:
def time(self):
return real_time.time() - age
try:
signing.time = Backdated()
return signing.dumps(
{"item": item.id, "user": self.users["sec-uploader"].id, "action": action},
salt=MEDIA_FILE_SALT,
)
finally:
signing.time = real_time
"""A signed media URL whose expiry is `age` seconds in the past."""
return signing.Signer(salt=MEDIA_FILE_SALT).sign_object(
{
"item": item.id,
"user": self.users["sec-uploader"].id,
"action": action,
"exp": int(time.time()) - age,
}
)
class GuestVisibilityTests(SecurityTestCase):
@@ -182,9 +178,9 @@ class GuestVisibilityTests(SecurityTestCase):
def test_authenticated_users_and_signed_urls_see_protected_items(self):
uploader = self.client_for("sec-uploader")
self.assertEqual(uploader.get(f"/api/files/J-{self.hidden.id}/").status_code, 200)
signed = signing.dumps(
signed = sign_payload(
{"item": self.hidden.id, "user": self.users["sec-uploader"].id, "action": "raw"},
salt=MEDIA_FILE_SALT,
MEDIA_FILE_SALT,
)
self.assertEqual(
self.guest.get(f"/api/files/J-{self.hidden.id}/raw/?sig={signed}").status_code,
@@ -192,9 +188,9 @@ class GuestVisibilityTests(SecurityTestCase):
)
def test_signature_integrity(self):
signed = signing.dumps(
signed = sign_payload(
{"item": self.hidden.id, "user": self.users["sec-uploader"].id, "action": "raw"},
salt=MEDIA_FILE_SALT,
MEDIA_FILE_SALT,
)
raw = f"/api/files/J-{self.hidden.id}/raw/"
thumbnail = f"/api/files/J-{self.hidden.id}/thumbnail/"
@@ -202,10 +198,29 @@ class GuestVisibilityTests(SecurityTestCase):
self.assertEqual(self.guest.get(f"{raw}?sig={signed[:-4]}AAAA").status_code, 404)
# Valid signature, wrong action.
self.assertEqual(self.guest.get(f"{thumbnail}?sig={signed}").status_code, 404)
# Expired signature (minted three days ago).
# Expired signature (expiry three days ago).
expired = self.old_signature(self.hidden)
self.assertEqual(self.guest.get(f"{raw}?sig={expired}").status_code, 404)
def test_signed_media_urls_are_stable_and_versioned(self):
"""The same item must keep the same URL across responses.
A per-second signature made browsers re-download every image on every
poll; the MD5 version parameter busts caches only when the file itself
changes (the optimize flow rewrites files under the same J-ID).
"""
item = self.visible
first = services.signed_media_url(item, self.users["sec-uploader"])
time.sleep(1.1)
second = services.signed_media_url(item, self.users["sec-uploader"])
self.assertEqual(first, second)
self.assertIn(f"v={item.md5}", first)
MediaItem.objects.filter(pk=item.pk).update(md5="b" * 32)
item.refresh_from_db()
self.assertNotEqual(
services.signed_media_url(item, self.users["sec-uploader"]), first
)
class RoleBoundaryTests(SecurityTestCase):
def test_non_uploader_is_read_only(self):
@@ -344,7 +359,13 @@ class PrivacyTests(SecurityTestCase):
)
self.assertEqual(self.guest.get(f"/api/uploads/{temp.id}/file/").status_code, 401)
self.assertNotIn(
str(temp.id), self.client_for("sec-uploader").get("/api/uploads/").content.decode()
str(temp.id),
self.client_for("sec-uploader").get("/api/uploads/").content.decode(),
)
# The board is per-user: staff only see their own staged uploads.
self.assertNotIn(
str(temp.id),
self.client_for("sec-staff").get("/api/uploads/").content.decode(),
)
def test_similarity_checks_are_private(self):
@@ -422,6 +443,36 @@ class ThrottleTests(SecurityTestCase):
{self.guest.get("/api/status/").status_code for _ in range(12)}, {200}
)
def test_signed_media_urls_are_not_throttled(self):
"""<img>/<video> tags fetch these without an Authorization header.
Regression: they were charged to the anonymous bucket, so galleries
and the fish-greeting download started returning 429 JSON instead of
the image bytes.
"""
item = self.make_item("throttle-media", owner=self.users["sec-uploader"])
codes = {
self.guest.get(f"/api/files/J-{item.id}/raw/").status_code
for _ in range(150)
}
self.assertEqual(codes, {200})
def test_staged_upload_files_are_not_throttled(self):
temp = TempUpload.objects.create(
user=self.users["sec-uploader"],
file=SimpleUploadedFile("throttle-temp.bin", b"staged"),
original_filename="throttle-temp.bin",
md5=hashlib.md5(b"throttle-temp").hexdigest(),
size=6,
)
signature = sign_payload(
{"temp": str(temp.id), "user": self.users["sec-uploader"].id},
services.UPLOAD_FILE_SALT,
)
url = f"/api/uploads/{temp.id}/file/?sig={signature}"
codes = {self.guest.get(url).status_code for _ in range(150)}
self.assertEqual(codes, {200})
class RemoteUrlTests(SecurityTestCase):
def test_allowlist(self):
@@ -433,6 +484,28 @@ class RemoteUrlTests(SecurityTestCase):
"https://static1.e621.net/data/x.png",
)
@mock.patch("requests.get")
def test_redirects_off_the_allowlist_are_refused(self, mocked_get):
redirect = mock.Mock(is_redirect=True, is_permanent_redirect=False)
redirect.headers = {"Location": "http://127.0.0.1:8000/health"}
mocked_get.return_value = redirect
with self.assertRaises(services.RemoteUrlError):
services.open_remote("https://static1.e621.net/x.png")
# The internal address was never requested: only the first hop was.
self.assertEqual(mocked_get.call_count, 1)
redirect.close.assert_called()
@mock.patch("requests.get")
def test_redirects_within_the_allowlist_are_followed(self, mocked_get):
redirect = mock.Mock(is_redirect=True, is_permanent_redirect=False)
redirect.headers = {"Location": "https://static2.e621.net/x.png"}
final = mock.Mock(is_redirect=False, is_permanent_redirect=False)
mocked_get.side_effect = [redirect, final]
self.assertIs(services.open_remote("https://static1.e621.net/x.png"), final)
self.assertEqual(mocked_get.call_count, 2)
def test_download_creation_rejects_internal_urls(self):
uploader = self.client_for("sec-uploader")
for url in ("http://127.0.0.1:1/", "http://192.168.1.1/", "file:///etc/passwd"):
@@ -0,0 +1,78 @@
"""The follow lists are complete sets, not pages.
The SPA uses `GET /api/follows/tags/` (and pools) for two things: deciding a
follow toggle's state, and rendering every card on the Followed page. With the
standard 48-item pagination that silently broke past 48 follows — a 49th tag
could not be followed (the toggle never saw it) and the page hid it. These
tests pin the unpaginated behaviour.
"""
from django.contrib.auth import get_user_model
from django.test import Client, TestCase
from rest_framework.authtoken.models import Token
from apps.follows.models import FollowedPool, FollowedTag
User = get_user_model()
class FollowListTests(TestCase):
def setUp(self):
self.user = User.objects.create_user(
username="follow-user", password="follow-pass-123456"
)
self.other = User.objects.create_user(
username="follow-other", password="follow-pass-123456"
)
self.client = self.api_client(self.user)
self.other_client = self.api_client(self.other)
def api_client(self, user):
client = Client()
client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=user).key}"
)
return client
def test_tag_list_is_not_paginated_past_48(self):
FollowedTag.objects.bulk_create(
[
FollowedTag(user=self.user, tag=f"tag_{index:03d}")
for index in range(60)
]
)
response = self.client.get("/api/follows/tags/")
self.assertEqual(response.status_code, 200)
data = response.json()
self.assertIsInstance(data, list, "the list must not be paginated")
self.assertEqual(len(data), 60)
self.assertEqual(data[0]["tag"], "tag_000")
self.assertEqual(data[-1]["tag"], "tag_059")
# The shape the SPA uses for its follow state.
self.assertIn("unseen_count", data[0])
self.assertIn("id", data[0])
def test_pool_list_is_not_paginated(self):
FollowedPool.objects.bulk_create(
[
FollowedPool(user=self.user, pool_id=59000 + index, name=f"pool {index}")
for index in range(55)
]
)
data = self.client.get("/api/follows/pools/").json()
self.assertIsInstance(data, list)
self.assertEqual(len(data), 55)
def test_lists_only_contain_the_callers_follows(self):
FollowedTag.objects.create(user=self.user, tag="mine")
FollowedTag.objects.create(user=self.other, tag="theirs")
mine = self.client.get("/api/follows/tags/").json()
theirs = self.other_client.get("/api/follows/tags/").json()
self.assertEqual([row["tag"] for row in mine], ["mine"])
self.assertEqual([row["tag"] for row in theirs], ["theirs"])
def test_anonymous_cannot_list(self):
self.assertEqual(Client().get("/api/follows/tags/").status_code, 401)
self.assertEqual(Client().get("/api/follows/pools/").status_code, 401)
+9 -2
View File
@@ -36,10 +36,16 @@ class FollowedTagViewSet(
mixins.DestroyModelMixin,
viewsets.GenericViewSet,
):
"""Tags the current user follows."""
"""Tags the current user follows.
Deliberately unpaginated: the SPA treats this as the complete set (follow
toggles read their state from it and the Followed page lists every card),
so a 48-item page silently broke following past 48 entries.
"""
serializer_class = FollowedTagSerializer
permission_classes = [IsAuthenticated]
pagination_class = None
http_method_names = ["get", "post", "delete", "head", "options"]
def get_queryset(self):
@@ -112,10 +118,11 @@ class FollowedPoolViewSet(
mixins.DestroyModelMixin,
viewsets.GenericViewSet,
):
"""Pools the current user follows."""
"""Pools the current user follows (unpaginated, like the tag list)."""
serializer_class = FollowedPoolSerializer
permission_classes = [IsAuthenticated]
pagination_class = None
http_method_names = ["get", "post", "delete", "head", "options"]
def get_queryset(self):
+264 -38
View File
@@ -1,19 +1,39 @@
"""Minimal e621 API client for server-side matching and metadata refresh.
The SPA talks to e621 directly for browsing; this client exists for work the
browser cannot do reliably: long batch scans, and requests tied to a library
item rather than an open page. It uses the requesting user's stored
credentials and a global throttle (e621 asks for at most two requests per
second).
browser cannot do reliably: long batch scans, staged-upload processing and
requests tied to a library item rather than an open page. It uses the
requesting user's stored credentials and a global throttle (e621 asks for at
most two requests per second, one per second sustained).
e621's load balancer also sheds load with 429s (sometimes with an HTML
"shedding" page instead of JSON) and the IQDB endpoint has its own, much
stricter throttle. Every call therefore retries with exponential backoff and
honours ``Retry-After``; only 401/403 are treated as fatal.
"""
import logging
import random
import threading
import time
from pathlib import Path
import requests
from django.conf import settings
REQUEST_INTERVAL = 0.5 # seconds between requests, per process
logger = logging.getLogger(__name__)
# Seconds between requests, per process. e621 allows 2/s hard and 1/s
# sustained; each gunicorn worker throttles on its own, so leave enough
# headroom that combined traffic does not trip the limit.
REQUEST_INTERVAL = 1.0
# IQDB is throttled far more aggressively than the rest of the API.
IQDB_INTERVAL = 2.0
MAX_ATTEMPTS = 4
BACKOFF_BASE = 2.0
BACKOFF_CAP = 60.0
RETRYABLE_STATUSES = {429, 500, 502, 503, 504}
class E621Error(Exception):
@@ -24,6 +44,14 @@ class E621NotFound(E621Error):
"""The requested post does not exist (HTTP 404)."""
class E621AuthError(E621Error):
"""e621 rejected the stored credentials (401/403)."""
class E621RateLimited(E621Error):
"""e621 shed load or throttled the request after every retry."""
_throttle_lock = threading.Lock()
_last_request_at = 0.0
@@ -32,56 +60,170 @@ def credentials_configured(user):
return bool(user is not None and getattr(user, "e621_configured", False))
def _wait_for_slot():
def _wait_for_slot(interval=REQUEST_INTERVAL):
global _last_request_at
with _throttle_lock:
delay = _last_request_at + REQUEST_INTERVAL - time.monotonic()
delay = _last_request_at + interval - time.monotonic()
if delay > 0:
time.sleep(delay)
_last_request_at = time.monotonic()
def _retry_delay(attempt, response=None):
"""Backoff for a retryable failure, honouring ``Retry-After``."""
if response is not None:
retry_after = response.headers.get("Retry-After")
if retry_after:
try:
return max(float(retry_after), 1.0)
except (TypeError, ValueError):
pass
delay = min(BACKOFF_BASE * (2**attempt), BACKOFF_CAP)
return delay + random.uniform(0, delay * 0.25)
def _request(
user,
method,
path,
*,
params=None,
data=None,
files=None,
timeout=30,
require_auth=True,
interval=REQUEST_INTERVAL,
attempts=MAX_ATTEMPTS,
):
"""One e621 call with retries; returns the parsed JSON payload.
``files`` may be a callable returning the multipart mapping, which is
called once per attempt: streamed uploads consume their file handle, so a
retry needs a freshly opened file.
Raises E621NotFound for 404s, E621AuthError for 401/403 and
E621RateLimited when e621 keeps shedding/throttling after every attempt.
"""
configured = credentials_configured(user)
if require_auth and not configured:
raise E621Error("Configure your e621 credentials in Account first.")
base = (getattr(user, "e621_base_url", "") or "https://e621.net").rstrip("/")
auth = (
(user.e621_username, user.e621_api_key_plain) if configured else None
)
url = f"{base}{path}"
last_error = None
for attempt in range(attempts):
request_files = files() if callable(files) else files
_wait_for_slot(interval)
response = None
try:
response = requests.request(
method,
url,
params=params,
data=data,
files=request_files,
auth=auth,
headers={"User-Agent": settings.USER_AGENT},
timeout=timeout,
)
except requests.RequestException as exc:
last_error = E621Error(f"Could not reach e621: {exc}")
else:
if response.status_code == 404:
raise E621NotFound(f"e621 returned 404 for {path}")
if response.status_code in {401, 403}:
raise E621AuthError(
f"e621 rejected the request ({response.status_code}). "
"Check the stored e621 credentials."
)
if response.status_code == 429:
# Throttles and load-shedding can arrive as JSON ({"message":
# "Throttled: ..."}) or as an HTML page.
message = ""
try:
payload = response.json()
except ValueError:
payload = None
if isinstance(payload, dict):
message = str(
payload.get("message") or payload.get("error") or ""
)
last_error = E621RateLimited(
message or f"e621 throttled the request for {path}"
)
elif response.status_code < 400:
try:
return response.json()
except ValueError:
# An HTML page with a 2xx status.
last_error = E621RateLimited(
f"e621 returned an unexpected {response.status_code} response."
)
elif response.status_code in RETRYABLE_STATUSES:
last_error = E621RateLimited(
f"e621 replied {response.status_code} for {path}"
)
else:
raise E621Error(f"e621 replied {response.status_code} for {path}")
finally:
_close_upload_files(request_files)
if attempt + 1 < attempts:
delay = _retry_delay(attempt, response)
logger.info(
"e621 %s %s failed (%s); retrying in %.1fs",
method,
path,
last_error,
delay,
)
time.sleep(delay)
if last_error is None:
last_error = E621Error("e621 request failed.")
raise last_error
def _close_upload_files(files):
"""Close the handles behind a multipart mapping (see _request)."""
if not isinstance(files, dict):
return
for value in files.values():
handle = value[1] if isinstance(value, tuple) and len(value) > 1 else value
close = getattr(handle, "close", None)
if close is not None:
try:
close()
except Exception: # noqa: BLE001 - closing must never mask errors
pass
def get(user, path, params=None, timeout=30, require_auth=True):
"""GET an e621 API path using the user's credentials.
Reads that e621 serves anonymously (searches, pools, tags) can pass
require_auth=False; matching endpoints keep requiring credentials.
Raises E621NotFound for 404s and E621Error for everything else that isn't
a 2xx, so callers never see requests exceptions.
"""
configured = credentials_configured(user)
if require_auth and not configured:
raise E621Error("Configure your e621 credentials in Account first.")
base = (getattr(user, "e621_base_url", "") or "https://e621.net").rstrip("/")
_wait_for_slot()
try:
response = requests.get(
f"{base}{path}",
params=params,
auth=(
(user.e621_username, user.e621_api_key_plain)
if configured
else None
),
headers={"User-Agent": settings.USER_AGENT},
timeout=timeout,
)
except requests.RequestException as exc:
raise E621Error(f"Could not reach e621: {exc}") from exc
if response.status_code == 404:
raise E621NotFound(f"e621 returned 404 for {path}")
if response.status_code >= 400:
raise E621Error(f"e621 replied {response.status_code} for {path}")
try:
return response.json()
except ValueError as exc:
raise E621Error("e621 returned an unexpected response.") from exc
return _request(
user,
"GET",
path,
params=params,
timeout=timeout,
require_auth=require_auth,
)
def find_post_by_md5(user, md5):
"""The e621 post with this exact MD5, or None."""
payload = get(user, "/posts.json", params={"tags": f"md5:{md5}", "limit": 1})
payload = get(
user,
"/posts.json",
params={"tags": f"md5:{md5}", "limit": 1},
require_auth=False,
)
posts = payload.get("posts") if isinstance(payload, dict) else None
if not posts:
return None
@@ -95,3 +237,87 @@ def fetch_post(user, post_id):
if not isinstance(post, dict):
raise E621Error("e621 returned an unexpected post payload.")
return post
def check_md5_batch(user, md5s):
"""Look many MD5s up in one posts.json query.
Returns ``{md5: post}`` for the ones e621 knows; missing MD5s are simply
absent. Works anonymously, like the original app's batch cache command.
"""
wanted = {str(value).strip().lower() for value in md5s if value}
if not wanted:
return {}
values = sorted(wanted)
payload = get(
user,
"/posts.json",
params={
"tags": f"md5:{','.join(values)}",
"limit": min(len(values), 320),
},
require_auth=False,
)
posts = payload.get("posts") if isinstance(payload, dict) else None
found = {}
for post in posts or []:
if not isinstance(post, dict):
continue
file_data = post.get("file") or {}
md5 = str(file_data.get("md5") or "").strip().lower()
if md5 in wanted:
found[md5] = post
return found
def fetch_posts_by_ids(user, ids):
"""Fetch many posts in one query (up to 320 ids). Missing ids are absent."""
values = sorted({int(value) for value in ids})
if not values:
return []
payload = get(
user,
"/posts.json",
params={
"tags": f"id:{','.join(str(value) for value in values)}",
"limit": min(len(values), 320),
},
require_auth=False,
)
posts = payload.get("posts") if isinstance(payload, dict) else None
return [post for post in posts or [] if isinstance(post, dict)]
def iqdb_search(user, path, timeout=60):
"""Reverse-image search one file through e621's IQDB endpoint.
Returns the legacy match list. Uses the extra-strict IQDB interval and
retries through e621's throttle; raises E621RateLimited when it persists.
"""
path = Path(path)
def open_file():
# A fresh handle per attempt: the stream is consumed by the request.
return {"search[file]": (path.name, open(path, "rb"))}
payload = _request(
user,
"POST",
"/iqdb_queries.json",
files=open_file,
timeout=timeout,
require_auth=False,
interval=IQDB_INTERVAL,
)
if isinstance(payload, list):
return payload
if isinstance(payload, dict):
matches = payload.get("matches")
if isinstance(matches, list):
return matches
# e621 answers its throttle with {"success": false, "message": ...}.
message = payload.get("message") or payload.get("error")
if message:
raise E621RateLimited(str(message))
raise E621Error("e621 returned an unexpected IQDB payload.")
return []
@@ -18,6 +18,9 @@ class Command(BaseCommand):
)
def handle(self, *args, **options):
from apps.library.upload_pipeline import reap_stale_claims
reap_stale_claims()
cutoff = timezone.now() - timedelta(hours=options["hours"])
queryset = TempUpload.objects.filter(created_at__lt=cutoff)
if not options["include_completed"]:
@@ -0,0 +1,68 @@
"""Process staged uploads: e621 MD5, visual similarity, IQDB.
Runs synchronously, unlike the daemon thread the API starts on demand. Useful
for tests, for a manual drain after an outage and for the scheduler if a
deployment wants a periodic safety net.
python manage.py process_uploads # every user with queued work, once
python manage.py process_uploads --user 3 # one user
python manage.py process_uploads --loop 60 # keep draining every 60s
"""
import time
from django.core.management.base import BaseCommand
from apps.library.models import TempUpload
from apps.library.upload_pipeline import (
MAX_ATTEMPTS,
OUTSTANDING_Q,
WORK_STATUSES,
reap_stale_claims,
run_pipeline,
)
class Command(BaseCommand):
help = "Run the staged-upload pipeline (MD5 -> visual similarity -> IQDB)."
def add_arguments(self, parser):
parser.add_argument(
"--user",
type=int,
default=None,
help="Only process this user id.",
)
parser.add_argument(
"--loop",
type=int,
default=0,
metavar="SECONDS",
help="Keep draining every SECONDS seconds instead of exiting.",
)
def handle(self, *args, **options):
interval = options["loop"] or 0
while True:
self.drain(user_id=options["user"])
if interval <= 0:
return
time.sleep(interval)
def drain(self, user_id=None):
reap_stale_claims()
queryset = (
TempUpload.objects.filter(status__in=WORK_STATUSES)
.filter(OUTSTANDING_Q)
.filter(attempts__lt=MAX_ATTEMPTS)
)
if user_id is not None:
queryset = queryset.filter(user_id=user_id)
user_ids = list(queryset.values_list("user_id", flat=True).distinct())
if not user_ids:
self.stdout.write("No staged uploads need processing.")
return
for value in user_ids:
self.stdout.write(f"Processing staged uploads for user {value}...")
run_pipeline(value)
self.stdout.write(self.style.SUCCESS(f"Processed {len(user_ids)} queue(s)."))
@@ -0,0 +1,81 @@
# Generated by Django 6.1.1 on 2026-09-21 13:27
import django.db.models.deletion
from django.conf import settings
from django.db import migrations, models
def backfill_pipeline_checks(apps, schema_editor):
"""Mark pre-pipeline rows as already MD5/visual-checked when they were.
Rows with an e621 post id (or already completed) clearly went through the
MD5 phase; rows with visual matches went through the visual phase.
Everything else stays unset so the new pipeline picks it up once after
deploy — a re-check of stale pending uploads is the desired behavior.
"""
TempUpload = apps.get_model("library", "TempUpload")
TempUpload.objects.filter(
models.Q(e621_post_id__isnull=False) | models.Q(status="completed")
).update(e621_checked_at=models.F("updated_at"))
TempUpload.objects.filter(visual_matches__isnull=False).update(
visual_checked_at=models.F("updated_at")
)
class Migration(migrations.Migration):
dependencies = [
('library', '0010_similaritycheck'),
migrations.swappable_dependency(settings.AUTH_USER_MODEL),
]
operations = [
migrations.AddField(
model_name='tempupload',
name='attempts',
field=models.PositiveSmallIntegerField(default=0),
),
migrations.AddField(
model_name='tempupload',
name='claimed_at',
field=models.DateTimeField(blank=True, db_index=True, null=True),
),
migrations.AddField(
model_name='tempupload',
name='e621_checked_at',
field=models.DateTimeField(blank=True, null=True),
),
migrations.AddField(
model_name='tempupload',
name='pipeline_error',
field=models.TextField(blank=True, default=''),
),
migrations.AddField(
model_name='tempupload',
name='visual_checked_at',
field=models.DateTimeField(blank=True, null=True),
),
migrations.CreateModel(
name='UploadRun',
fields=[
('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
('status', models.CharField(choices=[('idle', 'Idle'), ('running', 'Running'), ('paused', 'Paused'), ('error', 'Error')], default='idle', max_length=20)),
('phase', models.CharField(blank=True, choices=[('', 'None'), ('md5', 'e621 MD5'), ('visual', 'Visual similarity'), ('iqdb', 'IQDB')], default='', max_length=20)),
('total', models.IntegerField(default=0)),
('processed', models.IntegerField(default=0)),
('matched', models.IntegerField(default=0)),
('failed', models.IntegerField(default=0)),
('error', models.TextField(blank=True, default='')),
('started_at', models.DateTimeField(blank=True, null=True)),
('updated_at', models.DateTimeField(auto_now=True)),
('user', models.OneToOneField(on_delete=django.db.models.deletion.CASCADE, related_name='upload_run', to=settings.AUTH_USER_MODEL)),
],
options={
'ordering': ['-updated_at'],
},
),
migrations.RunPython(
code=backfill_pipeline_checks, reverse_code=migrations.RunPython.noop
),
]
@@ -0,0 +1,34 @@
# Generated by Django 6.1.1 on 2026-09-23
from django.db import migrations, models
def purge_completed_uploads(apps, schema_editor):
"""Completed staged uploads are notifications, not records.
The board used to keep every auto-uploaded/indexed row until it was
dismissed by hand, so they accumulated without bound (and bulk dismissal
is capped at 1000 ids). Completions now live in a small per-run feed, so
the old rows are removed here.
"""
TempUpload = apps.get_model("library", "TempUpload")
for temp in TempUpload.objects.filter(status="completed").iterator():
if temp.file:
temp.file.delete(save=False)
temp.delete()
class Migration(migrations.Migration):
dependencies = [
("library", "0011_upload_pipeline"),
]
operations = [
migrations.AddField(
model_name="uploadrun",
name="recent_completions",
field=models.JSONField(blank=True, default=list),
),
migrations.RunPython(purge_completed_uploads, migrations.RunPython.noop),
]
+70
View File
@@ -164,6 +164,16 @@ class TempUpload(models.Model):
on_delete=models.SET_NULL,
related_name="temp_uploads",
)
# Background pipeline bookkeeping (see apps/library/upload_pipeline.py).
# e621_checked_at/visual_checked_at are set once the corresponding phase
# ran, so "never checked" and "checked, nothing found" stay distinct.
e621_checked_at = models.DateTimeField(null=True, blank=True)
visual_checked_at = models.DateTimeField(null=True, blank=True)
# Worker claim for cross-process mutual exclusion; stale claims are
# reaped and the row queued again.
claimed_at = models.DateTimeField(null=True, blank=True, db_index=True)
pipeline_error = models.TextField(blank=True, default="")
attempts = models.PositiveSmallIntegerField(default=0)
created_at = models.DateTimeField(auto_now_add=True)
updated_at = models.DateTimeField(auto_now=True)
@@ -174,6 +184,66 @@ class TempUpload(models.Model):
return f"{self.original_filename} ({self.status})"
class UploadRun(models.Model):
"""Per-user state of the background upload pipeline.
One row per user acts as the cheap status source the SPA polls and as a
place for batch-level failures (broken e621 credentials, outages) that
would otherwise be repeated on every row.
"""
STATUS_IDLE = "idle"
STATUS_RUNNING = "running"
STATUS_PAUSED = "paused"
STATUS_ERROR = "error"
STATUS_CHOICES = [
(STATUS_IDLE, "Idle"),
(STATUS_RUNNING, "Running"),
(STATUS_PAUSED, "Paused"),
(STATUS_ERROR, "Error"),
]
PHASE_MD5 = "md5"
PHASE_VISUAL = "visual"
PHASE_IQDB = "iqdb"
PHASE_CHOICES = [
("", "None"),
(PHASE_MD5, "e621 MD5"),
(PHASE_VISUAL, "Visual similarity"),
(PHASE_IQDB, "IQDB"),
]
user = models.OneToOneField(
settings.AUTH_USER_MODEL,
on_delete=models.CASCADE,
related_name="upload_run",
)
status = models.CharField(
max_length=20, choices=STATUS_CHOICES, default=STATUS_IDLE
)
phase = models.CharField(
max_length=20, choices=PHASE_CHOICES, blank=True, default=""
)
total = models.IntegerField(default=0)
processed = models.IntegerField(default=0)
matched = models.IntegerField(default=0)
failed = models.IntegerField(default=0)
error = models.TextField(blank=True, default="")
# Rolling feed of recent completions for the upload board. Completed
# staged uploads are deleted as soon as they are indexed; this only tells
# the live page "filename -> J-x" while it watches. Bounded, never
# dismissed, and ignored by fresh page loads.
recent_completions = models.JSONField(default=list, blank=True)
started_at = models.DateTimeField(null=True, blank=True)
updated_at = models.DateTimeField(auto_now=True)
class Meta:
ordering = ["-updated_at"]
def __str__(self):
return f"Upload run for {self.user_id} ({self.status})"
class DownloadTask(models.Model):
"""A background 'Download to Library' job with progress tracking."""
+101 -5
View File
@@ -3,7 +3,6 @@ from datetime import timedelta
from pathlib import Path
from django.conf import settings
from django.core import signing
from rest_framework import serializers
from .models import (
@@ -20,6 +19,7 @@ from .services import (
VIDEO_EXTENSIONS,
signed_media_url,
)
from .signing_urls import sign_payload
class MediaLocationSerializer(serializers.ModelSerializer):
@@ -145,6 +145,12 @@ class TempUploadSerializer(serializers.ModelSerializer):
library_j_id = serializers.SerializerMethodField()
file_url = serializers.SerializerMethodField()
preview_url = serializers.SerializerMethodField()
md5_checked = serializers.SerializerMethodField()
visual_checked = serializers.SerializerMethodField()
iqdb_checked = serializers.SerializerMethodField()
processing = serializers.SerializerMethodField()
similar_count = serializers.SerializerMethodField()
visual_matches = serializers.SerializerMethodField()
class Meta:
model = TempUpload
@@ -165,6 +171,13 @@ class TempUploadSerializer(serializers.ModelSerializer):
"library_j_id",
"file_url",
"preview_url",
"pipeline_error",
"attempts",
"md5_checked",
"visual_checked",
"iqdb_checked",
"processing",
"similar_count",
"created_at",
"updated_at",
]
@@ -187,9 +200,9 @@ class TempUploadSerializer(serializers.ModelSerializer):
user = self._request_user()
if user is None:
return None
signature = signing.dumps(
signature = sign_payload(
{"temp": str(obj.id), "user": user.id},
salt=UPLOAD_FILE_SALT,
UPLOAD_FILE_SALT,
)
url = f"/api/uploads/{obj.id}/file/?sig={signature}"
request = self.context.get("request")
@@ -215,6 +228,89 @@ class TempUploadSerializer(serializers.ModelSerializer):
item, user, action, request=self.context.get("request")
)
def get_md5_checked(self, obj):
return obj.e621_checked_at is not None
def get_visual_checked(self, obj):
return obj.visual_checked_at is not None
def get_iqdb_checked(self, obj):
return obj.iqdb_data is not None
def get_processing(self, obj):
return obj.claimed_at is not None
def get_similar_count(self, obj):
return len(obj.iqdb_data or []) + len(obj.visual_matches or [])
@staticmethod
def _visual_item_id(entry):
if not isinstance(entry, dict):
return None
item_id = entry.get("item_id")
if item_id is None:
j_id = str(entry.get("j_id") or "")
if j_id.upper().startswith("J-"):
j_id = j_id[2:]
item_id = j_id if j_id.isdigit() else None
try:
return int(item_id)
except (TypeError, ValueError):
return None
def get_visual_matches(self, obj):
"""Rebuild match rows with fresh signed thumbnail URLs.
Storing the signed URL meant it aged out (or came from an older
signing scheme) and the "Already in your library" grid showed broken
tiles. The stored rows only carry the item reference now.
"""
entries = obj.visual_matches or []
if not entries:
return entries
wanted = {}
for entry in entries:
item_id = self._visual_item_id(entry)
if item_id is not None:
wanted[item_id] = None
items = MediaItem.objects.in_bulk(list(wanted))
user = self._request_user()
request = self.context.get("request")
matches = []
for entry in entries:
item_id = self._visual_item_id(entry)
item = items.get(item_id) if item_id is not None else None
if item is None:
continue
matches.append(
{
"j_id": f"J-{item.id}",
"filename": entry.get("filename") or item.md5,
"similarity": entry.get("similarity"),
"thumbnail_url": signed_media_url(
item, user, "thumbnail", request=request
),
}
)
return matches
class TempUploadListSerializer(TempUploadSerializer):
"""Compact staged-upload row for the board and the status polling.
Drops the heavy post/IQDB payloads (the metadata modal fetches the full
row) while keeping the pipeline flags the board renders per tile.
"""
class Meta(TempUploadSerializer.Meta):
fields = [
field
for field in TempUploadSerializer.Meta.fields
if field
not in {"e621_data", "iqdb_data", "visual_matches", "custom_tags", "custom_notes"}
]
read_only_fields = fields
class DownloadTaskSerializer(serializers.ModelSerializer):
task_id = serializers.UUIDField(source="id", read_only=True)
@@ -295,9 +391,9 @@ class SimilarityCheckSerializer(serializers.ModelSerializer):
user = self._request_user()
if user is None or not obj.file:
return None
signature = signing.dumps(
signature = sign_payload(
{"check": str(obj.id), "user": user.id},
salt=UPLOAD_FILE_SALT,
UPLOAD_FILE_SALT,
)
url = f"/api/similarity/{obj.id}/file/?sig={signature}"
request = self.context.get("request")
+138 -16
View File
@@ -6,16 +6,21 @@ import os
import re
import shutil
import subprocess
import uuid
from datetime import datetime, timezone
from pathlib import Path
from urllib.parse import urlencode
import imagehash
from django.conf import settings
from django.core import signing
from django.http import FileResponse, Http404, HttpResponse
from django.utils.cache import get_conditional_response
from django.utils.http import http_date
from django.utils.text import get_valid_filename
from PIL import Image
from PIL import Image, ImageOps
from .models import MediaItem, MediaLocation
from .signing_urls import sign_payload
logger = logging.getLogger(__name__)
@@ -37,6 +42,11 @@ UPLOAD_FILE_SALT = "j621.upload-file"
MEDIA_FILE_SALT = "j621.media-file"
CHUNK_SIZE = 1024 * 1024
RANGE_RE = re.compile(r"bytes=(\d*)-(\d*)$")
# Versioned media URLs are immutable, so they may sit in the browser cache for
# as long as the signature is guaranteed to stay valid (7 days).
MEDIA_CACHE_SECONDS = 6 * 86400
# Staged uploads and similarity files can be deleted at any moment.
TEMP_CACHE_SECONDS = 3600
def compute_md5(path):
@@ -78,14 +88,24 @@ def signed_media_url(item, user, action="raw", request=None):
With a ``request`` the URL is absolute, so the SPA also works when it is
served from a different origin; without one it stays relative.
The ``v`` parameter is the item's MD5: it busts the browser cache exactly
when the file is replaced (the optimize flow rewrites files under the same
J-ID), which is what lets the URL be cached for days instead of re-minted
on every response.
"""
path = f"/api/files/J-{item.id}/{action}/"
params = {}
if user is not None and getattr(user, "is_authenticated", False):
signature = signing.dumps(
{"item": item.id, "user": user.id, "action": action},
salt=MEDIA_FILE_SALT,
)
path = f"{path}?sig={signature}"
params = {
"v": item.md5,
"sig": sign_payload(
{"item": item.id, "user": user.id, "action": action},
MEDIA_FILE_SALT,
),
}
if params:
path = f"{path}?{urlencode(params)}"
if request is None:
return path
return request.build_absolute_uri(path)
@@ -207,12 +227,36 @@ class RangeFileWrapper:
self.file.close()
def serve_file(request, path, download=False):
"""Serve a file with HTTP range support (needed for video seeking)."""
def _apply_cache_headers(response, cache_control, etag, mtime):
response["Cache-Control"] = cache_control
response["ETag"] = etag
response["Last-Modified"] = http_date(mtime)
return response
def serve_file(request, path, download=False, *, max_age=TEMP_CACHE_SECONDS, immutable=False):
"""Serve a file with HTTP range support (needed for video seeking).
Responses carry validators (ETag/Last-Modified) and a private
``Cache-Control`` so browsers reuse media instead of re-downloading it on
every SPA poll. ``max_age``/``immutable`` are chosen by the caller: versioned
library media can be cached hard, staged files only briefly.
"""
path = Path(path)
if not path.is_file():
raise Http404
size = path.stat().st_size
stat = path.stat()
size = stat.st_size
etag = f'W/"{size:x}-{stat.st_mtime_ns:x}"'
last_modified = datetime.fromtimestamp(stat.st_mtime, tz=timezone.utc)
conditional = get_conditional_response(
request, etag=etag, last_modified=last_modified
)
if conditional is not None:
return conditional
cache_control = f"private, max-age={int(max_age)}"
if immutable:
cache_control += ", immutable"
content_type = mimetypes.guess_type(str(path))[0] or "application/octet-stream"
range_header = request.headers.get("Range", "").strip()
if range_header:
@@ -240,7 +284,9 @@ def serve_file(request, path, download=False):
response["Content-Length"] = str(length)
response["Content-Range"] = f"bytes {start}-{end}/{size}"
response["Accept-Ranges"] = "bytes"
return response
return _apply_cache_headers(
response, cache_control, etag, stat.st_mtime
)
response = FileResponse(
open(path, "rb"),
content_type=content_type,
@@ -248,7 +294,7 @@ def serve_file(request, path, download=False):
filename=path.name,
)
response["Accept-Ranges"] = "bytes"
return response
return _apply_cache_headers(response, cache_control, etag, stat.st_mtime)
class DownloadCancelled(Exception):
@@ -436,15 +482,38 @@ def sanitize_iqdb_results(results):
return cleaned
def _thumbnail_is_fresh(target, path):
"""True when the cached thumbnail exists and is at least as new as source."""
try:
stat = target.stat()
if stat.st_size <= 0:
return False
return stat.st_mtime >= os.path.getmtime(path)
except OSError:
return False
def _thumbs_dir():
thumbs_dir = Path(settings.MEDIA_ROOT) / "thumbs"
thumbs_dir.mkdir(parents=True, exist_ok=True)
return thumbs_dir
def generate_video_thumbnail(md5, path):
"""Extract a JPEG thumbnail from a video, cached under MEDIA_ROOT/thumbs."""
if not shutil.which("ffmpeg"):
return None
thumbs_dir = Path(settings.MEDIA_ROOT) / "thumbs"
thumbs_dir.mkdir(parents=True, exist_ok=True)
try:
thumbs_dir = _thumbs_dir()
except OSError:
logger.exception("Could not create the thumbnail folder")
return None
target = thumbs_dir / f"{md5}.jpg"
if target.exists() and target.stat().st_mtime >= os.path.getmtime(path):
if _thumbnail_is_fresh(target, path):
return target
# Write beside the target and move it into place, so a concurrent request
# can never read a half-written JPEG.
temp = thumbs_dir / f".{md5}.{uuid.uuid4().hex}.part.jpg"
command = [
"ffmpeg",
"-y",
@@ -458,10 +527,63 @@ def generate_video_thumbnail(md5, path):
"scale=480:-2",
"-loglevel",
"error",
str(target),
str(temp),
]
try:
subprocess.run(command, check=True, capture_output=True, timeout=60)
os.replace(temp, target)
except (subprocess.SubprocessError, OSError):
logger.exception("Could not build a video thumbnail for %s", path)
temp.unlink(missing_ok=True)
return None
return target if target.exists() else None
def generate_image_thumbnail(md5, path):
"""Downscale an image, cached under MEDIA_ROOT/thumbs like video thumbs.
The thumbnail action used to serve full-size originals for images; a
cached 480px JPEG keeps the library grid light without touching the
original file. Returns ``None`` when the source is missing or Pillow
cannot decode it, so callers can fall back to the original.
"""
try:
thumbs_dir = _thumbs_dir()
except OSError:
logger.exception("Could not create the thumbnail folder")
return None
target = thumbs_dir / f"{md5}.jpg"
if _thumbnail_is_fresh(target, path):
return target
temp = thumbs_dir / f".{md5}.{uuid.uuid4().hex}.part.jpg"
try:
with Image.open(path) as image:
# Animated formats: the first frame is the preview.
image.seek(0)
frame = ImageOps.exif_transpose(image) or image
frame = frame.convert("RGB")
frame.thumbnail((480, 480))
frame.save(temp, "JPEG", quality=82, optimize=True)
os.replace(temp, target)
except Exception: # noqa: BLE001 - previews must never break serving
logger.exception("Could not build an image thumbnail for %s", path)
temp.unlink(missing_ok=True)
return None
return target if target.exists() else None
def ensure_thumbnail(item):
"""Generate an item's cached thumbnail if it is missing or stale.
Warming thumbnails when a file is indexed keeps image decoding out of the
request path, where the upload pipeline's hashing used to starve it.
"""
location = item.locations.first()
if location is None:
return None
path = Path(location.path)
if not path.is_file():
return None
if path.suffix.lower() in VIDEO_EXTENSIONS:
return generate_video_thumbnail(item.md5, path)
return generate_image_thumbnail(item.md5, path)
+64
View File
@@ -0,0 +1,64 @@
"""Stable, expiring signatures for media URLs.
The SPA loads media with ``<img>``/``<video>`` tags, which cannot send the
API's ``Authorization`` header, so those URLs carry a signature instead. The
signature has to be *stable*: a URL that changes on every response makes the
browser treat every refetch as a new resource and re-download the file.
URLs are signed with a plain ``Signer`` (no per-second timestamp) plus an
explicit ``exp`` claim quantized to a bucket, so every request inside a bucket
mints the exact same URL. The URL rotates once per bucket and is valid for at
least ``URL_TTL_SECONDS`` and at most ``URL_TTL_SECONDS + URL_BUCKET_SECONDS``.
"""
import time
from django.core import signing
URL_TTL_SECONDS = 7 * 86400
URL_BUCKET_SECONDS = 24 * 3600
_BUCKETS = URL_TTL_SECONDS // URL_BUCKET_SECONDS
def _expiry(now=None):
current = time.time() if now is None else now
bucket = int(current // URL_BUCKET_SECONDS)
return (bucket + _BUCKETS + 1) * URL_BUCKET_SECONDS
def sign_payload(payload, salt, now=None):
"""Sign a payload with a stable, bucket-quantized expiry."""
return signing.Signer(salt=salt).sign_object(
{**payload, "exp": _expiry(now)}
)
def load_payload(signature, salt, legacy_max_age=86400):
"""Verify a signed payload; ``None`` when missing, tampered with or expired.
Legacy ``TimestampSigner`` values are still accepted for one release.
Detect them by their extra separator (``payload:timestamp:signature``):
a plain ``Signer`` accepts the HMAC a ``TimestampSigner`` computed over
``payload:timestamp`` and then chokes on the embedded timestamp while
decoding the JSON payload, which used to surface as a 500.
"""
if not signature:
return None
if signature.count(":") >= 2:
try:
return signing.TimestampSigner(salt=salt).unsign_object(
signature, max_age=legacy_max_age
)
except (signing.BadSignature, ValueError):
return None
try:
data = signing.Signer(salt=salt).unsign_object(signature)
except (signing.BadSignature, ValueError):
return None
if not isinstance(data, dict):
return None
try:
expired = int(data.get("exp", 0)) < time.time()
except (TypeError, ValueError):
return None
return None if expired else data
+8 -10
View File
@@ -11,7 +11,6 @@ from datetime import timedelta
from pathlib import Path
from django.conf import settings
from django.core import signing
from django.utils import timezone
from rest_framework import mixins, status, viewsets
from rest_framework.decorators import action
@@ -22,6 +21,7 @@ from rest_framework.response import Response
from . import services
from .models import MediaItem, SimilarityCheck
from .serializers import SimilarityCheckSerializer
from .signing_urls import load_payload
from .tools import item_brief
from .uploads import find_library_matches
@@ -136,7 +136,12 @@ class SimilarityCheckViewSet(
self.get_serializer(check).data, status=status.HTTP_201_CREATED
)
@action(detail=True, methods=["get", "head"], permission_classes=[AllowAny])
@action(
detail=True,
methods=["get", "head"],
permission_classes=[AllowAny],
throttle_classes=[],
)
def file(self, request, pk=None):
"""Serve the temp file; accepts a signed URL like staged uploads."""
check = None
@@ -144,14 +149,7 @@ class SimilarityCheckViewSet(
check = self.get_queryset().filter(pk=pk).first()
else:
signature = request.query_params.get("sig")
payload = None
if signature:
try:
payload = signing.loads(
signature, salt=services.UPLOAD_FILE_SALT, max_age=86400
)
except signing.BadSignature:
payload = None
payload = load_payload(signature, services.UPLOAD_FILE_SALT) if signature else None
if payload and payload.get("check") == str(pk):
check = SimilarityCheck.objects.filter(pk=pk).first()
if check is None or not check.file:
+99
View File
@@ -0,0 +1,99 @@
"""The server-side e621 client: batching, retries and IQDB stream handling."""
from pathlib import Path
from tempfile import TemporaryDirectory
from unittest import mock
from django.test import SimpleTestCase
from apps.library import e621
class FakeResponse:
def __init__(self, status_code, payload=None, headers=None):
self.status_code = status_code
self._payload = payload
self.headers = headers or {}
def json(self):
if self._payload is None:
raise ValueError("not json")
return self._payload
class E621ClientTests(SimpleTestCase):
def test_iqdb_search_reopens_the_file_on_retry(self):
bodies = []
def fake_request(method, url, **kwargs):
handle = kwargs["files"]["search[file]"][1]
bodies.append(handle.read())
if len(bodies) == 1:
return FakeResponse(
429, {"success": False, "message": "Throttled"}
)
return FakeResponse(200, [{"post_id": 1, "score": 90.0}])
with TemporaryDirectory() as tmp:
path = Path(tmp) / "x.png"
path.write_bytes(b"image-bytes")
with mock.patch.object(
e621.requests, "request", side_effect=fake_request
), mock.patch.object(e621.time, "sleep"), mock.patch.object(
e621, "_wait_for_slot"
):
result = e621.iqdb_search(None, path)
# Both attempts must send the full body, not the consumed handle.
self.assertEqual(bodies, [b"image-bytes", b"image-bytes"])
self.assertEqual(result, [{"post_id": 1, "score": 90.0}])
def test_check_md5_batch_keys_by_md5(self):
payload = {
"posts": [
{"id": 5, "file": {"md5": "a" * 32}},
{"id": 6, "file": {"md5": "b" * 32}},
]
}
with mock.patch.object(e621, "get", return_value=payload) as getter:
found = e621.check_md5_batch(None, ["A" * 32, "b" * 32])
self.assertEqual(set(found), {"a" * 32, "b" * 32})
params = getter.call_args.kwargs["params"]
self.assertTrue(params["tags"].startswith("md5:"))
self.assertEqual(params["limit"], 2)
def test_auth_errors_are_not_retried(self):
calls = []
def fake_request(*args, **kwargs):
calls.append(1)
return FakeResponse(403, {"error": "nope"})
with mock.patch.object(
e621.requests, "request", side_effect=fake_request
), mock.patch.object(e621, "_wait_for_slot"):
with self.assertRaises(e621.E621AuthError):
e621._request(None, "GET", "/posts.json", require_auth=False)
self.assertEqual(len(calls), 1)
def test_load_shedding_html_raises_rate_limited_after_retries(self):
with mock.patch.object(
e621.requests,
"request",
return_value=FakeResponse(200, None),
), mock.patch.object(e621.time, "sleep"), mock.patch.object(
e621, "_wait_for_slot"
):
with self.assertRaises(e621.E621RateLimited):
e621._request(
None, "GET", "/posts.json", require_auth=False, attempts=2
)
def test_fetch_posts_by_ids_queries_with_id_tag(self):
with mock.patch.object(
e621, "get", return_value={"posts": [{"id": 9}]}
) as getter:
posts = e621.fetch_posts_by_ids(None, [9])
self.assertEqual(posts, [{"id": 9}])
params = getter.call_args.kwargs["params"]
self.assertEqual(params["tags"], "id:9")
@@ -0,0 +1,188 @@
"""Signed media URLs must be stable, versioned and cacheable.
Regression: signatures embedded the current second, so every API response
re-minted every URL and browsers re-downloaded each image on every poll; the
file responses also carried no cache headers at all.
"""
import base64
import hashlib
import io
import shutil
import tempfile
import time
from pathlib import Path
from unittest import mock
from django.contrib.auth import get_user_model
from django.core import signing
from django.core.files.uploadedfile import SimpleUploadedFile
from django.test import Client, TestCase, override_settings
from PIL import Image
from rest_framework.authtoken.models import Token
from apps.library import services
from apps.library.models import MediaItem, MediaLocation, TempUpload
from apps.library.signing_urls import load_payload, sign_payload
User = get_user_model()
TINY_PNG = base64.b64decode(
"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg=="
)
def png_bytes(width=1200, height=800, color=(20, 120, 200)):
buffer = io.BytesIO()
Image.new("RGB", (width, height), color).save(buffer, format="PNG")
return buffer.getvalue()
class MediaCacheTests(TestCase):
@classmethod
def setUpClass(cls):
super().setUpClass()
cls._tmp = tempfile.mkdtemp(prefix="j621-cache-")
cls._media = Path(cls._tmp) / "media"
cls._watched = cls._media / "library"
cls._watched.mkdir(parents=True, exist_ok=True)
cls._settings = override_settings(
MEDIA_ROOT=str(cls._media), WATCHED_FOLDER=str(cls._watched)
)
cls._settings.enable()
@classmethod
def tearDownClass(cls):
cls._settings.disable()
shutil.rmtree(cls._tmp, ignore_errors=True)
super().tearDownClass()
def setUp(self):
self.user = User.objects.create_user(
username="cache-uploader", password="cache-pass-123456"
)
self.user.role = "uploader"
self.user.save(update_fields=["role"])
self.token = Token.objects.create(user=self.user).key
payload = png_bytes()
path = self._watched / "cache-image.png"
path.write_bytes(payload)
self.item = MediaItem.objects.create(
md5=hashlib.md5(payload).hexdigest(), size=len(payload)
)
MediaLocation.objects.create(
item=self.item, path=str(path), rel_path=path.name, mtime=time.time()
)
self.client = Client()
def signed(self, action):
return {
"sig": sign_payload(
{"item": self.item.id, "user": self.user.id, "action": action},
services.MEDIA_FILE_SALT,
)
}
def test_media_response_carries_cache_headers(self):
response = self.client.get(
f"/api/files/J-{self.item.id}/raw/", self.signed("raw")
)
self.assertEqual(response.status_code, 200)
self.assertIn("private", response["Cache-Control"])
self.assertIn(
f"max-age={services.MEDIA_CACHE_SECONDS}", response["Cache-Control"]
)
self.assertIn("immutable", response["Cache-Control"])
self.assertTrue(response["ETag"])
self.assertTrue(response["Last-Modified"])
def test_media_revalidation_returns_304(self):
url = f"/api/files/J-{self.item.id}/raw/"
first = self.client.get(url, self.signed("raw"))
second = self.client.get(
url, self.signed("raw"), HTTP_IF_NONE_MATCH=first["ETag"]
)
self.assertEqual(second.status_code, 304)
self.assertEqual(second.content, b"")
def test_image_thumbnail_is_generated_and_reused(self):
url = f"/api/files/J-{self.item.id}/thumbnail/"
response = self.client.get(url, self.signed("thumbnail"))
self.assertEqual(response.status_code, 200)
self.assertEqual(response["Content-Type"], "image/jpeg")
thumb = self._media / "thumbs" / f"{self.item.md5}.jpg"
self.assertTrue(thumb.exists())
with Image.open(thumb) as image:
self.assertLessEqual(max(image.size), 480)
before = thumb.stat().st_mtime_ns
self.client.get(url, self.signed("thumbnail"))
self.assertEqual(thumb.stat().st_mtime_ns, before)
def test_ensure_thumbnail_reuses_the_cache(self):
first = services.ensure_thumbnail(self.item)
self.assertIsNotNone(first)
self.assertTrue(first.exists())
mtime = first.stat().st_mtime_ns
second = services.ensure_thumbnail(self.item)
self.assertEqual(second, first)
self.assertEqual(second.stat().st_mtime_ns, mtime)
def test_thumbnail_of_a_missing_source_does_not_error(self):
"""Regression: getmtime() on a vanished source used to raise a 500."""
thumbs = self._media / "thumbs"
thumbs.mkdir(parents=True, exist_ok=True)
(thumbs / f"{self.item.md5}.jpg").write_bytes(b"stale")
Path(self.item.locations.first().path).unlink()
self.assertIsNone(services.ensure_thumbnail(self.item))
response = self.client.get(
f"/api/files/J-{self.item.id}/thumbnail/", self.signed("thumbnail")
)
self.assertEqual(response.status_code, 404)
def test_staged_files_cache_briefly(self):
temp = TempUpload.objects.create(
user=self.user,
file=SimpleUploadedFile("staged.png", TINY_PNG, content_type="image/png"),
original_filename="staged.png",
md5=hashlib.md5(b"staged").hexdigest(),
size=len(TINY_PNG),
)
signature = sign_payload(
{"temp": str(temp.id), "user": self.user.id}, services.UPLOAD_FILE_SALT
)
response = self.client.get(
f"/api/uploads/{temp.id}/file/", {"sig": signature}
)
self.assertEqual(response.status_code, 200)
self.assertIn(
f"max-age={services.TEMP_CACHE_SECONDS}", response["Cache-Control"]
)
self.assertNotIn("immutable", response["Cache-Control"])
def test_legacy_timestamp_signatures_are_accepted(self):
"""URLs minted before the stable scheme must not 500.
A TimestampSigner HMAC also passes a plain Signer's check, so the
embedded timestamp used to reach the JSON decoder and blow up.
"""
payload = {"item": self.item.id, "user": self.user.id, "action": "raw"}
legacy = signing.dumps(payload, salt=services.MEDIA_FILE_SALT)
self.assertEqual(
load_payload(legacy, services.MEDIA_FILE_SALT)["item"], self.item.id
)
response = self.client.get(
f"/api/files/J-{self.item.id}/raw/", {"sig": legacy}
)
self.assertEqual(response.status_code, 200)
def test_expired_and_malformed_signatures_return_none(self):
payload = {"item": self.item.id, "user": self.user.id, "action": "raw"}
with mock.patch.object(signing, "time") as clock:
clock.time.return_value = time.time() - 3 * 86400
expired = signing.dumps(payload, salt=services.MEDIA_FILE_SALT)
self.assertIsNone(load_payload(expired, services.MEDIA_FILE_SALT))
self.assertIsNone(load_payload("bogus", services.MEDIA_FILE_SALT))
self.assertIsNone(load_payload("a:b", services.MEDIA_FILE_SALT))
self.assertIsNone(load_payload("", services.MEDIA_FILE_SALT))
+137
View File
@@ -0,0 +1,137 @@
"""Tests for the random image endpoint (`/api/random/`, `/random`).
Used by the SPA's Random page and by shell greeting scripts (fish_greeting
with fastfetch), so the response contract matters:
* JSON with an absolute, directly fetchable URL,
* signed for authenticated callers (image viewers send no headers),
* fastfetch mode restricted to png/jpg/gif,
* rating filters and guest visibility applied server-side.
"""
import hashlib
import shutil
import tempfile
import time
from pathlib import Path
from django.contrib.auth import get_user_model
from django.test import Client, TestCase, override_settings
from rest_framework.authtoken.models import Token
from apps.library.models import MediaItem, MediaLocation
User = get_user_model()
IMAGE_EXTENSIONS = {"png", "jpg", "jpeg", "gif", "webp", "apng"}
FASTFETCH_EXTENSIONS = {"png", "jpg", "jpeg", "gif"}
class RandomItemTests(TestCase):
@classmethod
def setUpClass(cls):
super().setUpClass()
cls._tmp = tempfile.mkdtemp(prefix="j621-random-")
cls._watched = Path(cls._tmp) / "library"
cls._watched.mkdir(parents=True, exist_ok=True)
cls._settings = override_settings(
MEDIA_ROOT=cls._tmp, WATCHED_FOLDER=str(cls._watched)
)
cls._settings.enable()
@classmethod
def tearDownClass(cls):
cls._settings.disable()
shutil.rmtree(cls._tmp, ignore_errors=True)
super().tearDownClass()
def setUp(self):
self.user = User.objects.create_user(
username="random-user", password="random-pass-123456"
)
token = Token.objects.create(user=self.user)
self.authed = Client()
self.authed.defaults["HTTP_AUTHORIZATION"] = f"Token {token.key}"
self.guest = Client()
def make_item(self, label, extension, rating, *, hidden=False):
path = self._watched / f"{label}.{extension}"
path.write_bytes(b"random-" + label.encode())
item = MediaItem.objects.create(
md5=hashlib.md5(label.encode()).hexdigest(),
size=path.stat().st_size,
rating=rating,
uploaded_by=self.user,
)
MediaLocation.objects.create(
item=item, path=str(path), rel_path=path.name, mtime=time.time()
)
if hidden:
MediaItem.objects.filter(pk=item.pk).update(hidden_from_guests=True)
return item
def test_returns_image_with_signed_absolute_url(self):
item = self.make_item("plain", "png", "s")
response = self.authed.get("/api/random/")
self.assertEqual(response.status_code, 200)
data = response.json()
self.assertEqual(data["j_id"], f"J-{item.id}")
self.assertEqual(data["extension"], "png")
self.assertEqual(data["kind"], "image")
self.assertEqual(data["rating"], "s")
self.assertTrue(data["url"].startswith("http"))
self.assertIn("sig=", data["url"])
self.assertIn("download=1", data["download_url"])
self.assertFalse(data["fastfetch"])
def test_guest_url_is_unsigned_and_still_serves(self):
self.make_item("guest", "jpg", "s")
data = self.guest.get("/api/random/").json()
self.assertNotIn("sig=", data["url"])
path = data["url"].replace("http://testserver", "")
self.assertEqual(self.guest.get(path).status_code, 200)
def test_default_mode_returns_images_only(self):
self.make_item("movie", "mp4", "s")
self.make_item("picture", "webp", "s")
for _ in range(10):
extension = self.authed.get("/api/random/").json()["extension"]
self.assertIn(extension, IMAGE_EXTENSIONS)
def test_fastfetch_mode_flag_and_user_agent_restrict_formats(self):
self.make_item("movie", "mp4", "s")
self.make_item("modern", "webp", "s")
self.make_item("picture", "png", "s")
self.make_item("animation", "gif", "s")
attempts = [("flag", {"fastfetch": "1"}, {}), ("ua", {}, {"HTTP_USER_AGENT": "fastfetch/2.18.1"})]
for label, params, headers in attempts:
for _ in range(15):
response = self.authed.get("/api/random/", params, **headers)
self.assertEqual(response.status_code, 200, label)
data = response.json()
self.assertIn(data["extension"], FASTFETCH_EXTENSIONS, label)
self.assertTrue(data["fastfetch"], label)
def test_rating_filter(self):
self.make_item("safe", "png", "s")
explicit = self.make_item("explicit", "png", "e")
for _ in range(10):
data = self.authed.get("/api/random/", {"rating": "e"}).json()
self.assertEqual(data["j_id"], f"J-{explicit.id}")
self.assertEqual(data["rating"], "e")
self.assertEqual(self.authed.get("/api/random/", {"rating": "q"}).status_code, 404)
def test_guests_never_receive_hidden_items(self):
self.make_item("hidden", "png", "s", hidden=True)
self.assertEqual(self.guest.get("/api/random/").status_code, 404)
self.assertEqual(self.authed.get("/api/random/").status_code, 200)
def test_no_match_returns_404(self):
self.make_item("movie", "mp4", "s") # images only
self.assertEqual(self.authed.get("/api/random/").status_code, 404)
def test_short_top_level_alias(self):
self.make_item("alias", "gif", "s")
self.assertEqual(self.guest.get("/random/").status_code, 200)
self.assertEqual(self.guest.get("/random").status_code, 200)
+788
View File
@@ -0,0 +1,788 @@
"""Staged uploads: listing, phases (MD5/visual/IQDB) and the bulk tool."""
import base64
import hashlib
import io
import json
import shutil
import tempfile
from datetime import timedelta
from pathlib import Path
from unittest import mock
from django.contrib.auth import get_user_model
from django.core.files.uploadedfile import SimpleUploadedFile
from django.test import Client, TestCase, override_settings
from django.utils import timezone
from PIL import Image
from rest_framework.authtoken.models import Token
from apps.library import upload_pipeline
from apps.library.models import MediaItem, TempUpload, UploadRun
User = get_user_model()
# 1x1 transparent PNG so indexing/hashing has a real image to chew on.
TINY_PNG = base64.b64decode(
"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg=="
)
def similar_png(color=(200, 30, 40)):
"""A 1x1 PNG with the same visual hashes but a different MD5."""
buffer = io.BytesIO()
Image.new("RGB", (1, 1), color).save(buffer, format="PNG")
return buffer.getvalue()
def jpost(client, path, body=None):
return client.post(path, data=json.dumps(body or {}), content_type="application/json")
class TempUploadListTests(TestCase):
def setUp(self):
self.uploader = User.objects.create_user(
username="upload-user", password="upload-pass-123456"
)
self.uploader.role = "uploader"
self.uploader.save(update_fields=["role"])
self.other = User.objects.create_user(
username="upload-other", password="upload-pass-123456"
)
self.other.role = "uploader"
self.other.save(update_fields=["role"])
def api_client(self, user):
client = Client()
client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=user).key}"
)
return client
def test_list_is_not_paginated_past_48(self):
TempUpload.objects.bulk_create(
[
TempUpload(
user=self.uploader,
original_filename=f"file_{index:03d}.png",
md5=hashlib.md5(f"file-{index}".encode()).hexdigest(),
size=index,
)
for index in range(69)
]
)
response = self.api_client(self.uploader).get("/api/uploads/")
self.assertEqual(response.status_code, 200)
data = response.json()
self.assertIsInstance(data, list, "the list must not be paginated")
self.assertEqual(len(data), 69)
def test_list_only_contains_the_callers_uploads(self):
TempUpload.objects.create(
user=self.uploader,
original_filename="mine.png",
md5=hashlib.md5(b"mine").hexdigest(),
size=1,
)
TempUpload.objects.create(
user=self.other,
original_filename="theirs.png",
md5=hashlib.md5(b"theirs").hexdigest(),
size=1,
)
mine = self.api_client(self.uploader).get("/api/uploads/").json()
self.assertEqual([row["original_filename"] for row in mine], ["mine.png"])
def test_anonymous_cannot_list(self):
self.assertEqual(Client().get("/api/uploads/").status_code, 401)
@override_settings(UPLOAD_PIPELINE_AUTOSTART=False)
class StagedUploadWorkflowTests(TestCase):
@classmethod
def setUpClass(cls):
super().setUpClass()
cls._tmp = tempfile.mkdtemp(prefix="j621-bulk-")
cls._watched = Path(cls._tmp) / "library"
cls._watched.mkdir(parents=True, exist_ok=True)
cls._settings = override_settings(
MEDIA_ROOT=cls._tmp, WATCHED_FOLDER=str(cls._watched)
)
cls._settings.enable()
@classmethod
def tearDownClass(cls):
cls._settings.disable()
shutil.rmtree(cls._tmp, ignore_errors=True)
super().tearDownClass()
def setUp(self):
self.uploader = User.objects.create_user(
username="bulk-uploader", password="bulk-pass-123456"
)
self.uploader.role = "uploader"
self.uploader.save(update_fields=["role"])
self.other = User.objects.create_user(
username="bulk-other", password="bulk-pass-123456"
)
self.other.role = "uploader"
self.other.save(update_fields=["role"])
def api_client(self, user):
client = Client()
client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=user).key}"
)
return client
def make_temp(
self, user, label, status=TempUpload.STATUS_PENDING, payload=None
):
return TempUpload.objects.create(
user=user,
file=SimpleUploadedFile(
f"{label}.png", payload or TINY_PNG, content_type="image/png"
),
original_filename=f"{label}.png",
md5=hashlib.md5(label.encode()).hexdigest(),
size=len(payload or TINY_PNG),
status=status,
)
def resolve_bulk(self, client, ids, rating):
return jpost(
client,
"/api/uploads/resolve-bulk/",
{"temp_ids": [str(value) for value in ids], "rating": rating},
)
def test_resolves_selected_uploads_with_the_rating(self):
client = self.api_client(self.uploader)
first = self.make_temp(self.uploader, "one")
second = self.make_temp(self.uploader, "two")
untouched = self.make_temp(self.uploader, "three")
response = self.resolve_bulk(client, [first.id, second.id], "q")
self.assertEqual(response.status_code, 200)
body = response.json()
self.assertEqual(len(body["resolved"]), 2)
self.assertEqual(body["errors"], [])
# Completed uploads are notifications now: the staged rows are gone
# and the live feed carries the filename -> J-ID mapping.
self.assertFalse(
TempUpload.objects.filter(pk__in=[first.id, second.id]).exists()
)
feed = UploadRun.objects.get(user=self.uploader).recent_completions
self.assertEqual(
{entry["filename"] for entry in feed}, {"one.png", "two.png"}
)
for item in MediaItem.objects.all():
self.assertEqual(item.rating, "q")
self.assertEqual(item.uploaded_by_id, self.uploader.id)
untouched.refresh_from_db()
self.assertEqual(untouched.status, TempUpload.STATUS_PENDING)
def test_rejects_bad_input(self):
client = self.api_client(self.uploader)
temp = self.make_temp(self.uploader, "input")
self.assertEqual(self.resolve_bulk(client, [], "s").status_code, 400)
self.assertEqual(self.resolve_bulk(client, [temp.id], "").status_code, 400)
self.assertEqual(self.resolve_bulk(client, [temp.id], "x").status_code, 400)
self.assertEqual(
jpost(
client,
"/api/uploads/resolve-bulk/",
{"temp_ids": ["not-a-uuid"], "rating": "s"},
).status_code,
400,
)
def test_other_users_uploads_are_left_alone(self):
client = self.api_client(self.uploader)
theirs = self.make_temp(self.other, "theirs")
response = self.resolve_bulk(client, [theirs.id], "s")
self.assertEqual(response.status_code, 200)
body = response.json()
self.assertEqual(body["resolved"], [])
self.assertEqual(body["errors"][0]["error"], "not found")
theirs.refresh_from_db()
self.assertEqual(theirs.status, TempUpload.STATUS_PENDING)
def test_completed_uploads_report_an_error_but_others_resolve(self):
client = self.api_client(self.uploader)
done = self.make_temp(self.uploader, "done", status=TempUpload.STATUS_COMPLETED)
pending = self.make_temp(self.uploader, "pending")
response = self.resolve_bulk(client, [done.id, pending.id], "e")
body = response.json()
self.assertEqual(len(body["resolved"]), 1)
self.assertEqual(body["errors"][0]["error"], "already in the library")
self.assertEqual(MediaItem.objects.count(), 1)
def upload_via_api(self, client, label="phase.png"):
return client.post(
"/api/uploads/",
{
"file": SimpleUploadedFile(
label, similar_png(), content_type="image/png"
)
},
)
def seed_library_item(self, client, label):
seed = self.make_temp(self.uploader, label)
response = self.resolve_bulk(client, [seed.id], "s")
self.assertEqual(response.status_code, 200)
return seed
def test_duplicate_upload_returns_a_completion_without_a_record(self):
client = self.api_client(self.uploader)
first = self.upload_via_api(client, "same.png")
self.assertEqual(first.status_code, 201)
# Index the first upload so the second one is byte-identical.
seed = TempUpload.objects.get(pk=first.json()["temp_id"])
self.resolve_bulk(client, [seed.id], "s")
response = self.upload_via_api(client, "same.png")
self.assertEqual(response.status_code, 201)
body = response.json()
self.assertEqual(body["status"], TempUpload.STATUS_COMPLETED)
self.assertEqual(body["resolution"], TempUpload.RESOLUTION_DUPLICATE)
self.assertTrue(body["library_j_id"].startswith("J-"))
# Duplicates never become board records; the feed announces them.
self.assertFalse(TempUpload.objects.filter(pk=body["temp_id"]).exists())
feed = UploadRun.objects.get(user=self.uploader).recent_completions
self.assertEqual(feed[-1]["filename"], "same.png")
def test_upload_defers_visual_similarity_to_its_phase(self):
client = self.api_client(self.uploader)
self.seed_library_item(client, "seed-defer")
response = self.upload_via_api(client)
self.assertEqual(response.status_code, 201)
body = response.json()
# Uploading must not run the expensive library scan.
self.assertEqual(body["status"], TempUpload.STATUS_PENDING)
self.assertFalse(body["visual_matches"])
def test_visual_match_phase_flags_similar_library_items(self):
client = self.api_client(self.uploader)
self.seed_library_item(client, "seed-visual")
temp = self.make_temp(self.uploader, "check-visual")
response = client.post(f"/api/uploads/{temp.id}/visual-match/")
self.assertEqual(response.status_code, 200)
body = response.json()
self.assertEqual(body["status"], TempUpload.STATUS_VISUAL_MATCH)
self.assertGreaterEqual(len(body["visual_matches"]), 1)
temp.refresh_from_db()
self.assertEqual(temp.status, TempUpload.STATUS_VISUAL_MATCH)
def test_visual_match_phase_with_empty_library_stays_pending(self):
client = self.api_client(self.uploader)
temp = self.make_temp(self.uploader, "no-matches")
response = client.post(f"/api/uploads/{temp.id}/visual-match/")
self.assertEqual(response.status_code, 200)
body = response.json()
self.assertEqual(body["visual_matches"], [])
self.assertEqual(body["status"], TempUpload.STATUS_PENDING)
def test_detail_resigns_stored_visual_match_urls(self):
"""Stored matches carry only the item reference; URLs are re-minted.
Embedding the signed URL meant it expired (or used an older signing
scheme) and the modal showed alt text instead of thumbnails.
"""
client = self.api_client(self.uploader)
self.seed_library_item(client, "seed-resign")
item = MediaItem.objects.get()
temp = self.make_temp(self.uploader, "resign")
TempUpload.objects.filter(pk=temp.pk).update(
visual_matches=[
{
"item_id": item.id,
"j_id": f"J-{item.id}",
"filename": "seed-resign.png",
"similarity": 96.5,
"thumbnail_url": "/api/files/J-x/thumbnail/?sig=stale",
},
{"item_id": 999999, "j_id": "J-999999", "filename": "gone.png"},
]
)
body = client.get(f"/api/uploads/{temp.id}/").json()
self.assertEqual(len(body["visual_matches"]), 1)
match = body["visual_matches"][0]
self.assertEqual(match["j_id"], f"J-{item.id}")
self.assertEqual(match["similarity"], 96.5)
self.assertNotIn("stale", match["thumbnail_url"])
self.assertIn("/thumbnail/", match["thumbnail_url"])
self.assertIn(f"v={item.md5}", match["thumbnail_url"])
def test_visual_match_phase_rejects_completed_uploads(self):
client = self.api_client(self.uploader)
temp = self.make_temp(
self.uploader, "done-visual", status=TempUpload.STATUS_COMPLETED
)
response = client.post(f"/api/uploads/{temp.id}/visual-match/")
self.assertEqual(response.status_code, 400)
def test_link_bulk_moves_a_batch_with_one_call(self):
client = self.api_client(self.uploader)
first = self.make_temp(self.uploader, "bulk-link-1")
second = self.make_temp(
self.uploader, "bulk-link-2", payload=similar_png((10, 200, 30))
)
response = jpost(
client,
"/api/uploads/link-bulk/",
{
"links": [
{
"temp_id": str(first.id),
"post": {
"id": 900001,
"rating": "s",
"file": {
"md5": first.md5,
# Identical MD5: the staged file is indexed
# without downloading the URL.
"url": "https://static1.e621.net/data/fake-1.png",
},
},
},
{
"temp_id": str(second.id),
"post": {
"id": 900002,
"rating": "q",
"file": {
"md5": second.md5,
"url": "https://static1.e621.net/data/fake-2.png",
},
},
},
]
},
)
self.assertEqual(response.status_code, 200)
body = response.json()
self.assertEqual(body["errors"], [])
self.assertEqual(len(body["updated"]), 2)
for entry, post_id in zip(body["updated"], (900001, 900002)):
self.assertEqual(entry["resolution"], TempUpload.RESOLUTION_AUTO_MD5)
self.assertEqual(entry["e621_post_id"], post_id)
self.assertTrue(entry["library_j_id"].startswith("J-"))
# Indexed uploads no longer leave a board record; the completion feed
# carries them for the live page instead.
self.assertFalse(TempUpload.objects.exists())
self.assertEqual(
{item.e621_post_id for item in MediaItem.objects.all()},
{900001, 900002},
)
feed = UploadRun.objects.get(user=self.uploader).recent_completions
self.assertEqual(len(feed), 2)
self.assertEqual(
{entry["filename"] for entry in feed},
{"bulk-link-1.png", "bulk-link-2.png"},
)
class IqdbRecordingTests(TestCase):
"""The modal needs to tell "checked, no match" from "never checked"."""
@classmethod
def setUpClass(cls):
super().setUpClass()
cls._tmp = tempfile.mkdtemp(prefix="j621-iqdb-")
cls._settings = override_settings(MEDIA_ROOT=cls._tmp)
cls._settings.enable()
@classmethod
def tearDownClass(cls):
cls._settings.disable()
shutil.rmtree(cls._tmp, ignore_errors=True)
super().tearDownClass()
def setUp(self):
self.uploader = User.objects.create_user(
username="iqdb-uploader", password="iqdb-pass-123456"
)
self.uploader.role = "uploader"
self.uploader.save(update_fields=["role"])
self.client = Client()
self.client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=self.uploader).key}"
)
def make_temp(self):
return TempUpload.objects.create(
user=self.uploader,
file=SimpleUploadedFile(
"checked.png", TINY_PNG, content_type="image/png"
),
original_filename="checked.png",
md5=hashlib.md5(b"checked").hexdigest(),
size=len(TINY_PNG),
)
def test_empty_result_records_the_check_without_a_match(self):
temp = self.make_temp()
response = jpost(
self.client, f"/api/uploads/{temp.id}/iqdb/", {"results": []}
)
self.assertEqual(response.status_code, 200)
temp.refresh_from_db()
self.assertEqual(temp.iqdb_data, [])
# No candidates must not masquerade as a visual match.
self.assertEqual(temp.status, TempUpload.STATUS_PENDING)
def test_candidates_are_stored_and_flag_a_visual_match(self):
temp = self.make_temp()
response = jpost(
self.client,
f"/api/uploads/{temp.id}/iqdb/",
{
"results": [
{
"post_id": 123,
"score": 91.5,
"preview_url": "https://static1.e621.net/data/preview/ab/cd/x.jpg",
"rating": "q",
"md5": "a" * 32,
"score_total": 12,
"fav_count": 3,
"width": 800,
"height": 600,
"tags_preview": ["canine", "solo"],
}
]
},
)
self.assertEqual(response.status_code, 200)
temp.refresh_from_db()
self.assertEqual(len(temp.iqdb_data), 1)
self.assertEqual(temp.iqdb_data[0]["post_id"], 123)
self.assertEqual(temp.status, TempUpload.STATUS_VISUAL_MATCH)
def test_rejects_a_non_list_payload(self):
temp = self.make_temp()
response = jpost(
self.client, f"/api/uploads/{temp.id}/iqdb/", {"results": "nope"}
)
self.assertEqual(response.status_code, 400)
@override_settings(UPLOAD_PIPELINE_AUTOSTART=False)
class UploadPipelineTests(TestCase):
"""The server-side MD5 -> visual -> IQDB queue and its board API."""
@classmethod
def setUpClass(cls):
super().setUpClass()
cls._tmp = tempfile.mkdtemp(prefix="j621-pipeline-")
cls._watched = Path(cls._tmp) / "library"
cls._watched.mkdir(parents=True, exist_ok=True)
cls._settings = override_settings(
MEDIA_ROOT=cls._tmp, WATCHED_FOLDER=str(cls._watched)
)
cls._settings.enable()
@classmethod
def tearDownClass(cls):
cls._settings.disable()
shutil.rmtree(cls._tmp, ignore_errors=True)
super().tearDownClass()
def setUp(self):
self.uploader = User.objects.create_user(
username="pipe-uploader", password="pipe-pass-123456"
)
self.uploader.role = "uploader"
self.uploader.save(update_fields=["role"])
self.other = User.objects.create_user(
username="pipe-other", password="pipe-pass-123456"
)
self.other.role = "uploader"
self.other.save(update_fields=["role"])
def api_client(self, user):
client = Client()
client.defaults["HTTP_AUTHORIZATION"] = (
f"Token {Token.objects.create(user=user).key}"
)
return client
def stage(self, user=None, label="file", payload=None, filename=None):
user = user or self.uploader
payload = payload or TINY_PNG
name = filename or f"{label}.png"
return TempUpload.objects.create(
user=user,
file=SimpleUploadedFile(name, payload, content_type="image/png"),
original_filename=name,
md5=hashlib.md5(payload + label.encode()).hexdigest(),
size=len(payload),
)
def test_md5_match_auto_imports_the_file(self):
temp = self.stage(label="match")
temp_id = temp.id
post = {
"id": 123456,
"rating": "s",
"file": {
"md5": temp.md5,
"url": "https://static1.e621.net/data/m.png",
},
"tags": {"general": ["canine"]},
}
with mock.patch.object(
upload_pipeline.e621,
"check_md5_batch",
return_value={temp.md5: post},
):
upload_pipeline.run_pipeline(self.uploader.id)
# The indexed upload leaves no board record; the feed reports it.
self.assertFalse(TempUpload.objects.filter(pk=temp_id).exists())
item = MediaItem.objects.get(e621_post_id=123456)
self.assertEqual(item.uploaded_by_id, self.uploader.id)
run = UploadRun.objects.get(user=self.uploader)
self.assertEqual(run.status, UploadRun.STATUS_IDLE)
self.assertEqual(run.matched, 1)
self.assertEqual(run.processed, 1)
self.assertEqual(len(run.recent_completions), 1)
entry = run.recent_completions[0]
self.assertEqual(entry["id"], str(temp_id))
self.assertEqual(entry["item_id"], item.id)
self.assertEqual(entry["filename"], "match.png")
self.assertEqual(entry["resolution"], TempUpload.RESOLUTION_AUTO_MD5)
def test_status_reports_the_completion_feed(self):
temp = self.stage(label="feed")
client = self.api_client(self.uploader)
post = {
"id": 654321,
"rating": "s",
"file": {"md5": temp.md5, "url": "https://static1.e621.net/data/f.png"},
}
with mock.patch.object(
upload_pipeline.e621,
"check_md5_batch",
return_value={temp.md5: post},
):
upload_pipeline.run_pipeline(self.uploader.id)
body = client.get("/api/uploads/status/").json()
completions = body["recent_completions"]
self.assertEqual(len(completions), 1)
self.assertEqual(completions[0]["filename"], "feed.png")
self.assertEqual(completions[0]["j_id"], f"J-{MediaItem.objects.get().id}")
self.assertIn("/thumbnail/", completions[0]["thumbnail_url"])
def test_unmatched_file_runs_every_phase(self):
temp = self.stage(label="nomatch")
raw_iqdb = [
{
"post_id": 777,
"score": 91.0,
"post": {
"id": 777,
"rating": "q",
"md5": "a" * 32,
"score": 5,
"fav_count": 2,
"image_width": 800,
"image_height": 600,
},
}
]
modern = [
{
"id": 777,
"rating": "q",
"fav_count": 4,
"score": {"total": 9},
"preview": {"url": "https://static1.e621.net/data/preview/x.jpg"},
"file": {"md5": "a" * 32, "width": 801, "height": 601},
"tags": {"general": ["canine", "solo"]},
}
]
with mock.patch.object(
upload_pipeline.e621, "check_md5_batch", return_value={}
), mock.patch.object(
upload_pipeline.e621, "iqdb_search", return_value=raw_iqdb
), mock.patch.object(
upload_pipeline.e621, "fetch_posts_by_ids", return_value=modern
):
upload_pipeline.run_pipeline(self.uploader.id)
temp.refresh_from_db()
self.assertIsNotNone(temp.e621_checked_at)
self.assertIsNotNone(temp.visual_checked_at)
self.assertEqual(temp.status, TempUpload.STATUS_VISUAL_MATCH)
self.assertEqual(len(temp.iqdb_data), 1)
candidate = temp.iqdb_data[0]
self.assertEqual(candidate["post_id"], 777)
self.assertEqual(candidate["score_total"], 9)
self.assertEqual(candidate["fav_count"], 4)
self.assertEqual(candidate["width"], 801)
self.assertEqual(
candidate["preview_url"], "https://static1.e621.net/data/preview/x.jpg"
)
self.assertEqual(candidate["tags_preview"], ["canine", "solo"])
run = UploadRun.objects.get(user=self.uploader)
self.assertEqual(run.status, UploadRun.STATUS_IDLE)
self.assertEqual(run.processed, 1)
def test_visual_match_flags_similar_library_items(self):
from apps.library.uploads import complete_temp_upload
seed = self.stage(label="seed")
complete_temp_upload(seed)
self.assertTrue(MediaItem.objects.exists())
temp = self.stage(label="similar")
with mock.patch.object(
upload_pipeline.e621, "check_md5_batch", return_value={}
), mock.patch.object(
upload_pipeline.e621, "iqdb_search", return_value=[]
):
upload_pipeline.run_pipeline(self.uploader.id)
temp.refresh_from_db()
self.assertGreaterEqual(len(temp.visual_matches or []), 1)
self.assertEqual(temp.status, TempUpload.STATUS_VISUAL_MATCH)
def test_iqdb_rate_limit_pauses_the_run(self):
temp = self.stage(label="throttled")
with mock.patch.object(
upload_pipeline.e621, "check_md5_batch", return_value={}
), mock.patch.object(
upload_pipeline.e621,
"iqdb_search",
side_effect=upload_pipeline.e621.E621RateLimited("Throttled"),
):
upload_pipeline.run_pipeline(self.uploader.id)
run = UploadRun.objects.get(user=self.uploader)
self.assertEqual(run.status, UploadRun.STATUS_PAUSED)
self.assertIn("e621", run.error)
temp.refresh_from_db()
self.assertEqual(temp.attempts, 1)
self.assertNotEqual(temp.pipeline_error, "")
self.assertIsNone(temp.iqdb_data)
self.assertIsNone(temp.claimed_at)
client = self.api_client(self.uploader)
response = jpost(client, f"/api/uploads/{temp.id}/retry/", {})
self.assertEqual(response.status_code, 200)
temp.refresh_from_db()
self.assertEqual(temp.attempts, 0)
self.assertEqual(temp.pipeline_error, "")
def test_failed_rows_stop_after_the_attempt_cap(self):
temp = self.stage(label="broken")
for _ in range(upload_pipeline.MAX_ATTEMPTS):
with mock.patch.object(
upload_pipeline.e621, "check_md5_batch", return_value={}
), mock.patch.object(
upload_pipeline.e621,
"iqdb_search",
side_effect=upload_pipeline.e621.E621Error("boom"),
):
upload_pipeline.run_pipeline(self.uploader.id)
temp.refresh_from_db()
self.assertEqual(temp.attempts, upload_pipeline.MAX_ATTEMPTS)
self.assertEqual(upload_pipeline.count_outstanding(self.uploader), 0)
run = UploadRun.objects.get(user=self.uploader)
self.assertGreaterEqual(run.failed, 1)
def test_status_process_and_compact_board_payload(self):
temp = self.stage(label="board")
client = self.api_client(self.uploader)
status = client.get("/api/uploads/status/").json()
self.assertEqual(status["status"], UploadRun.STATUS_IDLE)
self.assertTrue(status["active"])
self.assertEqual(status["outstanding"], 1)
self.assertEqual(status["waiting"]["md5"], 1)
self.assertEqual(client.post("/api/uploads/process/").status_code, 200)
rows = client.get("/api/uploads/").json()
self.assertEqual(len(rows), 1)
row = rows[0]
for key in (
"md5_checked",
"visual_checked",
"iqdb_checked",
"processing",
"similar_count",
"pipeline_error",
):
self.assertIn(key, row)
self.assertNotIn("e621_data", row)
self.assertNotIn("iqdb_data", row)
detail = client.get(f"/api/uploads/{temp.id}/").json()
self.assertIn("e621_data", detail)
self.assertIn("iqdb_data", detail)
def test_discard_bulk_removes_only_own_rows(self):
client = self.api_client(self.uploader)
first = self.stage(label="discard-a")
second = self.stage(label="discard-b")
theirs = self.stage(self.other, label="discard-theirs")
paths = [Path(first.file.path), Path(second.file.path)]
response = jpost(
client,
"/api/uploads/discard-bulk/",
{"temp_ids": [str(first.id), str(second.id), str(theirs.id)]},
)
self.assertEqual(response.status_code, 200)
body = response.json()
self.assertEqual(len(body["discarded"]), 2)
self.assertEqual(len(body["errors"]), 1)
self.assertEqual(body["errors"][0]["error"], "not found")
for path in paths:
self.assertFalse(path.exists())
self.assertFalse(
TempUpload.objects.filter(pk__in=[first.id, second.id]).exists()
)
self.assertTrue(TempUpload.objects.filter(pk=theirs.id).exists())
def test_retry_with_phase_rechecks_iqdb(self):
temp = self.stage(label="recheck")
TempUpload.objects.filter(pk=temp.pk).update(
iqdb_data=[{"post_id": 1}],
status=TempUpload.STATUS_VISUAL_MATCH,
)
client = self.api_client(self.uploader)
response = jpost(
client, f"/api/uploads/{temp.id}/retry/", {"phase": "iqdb"}
)
self.assertEqual(response.status_code, 200)
temp.refresh_from_db()
self.assertIsNone(temp.iqdb_data)
self.assertEqual(temp.status, TempUpload.STATUS_VISUAL_MATCH)
def test_stale_claims_are_released(self):
temp = self.stage(label="stale")
TempUpload.objects.filter(pk=temp.pk).update(
claimed_at=timezone.now() - upload_pipeline.STALE_CLAIM_AFTER
- timedelta(minutes=1)
)
UploadRun.objects.create(user=self.uploader, status=UploadRun.STATUS_RUNNING)
UploadRun.objects.filter(user=self.uploader).update(
updated_at=timezone.now() - upload_pipeline.STALE_RUN_AFTER
- timedelta(minutes=1)
)
released, paused = upload_pipeline.reap_stale_claims()
self.assertEqual(released, 1)
self.assertEqual(paused, 1)
temp.refresh_from_db()
self.assertIsNone(temp.claimed_at)
run = UploadRun.objects.get(user=self.uploader)
self.assertEqual(run.status, UploadRun.STATUS_PAUSED)
+648
View File
@@ -0,0 +1,648 @@
"""Background processing for staged uploads.
Each staged upload runs through three phases, in batches of 75 (the same
lookup size the original app used for its e621 MD5 cache command):
1. e621 MD5 lookup — one ``posts.json`` query per round; byte-identical
matches are imported straight into the library (``auto_md5``).
2. Local visual similarity — perceptual hashes are compared against the
whole library once per round.
3. e621 IQDB — reverse-image search for whatever is still unresolved.
The pipeline runs in a daemon thread started on demand (like the download and
match scans), so the browser can navigate away and the work keeps going.
Progress and pause/error state live in the ``UploadRun`` row; per-file state
lives on ``TempUpload``. Rows are claimed with ``SELECT ... FOR UPDATE SKIP
LOCKED`` so several gunicorn workers cannot process the same file, and stale
claims left by a recycled worker are reaped and picked up again.
"""
import logging
import threading
import time
from datetime import timedelta
from django.conf import settings
from django.contrib.auth import get_user_model
from django.db import connection, transaction
from django.db.models import F, Q
from django.utils import timezone
from . import e621, services
from .models import MediaItem, TempUpload, UploadRun
logger = logging.getLogger(__name__)
# Same batch size as the original app's e621 cache command.
MD5_BATCH_SIZE = 75
# Rows one worker round claims; also the MD5 query size.
CLAIM_SIZE = MD5_BATCH_SIZE
# A claimed row is assumed dead after this long and is queued again.
STALE_CLAIM_AFTER = timedelta(minutes=15)
# A run whose heartbeat stopped this long ago can be taken over.
STALE_RUN_AFTER = timedelta(minutes=15)
# Per-row failures before the pipeline stops retrying automatically.
MAX_ATTEMPTS = 3
# Round-level e621 retries before the run is paused.
ROUND_ATTEMPTS = 3
ROUND_RETRY_SECONDS = 20
# Completions kept in the live feed. The board only shows what happened while
# the page was open, so a bounded rolling window is plenty.
COMPLETION_FEED_LIMIT = 200
WORK_STATUSES = (TempUpload.STATUS_PENDING, TempUpload.STATUS_VISUAL_MATCH)
VIDEO_RE = r"\.(mp4|webm)$"
# A staged upload still needs work when any phase has not run yet. IQDB is
# skipped for videos, which never get iqdb_data, so they must not stay
# "outstanding" forever.
OUTSTANDING_Q = (
Q(e621_checked_at__isnull=True)
| Q(visual_checked_at__isnull=True)
| (Q(iqdb_data__isnull=True) & ~Q(original_filename__iregex=VIDEO_RE))
)
_running_lock = threading.Lock()
_running_users: set[int] = set()
class PipelinePaused(Exception):
"""A round-level failure that should pause the run instead of failing rows."""
def is_video(filename):
return bool(filename) and filename.lower().endswith((".mp4", ".webm"))
def is_finished(temp):
"""True when every phase this file needs has run."""
if temp.status == TempUpload.STATUS_COMPLETED:
return True
if temp.e621_checked_at is None or temp.visual_checked_at is None:
return False
return is_video(temp.original_filename) or temp.iqdb_data is not None
def outstanding_queryset(user):
return (
TempUpload.objects.filter(user=user, status__in=WORK_STATUSES)
.filter(OUTSTANDING_Q)
.filter(attempts__lt=MAX_ATTEMPTS)
)
def count_outstanding(user):
return outstanding_queryset(user).count()
def count_failed(user):
return (
TempUpload.objects.filter(user=user, status__in=WORK_STATUSES)
.filter(attempts__gte=MAX_ATTEMPTS)
.count()
)
def waiting_counts(user):
"""How many files are left per phase (phases overlap by design)."""
base = TempUpload.objects.filter(
user=user, status__in=WORK_STATUSES, attempts__lt=MAX_ATTEMPTS
)
return {
"md5": base.filter(e621_checked_at__isnull=True).count(),
"visual": base.filter(visual_checked_at__isnull=True).count(),
"iqdb": (
base.filter(iqdb_data__isnull=True)
.exclude(original_filename__iregex=VIDEO_RE)
.count()
),
}
def status_payload(user, request=None):
"""Cheap state for the shell/upload page to poll."""
run = UploadRun.objects.filter(user=user).first()
outstanding = count_outstanding(user)
failed = count_failed(user)
status = run.status if run is not None else UploadRun.STATUS_IDLE
total = run.total if run is not None else 0
processed = run.processed if run is not None else 0
# A paused run with nothing left to do is not "active" (the user may have
# resolved or discarded the failed rows); failed rows stay visible until
# they are retried or dismissed.
active = (
outstanding > 0 or failed > 0 or status == UploadRun.STATUS_RUNNING
)
return {
"status": status,
"active": active,
"phase": run.phase if run is not None else "",
"total": max(total, processed + failed),
"processed": processed,
"matched": run.matched if run is not None else 0,
"failed": max(failed, run.failed if run is not None else 0),
"error": run.error if run is not None else "",
"outstanding": outstanding,
"waiting": waiting_counts(user),
"recent_completions": completion_payload(run, user, request=request),
"updated_at": run.updated_at.isoformat() if run is not None else None,
}
def completion_payload(run, user, request=None):
"""The live completion feed: filename -> J-ID for freshly indexed uploads.
Completed ``TempUpload`` rows are deleted, so this is the only place the
board learns about them. It is a notification feed, not durable state:
bounded, never dismissed, and ignored by fresh page loads.
"""
entries = list(run.recent_completions or []) if run is not None else []
if not entries:
return []
ids = [entry.get("item_id") for entry in entries if entry.get("item_id")]
items = MediaItem.objects.in_bulk(ids)
out = []
for entry in reversed(entries): # newest first
item = items.get(entry.get("item_id"))
if item is None:
continue
out.append(
{
"id": entry.get("id"),
"filename": entry.get("filename"),
"j_id": f"J-{item.id}",
"resolution": entry.get("resolution", ""),
"post_id": entry.get("post_id"),
"thumbnail_url": services.signed_media_url(
item, user, "thumbnail", request=request
),
"at": entry.get("at"),
}
)
return out
def record_completion(temp, item, resolution=""):
"""Append one completion to the owner's feed; never fails an import."""
entry = {
"id": str(temp.pk),
"filename": temp.original_filename,
"item_id": item.pk,
"resolution": resolution or temp.resolution or "",
"post_id": temp.e621_post_id,
"at": timezone.now().isoformat(),
}
try:
with transaction.atomic():
run, _ = UploadRun.objects.select_for_update().get_or_create(
user_id=temp.user_id
)
feed = list(run.recent_completions or [])
feed.append(entry)
run.recent_completions = feed[-COMPLETION_FEED_LIMIT:]
run.save(update_fields=["recent_completions", "updated_at"])
except Exception: # noqa: BLE001 - a notification must not break an import
logger.exception("Could not record the completion of %s", temp.pk)
def reap_stale_claims():
"""Queue rows left claimed by a recycled worker and pause dead runs."""
cutoff = timezone.now() - STALE_CLAIM_AFTER
released = TempUpload.objects.filter(claimed_at__lt=cutoff).update(
claimed_at=None
)
paused = UploadRun.objects.filter(
status=UploadRun.STATUS_RUNNING, updated_at__lt=cutoff
).update(
status=UploadRun.STATUS_PAUSED,
phase="",
error="The worker stopped before finishing. Retry to resume.",
updated_at=timezone.now(),
)
if released or paused:
logger.info(
"Reaped %s stale upload claims and %s dead upload runs", released, paused
)
return released, paused
def start_pipeline(user):
"""Start the pipeline for one user in a daemon thread (idempotent)."""
if not getattr(settings, "UPLOAD_PIPELINE_AUTOSTART", True):
return False
if user is None or not getattr(user, "can_upload", False):
return False
user_id = int(user.pk)
with _running_lock:
if user_id in _running_users:
return False
_running_users.add(user_id)
reap_stale_claims()
thread = threading.Thread(target=_thread_entry, args=(user_id,), daemon=True)
thread.start()
return True
def _thread_entry(user_id):
try:
run_pipeline(user_id)
except Exception: # noqa: BLE001 - a thread must never crash the worker
logger.exception("Upload pipeline for user %s crashed", user_id)
finally:
with _running_lock:
_running_users.discard(user_id)
connection.close()
def run_pipeline(user_id):
user = get_user_model().objects.filter(pk=user_id).first()
if user is None or not user.can_upload:
return
now = timezone.now()
# Claim the run row so two gunicorn workers cannot own the same queue.
with transaction.atomic():
run, _ = UploadRun.objects.select_for_update().get_or_create(user=user)
if (
run.status == UploadRun.STATUS_RUNNING
and run.updated_at is not None
and run.updated_at > now - STALE_RUN_AFTER
):
# Another worker owns this run.
return
# Reset the counters when a new queue starts cleanly; otherwise keep
# accumulating so failed rows from an earlier pass stay visible.
live_failed = count_failed(user)
outstanding = count_outstanding(user)
if run.status == UploadRun.STATUS_IDLE and live_failed == 0:
run.total = outstanding
run.processed = 0
run.matched = 0
run.failed = 0
else:
run.total = max(
run.total or 0, run.processed + run.failed + outstanding
)
run.failed = max(run.failed, live_failed)
run.status = UploadRun.STATUS_RUNNING
run.phase = ""
run.error = ""
run.started_at = now
run.save()
try:
while True:
rows = claim_round(user)
if not rows:
break
process_round(run, user, rows)
except PipelinePaused as exc:
_save_run(run, status=UploadRun.STATUS_PAUSED, phase="", error=str(exc))
except Exception as exc: # noqa: BLE001 - surface crashes as a run error
logger.exception("Upload pipeline for user %s failed", user_id)
_save_run(
run,
status=UploadRun.STATUS_ERROR,
phase="",
error=f"The upload pipeline stopped: {exc}",
)
else:
_save_run(run, status=UploadRun.STATUS_IDLE, phase="")
def claim_round(user, size=CLAIM_SIZE):
"""Claim up to ``size`` outstanding rows for this worker."""
now = timezone.now()
with transaction.atomic():
rows = list(
TempUpload.objects.select_for_update(skip_locked=True)
.filter(user=user, status__in=WORK_STATUSES)
.filter(OUTSTANDING_Q)
.filter(claimed_at__isnull=True, attempts__lt=MAX_ATTEMPTS)
.order_by("created_at")[:size]
)
if rows:
TempUpload.objects.filter(pk__in=[row.pk for row in rows]).update(
claimed_at=now
)
for row in rows:
row.claimed_at = now
return rows
def process_round(run, user, rows):
"""Run every phase for one claimed round, then release/account the rows."""
ids = [row.pk for row in rows]
matched = 0
try:
matched += md5_phase(run, user, rows)
rows = refresh(ids)
visual_phase(run, user, rows)
rows = refresh(ids)
iqdb_phase(run, user, rows)
finally:
finalize_round(run, ids, matched)
def refresh(ids):
return list(TempUpload.objects.filter(pk__in=ids))
def _save_run(run, **fields):
for key, value in fields.items():
setattr(run, key, value)
run.save(
update_fields=[*fields.keys(), "updated_at"]
)
def md5_phase(run, user, rows):
"""One e621 MD5 batch query; matches are imported into the library."""
targets = [row for row in rows if row.e621_checked_at is None]
if not targets:
return 0
_save_run(
run,
phase=UploadRun.PHASE_MD5,
total=run.processed + run.failed + count_outstanding(user),
)
posts = _e621_round(
lambda: e621.check_md5_batch(user, [row.md5 for row in targets])
)
by_md5 = {}
for post in posts.values():
file_data = post.get("file") or {}
md5 = str(file_data.get("md5") or "").strip().lower()
if md5:
by_md5[md5] = post
matched = 0
now = timezone.now()
for row in targets:
post = by_md5.get(str(row.md5).strip().lower())
if post is None:
TempUpload.objects.filter(pk=row.pk).update(
e621_checked_at=now, pipeline_error="", updated_at=now
)
continue
try:
trimmed = services.trim_e621_post(post)
if trimmed is None or not trimmed.get("id"):
raise e621.E621Error("e621 returned an unexpected post payload.")
TempUpload.objects.filter(pk=row.pk).update(
e621_post_id=int(trimmed["id"]),
e621_data=trimmed,
resolution=TempUpload.RESOLUTION_AUTO_MD5,
e621_checked_at=now,
pipeline_error="",
updated_at=now,
)
row.refresh_from_db()
from .uploads import complete_temp_upload
complete_temp_upload(row)
matched += 1
except Exception as exc: # noqa: BLE001 - keep going for other files
logger.exception("Could not auto-import staged upload %s", row.pk)
record_failure(row, f"Could not finish the upload: {exc}")
return matched
def visual_phase(run, user, rows):
"""Compare each row's perceptual hashes against the library once."""
from .uploads import build_hash_index, match_hashes
targets = [
row
for row in rows
if row.visual_checked_at is None
and row.status in WORK_STATUSES
and row.file
]
if not targets:
return
_save_run(run, phase=UploadRun.PHASE_VISUAL)
index = build_hash_index()
now = timezone.now()
for row in targets:
try:
hashes = services.compute_visual_hashes(row.file.path)
if not hashes:
TempUpload.objects.filter(pk=row.pk).update(
visual_checked_at=now, pipeline_error="", updated_at=now
)
continue
matches = match_hashes(hashes, index, user=user)
update = {
"visual_matches": matches,
"visual_checked_at": now,
"pipeline_error": "",
"updated_at": now,
}
if matches and row.status == TempUpload.STATUS_PENDING:
update["status"] = TempUpload.STATUS_VISUAL_MATCH
TempUpload.objects.filter(pk=row.pk).update(**update)
except Exception as exc: # noqa: BLE001 - keep going for other files
logger.exception("Visual similarity failed for %s", row.pk)
record_failure(row, f"Visual similarity failed: {exc}")
def iqdb_phase(run, user, rows):
"""Reverse-image search every unresolved image, one e621 query each."""
targets = [
row
for row in rows
if row.iqdb_data is None
and row.status in WORK_STATUSES
and row.file
and not is_video(row.original_filename)
]
if not targets:
return
_save_run(run, phase=UploadRun.PHASE_IQDB)
heartbeat_at = time.monotonic()
for row in targets:
# Keep the run row fresh: a 75-file IQDB round takes minutes and must
# not look like a dead worker to another request.
if time.monotonic() - heartbeat_at > 30:
UploadRun.objects.filter(pk=run.pk).update(updated_at=timezone.now())
heartbeat_at = time.monotonic()
try:
raw = e621.iqdb_search(user, row.file.path)
results = normalize_iqdb_results(user, raw)
except e621.E621AuthError as exc:
record_failure(row, str(exc))
raise PipelinePaused(
"e621 rejected the credentials — fix them in Account and retry."
) from exc
except e621.E621RateLimited as exc:
record_failure(row, str(exc))
raise PipelinePaused(
"e621 is throttling IQDB right now; the queue will resume."
) from exc
except Exception as exc: # noqa: BLE001 - keep going for other files
logger.exception("IQDB search failed for %s", row.pk)
record_failure(row, f"IQDB search failed: {exc}")
continue
now = timezone.now()
update = {
"iqdb_data": results,
"pipeline_error": "",
"updated_at": now,
}
if results and row.status == TempUpload.STATUS_PENDING:
update["status"] = TempUpload.STATUS_VISUAL_MATCH
TempUpload.objects.filter(pk=row.pk).update(**update)
def record_failure(row, message):
"""Count one failed attempt against a row and queue it for a retry."""
TempUpload.objects.filter(pk=row.pk).update(
attempts=F("attempts") + 1,
pipeline_error=str(message)[:2000],
claimed_at=None,
updated_at=timezone.now(),
)
def finalize_round(run, ids, matched):
"""Account finished/failed rows and release the rest of the claims."""
rows = refresh(ids)
finished = {row.pk for row in rows if is_finished(row)}
failed = {row.pk for row in rows if row.attempts >= MAX_ATTEMPTS}
# Release every claim: finished rows must not keep looking "processing"
# to the board, and unfinished rows are re-queued for the next run.
release = [row.pk for row in rows if row.claimed_at is not None]
if release:
TempUpload.objects.filter(pk__in=release).update(claimed_at=None)
_save_run(
run,
# Completed rows are deleted as they are imported, so they cannot be
# seen in the refreshed rows; count the matches explicitly.
processed=run.processed + len(finished) + matched,
failed=run.failed + len(failed - finished),
matched=run.matched + matched,
)
def _e621_round(task):
"""Run a round-level e621 call, retrying through rate limits."""
last_error = None
for attempt in range(ROUND_ATTEMPTS):
try:
return task()
except e621.E621AuthError:
raise
except (e621.E621RateLimited, e621.E621Error) as exc:
last_error = exc
if attempt + 1 >= ROUND_ATTEMPTS:
break
delay = ROUND_RETRY_SECONDS * (attempt + 1)
logger.info("e621 round failed (%s); retrying in %ss", exc, delay)
time.sleep(delay)
raise PipelinePaused(
f"e621 is not answering right now ({last_error}); the queue will resume."
) from last_error
def flatten_tag_preview(tags, limit=8):
"""First few tag names from a modern post payload, like the SPA shows."""
if not isinstance(tags, dict):
return []
out = []
for values in tags.values():
if not isinstance(values, list):
continue
for tag in values:
if isinstance(tag, str) and tag not in out:
out.append(tag)
if len(out) >= limit:
return out
return out
def _legacy_iqdb_post(entry):
"""Unwrap the post payload embedded in a legacy IQDB match."""
post = entry.get("post")
if not isinstance(post, dict):
return {}
inner = post.get("posts")
return inner if isinstance(inner, dict) else post
def normalize_iqdb_results(user, raw_results):
"""Shape legacy IQDB matches like the SPA's E621IqdbCandidate entries.
The IQDB payload carries little post data, so candidates are enriched
with one batched ``id:`` lookup before they are stored.
"""
candidates = []
for entry in (raw_results or [])[:10]:
if not isinstance(entry, dict):
continue
post = _legacy_iqdb_post(entry)
post_id = entry.get("post_id")
if not isinstance(post_id, int):
post_id = post.get("id")
score = entry.get("score")
candidates.append(
{
"post_id": post_id if isinstance(post_id, int) else None,
"score": float(score) if isinstance(score, (int, float)) else None,
"preview_url": None,
"rating": (
post.get("rating") if isinstance(post.get("rating"), str) else None
),
"md5": post.get("md5") if isinstance(post.get("md5"), str) else None,
"score_total": (
post.get("score") if isinstance(post.get("score"), int) else None
),
"fav_count": (
post.get("fav_count")
if isinstance(post.get("fav_count"), int)
else None
),
"width": (
post.get("image_width")
if isinstance(post.get("image_width"), int)
else None
),
"height": (
post.get("image_height")
if isinstance(post.get("image_height"), int)
else None
),
"tags_preview": [],
}
)
ids = [entry["post_id"] for entry in candidates if entry["post_id"]]
if not ids:
return services.sanitize_iqdb_results(candidates)
try:
posts = e621.fetch_posts_by_ids(user, ids)
except e621.E621Error as exc:
# Candidates without enrichment still show up; keep them.
logger.info("Could not enrich IQDB candidates: %s", exc)
posts = []
by_id = {post.get("id"): post for post in posts}
for entry in candidates:
post = by_id.get(entry["post_id"])
if not isinstance(post, dict):
continue
file_data = post.get("file") or {}
preview = post.get("preview") or {}
score = post.get("score") or {}
entry["preview_url"] = preview.get("url") or entry["preview_url"]
entry["rating"] = post.get("rating") or entry["rating"]
entry["md5"] = file_data.get("md5") or entry["md5"]
if isinstance(score, dict):
entry["score_total"] = score.get("total")
entry["fav_count"] = post.get("fav_count")
entry["width"] = file_data.get("width")
entry["height"] = file_data.get("height")
entry["tags_preview"] = flatten_tag_preview(post.get("tags"))
return services.sanitize_iqdb_results(candidates)
+401 -39
View File
@@ -16,7 +16,7 @@ from urllib.parse import urlparse
from django.conf import settings
from django.contrib.auth import get_user_model
from django.core import signing
from django.core.exceptions import ValidationError
from django.http import Http404
from django.utils import timezone
from rest_framework import mixins, status, viewsets
@@ -28,32 +28,39 @@ from rest_framework.response import Response
from . import services
from .models import MediaItem, TempUpload
from .permissions import CanUpload
from .serializers import TempUploadSerializer
from .tools import HASH_FIELDS, hashes_similarity
from .serializers import TempUploadListSerializer, TempUploadSerializer
from .signing_urls import load_payload
from .tools import HASH_FIELDS, hashed_items, hashes_similarity
logger = logging.getLogger(__name__)
def find_library_matches(path, limit=10, user=None, request=None):
"""Library items visually similar to a staged file."""
hashes = services.compute_visual_hashes(path)
if not hashes:
return []
def build_hash_index():
"""Library hash mappings for similarity scans, loaded once per batch.
Only items that actually carry perceptual hashes are included; the old
per-file scan walked every row (including videos and unchecked items).
"""
algorithms = list(HASH_FIELDS)
return [
(item, {field: getattr(item, field, "") for field in algorithms})
for item in hashed_items(algorithms)
]
def match_hashes(hashes, index, limit=10, user=None, request=None):
"""Library items whose perceptual hashes are close to ``hashes``."""
algorithms = list(HASH_FIELDS)
threshold = settings.VISUAL_MATCH_THRESHOLD
matches = []
for item in MediaItem.objects.prefetch_related("locations"):
similarity = hashes_similarity(
hashes,
{field: getattr(item, field, "") for field in algorithms},
algorithms,
threshold,
)
for item, item_hashes in index:
similarity = hashes_similarity(hashes, item_hashes, algorithms, threshold)
if similarity is None:
continue
location = item.locations.first()
matches.append(
{
"item_id": item.id,
"j_id": f"J-{item.id}",
"filename": Path(location.rel_path).name if location else item.md5,
"similarity": round(similarity * 100, 1),
@@ -66,6 +73,20 @@ def find_library_matches(path, limit=10, user=None, request=None):
return matches[:limit]
def find_library_matches(path, limit=10, user=None, request=None, index=None):
"""Library items visually similar to a staged file.
Pass a prebuilt ``index`` (see build_hash_index) to reuse it across a
whole batch instead of rescanning the library per file.
"""
hashes = services.compute_visual_hashes(path)
if not hashes:
return []
if index is None:
index = build_hash_index()
return match_hashes(hashes, index, limit=limit, user=user, request=request)
def complete_temp_upload(temp, download_url=None):
"""Index the upload into the library.
@@ -107,6 +128,12 @@ def complete_temp_upload(temp, download_url=None):
services.ensure_visual_hashes(item)
temp.file.delete(save=False)
# Warm the preview while the import is still off the request path.
try:
services.ensure_thumbnail(item)
except Exception: # noqa: BLE001 - a preview must not fail the import
logger.exception("Could not warm the thumbnail for J-%s", item.id)
temp.library_item = item
temp.status = TempUpload.STATUS_COMPLETED
@@ -138,6 +165,13 @@ def complete_temp_upload(temp, download_url=None):
item.save(update_fields=update_fields + ["updated_at"])
temp.save()
from .upload_pipeline import record_completion
record_completion(temp, item)
# The board learns about completions from the live feed, so the record is
# deleted as soon as the file is indexed: nothing left to dismiss. Delete
# through the queryset so callers keep ``temp.pk`` for their response.
TempUpload.objects.filter(pk=temp.pk).delete()
return item
@@ -150,14 +184,47 @@ class TempUploadViewSet(
serializer_class = TempUploadSerializer
permission_classes = [CanUpload]
parser_classes = [MultiPartParser, FormParser, JSONParser]
# Unpaginated: the board shows every staged upload (69-file batches were
# silently cut to the API's 48-item page).
pagination_class = None
http_method_names = ["get", "post", "delete", "head", "options"]
def get_serializer_class(self):
# The board polls the list, so its payload stays small; the metadata
# modal fetches the full row from the detail endpoint.
if self.action == "list":
return TempUploadListSerializer
return TempUploadSerializer
def get_queryset(self):
queryset = TempUpload.objects.select_related("library_item")
user = self.request.user
if not user.is_app_staff:
queryset = queryset.filter(user=user)
return queryset
# Staged uploads are private: everyone, staff included, only sees
# their own board. (The file action still lets staff read bytes by id
# for support purposes.)
return TempUpload.objects.select_related("library_item").filter(
user=self.request.user
)
def _completed_payload(self, temp, item):
"""Synthetic row for an upload that is indexed immediately.
Duplicates and resolved uploads never leave a board record; the SPA
turns this response (or the live completion feed) into a "J-x
uploaded" card that lives only in the page session.
"""
return {
"temp_id": str(temp.pk),
"original_filename": temp.original_filename,
"md5": temp.md5,
"size": temp.size,
"status": TempUpload.STATUS_COMPLETED,
"resolution": temp.resolution,
"e621_post_id": temp.e621_post_id,
"library_j_id": f"J-{item.id}",
"file_url": None,
"preview_url": services.signed_media_url(
item, self.request.user, "thumbnail", request=self.request
),
}
def create(self, request):
upload = request.FILES.get("file")
@@ -186,33 +253,190 @@ class TempUploadViewSet(
temp.resolution = TempUpload.RESOLUTION_DUPLICATE
temp.library_item = existing
temp.file.delete(save=False)
else:
matches = find_library_matches(
temp.file.path, user=request.user, request=request
)
if matches:
temp.visual_matches = matches
temp.status = TempUpload.STATUS_VISUAL_MATCH
temp.save()
from .upload_pipeline import record_completion
record_completion(temp, existing)
payload = self._completed_payload(temp, existing)
TempUpload.objects.filter(pk=temp.pk).delete()
return Response(payload, status=status.HTTP_201_CREATED)
# Visual similarity and IQDB run in the background pipeline so a large
# batch uploads at full speed and the work survives the browser.
temp.save()
# Kick the server-side pipeline; staging no longer waits on e621 and
# the work continues even if the browser navigates away.
from .upload_pipeline import start_pipeline
start_pipeline(request.user)
return Response(
self.get_serializer(temp).data, status=status.HTTP_201_CREATED
)
@action(detail=True, methods=["get", "head"], permission_classes=[AllowAny])
@action(detail=True, methods=["post"], url_path="visual-match")
def visual_match(self, request, pk=None):
"""Run the local visual-similarity pass for one staged upload."""
temp = self.get_object()
if temp.status == TempUpload.STATUS_COMPLETED:
return Response(
{"detail": "This upload is already in the library."},
status=status.HTTP_400_BAD_REQUEST,
)
if not temp.file:
return Response(
{"detail": "The staged file is missing."},
status=status.HTTP_400_BAD_REQUEST,
)
matches = find_library_matches(
temp.file.path, user=request.user, request=request
)
temp.visual_matches = matches
if matches and temp.status == TempUpload.STATUS_PENDING:
temp.status = TempUpload.STATUS_VISUAL_MATCH
temp.save(update_fields=["visual_matches", "status", "updated_at"])
return Response(self.get_serializer(temp).data)
@action(detail=False, methods=["get"])
def status(self, request):
"""Cheap pipeline state for the shell indicator and the upload page."""
from .upload_pipeline import status_payload
return Response(status_payload(request.user, request=request))
@action(detail=False, methods=["post"])
def process(self, request):
"""Start (or resume) the pipeline for the caller's staged uploads.
Idempotent: the client calls this after staging files, on page load
and when a paused run should be retried.
"""
from .upload_pipeline import start_pipeline, status_payload
start_pipeline(request.user)
return Response(status_payload(request.user, request=request))
@action(detail=True, methods=["post"])
def retry(self, request, pk=None):
"""Queue one staged upload for another pipeline pass."""
from .upload_pipeline import start_pipeline, status_payload
temp = self.get_object()
if temp.status == TempUpload.STATUS_COMPLETED:
return Response(
{"detail": "This upload is already in the library."},
status=status.HTTP_400_BAD_REQUEST,
)
if not temp.file:
return Response(
{"detail": "The staged file is missing."},
status=status.HTTP_400_BAD_REQUEST,
)
update = {
"claimed_at": None,
"attempts": 0,
"pipeline_error": "",
"updated_at": timezone.now(),
}
# An explicit phase re-runs that one check even if it already ran.
phase = str(request.data.get("phase") or "").strip()
if phase == "md5":
update["e621_checked_at"] = None
elif phase == "visual":
update["visual_matches"] = None
update["visual_checked_at"] = None
elif phase == "iqdb":
update["iqdb_data"] = None
if temp.status == TempUpload.STATUS_ERROR:
# A failed import has to go through the MD5 phase again so the
# completion is retried; other errors only re-run missing phases.
update["status"] = TempUpload.STATUS_PENDING
update["e621_checked_at"] = None
elif temp.status not in (
TempUpload.STATUS_PENDING,
TempUpload.STATUS_VISUAL_MATCH,
):
update["status"] = TempUpload.STATUS_PENDING
TempUpload.objects.filter(pk=temp.pk).update(**update)
start_pipeline(request.user)
return Response(status_payload(request.user, request=request))
@action(detail=False, methods=["post"], url_path="retry-all")
def retry_all(self, request):
"""Queue every retryable staged upload for another pipeline pass."""
from .upload_pipeline import start_pipeline, status_payload
now = timezone.now()
retryable = self.get_queryset().filter(
status__in=[
TempUpload.STATUS_PENDING,
TempUpload.STATUS_VISUAL_MATCH,
TempUpload.STATUS_ERROR,
]
)
retryable.exclude(file="").update(
claimed_at=None,
attempts=0,
pipeline_error="",
updated_at=now,
)
retryable.exclude(file="").filter(status=TempUpload.STATUS_ERROR).update(
status=TempUpload.STATUS_PENDING,
e621_checked_at=None,
)
start_pipeline(request.user)
return Response(status_payload(request.user, request=request))
@action(detail=False, methods=["post"], url_path="discard-bulk")
def discard_bulk(self, request):
"""Discard many staged uploads in one request (the board's "all")."""
ids = request.data.get("temp_ids")
if not isinstance(ids, list) or not ids:
return Response(
{"detail": "temp_ids must be a non-empty list."},
status=status.HTTP_400_BAD_REQUEST,
)
if len(ids) > 1000:
return Response(
{"detail": "Too many ids in one request (max 1000)."},
status=status.HTTP_400_BAD_REQUEST,
)
values = list(dict.fromkeys(str(value) for value in ids))
try:
queryset = self.get_queryset().filter(pk__in=values)
except (ValidationError, ValueError):
return Response(
{"detail": "One or more temp_ids are not valid upload ids."},
status=status.HTTP_400_BAD_REQUEST,
)
by_id = {str(temp.pk): temp for temp in queryset}
discarded: list[str] = []
errors: list[dict[str, str]] = []
for value in values:
temp = by_id.get(value)
if temp is None:
errors.append({"temp_id": value, "error": "not found"})
continue
try:
self.perform_destroy(temp)
discarded.append(value)
except Exception as exc: # noqa: BLE001 - report per-file failures
logger.exception("Could not discard staged upload %s", value)
errors.append({"temp_id": value, "error": str(exc)})
return Response({"discarded": discarded, "errors": errors})
@action(
detail=True,
methods=["get", "head"],
permission_classes=[AllowAny],
throttle_classes=[],
)
def file(self, request, pk=None):
"""Serve the staged file; accepts a signed URL for media tags."""
user = request.user if request.user.is_authenticated else None
if user is None:
signature = request.query_params.get("sig")
if signature:
try:
payload = signing.loads(
signature,
salt=services.UPLOAD_FILE_SALT,
max_age=86400,
)
except signing.BadSignature:
payload = None
payload = load_payload(signature, services.UPLOAD_FILE_SALT)
if payload and str(payload.get("temp")) == str(pk):
user = (
get_user_model()
@@ -313,7 +537,7 @@ class TempUploadViewSet(
temp.save()
try:
complete_temp_upload(temp, download_url=download_url)
item = complete_temp_upload(temp, download_url=download_url)
except Exception as exc: # noqa: BLE001 - report completion failures
logger.exception("Could not complete staged upload %s", temp.id)
temp.status = TempUpload.STATUS_ERROR
@@ -322,8 +546,146 @@ class TempUploadViewSet(
{"detail": f"Could not finish the upload: {exc}"},
status=status.HTTP_400_BAD_REQUEST,
)
temp.refresh_from_db()
return Response(self.get_serializer(temp).data)
return Response(self._completed_payload(temp, item))
@action(detail=False, methods=["post"], url_path="link-bulk")
def link_bulk(self, request):
"""Attach e621 posts to many staged uploads in one request.
The upload board's MD5 phase sends one posts.json query per 75 files
and hands the matches over here, so a whole batch moves into the
library in one update instead of one request per file.
"""
links = request.data.get("links")
if not isinstance(links, list) or not links:
return Response(
{"detail": "links must be a non-empty list."},
status=status.HTTP_400_BAD_REQUEST,
)
ids = []
for link in links:
if not isinstance(link, dict) or link.get("temp_id") is None:
return Response(
{"detail": "Each link needs a temp_id."},
status=status.HTTP_400_BAD_REQUEST,
)
ids.append(str(link["temp_id"]))
try:
queryset = self.get_queryset().filter(pk__in=ids)
except (ValidationError, ValueError):
return Response(
{"detail": "One or more temp_ids are not valid upload ids."},
status=status.HTTP_400_BAD_REQUEST,
)
by_id = {str(temp.id): temp for temp in queryset}
updated = []
errors = []
for link in links:
temp_id = str(link["temp_id"])
temp = by_id.get(temp_id)
if temp is None:
errors.append({"temp_id": temp_id, "error": "not found"})
continue
if temp.status == TempUpload.STATUS_COMPLETED:
errors.append({"temp_id": temp_id, "error": "already in the library"})
continue
trimmed = services.trim_e621_post(link.get("post"))
post_id = link.get("post_id")
has_numeric_id = post_id is not None and str(post_id).isdigit()
if trimmed is None and not has_numeric_id:
errors.append({"temp_id": temp_id, "error": "no post payload"})
continue
if trimmed is not None and trimmed.get("id") is not None:
temp.e621_post_id = int(trimmed["id"])
elif has_numeric_id:
temp.e621_post_id = int(post_id)
temp.e621_data = trimmed
temp.resolution = (
TempUpload.RESOLUTION_AUTO_MD5
if link.get("auto", True)
else TempUpload.RESOLUTION_LINKED
)
# An MD5 match means the staged file is byte-identical, so there
# is nothing to download from e621; only link a remote URL when
# the post's file differs.
file_data = (trimmed or {}).get("file") or {}
candidate_url = str(
link.get("file_url") or file_data.get("url") or ""
).strip()
post_md5 = str(file_data.get("md5") or "").strip().lower()
if candidate_url and post_md5 and post_md5 == temp.md5.lower():
candidate_url = ""
temp.save()
try:
item = complete_temp_upload(temp, download_url=candidate_url or None)
except Exception as exc: # noqa: BLE001 - report per-file failures
logger.exception("Could not complete staged upload %s", temp.id)
temp.status = TempUpload.STATUS_ERROR
temp.save(update_fields=["status", "updated_at"])
errors.append({"temp_id": temp_id, "error": str(exc)})
continue
updated.append(self._completed_payload(temp, item))
return Response({"updated": updated, "errors": errors})
@action(detail=False, methods=["post"], url_path="resolve-bulk")
def resolve_bulk(self, request):
"""Move many staged uploads into the library with one rating.
The upload board's bulk tool: pick a rating, tick pending files, and
they are all resolved as custom entries (their staged tags/notes are
kept). Own rows only, like the rest of the viewset.
"""
ids = request.data.get("temp_ids")
if not isinstance(ids, list) or not ids:
return Response(
{"detail": "temp_ids must be a non-empty list."},
status=status.HTTP_400_BAD_REQUEST,
)
rating = str(request.data.get("rating") or "").strip()
if rating not in {"s", "q", "e"}:
return Response(
{"detail": "rating must be one of s, q or e."},
status=status.HTTP_400_BAD_REQUEST,
)
try:
queryset = self.get_queryset().filter(
pk__in=[str(value) for value in ids]
)
except (ValidationError, ValueError):
return Response(
{"detail": "One or more ids are not valid upload ids."},
status=status.HTTP_400_BAD_REQUEST,
)
resolved: list[str] = []
errors: list[dict[str, str]] = []
for temp in queryset:
if temp.status == TempUpload.STATUS_COMPLETED:
errors.append(
{"temp_id": str(temp.id), "error": "already in the library"}
)
continue
temp.custom_rating = rating
temp.resolution = TempUpload.RESOLUTION_CUSTOM
temp.save(update_fields=["custom_rating", "resolution", "updated_at"])
try:
complete_temp_upload(temp)
except Exception as exc: # noqa: BLE001 - report per-file failures
logger.exception("Could not complete staged upload %s", temp.id)
temp.status = TempUpload.STATUS_ERROR
temp.save(update_fields=["status", "updated_at"])
errors.append({"temp_id": str(temp.id), "error": str(exc)})
continue
resolved.append(str(temp.id))
found = {str(temp.id) for temp in queryset}
for value in sorted({str(value) for value in ids} - found):
errors.append({"temp_id": value, "error": "not found"})
return Response({"resolved": resolved, "errors": errors})
def perform_destroy(self, instance):
if instance.file:
+2
View File
@@ -17,6 +17,7 @@ from .views import (
DownloadTaskViewSet,
MatchTaskViewSet,
MediaItemViewSet,
RandomItemView,
)
router = DefaultRouter()
@@ -28,6 +29,7 @@ router.register("similarity", SimilarityCheckViewSet, basename="similarity")
urlpatterns = [
path("", include(router.urls)),
path("random/", RandomItemView.as_view(), name="random_item"),
path("online/file/", ClientDownloadView.as_view(), name="client_download"),
path(
"duplicates/md5/",
+131 -17
View File
@@ -5,19 +5,21 @@ from pathlib import Path
from urllib.parse import urlparse
from django.conf import settings
from django.core import signing
from django.db.models import Min, Q
from django.http import Http404, StreamingHttpResponse
from django.shortcuts import get_object_or_404
from django.utils import timezone
from django.utils.text import get_valid_filename
from rest_framework import mixins, status, viewsets
from rest_framework.authentication import TokenAuthentication
from rest_framework.decorators import action
from rest_framework.permissions import AllowAny, IsAuthenticatedOrReadOnly
from rest_framework.response import Response
from rest_framework.throttling import ScopedRateThrottle
from rest_framework.views import APIView
from apps.accounts.auth import GreetingTokenAuthentication
from . import e621, matching, services
from .downloads import reap_stale_downloads, start_download_task
from .matching import reap_stale_match_tasks, start_match_task
@@ -28,6 +30,7 @@ from .serializers import (
MatchTaskSerializer,
MediaItemSerializer,
)
from .signing_urls import load_payload
LIST_ORDERINGS = {"name", "-name", "size", "-size", "created_at", "-created_at"}
MD5_RE = re.compile(r"[0-9a-fA-F]{32}")
@@ -129,11 +132,8 @@ class MediaItemViewSet(
signature = request.query_params.get("sig")
if not signature:
return None
try:
payload = signing.loads(
signature, salt=services.MEDIA_FILE_SALT, max_age=86400
)
except signing.BadSignature:
payload = load_payload(signature, services.MEDIA_FILE_SALT)
if payload is None:
return None
if payload.get("action") != action_name:
return None
@@ -145,7 +145,7 @@ class MediaItemViewSet(
return item
return self.get_object()
@action(detail=True, methods=["get"])
@action(detail=True, methods=["get"], throttle_classes=[])
def raw(self, request, pk=None):
item = self._media_object(request, "raw")
location = item.locations.first()
@@ -155,10 +155,14 @@ class MediaItemViewSet(
status=status.HTTP_404_NOT_FOUND,
)
return services.serve_file(
request, location.path, download=request.query_params.get("download") == "1"
request,
location.path,
download=request.query_params.get("download") == "1",
max_age=services.MEDIA_CACHE_SECONDS,
immutable=True,
)
@action(detail=True, methods=["get"])
@action(detail=True, methods=["get"], throttle_classes=[])
def thumbnail(self, request, pk=None):
item = self._media_object(request, "thumbnail")
location = item.locations.first()
@@ -170,13 +174,26 @@ class MediaItemViewSet(
path = Path(location.path)
if path.suffix.lower() in services.VIDEO_EXTENSIONS:
thumbnail = services.generate_video_thumbnail(item.md5, path)
if thumbnail is None:
return Response(
{"detail": "Thumbnail unavailable."},
status=status.HTTP_404_NOT_FOUND,
)
return services.serve_file(request, thumbnail)
return services.serve_file(request, path)
else:
thumbnail = services.generate_image_thumbnail(item.md5, path)
if thumbnail is not None:
return services.serve_file(
request,
thumbnail,
max_age=services.MEDIA_CACHE_SECONDS,
immutable=True,
)
if path.suffix.lower() in services.VIDEO_EXTENSIONS:
return Response(
{"detail": "Thumbnail unavailable."},
status=status.HTTP_404_NOT_FOUND,
)
return services.serve_file(
request,
path,
max_age=services.MEDIA_CACHE_SECONDS,
immutable=True,
)
@action(detail=False, methods=["post"], permission_classes=[AllowAny])
def lookup(self, request):
@@ -520,6 +537,103 @@ class MatchTaskViewSet(
return Response({"success": True})
class RandomItemView(APIView):
"""A random library image, optionally filtered by rating.
Two kinds of clients use this:
* the SPA's Random page, which renders the returned URL, and
* shell greeting scripts (fish_greeting) that fetch the URL with
fastfetch in a terminal.
Fastfetch mode — ``?fastfetch=1`` or a User-Agent containing "fastfetch"
— only considers png/jpg/gif files, because that is what those terminals
display. Responses always carry a signed absolute URL (minted for the
requesting user) so image viewers can load it without auth headers;
guests get unsigned URLs for guest-visible items only.
Accepts the normal API token *and* the scope-limited greeting tokens
(``j621r_…``), which work here and nowhere else.
"""
authentication_classes = [GreetingTokenAuthentication, TokenAuthentication]
permission_classes = [AllowAny]
FASTFETCH_EXTENSIONS = {".png", ".jpg", ".jpeg", ".gif"}
def get(self, request):
fastfetch = request.query_params.get("fastfetch", "").lower() in {
"1",
"true",
"yes",
} or "fastfetch" in (request.META.get("HTTP_USER_AGENT") or "").lower()
extensions = (
self.FASTFETCH_EXTENSIONS
if fastfetch
else services.IMAGE_EXTENSIONS
)
queryset = MediaItem.objects.prefetch_related("locations")
if not request.user.is_authenticated:
queryset = queryset.filter(hidden_from_guests=False)
ratings = [
value
for value in request.query_params.get("rating", "").split(",")
if value in {"s", "q", "e"}
]
if ratings:
queryset = queryset.filter(rating__in=ratings)
# Any copy with an allowed extension qualifies. ORDER BY RAND() is
# fine for a personal library (same trade-off as the duplicates page).
suffixes = "|".join(extension.lstrip(".") for extension in sorted(extensions))
item = (
queryset.filter(locations__rel_path__iregex=rf"\.({suffixes})$")
.distinct()
.order_by("?")
.first()
)
if item is None:
return Response(
{"detail": "No image matches those filters."},
status=status.HTTP_404_NOT_FOUND,
)
location = next(
(
candidate
for candidate in item.locations.all()
if Path(candidate.rel_path).suffix.lower() in extensions
),
item.locations.first(),
)
url = services.signed_media_url(item, request.user, "raw", request=request)
return Response(
{
"j_id": f"J-{item.id}",
"md5": item.md5,
"filename": Path(location.rel_path).name if location else item.md5,
"extension": (
Path(location.rel_path).suffix.lower().lstrip(".")
if location
else ""
),
"kind": "image",
"rating": item.rating or "",
"size": item.size,
"e621_post_id": item.e621_post_id,
"url": url,
"download_url": f"{url}{'&' if '?' in url else '?'}download=1",
"thumbnail_url": services.signed_media_url(
item, request.user, "thumbnail", request=request
),
"fastfetch": fastfetch,
}
)
class ClientDownloadView(APIView):
"""Stream an e621 file straight to the browser (no library write)."""
@@ -562,7 +676,7 @@ class ClientDownloadView(APIView):
"Content-Type", "application/octet-stream"
)
name = get_valid_filename(
filename or Path(parsed.path).name or "download"
filename or Path(urlparse(url).path).name or "download"
)
def stream():
+32 -7
View File
@@ -206,7 +206,9 @@ WATCHED_FOLDER = str(WATCHED_FOLDER)
# e621 integration
E621_BASE_URL = os.getenv("E621_BASE_URL", "https://e621.net").rstrip("/")
USER_AGENT = os.getenv("USER_AGENT", "J621/0.1 (by J621 on e621)")
# e621 asks for "Application name/version (developer)". Browser clients cannot
# set a User-Agent, so the SPA sends the same string in its `_client` parameter.
USER_AGENT = os.getenv("USER_AGENT", f"J621/{GIT_COMMIT_HASH} (JakeBreath)")
# Hosts the client-download proxy is allowed to stream from.
E621_MEDIA_HOSTS = [
host.strip()
@@ -232,6 +234,12 @@ GUEST_BLACKLIST_TTL = int(os.getenv("GUEST_BLACKLIST_TTL", "3600"))
# Similarity threshold for flagging staged uploads that match library items.
VISUAL_MATCH_THRESHOLD = float(os.getenv("VISUAL_MATCH_THRESHOLD", "0.9"))
# Start the staged-upload pipeline when a file is staged (daemon thread in the
# worker). Tests turn this off and drive the pipeline synchronously.
UPLOAD_PIPELINE_AUTOSTART = os.getenv(
"UPLOAD_PIPELINE_AUTOSTART", "true"
).strip().lower() not in {"0", "false", "no", "off"}
# Ephemeral similarity-check uploads are deleted after this many minutes
# (and always on startup).
SIMILARITY_TTL_MINUTES = int(os.getenv("SIMILARITY_TTL_MINUTES", "30"))
@@ -240,13 +248,23 @@ SIMILARITY_TTL_MINUTES = int(os.getenv("SIMILARITY_TTL_MINUTES", "30"))
# workers and management commands (e.g. the mirrored guest blacklist).
CACHES = {
"default": {
"BACKEND": "django.core.cache.backends.redis.RedisCache",
"BACKEND": "apps.core.cache.ResilientRedisCache",
"LOCATION": os.getenv("REDIS_URL", "redis://127.0.0.1:6380/1"),
}
}
# Django REST Framework
# Private / tailnet-only deployments can drop the general anon+user limits
# entirely (THROTTLE_ENABLED=false). The scoped guards below (login, register,
# e621 proxy) and the media endpoints' own protections stay active either way.
THROTTLE_ENABLED = os.getenv("THROTTLE_ENABLED", "true").strip().lower() not in {
"0",
"false",
"no",
"off",
}
REST_FRAMEWORK = {
"DEFAULT_AUTHENTICATION_CLASSES": [
"rest_framework.authentication.TokenAuthentication",
@@ -261,11 +279,18 @@ REST_FRAMEWORK = {
],
"DEFAULT_PAGINATION_CLASS": "config.pagination.StandardPagination",
"PAGE_SIZE": 48,
# Per-IP/per-user rate limits (counted in the shared Redis cache).
"DEFAULT_THROTTLE_CLASSES": [
"rest_framework.throttling.AnonRateThrottle",
"rest_framework.throttling.UserRateThrottle",
],
# Per-IP/per-user rate limits (counted in the shared Redis cache). Signed
# media URLs are deliberately excluded at the view level: <img>/<video>
# tags fetch them without an Authorization header, so a library page would
# otherwise burn the anonymous bucket and start returning JSON 429s.
"DEFAULT_THROTTLE_CLASSES": (
[
"rest_framework.throttling.AnonRateThrottle",
"rest_framework.throttling.UserRateThrottle",
]
if THROTTLE_ENABLED
else []
),
"DEFAULT_THROTTLE_RATES": {
# Generous enough for the shell polling (status every 5s, stats every 2s).
"anon": os.getenv("THROTTLE_ANON", "120/min"),
+5
View File
@@ -4,6 +4,7 @@ from django.contrib import admin
from django.urls import include, path
from apps.core.views import HealthView
from apps.library.views import RandomItemView
urlpatterns = [
path("admin/", admin.site.urls),
@@ -15,6 +16,10 @@ urlpatterns = [
# Liveness probe for uptime monitors (no /api prefix, no auth).
path("health", HealthView.as_view(), name="health"),
path("health/", HealthView.as_view()),
# Short alias for shell greeting scripts (fish_greeting + fastfetch);
# /api/random/ is the canonical path.
path("random", RandomItemView.as_view(), name="random"),
path("random/", RandomItemView.as_view()),
]
if settings.DEBUG:
+14 -1
View File
@@ -30,7 +30,10 @@ ALLOWED_HOSTS=j621.rainbow-herring.ts.net,localhost,127.0.0.1
# ---------------------------------------------------------------------------
# Cross-origin access (needed for the separate frontend + backend deploys)
# ---------------------------------------------------------------------------
# The origin the SPA is served from, e.g. https://j621-frontend.<tailnet>.ts.net
# The origin the SPA is served from for split deploys, e.g.
# https://j621-frontend.<tailnet>.ts.net. gen_env.sh writes this plus the
# desktop shell's app://j621 origin into the line below (and keeps existing
# entries on --update).
# CORS_ALLOWED_ORIGINS=
# CSRF_TRUSTED_ORIGINS=
@@ -49,12 +52,22 @@ DB_ROOT_PASSWORD=j621root
# GUNICORN_THREADS=4
# GUNICORN_TIMEOUT=120
# Scheduler intervals in seconds (the "scheduler" service runs the periodic
# management commands; see deploy/README.md)
# J621_SYNC_EVERY=1800
# J621_CLEAN_EVERY=3600
# J621_BLACKLIST_EVERY=86400
# Rate limits (per IP anonymous, per account signed in)
# THROTTLE_ANON=120/min
# THROTTLE_USER=600/min
# THROTTLE_LOGIN=5/min
# THROTTLE_REGISTER=20/hour
# THROTTLE_E621_PROXY=60/hour
# Tailnet-only / private deployments can drop the general limits entirely.
# Signed media URLs (<img>/<video>) and the login/register/proxy guards are
# exempt from this switch either way.
# THROTTLE_ENABLED=false
# e621 media hosts the backend may fetch from (downloads, proxies)
# E621_MEDIA_HOSTS=static1.e621.net,static2.e621.net,static3.e621.net
+10 -1
View File
@@ -28,8 +28,17 @@ RUN pip install --no-cache-dir -r requirements.txt
COPY backend/ ./
COPY deploy/backend-entrypoint.sh /usr/local/bin/j621-entrypoint
COPY deploy/scheduler-entrypoint.sh /usr/local/bin/j621-scheduler
RUN chmod +x /usr/local/bin/j621-entrypoint \
# Guard: fail the build if the context leaked secrets or runtime data
# (.dockerignore excludes them — see the repository root).
RUN test ! -e /app/.env \
&& test ! -d /app/venv \
&& test ! -d /app/media/library \
&& test ! -e /app/db.sqlite3 \
&& echo "build context clean"
RUN chmod +x /usr/local/bin/j621-entrypoint /usr/local/bin/j621-scheduler \
&& mkdir -p /app/media /app/logs \
&& python manage.py collectstatic --noinput
+2
View File
@@ -10,6 +10,8 @@
# syntax=docker/dockerfile:1
FROM node:22-alpine AS build
ARG GIT_HASH=dev
ENV GIT_HASH=$GIT_HASH
WORKDIR /app
COPY frontend/package.json frontend/package-lock.json ./
RUN npm ci --no-audit --no-fund
+119 -3
View File
@@ -5,7 +5,9 @@ sidecar. Nothing is published on the host's ports and no system nginx or
reverse proxy is involved: the sidecar shares the nginx service's network
namespace and Tailscale Serve/Funnel exposes it.
| Compose | Services | Funnel | Serve config |
## Funnel set (public HTTPS)
| Compose | Services | Entry point | Serve config |
| --- | --- | --- | --- |
| `compose.yml` (default) | mariadb, redis, backend, frontend, nginx, tailscale | `https://<host>.<tailnet>.ts.net` → nginx → backend + SPA | `serve.default.json` |
| `compose.frontend.yml` | frontend, nginx, tailscale | `https://<host>.<tailnet>.ts.net` → nginx → SPA only | `serve.frontend.json` |
@@ -15,15 +17,64 @@ Funnel can only expose ports **443**, **8443** and **10000**, so the separate
backend uses 8443. Change it in `serve.backend.json` (and `.env`) if you
prefer 10000.
## Tailnet-only set (no Funnel)
The same three stacks without `AllowFunnel` in the serve config: the sidecar
still registers the node and serves over HTTPS with a tailnet certificate,
but only devices on your tailnet can reach it — nothing is exposed to the
public internet.
| Compose | Serve config | Reachable at |
| --- | --- | --- |
| `compose.tailnet.yml` (both) | `serve.default.tailnet.json` | `https://<host>.<tailnet>.ts.net` |
| `compose.tailnet.frontend.yml` | `serve.frontend.tailnet.json` | `https://<host>.<tailnet>.ts.net` |
| `compose.tailnet.backend.yml` | `serve.backend.tailnet.json` | `https://<host>.<tailnet>.ts.net:8443` |
```bash
docker compose -f compose.tailnet.yml up -d
# or compose.tailnet.frontend.yml / compose.tailnet.backend.yml
```
Notes:
- Project names are `…-tailnet` so both sets can be installed side by side.
- Serve needs no tailnet policy change (Funnel requires the `funnel` node
attribute in your ACLs), so this set works as soon as the sidecar joins.
- Both sets use the same `deploy/data` directory (library, database, logs).
Run one set per data directory — two MariaDB instances on one data dir would
corrupt it. Giving the tailnet set its own `TS_HOSTNAME` (or running it on a
second host) is the way to run both.
- Switching a host from funnel to tailnet (or back) is just starting the other
compose file with the same `.env`.
## Environment file
`gen_env.sh` builds `deploy/.env` from `.env.example`, generating `SECRET_KEY`
and both database passwords with `openssl`:
```bash
./gen_env.sh # asks for the Tailscale auth key + hostname(s)
./gen_env.sh --no-prompt # secrets and defaults only; fill TS_AUTHKEY later
./gen_env.sh --update # refresh hostnames/authkey, keep the existing secrets
./gen_env.sh --force # regenerate everything, including SECRET_KEY
```
It derives `TS_HOSTNAME` and `ALLOWED_HOSTS` from the tailnet hostname you
give it, and can set `CORS_ALLOWED_ORIGINS`/`CSRF_TRUSTED_ORIGINS` when you
provide the frontend's hostname for a split deployment. `--update` is the safe
way to add hostnames later; `--force` rotates SECRET_KEY, which invalidates
signed media URLs and stored e621 API keys. The file is written with mode 600
and is git-ignored.
## Usage
```bash
cd deploy
cp .env.example .env # fill in TS_AUTHKEY, SECRET_KEY, ALLOWED_HOSTS, ...
docker compose up -d # default: everything on one host
./gen_env.sh # or: cp .env.example .env and edit it yourself
docker compose up -d # default: everything on one host, public funnel
# or
docker compose -f compose.frontend.yml up -d
docker compose -f compose.backend.yml up -d
# tailnet-only (no public funnel): compose.tailnet*.yml, see below
```
The images are pulled from the Gitea registry; append `--build` (or run the
@@ -41,6 +92,10 @@ The SPA asks where the backend is (`/setup`) in production builds:
backend's funnel URL, e.g. `https://j621-backend.<tailnet>.ts.net:8443`,
and give the backend `CORS_ALLOWED_ORIGINS=https://<frontend-host>.<tailnet>.ts.net`
(plus `CSRF_TRUSTED_ORIGINS` for the admin).
- **Desktop app** — the Electron shell (`desktop/`) is served from the
`app://j621` origin, which `gen_env.sh` includes in `CORS_ALLOWED_ORIGINS`
automatically (add it by hand if you wrote `.env` yourself). The shell's
setup screen prints the exact origin when the connection test fails.
- Without a backend at all, choose **"Continue without a backend"** to run in
local mode (e621 browsing only).
@@ -63,6 +118,15 @@ ffmpeg for video thumbnails. The `GIT_HASH` build arg is baked into the
backend image so the shell's version pill shows the commit (images have no
`.git` directory); the push scripts pass it automatically.
Both build from the repository root, which is filtered by `.dockerignore`:
`backend/venv`, `backend/media`, `backend/logs`, `backend/staticfiles`,
`backend/.env`, `frontend/node_modules`, `frontend/dist` and the deploy
runtime state never enter the images — the database, media and logs come
from the compose volumes and `deploy/.env` at runtime. The backend build
also asserts that (`.env`, `venv`, `media/library`, `db.sqlite3` absent), so
a missing ignore file fails the build instead of shipping secrets. Rebuild
(and `docker image prune`) if you built before that guard existed.
Push multi-arch images to the Gitea registry:
```bash
@@ -72,6 +136,58 @@ Push multi-arch images to the Gitea registry:
They tag `:latest` and `:<commit-sha>` and expect `docker login
gitea.rainbow-herring.ts.net` to succeed.
Build the desktop installers without publishing anything:
```bash
./build_desktop.sh # Arch + Debian + Windows -> desktop/release/
./build_desktop.sh --linux # Arch + Debian only
./build_desktop.sh --win # Windows installer only
```
Hand the files out or attach them to a Gitea release manually — the script
prints sizes and SHA-256 sums for the release notes. The release assets are
also the desktop update feed; the app resolves the newest `desktop-v*`
release on Gitea at check time (see `desktop/README.md`).
The frontend's `/desktop/` feed is optional now — kept for manual downloads
and for installs older than 0.1.2. To publish it:
```bash
./push_desktop.sh # build Linux packages + copy the feed to jakerasp
./push_desktop.sh --win # also cross-build the NSIS installer (wine)
./push_desktop.sh --local # publish locally only, skip the remote copy
./push_desktop.sh --host other:/path/to/J621
```
The remote copy defaults to `jakerasp:/home/jake/servers/J621`, or
`$J621_DESKTOP_FEED_HOST` when set. Artifacts land in `deploy/data/desktop/`,
which the frontend nginx mounts read-only and serves at `/desktop/`. The
remote copy uses rsync when both ends have it, tar over ssh when the server
does not. Backend-only composes have no frontend, so no website feed.
The manual **CD** workflow (Actions tab) builds the desktop packages on the
runner and attaches them plus the update metadata to the Gitea release
`desktop-v<version>`; that is what the desktop updater reads. The website feed
is not touched by CI (runtime state on the deploy host, no SSH key there); use
`push_desktop.sh` when it needs refreshing for old installs.
## Scheduled jobs
Compose files with a backend also run a **`scheduler`** service — the same
backend image with a different entrypoint, so no host cron is involved:
| Job | Default interval | Env override |
| --- | --- | --- |
| `sync_followed_tags` + `sync_followed_pools` | every 30 minutes | `J621_SYNC_EVERY` |
| `cleanup_similarity` | hourly | `J621_CLEAN_EVERY` |
| `refresh_guest_blacklist` | daily | `J621_BLACKLIST_EVERY` |
It waits for the database and migrations before its first run, runs every job
once on start, then keeps to the intervals (failures are logged and retried
next round). Output goes to `docker compose logs scheduler`. Intervals are
seconds, set in `deploy/.env`. The frontend-only composes have no backend, so
no scheduler.
## Tests
The security/permission suite lives in `backend/apps/core/tests/`:
+79
View File
@@ -0,0 +1,79 @@
#!/bin/bash
# Build the J621 desktop packages into desktop/release/ — Arch (.pkg.tar.zst),
# Debian (.deb) and the Windows NSIS installer. Nothing is published: install
# the package locally, hand the files out, or attach them to a Gitea release
# manually. Use push_desktop.sh when the in-app update feed should get them.
#
# Usage: ./build_desktop.sh [--linux | --win | --all]
# --linux Arch + Debian packages only
# --win Windows installer only (cross-built with wine)
# --all everything (default)
set -euo pipefail
cd "$(dirname "$0")/.."
TARGET="${1:---all}"
case "$TARGET" in
--linux) TARGET=linux ;;
--win) TARGET=win ;;
--all) TARGET=all ;;
*)
echo "usage: $0 [--linux|--win|--all]" >&2
exit 2
;;
esac
VERSION="$(node -p "require('./desktop/package.json').version")"
if [ ! -d desktop/node_modules ]; then
echo "==> Installing desktop dependencies ..."
npm --prefix desktop ci --no-audit --no-fund
fi
if [ "$TARGET" != "linux" ] && ! command -v wine >/dev/null 2>&1; then
echo "==> wine is not installed; the Windows installer needs it." >&2
exit 1
fi
# release/ should only ever hold the current version: old installers are
# rebuilt from git when needed, and the unpacked trees are regenerated.
shopt -s nullglob
stale=(desktop/release/*.deb desktop/release/*.pkg.tar.zst desktop/release/"J621 Setup "*.exe
desktop/release/*.blockmap desktop/release/latest*.yml)
for file in "${stale[@]}"; do
case "$(basename "$file")" in
*"$VERSION"*) ;;
*) echo " removing $(basename "$file")"; rm -f "$file" ;;
esac
done
rm -rf desktop/release/linux-unpacked desktop/release/win-unpacked
case "$TARGET" in
linux) npm --prefix desktop run dist:linux ;;
win) npm --prefix desktop run dist:win ;;
all) npm --prefix desktop run dist:all ;;
esac
case "$TARGET" in
linux) ARTIFACTS=(-name "*${VERSION}*.deb" -o -name "*${VERSION}*.pkg.tar.zst") ;;
win) ARTIFACTS=(-name "*${VERSION}*.exe") ;;
all) ARTIFACTS=(-name "*${VERSION}*.deb" -o -name "*${VERSION}*.pkg.tar.zst" -o -name "*${VERSION}*.exe") ;;
esac
echo
echo "==> Artifacts in desktop/release/:"
find desktop/release -maxdepth 1 -type f \( "${ARTIFACTS[@]}" \) \
-printf '%s\t%p\n' |
sort -rn |
while IFS=$'\t' read -r size file; do
printf ' %8s %s\n' "$(numfmt --to=iec "$size")" "$file"
done
echo
echo "==> SHA-256 (for release notes):"
find desktop/release -maxdepth 1 -type f \( "${ARTIFACTS[@]}" \) \
-exec sha256sum {} + | sed 's/^/ /'
echo
echo "==> J621 desktop $VERSION built."
echo " Install here: sudo pacman -U desktop/release/j621-desktop-*.pkg.tar.zst"
echo " Gitea release tag: desktop-v$VERSION"
+27
View File
@@ -68,6 +68,33 @@ services:
redis:
condition: service_healthy
# Periodic maintenance inside the deployment (no host cron): follow syncs,
# similarity cleanup and the guest blacklist refresh.
scheduler:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-backend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Backend
args:
GIT_HASH: ${GIT_HASH:-unknown}
restart: unless-stopped
entrypoint: ["j621-scheduler"]
env_file: [./.env]
environment:
DB_HOST: mariadb
DB_PORT: "3306"
REDIS_URL: redis://redis:6379/1
SYNC_EVERY: ${J621_SYNC_EVERY:-1800}
CLEAN_EVERY: ${J621_CLEAN_EVERY:-3600}
BLACKLIST_EVERY: ${J621_BLACKLIST_EVERY:-86400}
volumes:
- ./data/media:/app/media
depends_on:
mariadb:
condition: service_healthy
redis:
condition: service_healthy
nginx:
image: nginx:1.29-alpine
restart: unless-stopped
+3
View File
@@ -20,6 +20,9 @@ services:
context: ..
dockerfile: deploy/J621-Frontend
restart: unless-stopped
volumes:
# Desktop update feed (deploy/push_desktop.sh): latest*.yml + installers.
- ./data/desktop:/usr/share/nginx/html/desktop:ro
nginx:
image: nginx:1.29-alpine
+136
View File
@@ -0,0 +1,136 @@
# J621 — backend-only deployment: API + database + nginx proxy + Tailscale.
#
# cd deploy && cp .env.example .env # TS_AUTHKEY, SECRET_KEY, hosts, CORS
# docker compose -f compose.tailnet.backend.yml up -d
#
# Tailnet-only: https://<TS_HOSTNAME>.<tailnet>.ts.net:8443 reaches the
# nginx service from your tailnet, which routes /api, /admin, /static and /health to
# Django. "/" answers a small JSON hint because no frontend is attached.
#
# Because the frontend lives elsewhere it is cross-origin — set in .env:
# CORS_ALLOWED_ORIGINS=https://<frontend-host>.<tailnet>.ts.net
# CSRF_TRUSTED_ORIGINS=... (same value; only needed for the admin)
# and add this host to ALLOWED_HOSTS (comma separated list).
name: j621-backend-deploy-tailnet
services:
mariadb:
image: mariadb:11.4
restart: unless-stopped
environment:
MARIADB_ROOT_PASSWORD: ${DB_ROOT_PASSWORD:-j621root}
MARIADB_DATABASE: ${DB_NAME:-j621}
MARIADB_USER: ${DB_USER:-j621}
MARIADB_PASSWORD: ${DB_PASSWORD:-j621}
volumes:
- ./data/mariadb:/var/lib/mysql
healthcheck:
test: ["CMD", "healthcheck.sh", "--connect", "--innodb_initialized"]
interval: 5s
timeout: 5s
retries: 20
redis:
image: redis:7-alpine
restart: unless-stopped
command: ["redis-server", "--appendonly", "yes"]
volumes:
- ./data/redis:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 5s
timeout: 5s
retries: 20
backend:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-backend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Backend
# Bake the commit into the image so the shell's version pill shows it;
# the push scripts pass --build-arg themselves.
args:
GIT_HASH: ${GIT_HASH:-unknown}
restart: unless-stopped
env_file: [./.env]
environment:
DB_HOST: mariadb
DB_PORT: "3306"
REDIS_URL: redis://redis:6379/1
TRUST_PROXY_HEADERS: "true"
volumes:
- ./data/media:/app/media
- ./data/logs:/app/logs
depends_on:
mariadb:
condition: service_healthy
redis:
condition: service_healthy
# Periodic maintenance inside the deployment (no host cron): follow syncs,
# similarity cleanup and the guest blacklist refresh.
scheduler:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-backend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Backend
args:
GIT_HASH: ${GIT_HASH:-unknown}
restart: unless-stopped
entrypoint: ["j621-scheduler"]
env_file: [./.env]
environment:
DB_HOST: mariadb
DB_PORT: "3306"
REDIS_URL: redis://redis:6379/1
SYNC_EVERY: ${J621_SYNC_EVERY:-1800}
CLEAN_EVERY: ${J621_CLEAN_EVERY:-3600}
BLACKLIST_EVERY: ${J621_BLACKLIST_EVERY:-86400}
volumes:
- ./data/media:/app/media
depends_on:
mariadb:
condition: service_healthy
redis:
condition: service_healthy
nginx:
image: nginx:1.29-alpine
restart: unless-stopped
volumes:
- ./nginx-proxy.conf:/etc/nginx/conf.d/default.conf:ro
depends_on:
backend:
condition: service_healthy
healthcheck:
test: ["CMD-SHELL", "wget -q --spider http://127.0.0.1/nginx-health || exit 1"]
interval: 30s
timeout: 3s
retries: 3
start_period: 10s
tailscale:
image: tailscale/tailscale:latest
restart: unless-stopped
# Shares the nginx service's network namespace: 127.0.0.1:80 is the proxy.
network_mode: "service:nginx"
environment:
TS_AUTHKEY: ${TS_AUTHKEY:?Set TS_AUTHKEY in deploy/.env}
TS_HOSTNAME: ${TS_HOSTNAME:-j621-backend}
TS_AUTH_ONCE: "true"
TS_STATE_DIR: /var/lib/tailscale
TS_SERVE_CONFIG: /config/serve.json
volumes:
- ./tailscale-state:/var/lib/tailscale
- ./serve.backend.tailnet.json:/config/serve.json:ro
- /etc/ssl/certs:/etc/ssl/certs:ro
depends_on:
nginx:
condition: service_healthy
healthcheck:
test: ["CMD", "tailscale", "status"]
interval: 30s
timeout: 5s
retries: 3
start_period: 30s
+66
View File
@@ -0,0 +1,66 @@
# J621 — frontend-only deployment: SPA + nginx proxy + Tailscale sidecar.
#
# cd deploy && cp .env.example .env # TS_AUTHKEY at minimum
# docker compose -f compose.tailnet.frontend.yml up -d
#
# Tailnet-only: https://<TS_HOSTNAME>.<tailnet>.ts.net (443) serves the SPA
# for devices on your tailnet; nothing is exposed publicly.
# On first start the SPA asks for a backend (/setup): point it at a
# backend-only deployment (e.g. https://j621-backend.<tailnet>.ts.net:8443),
# leave it blank to serve one yourself, or stay in local mode.
#
# There is no backend in this compose, so /api answers 502 here until the SPA
# is configured to call one elsewhere.
name: j621-frontend-deploy-tailnet
services:
frontend:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-frontend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Frontend
restart: unless-stopped
volumes:
# Desktop update feed (deploy/push_desktop.sh): latest*.yml + installers.
- ./data/desktop:/usr/share/nginx/html/desktop:ro
nginx:
image: nginx:1.29-alpine
restart: unless-stopped
volumes:
- ./nginx-proxy.conf:/etc/nginx/conf.d/default.conf:ro
depends_on:
frontend:
condition: service_healthy
healthcheck:
test: ["CMD-SHELL", "wget -q --spider http://127.0.0.1/nginx-health || exit 1"]
interval: 30s
timeout: 3s
retries: 3
start_period: 10s
tailscale:
image: tailscale/tailscale:latest
restart: unless-stopped
# Shares the nginx service's network namespace: 127.0.0.1:80 is the proxy.
network_mode: "service:nginx"
environment:
TS_AUTHKEY: ${TS_AUTHKEY:?Set TS_AUTHKEY in deploy/.env}
TS_HOSTNAME: ${TS_HOSTNAME:-j621-frontend}
TS_AUTH_ONCE: "true"
TS_STATE_DIR: /var/lib/tailscale
TS_SERVE_CONFIG: /config/serve.json
volumes:
- ./tailscale-state:/var/lib/tailscale
- ./serve.frontend.tailnet.json:/config/serve.json:ro
- /etc/ssl/certs:/etc/ssl/certs:ro
depends_on:
nginx:
condition: service_healthy
healthcheck:
test: ["CMD", "tailscale", "status"]
interval: 30s
timeout: 5s
retries: 3
start_period: 30s
+147
View File
@@ -0,0 +1,147 @@
# J621 — default deployment: frontend + backend + database on one Tailscale
# host, behind the nginx proxy service (no host nginx, nothing published on
# the host's ports).
#
# cd deploy && cp .env.example .env # fill in TS_AUTHKEY, SECRET_KEY, ...
# docker compose up -d # or: docker compose -f compose.yml up -d
#
# Tailnet-only: no Funnel, so the service is reachable from your tailnet
# (https://<TS_HOSTNAME>.<tailnet>.ts.net) but not from the public internet.
# The nginx service routes /api, /admin, /static and /health to the
# backend and everything else to the SPA. Same origin, so no CORS needed.
name: j621-deploy-tailnet
services:
mariadb:
image: mariadb:11.4
restart: unless-stopped
environment:
MARIADB_ROOT_PASSWORD: ${DB_ROOT_PASSWORD:-j621root}
MARIADB_DATABASE: ${DB_NAME:-j621}
MARIADB_USER: ${DB_USER:-j621}
MARIADB_PASSWORD: ${DB_PASSWORD:-j621}
volumes:
- ./data/mariadb:/var/lib/mysql
healthcheck:
test: ["CMD", "healthcheck.sh", "--connect", "--innodb_initialized"]
interval: 5s
timeout: 5s
retries: 20
redis:
image: redis:7-alpine
restart: unless-stopped
command: ["redis-server", "--appendonly", "yes"]
volumes:
- ./data/redis:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 5s
timeout: 5s
retries: 20
backend:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-backend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Backend
# Bake the commit into the image so the shell's version pill shows it;
# the push scripts pass --build-arg themselves.
args:
GIT_HASH: ${GIT_HASH:-unknown}
restart: unless-stopped
env_file: [./.env]
environment:
DB_HOST: mariadb
DB_PORT: "3306"
REDIS_URL: redis://redis:6379/1
# The tailnet funnel terminates TLS and forwards the original host/proto.
TRUST_PROXY_HEADERS: "true"
volumes:
- ./data/media:/app/media
- ./data/logs:/app/logs
depends_on:
mariadb:
condition: service_healthy
redis:
condition: service_healthy
# Periodic maintenance inside the deployment (no host cron): follow syncs,
# similarity cleanup and the guest blacklist refresh.
scheduler:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-backend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Backend
args:
GIT_HASH: ${GIT_HASH:-unknown}
restart: unless-stopped
entrypoint: ["j621-scheduler"]
env_file: [./.env]
environment:
DB_HOST: mariadb
DB_PORT: "3306"
REDIS_URL: redis://redis:6379/1
SYNC_EVERY: ${J621_SYNC_EVERY:-1800}
CLEAN_EVERY: ${J621_CLEAN_EVERY:-3600}
BLACKLIST_EVERY: ${J621_BLACKLIST_EVERY:-86400}
volumes:
- ./data/media:/app/media
depends_on:
mariadb:
condition: service_healthy
redis:
condition: service_healthy
frontend:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-frontend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Frontend
restart: unless-stopped
volumes:
# Desktop update feed (deploy/push_desktop.sh): latest*.yml + installers.
- ./data/desktop:/usr/share/nginx/html/desktop:ro
nginx:
image: nginx:1.29-alpine
restart: unless-stopped
volumes:
- ./nginx-proxy.conf:/etc/nginx/conf.d/default.conf:ro
depends_on:
backend:
condition: service_healthy
frontend:
condition: service_healthy
healthcheck:
test: ["CMD-SHELL", "wget -q --spider http://127.0.0.1/nginx-health || exit 1"]
interval: 30s
timeout: 3s
retries: 3
start_period: 10s
tailscale:
image: tailscale/tailscale:latest
restart: unless-stopped
# Shares the nginx service's network namespace: 127.0.0.1:80 is the proxy.
network_mode: "service:nginx"
environment:
TS_AUTHKEY: ${TS_AUTHKEY:?Set TS_AUTHKEY in deploy/.env}
TS_HOSTNAME: ${TS_HOSTNAME:-j621}
TS_AUTH_ONCE: "true"
TS_STATE_DIR: /var/lib/tailscale
TS_SERVE_CONFIG: /config/serve.json
volumes:
- ./tailscale-state:/var/lib/tailscale
- ./serve.default.tailnet.json:/config/serve.json:ro
- /etc/ssl/certs:/etc/ssl/certs:ro
depends_on:
nginx:
condition: service_healthy
healthcheck:
test: ["CMD", "tailscale", "status"]
interval: 30s
timeout: 5s
retries: 3
start_period: 30s
+30
View File
@@ -66,12 +66,42 @@ services:
redis:
condition: service_healthy
# Periodic maintenance inside the deployment (no host cron): follow syncs,
# similarity cleanup and the guest blacklist refresh.
scheduler:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-backend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Backend
args:
GIT_HASH: ${GIT_HASH:-unknown}
restart: unless-stopped
entrypoint: ["j621-scheduler"]
env_file: [./.env]
environment:
DB_HOST: mariadb
DB_PORT: "3306"
REDIS_URL: redis://redis:6379/1
SYNC_EVERY: ${J621_SYNC_EVERY:-1800}
CLEAN_EVERY: ${J621_CLEAN_EVERY:-3600}
BLACKLIST_EVERY: ${J621_BLACKLIST_EVERY:-86400}
volumes:
- ./data/media:/app/media
depends_on:
mariadb:
condition: service_healthy
redis:
condition: service_healthy
frontend:
image: ${J621_REGISTRY:-gitea.rainbow-herring.ts.net/jakebreath}/j621-frontend:${J621_TAG:-latest}
build:
context: ..
dockerfile: deploy/J621-Frontend
restart: unless-stopped
volumes:
# Desktop update feed (deploy/push_desktop.sh): latest*.yml + installers.
- ./data/desktop:/usr/share/nginx/html/desktop:ro
nginx:
image: nginx:1.29-alpine
+163
View File
@@ -0,0 +1,163 @@
#!/bin/bash
# Generate deploy/.env from .env.example with fresh secrets.
#
# ./gen_env.sh # interactive: asks for the auth key/hostnames
# ./gen_env.sh --no-prompt # secrets + defaults only
# ./gen_env.sh --update # refresh hostnames/authkey, KEEP existing secrets
# ./gen_env.sh --force # regenerate everything (new SECRET_KEY!)
#
# SECRET_KEY and the database passwords come from openssl. Rotating
# SECRET_KEY invalidates signed media URLs and stored e621 API keys, which is
# why --update keeps it.
set -euo pipefail
cd "$(dirname "$0")"
FORCE=0
UPDATE=0
NO_PROMPT=0
TS_AUTHKEY=""
FQDN=""
FRONTEND=""
DEFAULT_FQDN="j621.rainbow-herring.ts.net"
usage() {
sed -n '2,12p' "$0" | sed 's/^# \{0,1\}//'
}
while [ $# -gt 0 ]; do
case "$1" in
--force) FORCE=1 ;;
--update) UPDATE=1 ;;
--no-prompt) NO_PROMPT=1 ;;
--ts-authkey) TS_AUTHKEY="${2:?missing value}"; shift ;;
--hostname) FQDN="${2:?missing value}"; shift ;;
--frontend) FRONTEND="${2:?missing value}"; shift ;;
-h|--help) usage; exit 0 ;;
*) echo "Unknown option: $1" >&2; usage >&2; exit 1 ;;
esac
shift
done
command -v openssl >/dev/null || { echo "openssl is required." >&2; exit 1; }
command -v python3 >/dev/null || { echo "python3 is required." >&2; exit 1; }
[ -f .env.example ] || { echo "Run me from the deploy/ directory." >&2; exit 1; }
if [ -f .env ] && [ "$FORCE" -eq 0 ] && [ "$UPDATE" -eq 0 ]; then
echo "deploy/.env already exists."
echo " --update keeps the existing secrets and just refreshes the rest"
echo " --force regenerates everything, including SECRET_KEY and DB passwords"
exit 1
fi
existing() { grep -E "^$1=" .env 2>/dev/null | head -1 | cut -d= -f2- || true; }
gen_key() { openssl rand -base64 48 | tr -d '\n'; }
gen_password() { openssl rand -hex 24; }
if [ "$UPDATE" -eq 1 ] && [ -f .env ]; then
SECRET_KEY="$(existing SECRET_KEY)"
DB_PASSWORD="$(existing DB_PASSWORD)"
DB_ROOT_PASSWORD="$(existing DB_ROOT_PASSWORD)"
TS_AUTHKEY="${TS_AUTHKEY:-$(existing TS_AUTHKEY)}"
# TS_HOSTNAME is the short label; the full FQDN lives in ALLOWED_HOSTS.
EXISTING_HOSTS="$(existing ALLOWED_HOSTS)"
FQDN="${FQDN:-${EXISTING_HOSTS%%,*}}"
fi
# Fill whatever is still missing.
SECRET_KEY="${SECRET_KEY:-$(gen_key)}"
DB_PASSWORD="${DB_PASSWORD:-$(gen_password)}"
DB_ROOT_PASSWORD="${DB_ROOT_PASSWORD:-$(gen_password)}"
if [ "$NO_PROMPT" -eq 0 ]; then
if [ -z "$TS_AUTHKEY" ]; then
read -rp "Tailscale auth key (tskey-..., Enter to fill in later): " TS_AUTHKEY
fi
if [ -z "$FQDN" ]; then
read -rp "Tailnet hostname of this deployment [$DEFAULT_FQDN]: " FQDN
fi
FQDN="${FQDN:-$DEFAULT_FQDN}"
if [ -z "$FRONTEND" ]; then
read -rp "Frontend hostname for the split deploys (optional, Enter to skip): " FRONTEND
fi
fi
FQDN="${FQDN:-$DEFAULT_FQDN}"
# Derive the rest from the tailnet hostname.
TS_HOSTNAME="${FQDN%%.*}"
ALLOWED_HOSTS="$FQDN,localhost,127.0.0.1"
# Cross-origin access: the optional split-deploy frontend plus the desktop
# shell, which is always a different origin from the backend. On --update the
# existing list is kept, so hand-added origins survive.
CORS_ALLOWED_ORIGINS=""
add_origin() {
[ -n "${1:-}" ] || return 0
case ",$CORS_ALLOWED_ORIGINS," in
*",$1,"*) ;;
*) CORS_ALLOWED_ORIGINS="${CORS_ALLOWED_ORIGINS:+$CORS_ALLOWED_ORIGINS,}$1" ;;
esac
}
if [ "$UPDATE" -eq 1 ]; then
while IFS= read -r origin; do
add_origin "$origin"
done < <(existing CORS_ALLOWED_ORIGINS | tr ',' '\n')
fi
[ -n "$FRONTEND" ] && add_origin "https://$FRONTEND"
add_origin "app://j621"
CSRF_TRUSTED_ORIGINS="${FRONTEND:+https://$FRONTEND}"
J621_SECRET_KEY="$SECRET_KEY" \
J621_DB_PASSWORD="$DB_PASSWORD" \
J621_DB_ROOT_PASSWORD="$DB_ROOT_PASSWORD" \
J621_TS_AUTHKEY="$TS_AUTHKEY" \
J621_TS_HOSTNAME="$TS_HOSTNAME" \
J621_ALLOWED_HOSTS="$ALLOWED_HOSTS" \
J621_CORS_ALLOWED_ORIGINS="$CORS_ALLOWED_ORIGINS" \
J621_CSRF_TRUSTED_ORIGINS="$CSRF_TRUSTED_ORIGINS" \
python3 - <<'PY'
import os
import re
from pathlib import Path
text = Path(".env.example").read_text()
def apply(name):
value = os.environ.get(f"J621_{name}")
if not value:
return text
line = f"{name}={value}"
pattern = re.compile(rf"^(?:# )?{re.escape(name)}=.*$", re.M)
if pattern.search(text):
return pattern.sub(line, text, count=1)
return text + f"\n{line}\n"
for name in (
"SECRET_KEY",
"TS_AUTHKEY",
"TS_HOSTNAME",
"ALLOWED_HOSTS",
"CORS_ALLOWED_ORIGINS",
"CSRF_TRUSTED_ORIGINS",
"DB_PASSWORD",
"DB_ROOT_PASSWORD",
):
text = apply(name)
text = re.sub(r"^DEBUG=.*$", "DEBUG=False", text, count=1, flags=re.M)
Path(".env").write_text(text)
PY
chmod 600 .env
echo
echo "Wrote deploy/.env (mode 600):"
echo " SECRET_KEY $([ "$UPDATE" -eq 1 ] && echo 'kept from the existing file' || echo 'generated with openssl')"
echo " DB passwords $([ "$UPDATE" -eq 1 ] && echo 'kept from the existing file' || echo 'generated with openssl')"
echo " TS_HOSTNAME $TS_HOSTNAME"
echo " ALLOWED_HOSTS $ALLOWED_HOSTS"
[ -n "$CORS_ALLOWED_ORIGINS" ] && echo " cross-origin $CORS_ALLOWED_ORIGINS"
[ -z "$TS_AUTHKEY" ] && echo " TS_AUTHKEY still empty - paste your Tailscale auth key before starting"
echo
echo "Next: docker compose -f compose.yml up -d (or a compose.tailnet*.yml variant)"
+34
View File
@@ -17,6 +17,13 @@ map $http_x_forwarded_proto $j621_forwarded_proto {
"" $scheme;
}
# /random is both the SPA route and the shell-greeting shortcut: browsers
# (Accept: text/html) get the SPA, scripts (curl/wget/fetch) get the API JSON.
map $http_accept $j621_random_target {
default @j621_random_api;
~*text/html @j621_random_spa;
}
server {
listen 80;
server_name _;
@@ -26,6 +33,33 @@ server {
return 200 "ok\n";
}
location ~ ^/random/?$ {
error_page 418 = $j621_random_target;
return 418;
}
location @j621_random_api {
set $j621_backend http://backend:8000;
proxy_pass $j621_backend$request_uri;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Host $host;
proxy_set_header X-Forwarded-Proto $j621_forwarded_proto;
}
location @j621_random_spa {
set $j621_frontend http://frontend:80;
proxy_pass $j621_frontend$request_uri;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Host $host;
proxy_set_header X-Forwarded-Proto $j621_forwarded_proto;
}
location ~ ^/(api|admin|static|health)(/|$) {
set $j621_backend http://backend:8000;
proxy_pass $j621_backend$request_uri;
+8 -2
View File
@@ -10,10 +10,16 @@ REGISTRY="gitea.rainbow-herring.ts.net/jakebreath/j621-backend"
REGISTRY_HOST="$(printf '%s' "$REGISTRY" | cut -d/ -f1)"
SHA="${1:-$(git rev-parse --short HEAD)}"
BUILDER=multiarch
PLATFORMS="linux/amd64,linux/arm64"
PLATFORMS="${PLATFORMS:-linux/amd64,linux/arm64}"
echo "==> Logging in to $REGISTRY_HOST ..."
docker login "$REGISTRY_HOST"
if [ -n "${REGISTRY_USER:-}" ] && [ -n "${REGISTRY_TOKEN:-}" ]; then
# Non-interactive login for CI (workflow passes GITHUB_TOKEN).
printf '%s' "$REGISTRY_TOKEN" | docker login "$REGISTRY_HOST" \
-u "$REGISTRY_USER" --password-stdin
else
docker login "$REGISTRY_HOST"
fi
if ! docker buildx inspect "$BUILDER" >/dev/null 2>&1; then
echo "==> Creating buildx builder '$BUILDER' ..."
+126
View File
@@ -0,0 +1,126 @@
#!/bin/bash
# Build the J621 desktop packages, optionally publish them to the website
# feed.
#
# Desktop updates no longer depend on this: the app resolves the newest
# `desktop-v*` release on Gitea at check time (see desktop/README.md). This
# script builds the packages and can copy them to deploy/data/desktop, which
# the frontend nginx mounts read-only and serves at /desktop/ for manual
# downloads and for pre-0.1.2 installs.
#
# Usage: ./push_desktop.sh [--win] [--no-build] [--local] [--host user@server:/path]
# --win also cross-build the Windows NSIS installer (needs wine)
# --no-build publish what is already in desktop/release/
# --local only publish locally, skip the remote copy
# --host override the remote deploy checkout
#
# The remote copy defaults to $J621_DESKTOP_FEED_HOST, or
# jakerasp:/home/jake/servers/J621 when that is not set.
set -euo pipefail
cd "$(dirname "$0")/.."
DEFAULT_REMOTE="jakerasp:/home/jake/servers/J621"
BUILD=1
WIN=0
REMOTE="${J621_DESKTOP_FEED_HOST:-$DEFAULT_REMOTE}"
while [ $# -gt 0 ]; do
case "$1" in
--win) WIN=1 ;;
--no-build) BUILD=0 ;;
--local) REMOTE="" ;;
--host) REMOTE="${2:?--host needs user@server:/path}"; shift ;;
*)
echo "usage: $0 [--win] [--no-build] [--local] [--host user@server:/path]" >&2
exit 2
;;
esac
shift
done
if [ -n "$REMOTE" ] && [ "$REMOTE" = "${REMOTE#*:}" ]; then
echo "--host needs user@server:/path (got '$REMOTE')" >&2
exit 2
fi
if [ "$BUILD" = 1 ]; then
echo "==> Building Linux packages (deb + pacman) ..."
npm --prefix desktop run dist:linux
if [ "$WIN" = 1 ]; then
echo "==> Cross-building the Windows installer (wine) ..."
npm --prefix desktop run dist:win
fi
fi
FEED=deploy/data/desktop
mkdir -p "$FEED"
shopt -s nullglob
VERSION="$(node -p "require('./desktop/package.json').version")"
deb=(desktop/release/*"$VERSION"*.deb)
zst=(desktop/release/*"$VERSION"*.pkg.tar.zst)
linux_meta=(desktop/release/latest-linux.yml)
win_exe=("desktop/release/J621 Setup ${VERSION}"*.exe)
win_meta=(desktop/release/latest.yml)
blockmaps=(desktop/release/*"$VERSION"*.exe.blockmap)
if [ "${#deb[@]}" -eq 0 ] && [ "${#zst[@]}" -eq 0 ]; then
echo "No $VERSION artifacts in desktop/release/ — run without --no-build first." >&2
exit 1
fi
# Replace the previous release's metadata and drop older installers from the
# feed: latest*.yml only ever points at the current version.
rm -f "$FEED"/latest-linux.yml "$FEED"/latest.yml
old=("$FEED"/*.deb "$FEED"/*.pkg.tar.zst "$FEED"/"J621 Setup "*.exe "$FEED"/*.exe.blockmap)
for file in "${old[@]}"; do
case "$(basename "$file")" in
*"$VERSION"*) ;;
*) echo " pruning $(basename "$file")"; rm -f "$file" ;;
esac
done
cp -f "${deb[@]}" "${zst[@]}" "${linux_meta[@]}" "$FEED"/ 2>/dev/null || true
if [ "${#win_exe[@]}" -gt 0 ]; then
cp -f "${win_exe[@]}" "${win_meta[@]}" "${blockmaps[@]}" "$FEED"/ 2>/dev/null || true
fi
echo "==> Published to $FEED:"
ls -1sh "$FEED" | sed 's/^/ /'
if [ -n "$REMOTE" ]; then
REMOTE_TARGET="${REMOTE%%:*}"
REMOTE_ROOT="${REMOTE#*:}"
REMOTE_FEED="${REMOTE_ROOT%/}/deploy/data/desktop"
echo
echo "==> Copying the feed to $REMOTE_TARGET:$REMOTE_FEED ..."
if ! ssh "$REMOTE_TARGET" "mkdir -p '$REMOTE_FEED'"; then
echo "ssh to $REMOTE_TARGET failed; nothing was copied." >&2
exit 1
fi
# The remote feed keeps only the current version, like the local one.
if ! ssh "$REMOTE_TARGET" \
"find '$REMOTE_FEED' -maxdepth 1 -type f \\( -name '*.deb' -o -name '*.pkg.tar.zst' -o -name '*.exe' -o -name '*.exe.blockmap' \\) ! -name '*$VERSION*' -exec rm -f {} +"; then
echo "==> Could not prune old installers on $REMOTE_TARGET (continuing)." >&2
fi
transferred=0
if command -v rsync >/dev/null 2>&1; then
if rsync -a --info=progress2 "$FEED"/ "$REMOTE_TARGET:$REMOTE_FEED/"; then
transferred=1
else
echo "==> rsync did not work (missing on the server?); trying tar over ssh ..."
fi
fi
if [ "$transferred" -eq 0 ]; then
if ! tar -C "$FEED" -cf - . |
ssh "$REMOTE_TARGET" "tar -C '$REMOTE_FEED' -xf -"; then
echo "Copy to $REMOTE_TARGET:$REMOTE_FEED failed." >&2
exit 1
fi
fi
echo "==> Feed live at $REMOTE_TARGET:$REMOTE_FEED (no restart needed)."
fi
echo
echo "Served read-only by the frontend nginx at /desktop/ (no restart needed):"
echo " https://<frontend-host>/desktop/latest-linux.yml"
+9 -2
View File
@@ -10,10 +10,16 @@ REGISTRY="gitea.rainbow-herring.ts.net/jakebreath/j621-frontend"
REGISTRY_HOST="$(printf '%s' "$REGISTRY" | cut -d/ -f1)"
SHA="${1:-$(git rev-parse --short HEAD)}"
BUILDER=multiarch
PLATFORMS="linux/amd64,linux/arm64"
PLATFORMS="${PLATFORMS:-linux/amd64,linux/arm64}"
echo "==> Logging in to $REGISTRY_HOST ..."
docker login "$REGISTRY_HOST"
if [ -n "${REGISTRY_USER:-}" ] && [ -n "${REGISTRY_TOKEN:-}" ]; then
# Non-interactive login for CI (workflow passes GITHUB_TOKEN).
printf '%s' "$REGISTRY_TOKEN" | docker login "$REGISTRY_HOST" \
-u "$REGISTRY_USER" --password-stdin
else
docker login "$REGISTRY_HOST"
fi
if ! docker buildx inspect "$BUILDER" >/dev/null 2>&1; then
echo "==> Creating buildx builder '$BUILDER' ..."
@@ -26,6 +32,7 @@ fi
echo "==> Building + pushing $PLATFORMS -> $REGISTRY:{latest,$SHA} ..."
docker buildx build --builder "$BUILDER" --push \
--platform "$PLATFORMS" \
--build-arg "GIT_HASH=$SHA" \
-f deploy/J621-Frontend \
-t "$REGISTRY:latest" \
-t "$REGISTRY:$SHA" \
+62
View File
@@ -0,0 +1,62 @@
#!/bin/sh
# Periodic maintenance for a J621 deployment (runs in the "scheduler"
# service; no host cron involved).
#
# sync_followed_tags + sync_followed_pools every SYNC_EVERY seconds (30 min)
# cleanup_similarity every CLEAN_EVERY seconds (1 h)
# refresh_guest_blacklist every BLACKLIST_EVERY (24 h)
#
# Waits for the database and migrations to be ready, runs everything once,
# then keeps checking. A failing command is logged and retried next interval,
# so a temporary e621 outage is not fatal.
set -u
SYNC_EVERY="${SYNC_EVERY:-1800}"
CLEAN_EVERY="${CLEAN_EVERY:-3600}"
BLACKLIST_EVERY="${BLACKLIST_EVERY:-86400}"
log() { echo "[scheduler] $(date -Iseconds) $*"; }
run() {
log "running: $*"
if "$@"; then
log "done: $*"
else
log "FAILED (retrying at the next interval): $*"
fi
}
command -v python >/dev/null || { log "python not found"; exit 1; }
log "waiting for the database and migrations..."
until python manage.py migrate --check >/dev/null 2>&1; do
sleep 10
done
log "database ready; intervals: sync=${SYNC_EVERY}s cleanup=${CLEAN_EVERY}s blacklist=${BLACKLIST_EVERY}s"
now=$(date +%s)
last_sync=$((now - SYNC_EVERY))
last_clean=$((now - CLEAN_EVERY))
last_blacklist=$((now - BLACKLIST_EVERY))
while true; do
now=$(date +%s)
if [ $((now - last_sync)) -ge "$SYNC_EVERY" ]; then
last_sync=$now
run python manage.py sync_followed_tags
run python manage.py sync_followed_pools
fi
if [ $((now - last_clean)) -ge "$CLEAN_EVERY" ]; then
last_clean=$now
run python manage.py cleanup_similarity
fi
if [ $((now - last_blacklist)) -ge "$BLACKLIST_EVERY" ]; then
last_blacklist=$now
run python manage.py refresh_guest_blacklist
fi
sleep 30
done
+16
View File
@@ -0,0 +1,16 @@
{
"TCP": {
"8443": {
"HTTPS": true
}
},
"Web": {
"${TS_CERT_DOMAIN}:8443": {
"Handlers": {
"/": {
"Proxy": "http://127.0.0.1:80"
}
}
}
}
}
+16
View File
@@ -0,0 +1,16 @@
{
"TCP": {
"443": {
"HTTPS": true
}
},
"Web": {
"${TS_CERT_DOMAIN}:443": {
"Handlers": {
"/": {
"Proxy": "http://127.0.0.1:80"
}
}
}
}
}
+16
View File
@@ -0,0 +1,16 @@
{
"TCP": {
"443": {
"HTTPS": true
}
},
"Web": {
"${TS_CERT_DOMAIN}:443": {
"Handlers": {
"/": {
"Proxy": "http://127.0.0.1:80"
}
}
}
}
}
+4
View File
@@ -0,0 +1,4 @@
node_modules/
dist/
release/
*.tsbuildinfo
+119
View File
@@ -0,0 +1,119 @@
# J621 desktop (Electron)
An Electron shell around the web build. The main process serves the SPA from
the privileged `app://j621` scheme — a real origin, so `localStorage`, OPFS,
Web Workers and WebCodecs all behave exactly like they do in Chrome — and the
SPA keeps its normal first-run flow: on launch it asks where the backend is
(or runs in local mode with no backend at all).
The renderer code is not forked: the shell loads `frontend/dist` as built by
`npm run build` in `frontend/`.
Licensed under the Jake Labs Non-Commercial Software Licence — see
`../LICENSE`.
## Layout
```
src/main.ts window, app:// protocol handler, menu, external links
src/preload.ts window.j621Desktop bridge
electron-builder.yml deb + pacman + nsis packaging
```
The main process attaches a `Referer` to requests for e621 hosts: Chromium
sends no referrer from a custom-scheme page, and e621's CDN answers
cross-site image loads without one with a 403 (which Chromium then blocks as
ORB). API calls work either way.
## Development
```bash
# terminal 1: the SPA with its /api proxy (needs the backend for library pages)
cd frontend && npm run dev
# terminal 2
cd desktop && npm run dev
```
`npm run dev` builds the main process and points the window at
`http://localhost:5173` (`J621_DEV_SERVER`). Without the dev server,
`npm start` builds both halves and serves the bundled SPA over `app://j621`.
## Builds
```bash
npm run dist:linux # release/j621-desktop_0.1.0_amd64.deb + j621-desktop-0.1.0.pkg.tar.zst
npm run dist:win # release/J621 Setup 0.1.0.exe (cross-built with wine; untested)
npm run dist:all
```
`deploy/build_desktop.sh` wraps the same commands, installs dependencies on
first run and prints sizes plus SHA-256 sums for release notes. Nothing is
published by it; the CD workflow attaches the artifacts to the Gitea release,
which is also the update feed.
The Arch package can be installed and removed with pacman:
```bash
sudo pacman -U release/j621-desktop-*.pkg.tar.zst
sudo pacman -R j621-desktop
```
It installs to `/opt/J621` with a `/usr/bin/j621-desktop` symlink and a
`io.j621.desktop` launcher entry (`StartupWMClass` matches the shell's
`desktopName`). The Windows installer is a per-user NSIS build, so no admin
rights are needed to install or update it — but it is unsigned, so SmartScreen
will warn, and it has not been smoke-tested on real Windows.
## Updates
The app checks only when asked (**J621 → Check for updates…** in the menu):
Linux packages install through pacman/dpkg, which needs administrator rights,
and the Windows build is unsigned, so nothing installs silently.
The check resolves the feed itself: it asks the Gitea API for the newest
non-draft `desktop-v*` release (`J621_UPDATE_REPO`, default
`https://gitea.rainbow-herring.ts.net/jakebreath/j621` — lowercase on purpose,
the API path is case-sensitive), picks the `latest-linux.yml` / `latest.yml`
asset for the platform and uses that release as an electron-updater generic
feed. The baked `publish.url` in `electron-builder.yml` is metadata only.
`J621_UPDATE_URL` overrides the whole lookup (the smoke test uses this).
Publishing a release:
1. Bump `version` in `desktop/package.json` — that is what the updater compares.
2. Run the manual CD workflow, which builds the packages and attaches them
plus both channel files to the Gitea release `desktop-v<version>`.
`./deploy/push_desktop.sh --win` does the same build locally (and can also
copy the files to the website feed, which is optional now).
3. Existing installs find the new version on their next manual check.
Note for the 0.1.1 → 0.1.2 step: 0.1.1 only knows the old `/desktop/` feed, so
publish 0.1.2 there once (`./deploy/push_desktop.sh --no-build`, or install it
manually). From 0.1.2 on, updates come from Gitea.
`package-type` in the app resources tells electron-updater whether to run
`pacman -U` or `dpkg -i` (both via pkexec/sudo); the per-user NSIS install
updates without elevation.
## Smoke test
`npm run smoke` builds everything, runs Electron headless through `xvfb-run`
and checks that the bundled SPA loads over `app://j621`: the SPA fallback, an
asset fetch, `localStorage`, OPFS, WebCodecs and the preload bridge. It exits
non-zero on failure. With `J621_UPDATE_URL` set it also checks that the feed
reports the expected version. Pointing `J621_DEV_SERVER` at the Vite dev
server checks the development path instead of the bundled one.
## Backend CORS
A desktop app is cross-origin to the backend, exactly like the frontend-only
Docker deployment. Add the shell's origin to the backend:
```
CORS_ALLOWED_ORIGINS=app://j621
```
The setup screen shows this hint (with the actual origin) when the connection
test fails. Without a backend, choose "Continue without a backend" — the
e621-facing pages work the same as in the browser.
Binary file not shown.

After

Width:  |  Height:  |  Size: 361 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 19 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 19 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.0 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 302 B

Binary file not shown.

After

Width:  |  Height:  |  Size: 405 B

Binary file not shown.

After

Width:  |  Height:  |  Size: 3.9 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 527 B

Binary file not shown.

After

Width:  |  Height:  |  Size: 758 B

Binary file not shown.

After

Width:  |  Height:  |  Size: 8.7 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 941 B

Binary file not shown.

After

Width:  |  Height:  |  Size: 1.4 KiB

+71
View File
@@ -0,0 +1,71 @@
appId: io.j621.desktop
productName: J621
copyright: Copyright (c) 2026 JakeBreath — Jake Labs Non-Commercial Software Licence
# Update feed: desktop/src/main.ts resolves the newest desktop-v* release on
# Gitea at check time (J621_UPDATE_REPO). This block only tells electron-builder
# to emit latest.yml/latest-linux.yml next to the installers; J621_UPDATE_URL
# overrides the feed for a fork or a test.
publish:
provider: generic
url: https://gitea.rainbow-herring.ts.net/JakeBreath/J621/releases
directories:
output: release
buildResources: build
# The Electron main process and preload live in dist/ (tsc output). The SPA is
# copied from the web build as an extra resource and served over app://j621.
files:
- dist/**
- package.json
extraResources:
- from: ../frontend/dist
to: dist
filter:
- "**/*"
linux:
target:
- deb
- pacman
category: AudioVideo
synopsis: Desktop client for the J621 self-hosted media archive
description: >-
J621 is a self-hosted browser for an e621-linked media library. This
desktop build talks to a J621 backend over the network, the same way the
web app does.
icon: build/icons
syncDesktopName: true
pacman:
compression: zstd
artifactName: ${name}-${version}.pkg.tar.zst
# Electron's shared-library requirements on Arch (the electron-builder
# defaults still list an ancient Electron 2 dependency set).
depends:
- gtk3
- nss
- libxss
- libxtst
- xdg-utils
- at-spi2-core
- libsecret
- libnotify
- alsa-lib
- libcups
- libdrm
- mesa
- libxkbcommon
win:
target:
- nsis
icon: build/icon.ico
nsis:
oneClick: false
perMachine: false
allowToChangeInstallationDirectory: true
shortcutName: J621
+3656
View File
File diff suppressed because it is too large Load Diff
+35
View File
@@ -0,0 +1,35 @@
{
"name": "j621-desktop",
"productName": "J621",
"version": "0.1.2",
"private": true,
"description": "Desktop shell for the J621 self-hosted media archive",
"author": {
"name": "JakeBreath",
"email": "tolozacdcmo@gmail.com"
},
"homepage": "https://gitea.rainbow-herring.ts.net/JakeBreath/J621",
"license": "LicenseRef-Jake-Labs-Non-Commercial",
"desktopName": "io.j621.desktop",
"main": "dist/main.js",
"scripts": {
"build:main": "tsc -p tsconfig.json",
"build:renderer": "npm --prefix ../frontend run build",
"build": "npm run build:main && npm run build:renderer",
"dev": "npm run build:main && J621_DEV_SERVER=http://localhost:5173 electron .",
"start": "npm run build && electron .",
"smoke": "npm run build && xvfb-run -a electron . --j621-smoke",
"dist:linux": "npm run build && electron-builder --linux",
"dist:win": "npm run build && electron-builder --win",
"dist:all": "npm run build && electron-builder --linux --win"
},
"devDependencies": {
"@types/node": "^24.13.6",
"electron": "^44.4.3",
"electron-builder": "^26.15.3",
"typescript": "^6.0.3"
},
"dependencies": {
"electron-updater": "^6.8.9"
}
}
+681
View File
@@ -0,0 +1,681 @@
/**
* Electron shell for the J621 SPA.
*
* The renderer is the normal web build (`frontend/dist`), served from the
* privileged `app://j621` scheme so it is a real secure origin: localStorage,
* OPFS, Web Workers, WebCodecs and `history.pushState` all behave exactly
* like they do in Chrome. In development `J621_DEV_SERVER` points the window
* at the Vite dev server instead, which keeps HMR and the `/api` proxy.
*
* The shell is deliberately thin: the SPA still asks for its backend URL on
* first launch (or runs in local e621-only mode) and talks to it over HTTP
* the same way it does in a browser. The only desktop-specific bits are the
* origin, external-link handling and (later) updates.
*/
import { app, BrowserWindow, dialog, ipcMain, Menu, protocol, session, shell } from "electron";
import { autoUpdater } from "electron-updater";
import { readFileSync, rmSync } from "node:fs";
import { readFile, stat, writeFile } from "node:fs/promises";
import path from "node:path";
const SCHEME = "app";
const HOST = "j621";
const APP_ORIGIN = `${SCHEME}://${HOST}`;
const DEV_SERVER = (process.env.J621_DEV_SERVER ?? "").replace(/\/+$/, "");
const SMOKE = process.argv.includes("--j621-smoke");
if (SMOKE) {
// A throwaway profile keeps the checks deterministic (always the first-run
// setup screen) and never touches real state or leaves a stale
// single-instance lock behind.
const smokeDir = path.join(app.getPath("temp"), "j621-smoke");
rmSync(smokeDir, { recursive: true, force: true });
app.setPath("userData", smokeDir);
}
protocol.registerSchemesAsPrivileged([
{
scheme: SCHEME,
privileges: {
standard: true,
secure: true,
supportFetchAPI: true,
corsEnabled: true,
stream: true,
},
},
]);
const MIME_TYPES: Record<string, string> = {
".html": "text/html; charset=utf-8",
".js": "text/javascript; charset=utf-8",
".mjs": "text/javascript; charset=utf-8",
".css": "text/css; charset=utf-8",
".json": "application/json; charset=utf-8",
".map": "application/json; charset=utf-8",
".svg": "image/svg+xml",
".png": "image/png",
".jpg": "image/jpeg",
".jpeg": "image/jpeg",
".webp": "image/webp",
".gif": "image/gif",
".ico": "image/x-icon",
".woff": "font/woff",
".woff2": "font/woff2",
".wasm": "application/wasm",
".mp4": "video/mp4",
".webm": "video/webm",
".txt": "text/plain; charset=utf-8",
};
/** Where `vite build` output lives, packaged or not. */
function rendererRoot(): string {
return app.isPackaged
? path.join(process.resourcesPath, "dist")
: path.resolve(__dirname, "..", "..", "frontend", "dist");
}
async function existingFile(candidate: string): Promise<boolean> {
try {
return (await stat(candidate)).isFile();
} catch {
return false;
}
}
async function fileResponse(file: string): Promise<Response> {
const body = await readFile(file);
const type =
MIME_TYPES[path.extname(file).toLowerCase()] ?? "application/octet-stream";
return new Response(body, {
headers: { "content-type": type, "cache-control": "no-cache" },
});
}
/** Paths the nginx proxy sends to Django; there is no backend behind app://. */
const BACKEND_PREFIXES = ["/api/", "/admin/", "/static/", "/health"];
/**
* e621's CDN refuses cross-site image loads that carry no Referer
* (`Sec-Fetch-Site: cross-site` with an empty referrer returns 403), and
* Chromium never sends one for pages on a custom scheme like app://j621.
* Attach a normal e621 referrer to its hosts so images load; API calls are
* unaffected.
*/
function installRefererFix(targetSession: Electron.Session): void {
targetSession.webRequest.onBeforeSendHeaders(
{ urls: ["*://e621.net/*", "*://*.e621.net/*"] },
(details, callback) => {
const headers = details.requestHeaders;
if (!headers.Referer && !headers.referer) {
headers.Referer = "https://e621.net/";
}
callback({ requestHeaders: headers });
},
);
}
async function handleAppRequest(request: Request): Promise<Response> {
const url = new URL(request.url);
if (url.host !== HOST) return new Response("Not found", { status: 404 });
const root = rendererRoot();
let pathname = decodeURIComponent(url.pathname);
if (!pathname || pathname === "/") pathname = "/index.html";
// Never answer backend paths with the SPA: that made the setup screen's
// connection test "succeed" against the shell's own origin.
if (
BACKEND_PREFIXES.some(
(prefix) => pathname === prefix.replace(/\/$/, "") || pathname.startsWith(prefix),
)
) {
return new Response(
JSON.stringify({
detail: "No backend is attached to the desktop app; set one in /setup.",
}),
{ status: 404, headers: { "content-type": "application/json" } },
);
}
const target = path.resolve(root, "." + pathname);
if (target !== root && !target.startsWith(root + path.sep)) {
return new Response("Forbidden", { status: 403 });
}
try {
if (await existingFile(target)) return await fileResponse(target);
// Missing assets keep their 404; anything else is a client route and
// falls back to the SPA entry point (BrowserRouter).
if (path.extname(pathname)) return new Response("Not found", { status: 404 });
return await fileResponse(path.join(root, "index.html"));
} catch (error) {
const hint = app.isPackaged
? "The bundled frontend is missing from the application resources."
: "Build the frontend first: npm --prefix ../frontend run build";
console.error("[app://]", error);
return new Response(`${hint}\n`, { status: 500 });
}
}
interface WindowState {
x?: number;
y?: number;
width: number;
height: number;
maximized: boolean;
}
const DEFAULT_WINDOW_STATE: WindowState = {
width: 1440,
height: 900,
maximized: false,
};
function stateFile(): string {
return path.join(app.getPath("userData"), "window-state.json");
}
function readWindowState(): WindowState {
try {
const parsed = JSON.parse(
readFileSync(stateFile(), "utf8"),
) as Partial<WindowState>;
return { ...DEFAULT_WINDOW_STATE, ...parsed };
} catch {
return DEFAULT_WINDOW_STATE;
}
}
function saveWindowState(win: BrowserWindow): void {
if (win.isDestroyed()) return;
const { x, y, width, height } = win.getNormalBounds();
const state: WindowState = {
x,
y,
width,
height,
maximized: win.isMaximized(),
};
void writeFile(stateFile(), JSON.stringify(state)).catch(() => {
// Window geometry is a nice-to-have; never fail a quit over it.
});
}
function openExternal(rawUrl: string): void {
if (/^https?:/i.test(rawUrl)) void shell.openExternal(rawUrl);
}
/** File extensions the SPA navigates to when it wants a save dialog. */
const DOWNLOAD_EXTENSIONS = new Set([
".jpg",
".jpeg",
".jpe",
".png",
".gif",
".webp",
".avif",
".mp4",
".webm",
".mov",
".swf",
".zip",
".pdf",
".bin",
]);
/**
* Downloads in the SPA are same-window navigations (`location.href =
* url?download=1`, `<a href=...>`), and Chromium cancels those when
* `will-navigate` is prevented. Recognise them so they become real
* downloads instead of being opened in the browser.
*/
function isDownloadNavigation(url: URL): boolean {
if (/(?:^|&)download=1(?:&|$)/.test(url.search.slice(1))) return true;
return DOWNLOAD_EXTENSIONS.has(path.extname(url.pathname).toLowerCase());
}
/**
* Updates are manual by design: Linux packages install through pacman/dpkg
* (pkexec/sudo) and the Windows build is unsigned, so the app asks before
* downloading and again before installing. The feed is resolved at check time
* from the newest `desktop-v*` release on Gitea (`J621_UPDATE_REPO`);
* `J621_UPDATE_URL` overrides it for forks and the smoke test.
*/
type UpdateEvent =
| { state: "available"; version: string }
| { state: "not-available" }
| { state: "downloaded"; version: string }
| { state: "error"; message: string };
let updateReporter: ((event: UpdateEvent) => void) | null = null;
let updateCheckRunning = false;
const UPDATE_REPO =
process.env.J621_UPDATE_REPO?.trim() ||
"https://gitea.rainbow-herring.ts.net/jakebreath/j621";
interface ReleaseAsset {
name: string;
browser_download_url: string;
}
interface Release {
tag_name: string;
draft: boolean;
prerelease: boolean;
assets?: ReleaseAsset[];
}
function releaseVersion(tag: string): [number, number, number] | null {
const match = /^desktop-v(\d+)\.(\d+)\.(\d+)$/.exec(tag);
if (!match) return null;
return [Number(match[1]), Number(match[2]), Number(match[3])];
}
function compareVersions(
a: [number, number, number],
b: [number, number, number],
): number {
for (let index = 0; index < 3; index += 1) {
if (a[index] !== b[index]) return a[index] - b[index];
}
return 0;
}
/**
* Resolve the generic feed base electron-updater should use.
*
* Gitea's API path is case-sensitive (owner/repo must match the login), while
* the asset URLs it returns are canonical, so the base is derived from the
* platform's metadata asset (`latest-linux.yml` / `latest.yml`).
*/
async function resolveReleaseFeed(): Promise<string> {
const override = process.env.J621_UPDATE_URL?.trim();
if (override) return override;
const match = /^(https?:\/\/[^/]+)\/([^/]+)\/([^/]+?)\/?$/.exec(UPDATE_REPO);
if (!match) {
throw new Error(
`J621_UPDATE_REPO must be <origin>/<owner>/<repo> (got ${UPDATE_REPO}).`,
);
}
const [, origin, owner, repo] = match;
const response = await fetch(
`${origin}/api/v1/repos/${owner}/${repo}/releases?limit=50`,
{ headers: { Accept: "application/json" } },
);
if (!response.ok) {
throw new Error(`Release lookup on Gitea failed (HTTP ${response.status}).`);
}
const releases = (await response.json()) as Release[];
const assetName =
process.platform === "win32" ? "latest.yml" : "latest-linux.yml";
let best: { version: [number, number, number]; asset: ReleaseAsset } | null =
null;
for (const release of releases) {
if (release.draft || release.prerelease) continue;
const version = releaseVersion(release.tag_name);
if (!version) continue;
const asset = release.assets?.find((entry) => entry.name === assetName);
if (!asset) continue;
if (!best || compareVersions(version, best.version) > 0) {
best = { version, asset };
}
}
if (!best) {
throw new Error(`No ${assetName} asset found in ${UPDATE_REPO} releases.`);
}
return best.asset.browser_download_url.replace(/\/[^/]*$/, "");
}
function setUpUpdates(win: BrowserWindow): void {
if (!app.isPackaged) autoUpdater.forceDevUpdateConfig = true;
autoUpdater.autoDownload = false;
autoUpdater.autoInstallOnAppQuit = false;
autoUpdater.on("error", (error) => {
updateReporter?.({ state: "error", message: error.message });
});
autoUpdater.on("update-not-available", () => {
if (SMOKE) return updateReporter?.({ state: "not-available" });
void dialog.showMessageBox(win, {
type: "info",
title: "Updates",
message: `J621 ${app.getVersion()} is up to date.`,
});
});
autoUpdater.on("update-available", (info) => {
if (SMOKE) {
return updateReporter?.({ state: "available", version: info.version });
}
void (async () => {
const { response } = await dialog.showMessageBox(win, {
type: "info",
title: "Updates",
message: `J621 ${info.version} is available.`,
detail:
"Download it now? Installing it later needs administrator rights " +
"on Linux (pacman/dpkg); the Windows installer runs without them.",
buttons: ["Download", "Later"],
defaultId: 0,
cancelId: 1,
});
if (response !== 0) return;
try {
await autoUpdater.downloadUpdate();
} catch (error) {
await dialog.showMessageBox(win, {
type: "error",
title: "Update failed",
message: "The update could not be downloaded.",
detail: error instanceof Error ? error.message : String(error),
});
}
})();
});
autoUpdater.on("download-progress", (progress) => {
win.setProgressBar(Math.min(progress.percent / 100, 1));
});
autoUpdater.on("update-downloaded", (info) => {
win.setProgressBar(-1);
if (SMOKE) {
return updateReporter?.({ state: "downloaded", version: info.version });
}
void (async () => {
const { response } = await dialog.showMessageBox(win, {
type: "info",
title: "Updates",
message: `J621 ${info.version} is ready to install.`,
detail:
process.platform === "linux"
? "Installing asks for administrator rights and then restarts J621."
: "J621 will restart to finish installing.",
buttons: ["Restart and install", "Later"],
defaultId: 0,
cancelId: 1,
});
if (response === 0) {
autoUpdater.quitAndInstall(false, true);
} else {
// Linux packages elevate, so never install silently on quit there.
autoUpdater.autoInstallOnAppQuit = process.platform !== "linux";
}
})();
});
}
async function checkForUpdates(win: BrowserWindow): Promise<void> {
if (updateCheckRunning) return;
updateCheckRunning = true;
try {
const feed = await resolveReleaseFeed();
autoUpdater.setFeedURL({ provider: "generic", url: feed });
await autoUpdater.checkForUpdates();
} catch (error) {
const message = error instanceof Error ? error.message : String(error);
updateReporter?.({ state: "error", message });
if (!SMOKE) {
await dialog.showMessageBox(win, {
type: "error",
title: "Updates",
message: "Could not check for updates.",
detail: message,
});
}
} finally {
updateCheckRunning = false;
}
}
let mainWindow: BrowserWindow | null = null;
function buildMenu(win: BrowserWindow): void {
const template: Electron.MenuItemConstructorOptions[] = [
{
label: "J621",
submenu: [
{
label: "Backend setup…",
click: () => void win.loadURL(`${APP_ORIGIN}/setup`),
},
{
label: "Open backend in browser",
click: () => win.webContents.send("j621:open-backend"),
},
{
label: "Check for updates…",
click: () => void checkForUpdates(win),
},
{ type: "separator" },
{ role: "quit" },
],
},
{ role: "editMenu" },
{
label: "View",
submenu: [
{ role: "reload" },
{ role: "forceReload" },
{ role: "toggleDevTools" },
{ type: "separator" },
{ role: "resetZoom" },
{ role: "zoomIn" },
{ role: "zoomOut" },
{ type: "separator" },
{ role: "togglefullscreen" },
],
},
];
Menu.setApplicationMenu(Menu.buildFromTemplate(template));
}
function createWindow(): BrowserWindow {
const state = readWindowState();
const win = new BrowserWindow({
x: state.x,
y: state.y,
width: state.width,
height: state.height,
minWidth: 960,
minHeight: 600,
backgroundColor: "#11111b",
show: false,
webPreferences: {
preload: path.join(__dirname, "preload.js"),
contextIsolation: true,
nodeIntegration: false,
sandbox: true,
spellcheck: false,
},
});
if (state.maximized) win.maximize();
win.once("ready-to-show", () => win.show());
win.on("close", () => saveWindowState(win));
win.on("closed", () => {
mainWindow = null;
});
// Links to the web open in the user's browser, never in this window.
win.webContents.setWindowOpenHandler(({ url }) => {
openExternal(url);
return { action: "deny" };
});
win.webContents.on("will-navigate", (event, url) => {
const allowed = DEV_SERVER
? url.startsWith(DEV_SERVER)
: url.startsWith(APP_ORIGIN);
if (allowed) return;
event.preventDefault();
let parsed: URL;
try {
parsed = new URL(url);
} catch {
return;
}
if (/^https?:$/.test(parsed.protocol) && isDownloadNavigation(parsed)) {
win.webContents.downloadURL(url);
return;
}
openExternal(url);
});
buildMenu(win);
setUpUpdates(win);
void win.loadURL(DEV_SERVER || `${APP_ORIGIN}/`);
if (SMOKE) runSmokeTest(win);
return win;
}
/**
* `npm run smoke`: load the bundled SPA and check the things that are only
* true under the custom scheme. Exits non-zero on the first broken promise.
*/
function runSmokeTest(win: BrowserWindow): void {
win.webContents.once("did-finish-load", () => {
void (async () => {
try {
const result = (await win.webContents.executeJavaScript(`(async () => {
const waitFor = async (probe, timeout = 8000) => {
const start = Date.now();
while (Date.now() - start < timeout) {
if (probe()) return true;
await new Promise((resolve) => setTimeout(resolve, 100));
}
return false;
};
const asset = await fetch("/favicon.svg");
const route = await fetch("/gallery/some/deep/route");
const routeBody = await route.text();
const health = await fetch("/health");
localStorage.setItem("j621.smoke", "ok");
history.pushState({}, "", "/gallery");
return {
origin: location.origin,
title: document.title,
assetOk: asset.ok && (await asset.text()).includes("<svg"),
routeOk: route.ok && routeBody.includes('id="root"'),
// /health must not fall back to the SPA (false "connected" test).
healthOk: health.status === 404,
// Testing an empty backend URL on the desktop must say so instead
// of "Connected" against the shell's own origin.
emptyTestOk: await (async () => {
const button = [...document.querySelectorAll("button")].find((b) =>
b.textContent.includes("Test connection"),
);
if (!button) return false;
button.click();
return waitFor(() => document.body.innerText.includes("no backend of its own"));
})(),
mounted: await waitFor(() => (document.querySelector("#root")?.childElementCount ?? 0) > 0),
setupOk: document.body.innerText.includes("Where is your backend?"),
storageOk: localStorage.getItem("j621.smoke") === "ok",
historyOk: location.pathname === "/gallery",
opfsOk: typeof navigator.storage?.getDirectory === "function",
webcodecsOk: typeof VideoEncoder !== "undefined",
bridgeOk: window.j621Desktop?.isDesktop === true,
};
})()`, true)) as Record<string, unknown>;
const expected: Record<string, unknown> = {
// In dev the window hosts the Vite dev server, which has no /setup
// screen (the SPA only asks for a backend in production builds).
origin: DEV_SERVER || APP_ORIGIN,
title: "J621",
assetOk: true,
routeOk: true,
mounted: true,
setupOk: !DEV_SERVER,
// Both are app://-only checks: the Vite dev server serves the SPA
// for /health and never shows the setup screen.
healthOk: !DEV_SERVER,
emptyTestOk: !DEV_SERVER,
storageOk: true,
historyOk: true,
opfsOk: true,
webcodecsOk: true,
bridgeOk: true,
};
const failures = Object.entries(expected).filter(
([key, value]) => result[key] !== value,
);
console.log(
"[smoke] electron",
process.versions.electron,
"chrome",
process.versions.chrome,
);
console.log("[smoke]", JSON.stringify(result));
if (failures.length > 0) {
console.error(
"[smoke] FAILED:",
failures.map(([key]) => key).join(", "),
);
app.exit(1);
return;
}
if (process.env.J621_UPDATE_URL) {
const update = await new Promise<UpdateEvent>((resolve) => {
updateReporter = resolve;
const timer = setTimeout(
() => resolve({ state: "error", message: "update check timed out" }),
60_000,
);
void checkForUpdates(win).finally(() => clearTimeout(timer));
});
updateReporter = null;
console.log("[smoke] update:", JSON.stringify(update));
if (update.state === "error") {
console.error("[smoke] FAILED: update check");
app.exit(1);
return;
}
}
console.log("[smoke] OK");
app.exit(0);
} catch (error) {
console.error("[smoke] FAILED:", error);
app.exit(1);
}
})();
});
}
const gotLock = app.requestSingleInstanceLock();
if (!gotLock) {
app.quit();
} else {
app.on("second-instance", () => {
if (!mainWindow) return;
if (mainWindow.isMinimized()) mainWindow.restore();
mainWindow.focus();
});
app.setAppUserModelId("io.j621.desktop");
app.whenReady().then(() => {
protocol.handle(SCHEME, handleAppRequest);
installRefererFix(session.defaultSession);
ipcMain.handle("j621:version", () => app.getVersion());
ipcMain.handle("j621:open-external", (_event, url: unknown) => {
if (typeof url === "string") openExternal(url);
});
mainWindow = createWindow();
app.on("activate", () => {
if (BrowserWindow.getAllWindows().length === 0) {
mainWindow = createWindow();
}
});
});
app.on("window-all-closed", () => {
if (process.platform !== "darwin") app.quit();
});
}
+27
View File
@@ -0,0 +1,27 @@
/**
* The only bridge between the desktop shell and the SPA. Keep it tiny: the
* renderer is the normal web build and must not need Electron to work.
*/
import { contextBridge, ipcRenderer } from "electron";
contextBridge.exposeInMainWorld("j621Desktop", {
isDesktop: true as const,
platform: process.platform,
// The SPA's own origin (`app://j621`), shown by the setup screen when it
// has to explain which origin to allow in the backend's CORS settings.
origin: window.location.origin,
getVersion: (): Promise<string> => ipcRenderer.invoke("j621:version"),
});
// Menu → "Open backend in browser": read the SPA's stored backend URL and
// hand it to the main process after checking it is a real web URL.
ipcRenderer.on("j621:open-backend", () => {
let url = "";
try {
url = window.localStorage.getItem("j621.backend") ?? "";
} catch {
// Storage can be unavailable in rare cases; nothing to open then.
}
if (url === "none") url = "";
if (url) void ipcRenderer.invoke("j621:open-external", url);
});
+19
View File
@@ -0,0 +1,19 @@
{
"compilerOptions": {
"target": "ES2022",
"module": "node16",
"moduleResolution": "node16",
"lib": ["ES2023", "DOM"],
"types": ["node"],
"outDir": "dist",
"rootDir": "src",
"sourceMap": true,
"strict": true,
"noUnusedLocals": true,
"noUnusedParameters": true,
"esModuleInterop": true,
"skipLibCheck": true,
"forceConsistentCasingInFileNames": true
},
"include": ["src"]
}
+119
View File
@@ -0,0 +1,119 @@
# fish_greeting (updated for the J621 rewrite)
Shows a random library image as your shell greeting, using fastfetch. This is
the updated version of the script from
`J621-Django/extras/fish_greeting system/fish_greeting.fish`, adapted to the
new REST API.
## What changed from the old script
| | Old J621-Django | This version |
| --- | --- | --- |
| Endpoint | `GET /random/?rating=X` returned **image bytes** | `GET /api/random/?fastfetch=1&rating=X` returns **JSON** |
| Image access | same response, `X-File-MD5`/`X-File-Name` headers | signed `url` from the JSON, downloaded separately |
| Unsupported types | rolled again (recursion) | server only returns png/jpg/gif in fastfetch mode |
| Config | hardcoded `https://j621.jake.i` | `J621_BASE` / `J621_TOKEN` / `J621_WEB` |
| Printed link | `/view/<md5>` on the old host | `/detail/<J-ID>` on `J621_WEB` |
| Modes, logging, gifsicle, fastfetch flags | same | same (`~/.config/fish/greeting_mode`, `~/.config/j621Logos/logs.log`) |
The `gm-switch` helper and the `.desktop` launchers from the old repo keep
working unchanged: they write the same `~/.config/fish/greeting_mode` file
(0 = NSFW, 1 = SFW, 2 = Questionable).
## Install
```fish
cd extras/fish_greeting
fish install.fish
```
The installer copies the function into `~/.config/fish/functions/`, **asks for
your API origin** (and an optional scoped token — see below), writes
`~/.config/j621Greeting/config.fish` (mode 600, it may hold the token), checks
the tools it needs and then verifies the backend: `/health` must answer and a
random roll is attempted. It exits non-zero when the backend cannot be
reached.
It also appends a guarded block to `~/.config/fish/config.fish` that sources
the function. That is needed on setups whose distro/theme config (CachyOS,
Oh My Fish, hand-rolled `config.fish`) defines `fish_greeting` inline while
the shell starts — such a definition wins over autoloading from
`functions/`, so ours has to be loaded afterwards. The block is written once
and repeated installs leave it alone; pass `--no-config` to skip it.
Non-interactive / re-install:
```fish
fish install.fish --url https://j621.example.ts.net --token <api-token>
fish install.fish --no-prompt # defaults, never asks
fish install.fish --force # rewrite an existing config
```
Then test:
```fish
fish_greeting
```
Requirements: `curl` (or `wget`), `fastfetch`, `file`. Optional: `gifsicle`
(GIF downscaling), `jq` or `python3` (JSON parsing — without either it falls
back to grep/sed).
## No token? That is fine
The greeting is meant to run **without** a token: it then behaves like a
guest of your instance — unsigned image links and only items that are visible
to guests. Adding a token to the config unlocks signed links (useful when
something else fetches the URL for you) and items that are hidden from
guests.
## Configuration
Any of these work; the config file is read by the function on every run:
```fish
# ~/.config/j621Greeting/config.fish, or universal variables
set -g J621_BASE https://j621.rainbow-herring.ts.net # API origin
set -g J621_WEB https://j621.example.ts.net # link origin (split deploys)
set -g J621_TOKEN <api token> # signed URLs + hidden items
set -g J621_FASTFETCH_CONFIG jake # fastfetch config name
```
`J621_TOKEN` is optional. The recommended value is a **scoped greeting
token**: open the app, go to *Account → Shell tokens* (or `/tokens`), create
one and copy the `j621r_…` key — it only works with `/api/random/`, so it is
safe to keep in this config. It also gives you signed image URLs and access
to items that are hidden from guests. Without a token the greeting runs as a
guest and sees the public library only.
## Rating filters
| `greeting_mode` | Rating requested | Label |
| --- | --- | --- |
| `0` | `e` | NSFW |
| `1` | `s` | SFW |
| `2` (default) | `q` | Questionable |
## Troubleshooting
- **The distro/theme greeting still shows** (CachyOS, Oh My Fish, …): those
configs define `fish_greeting` while the shell starts, which beats
autoloading. Check what fish resolved:
`functions --details fish_greeting` — it must point at
`~/.config/fish/functions/fish_greeting.fish`. If it points at a distro
file, run `fish install.fish` again (it adds the source block to
`config.fish`) or add it yourself:
```fish
if test -f ~/.config/fish/functions/fish_greeting.fish
source ~/.config/fish/functions/fish_greeting.fish
end
```
at the **end** of `~/.config/fish/config.fish`.
- `No logo (No image matches those filters.)` — the library has no images for
that rating; try another mode or add files.
- `No logo (API error)` — check `J621_BASE`, and that the deployment is up
(`https://<host>/health`).
- `No logo (download failed)` — the signed URL expired (they last 24 h) or the
item was deleted between the two requests; just run it again.
- The greeting is slow — the API and image fetch have `--max-time` guards; a
slow tailnet link is usually the cause.
+19
View File
@@ -0,0 +1,19 @@
# Copy to ~/.config/j621Greeting/config.fish and adjust. All of these can also
# be universal variables, e.g. `set -Ux J621_BASE https://j621.example.ts.net`.
# API origin (no trailing slash needed).
set -g J621_BASE https://j621.rainbow-herring.ts.net
# Web origin used in the printed link. Only needed when the SPA is served
# from a different host than the API (split deployment).
# set -g J621_WEB https://j621-frontend.rainbow-herring.ts.net
# API token. Optional: gives you signed image URLs and access to items that
# are hidden from guests. Use a scoped greeting token (Account -> Shell
# tokens, or /tokens in the app): those only work with /api/random/, so they
# are safe to keep in this file. The full API token works too, but grants
# everything.
# set -g J621_TOKEN paste-your-token-here
# fastfetch config name used for the greeting logo.
# set -g J621_FASTFETCH_CONFIG jake
+219
View File
@@ -0,0 +1,219 @@
# fish_greeting — show a random library image as the shell greeting.
#
# Updated for the J621 Docker/REST rewrite: the API answers with JSON that
# carries a signed URL instead of streaming the image, so this function asks
# for one random png/jpg/gif (fastfetch mode), downloads the signed link and
# hands the file to fastfetch. Based on the original script from
# J621-Django/extras/fish_greeting system/fish_greeting.fish.
#
# Requires: curl (or wget) and fastfetch; gifsicle is used for animated GIFs
# (without it, the GIF is shown as-is).
#
# Rating modes (same file as the original script):
# ~/.config/fish/greeting_mode 0 = NSFW (explicit), 1 = SFW (safe),
# 2 = Questionable (default)
#
# Configuration, any of these (all optional):
# ~/.config/j621Greeting/config.fish with `set -g VAR value` lines, or
# universal variables, e.g. `set -Ux J621_BASE https://j621.example.ts.net`
# J621_BASE API origin (default https://j621.rainbow-herring.ts.net)
# J621_TOKEN API token, sent as `Authorization: Token …`; needed for
# signed URLs and for hidden-from-guests items (optional)
# J621_WEB Web origin used in the printed link (defaults to J621_BASE)
# J621_FASTFETCH_CONFIG fastfetch config name (default "jake")
function __j621_json_field --argument-names json field
if command -v jq >/dev/null 2>&1
printf '%s' "$json" | jq -r --arg field "$field" '.[$field] // empty'
else if command -v python3 >/dev/null 2>&1
printf '%s' "$json" | python3 -c "import json,sys; value = json.load(sys.stdin).get(sys.argv[1], ''); print(value if value is not None else '')" $field
else
printf '%s' "$json" | grep -o "\"$field\"[[:space:]]*:[[:space:]]*\"[^\"]*\"" | head -1 | sed -E 's/.*:[[:space:]]*"//; s/"$//'
end
end
function __j621_fetch_random --argument-names rating
set base (string trim --right --chars=/ "$J621_BASE")
set api_url "$base/api/random/?fastfetch=1&rating=$rating"
set curl_args -s -L --max-time 20
if set -q J621_TOKEN; and test -n "$J621_TOKEN"
set -a curl_args -H "Authorization: Token $J621_TOKEN"
end
if command -v curl >/dev/null 2>&1
curl $curl_args "$api_url" | string collect
else if command -v wget >/dev/null 2>&1
wget -qO- "$api_url" | string collect
end
end
function fish_greeting --argument-names depth
# Initialize depth to 0 if not provided or empty
if not set -q depth; or test -z "$depth"
set depth 0
end
set MAX_DEPTH 3
# --- Configuration ---
set config_file ~/.config/j621Greeting/config.fish
if test -f $config_file
source $config_file
end
set -q J621_BASE; or set -g J621_BASE https://j621.rainbow-herring.ts.net
set -q J621_WEB; or set -g J621_WEB $J621_BASE
set -q J621_FASTFETCH_CONFIG; or set -g J621_FASTFETCH_CONFIG jake
set web_base (string trim --right --chars=/ "$J621_WEB")
# --- Setup logging ---
set log_dir ~/.config/j621Logos
if not test -d $log_dir
mkdir -p $log_dir
end
set log_file $log_dir/logs.log
# --- Mode selection ---
if test -f ~/.config/fish/greeting_mode
set mode (cat ~/.config/fish/greeting_mode | string trim)
else
set mode 2
end
switch $mode
case "0"
set rating e
set modename "NSFW"
case "1"
set rating s
set modename "SFW"
case "2"
set rating q
set modename "Questionable"
case "*"
echo "Unknown mode, using default NSFW"
set rating e
set modename "NSFW"
end
# --- Ask the API for a random image (JSON with a signed URL) ---
set json (__j621_fetch_random $rating)
set image_url (__j621_json_field "$json" url)
set j_id (__j621_json_field "$json" j_id)
set md5 (__j621_json_field "$json" md5)
set filename (__j621_json_field "$json" filename)
if test -z "$image_url"
set detail (__j621_json_field "$json" detail)
fastfetch --config "$J621_FASTFETCH_CONFIG"
if test -n "$detail"
printf '\e[4;2;38;2;138;0;222;49mNo logo (%s)\e[0m\n' "$detail"
else
printf '\e[4;2;38;2;138;0;222;49mNo logo (API error)\e[0m\n'
end
echo "$(date '+%Y-%m-%d %H:%M:%S') No logo for rating=$rating: $detail" >> "$log_file"
return
end
# --- Download the signed image ---
set tempfile (mktemp /tmp/fastfetch_logo_XXXXXX)
set download_success 0
if command -v curl >/dev/null 2>&1
curl -s -L --max-time 60 -o "$tempfile" "$image_url"
if test $status -eq 0 -a -s "$tempfile"
set download_success 1
end
else if command -v wget >/dev/null 2>&1
wget -q -O "$tempfile" "$image_url"
if test $status -eq 0 -a -s "$tempfile"
set download_success 1
end
end
if test $download_success -ne 1
fastfetch --config "$J621_FASTFETCH_CONFIG"
printf '\e[4;2;38;2;138;0;222;49mNo logo (download failed)\e[0m\n'
echo "$(date '+%Y-%m-%d %H:%M:%S') Download failed for $j_id" >> "$log_file"
rm -f "$tempfile"
return
end
# --- Determine MIME type ---
set mime (file --mime-type -b "$tempfile" 2>/dev/null | string trim | string collect)
# --- Accept only PNG, JPEG, GIF; everything else causes one more roll ---
if not string match -q "image/png" "$mime"; and not string match -q "image/jpeg" "$mime"; and not string match -q "image/gif" "$mime"
if test $depth -lt $MAX_DEPTH
fish_greeting (math $depth + 1)
else
printf '\e[31mMax recursion depth reached, skipping unsupported file type: %s\e[0m\n' "$mime"
end
if test $depth -eq 0
printf '\e[31mForked (Got unsupported type: %s - %s)\e[0m\n' "$mime" "$filename"
if test -n "$md5"
echo "Link: $web_base/view/$md5"
end
end
rm -f "$tempfile" "$tempfile.tmp" 2>/dev/null
return
end
# --- Process the file (only PNG, JPEG, GIF reach here) ---
set logo_type "kitty"
set processing_success 1
switch "$mime"
case "image/png" "image/jpeg"
set logo_type "kitty"
case "image/gif"
set processed_ok 0
set simple_file (mktemp /tmp/fastfetch_simple_XXXXXX)
set gif_err (mktemp)
if command -v gifsicle >/dev/null 2>&1
gifsicle --colors 255 "$tempfile" > "$simple_file" 2> "$gif_err"
set gif_exit $status
if test $gif_exit -eq 0 -a -s "$simple_file"
gifsicle --unoptimize --resize-fit 360x360 "$simple_file" > "$tempfile.tmp" 2> "$gif_err"
set final_exit $status
set gif_warnings (cat "$gif_err" | string trim)
if test $final_exit -eq 0 -a -s "$tempfile.tmp" -a -z "$gif_warnings"
mv "$tempfile.tmp" "$tempfile"
set processed_ok 1
set logo_type "kitty-icat"
echo "$(date '+%Y-%m-%d %H:%M:%S') GIF processed with gifsicle" >> "$log_file"
else
echo "$(date '+%Y-%m-%d %H:%M:%S') gifsicle final failed: $gif_warnings" >> "$log_file"
end
else
echo "$(date '+%Y-%m-%d %H:%M:%S') gifsicle --colors failed" >> "$log_file"
end
end
rm -f "$simple_file" "$gif_err"
if test $processed_ok -eq 0
set logo_type "kitty-icat"
echo "$(date '+%Y-%m-%d %H:%M:%S') Using original GIF (processing failed)" >> "$log_file"
end
end
# --- Display result ---
if test $processing_success -eq 1
fastfetch --config "$J621_FASTFETCH_CONFIG" --logo-type "$logo_type" --logo "$tempfile" --logo-width 35
set timestamp (date '+%Y-%m-%d %H:%M:%S')
echo "$timestamp Downloaded: $j_id $filename [$modename]" >> "$log_file"
if test -n "$filename"
printf '\e[4;2;38;2;138;0;222;49mLogo from API: %s (%s, %s)\e[0m\n' "$filename" "$modename" "$j_id"
echo "Link: $web_base/detail/$j_id"
else
printf '\e[4;2;38;2;138;0;222;49mLogo from API (%s)\e[0m\n' "$modename"
end
else
fastfetch --config "$J621_FASTFETCH_CONFIG"
printf '\e[4;2;38;2;138;0;222;49mNo logo (Image processing error)\e[0m\n'
end
# --- Keep only the last 1000 lines of the log ---
set log_tmp (mktemp)
tail -n 1000 "$log_file" > "$log_tmp" 2>/dev/null
mv "$log_tmp" "$log_file"
# --- Cleanup ---
rm -f "$tempfile" "$tempfile.tmp" 2>/dev/null
end
+178
View File
@@ -0,0 +1,178 @@
#!/usr/bin/env fish
# Install the J621 fish greeting:
# * copies fish_greeting.fish into ~/.config/fish/functions/
# * writes ~/.config/j621Greeting/config.fish (asks for the API origin and an
# optional token, or takes them as flags)
# * checks the tools it needs and probes the configured backend
#
# Usage:
# fish install.fish
# fish install.fish --url https://j621.example.ts.net --token <api-token>
# fish install.fish --no-prompt # never ask, keep/derive defaults
# fish install.fish --force # rewrite an existing config
# fish install.fish --no-config # do not touch ~/.config/fish/config.fish
argparse 'url=' 'token=' 'no-prompt' 'force' 'no-config' 'help' -- $argv
or begin
echo "Try: fish install.fish --help"
exit 1
end
if set -q _flag_help
sed -n '2,12p' (status --current-filename) | sed 's/^# \{0,1\}//'
exit 0
end
set -l here (path resolve (dirname (status --current-filename)))
set -l functions_dir ~/.config/fish/functions
set -l config_dir ~/.config/j621Greeting
set -l config_file $config_dir/config.fish
set -l default_url https://j621.rainbow-herring.ts.net
set -l url $default_url
set -l token ""
set -l problems 0
if set -q _flag_url
set url $_flag_url
end
if set -q _flag_token
set token $_flag_token
end
# --- Install the function -----------------------------------------------------
mkdir -p $functions_dir
cp "$here/fish_greeting.fish" "$functions_dir/fish_greeting.fish"
echo "Installed $functions_dir/fish_greeting.fish"
# --- Override distro/OMF greetings -------------------------------------------
# Some setups (CachyOS, OMF themes, hand-rolled configs) define fish_greeting
# inline while config.fish is being read. A function defined that way wins over
# autoloading, so our file would never load. Source it from the end of
# config.fish to define ours last.
set -l marker "# >>> J621 greeting >>>"
set -l user_config ~/.config/fish/config.fish
if set -q _flag_no_config
echo "Skipping $user_config (--no-config)"
else if test -f $user_config; and grep -qF "$marker" $user_config
echo "Config already sources the greeting (marker present)"
else
mkdir -p (dirname $user_config)
begin
echo ""
echo "$marker"
echo "# Loaded after the distro config so it wins over an inline fish_greeting."
echo "if test -f $functions_dir/fish_greeting.fish"
echo " source $functions_dir/fish_greeting.fish"
echo "end"
echo "# <<< J621 greeting <<<"
end >> $user_config
echo "Added the J621 greeting to $user_config (removes any distro greeting override)"
end
# --- Configuration ------------------------------------------------------------
mkdir -p $config_dir
set -l write_config 0
if not test -f $config_file
set write_config 1
else if set -q _flag_force; or set -q _flag_url
set write_config 1
else
echo "Keeping existing $config_file (use --force to rewrite it)"
# Read the current origin back so the probe below uses it.
set -l current (grep -E '^set -g J621_BASE ' $config_file 2>/dev/null | head -1 | string replace -r '^set -g J621_BASE +' '')
test -n "$current"; and set url $current
end
if test $write_config -eq 1
if not set -q _flag_no_prompt
if not set -q _flag_url
read -P "API origin [$default_url]: " answer
test -n "$answer"; and set url $answer
end
if not set -q _flag_token
read -P "API token (optional, Enter to run as a guest): " answer
test -n "$answer"; and set token $answer
end
end
set url (string trim --right --chars=/ "$url")
printf '# Written by install.fish on %s\n' (date '+%Y-%m-%d') > $config_file
printf 'set -g J621_BASE %s\n' "$url" >> $config_file
if test -n "$token"
printf 'set -g J621_TOKEN %s\n' "$token" >> $config_file
else
printf '# set -g J621_TOKEN paste-a-token-here\n' >> $config_file
end
printf '# set -g J621_WEB %s\n' "$url" >> $config_file
printf '# set -g J621_FASTFETCH_CONFIG jake\n' >> $config_file
chmod 600 $config_file
echo "Wrote $config_file"
end
# --- Dependencies -------------------------------------------------------------
for tool in fastfetch file
if not command -v $tool >/dev/null 2>&1
set problems 1
echo "Missing required tool: $tool"
end
end
if not command -v curl >/dev/null 2>&1; and not command -v wget >/dev/null 2>&1
set problems 1
echo "Missing required tool: curl (or wget)"
end
for tool in gifsicle jq python3
if not command -v $tool >/dev/null 2>&1
echo "Optional tool not found: $tool (the greeting still works)"
end
end
# --- Probe the configured backend --------------------------------------------
echo
echo "Checking $url ..."
set -l health ""
if command -v curl >/dev/null 2>&1
set health (curl -s -m 10 "$url/health" | string collect)
else
set health (wget -qO- -T 10 "$url/health" | string collect)
end
if string match -q '*"status":"ok"*' "$health"
echo " backend: ok"
set -l random_args -s -m 10
if test -n "$token"
set -a random_args -H "Authorization: Token $token"
end
set -l probe ""
if command -v curl >/dev/null 2>&1
set probe (curl $random_args "$url/api/random/?fastfetch=1" | string collect)
else
set probe (wget -qO- -T 10 "$url/api/random/?fastfetch=1" | string collect)
end
set -l detail (printf '%s' "$probe" | grep -o '"detail"[[:space:]]*:[[:space:]]*"[^"]*"' | head -1 | sed -E 's/.*:[[:space:]]*"//; s/"$//')
if test -n "$detail"
echo " random: $detail"
if test -z "$token"
echo " (a token would also let you see items that are hidden from guests)"
end
else
echo " random: ok"
end
else
set problems 1
echo " backend did not answer at $url/health"
echo " got: $health"
echo " Fix J621_BASE in $config_file (or pass --url) and run again."
end
# --- Wrap up ------------------------------------------------------------------
echo
if test $problems -eq 0
echo "Done. Test it now with: fish_greeting"
else
echo "Finished with problems above; the greeting will report them too."
end
echo "Rating mode lives in ~/.config/fish/greeting_mode (0=NSFW, 1=SFW, 2=Questionable)."
echo "The old gm-switch tool / .desktop launchers keep working — same file."
exit $problems
+20
View File
@@ -23,9 +23,11 @@ import Md5Redirect from "@/features/library/Md5Redirect";
import OnlinePage from "@/features/online/OnlinePage";
import PoolDetailPage from "@/features/pools/PoolDetailPage";
import PoolsPage from "@/features/pools/PoolsPage";
import RandomPage from "@/features/random/RandomPage";
import SimilarPage from "@/features/similar/SimilarPage";
import { SetupPage } from "@/features/setup/SetupPage";
import StatsPage from "@/features/stats/StatsPage";
import TokensPage from "@/features/tokens/TokensPage";
import UploadPage from "@/features/upload/UploadPage";
import UsersPage from "@/features/users/UsersPage";
import { isAgeVerified, markAgeVerified } from "@/lib/age";
@@ -99,6 +101,14 @@ export default function App() {
</RequireBackend>
}
/>
<Route
path="/random"
element={
<RequireBackend>
<RandomPage />
</RequireBackend>
}
/>
<Route path="/online" element={<OnlinePage />} />
<Route path="/online/view/:postId" element={<PostRedirect />} />
<Route path="/pools" element={<PoolsPage />} />
@@ -174,6 +184,16 @@ export default function App() {
}
/>
<Route path="/account" element={<AccountPage />} />
<Route
path="/tokens"
element={
<RequireBackend>
<RequireAuth>
<TokensPage />
</RequireAuth>
</RequireBackend>
}
/>
<Route path="*" element={<Navigate to="/" replace />} />
</Route>
<Route
+4 -1
View File
@@ -15,6 +15,7 @@ import { ConfirmDialog } from "@/components/ConfirmDialog";
import { StatusFooter } from "@/components/StatusFooter";
import { StatusPill } from "@/components/StatusPill";
import { Toasts } from "@/components/Toasts";
import { UploadIndicator } from "@/components/UploadIndicator";
import { api } from "@/lib/api";
import { hasBackend } from "@/lib/backend";
import { cn } from "@/lib/cn";
@@ -82,6 +83,7 @@ export function AppShell() {
const navItems = [
...(backend ? [{ to: "/", label: "Library" }] : []),
...(backend ? [{ to: "/random", label: "Random" }] : []),
{ to: "/online", label: "Online" },
{ to: "/pools", label: "Pools" },
...(backend && user ? [{ to: "/followed", label: "Followed" }] : []),
@@ -111,7 +113,7 @@ export function AppShell() {
return (
<div className="flex min-h-screen flex-col">
<header className="sticky top-0 z-40 border-b border-ctp-surface0 bg-ctp-crust/95 backdrop-blur">
<div className="mx-auto flex h-14 w-full max-w-[1600px] items-center gap-4 px-4">
<div className="flex h-14 w-full items-center gap-4 px-3 sm:px-4">
<Link to="/" className="flex shrink-0 items-center">
<span className="rounded bg-ctp-mauve px-2 py-1 font-mono text-xs font-bold tracking-wide text-ctp-crust">
J621
@@ -152,6 +154,7 @@ export function AppShell() {
<div className="ml-auto flex items-center gap-3">
{backend ? (
<>
<UploadIndicator />
<StatusPill status={status} />
{user ? (
<>
@@ -2,11 +2,13 @@ import { useQuery } from "@tanstack/react-query";
import {
Cable,
Copy,
Dices,
ExternalLink,
FolderOpen,
Globe,
History,
Images,
KeyRound,
Layers,
LogIn,
LogOut,
@@ -124,6 +126,12 @@ function CommandPaletteDialog({ onClose }: { onClose: () => void }) {
icon: FolderOpen,
run: () => navigate("/"),
},
{
id: "random",
label: "Roll a random image",
icon: Dices,
run: () => navigate("/random"),
},
]
: []),
{
@@ -213,6 +221,12 @@ function CommandPaletteDialog({ onClose }: { onClose: () => void }) {
icon: Settings,
run: () => navigate("/account"),
});
list.push({
id: "tokens",
label: "API tokens",
icon: KeyRound,
run: () => navigate("/tokens"),
});
if (user.is_staff || user.is_superuser || user.role === "staff") {
list.push({
id: "users",
+4 -2
View File
@@ -1,5 +1,6 @@
import { Link } from "react-router-dom";
import { apiUrl } from "@/lib/api";
import { cn } from "@/lib/cn";
import { formatBytes } from "@/lib/format";
import type { MediaItem } from "@/lib/types";
@@ -17,8 +18,9 @@ const ratingLabels: Record<string, string> = {
};
export function MediaCard({ item }: { item: MediaItem }) {
const preview =
item.kind === "video" ? item.thumbnail_url : item.raw_url;
// The thumbnail endpoint now builds real 480px previews for images too, so
// the grid no longer pulls full-size originals.
const preview = apiUrl(item.thumbnail_url);
const rating = item.display_rating;
return (
@@ -0,0 +1,56 @@
import { Link } from "react-router-dom";
import { cn } from "@/lib/cn";
import { useUploadStatus } from "@/lib/uploadStatus";
const PHASE_LABELS: Record<string, string> = {
md5: "e621 MD5",
visual: "Visual similarity",
iqdb: "IQDB",
};
/**
* Shell-wide progress for the server-side upload pipeline.
*
* Staged uploads keep processing after the Upload page is closed, so this
* pill is the "leave and keep an eye on it" view: it lives in the header on
* every page and links back to the board.
*/
export function UploadIndicator() {
const { data: status } = useUploadStatus();
if (!status || !status.active) return null;
const total = Math.max(status.total, status.processed + status.failed);
const done = status.processed + status.failed;
const percent = total > 0 ? Math.min(100, Math.round((done / total) * 100)) : 0;
const broken = status.status === "paused" || status.status === "error";
const label =
status.status === "paused"
? "uploads paused"
: status.status === "error"
? "uploads failed"
: (PHASE_LABELS[status.phase] ?? "processing uploads");
const count = total > 0 ? `${done}/${total}` : `${status.outstanding} left`;
return (
<Link
to="/upload"
title={status.error || "Staged uploads are being processed in the background"}
className={cn(
"flex items-center gap-2 rounded-md border px-2 py-1 font-mono text-[10px] transition",
broken
? "border-ctp-red/40 text-ctp-red hover:bg-ctp-red/10"
: "border-ctp-surface1 text-ctp-subtext0 hover:bg-ctp-surface0 hover:text-ctp-text",
)}
>
<span className="hidden sm:inline">{label}</span>
<span>{count}</span>
<span className="h-1 w-12 overflow-hidden rounded-full bg-ctp-surface0">
<span
className={cn("block h-full", broken ? "bg-ctp-red" : "bg-ctp-teal")}
style={{ width: `${percent}%` }}
/>
</span>
</Link>
);
}
@@ -17,6 +17,7 @@ import { toast } from "@/store/toasts";
import { AvatarCard } from "./AvatarCard";
import { PreferencesCard } from "./PreferencesCard";
import { TokensCard } from "./TokensCard";
const BASE_URL_OPTIONS = [
{ value: "https://e621.net", label: "e621.net — main site" },
@@ -247,6 +248,7 @@ export default function AccountPage() {
</header>
<AvatarCard />
<PreferencesCard />
<TokensCard />
{loading && !credentials ? (
<div className="flex justify-center py-12">
<Spinner className="h-6 w-6" />
@@ -9,6 +9,8 @@ import type { UserPreferences } from "@/lib/types";
import { useAuth } from "@/store/auth";
import { toast } from "@/store/toasts";
const E621_PER_PAGE_OPTIONS = [48, 100, 200, 320];
const LANDING_OPTIONS: { value: NonNullable<UserPreferences["landing_page"]>; label: string }[] = [
{ value: "library", label: "Library" },
{ value: "online", label: "Online" },
@@ -26,6 +28,7 @@ export function PreferencesCard() {
per_page: user?.preferences?.per_page ?? 48,
zoom: user?.preferences?.zoom ?? 190,
online_hot_default: user?.preferences?.online_hot_default ?? true,
e621_per_page: user?.preferences?.e621_per_page ?? 48,
});
const [saving, setSaving] = useState(false);
@@ -149,8 +152,32 @@ export function PreferencesCard() {
))}
</select>
</Field>
<Field label="e621 posts per page">
<select
className={inputClass}
value={values.e621_per_page ?? 48}
onChange={(event) =>
setValues((current) => ({
...current,
e621_per_page: Number(event.target.value),
}))
}
>
{E621_PER_PAGE_OPTIONS.map((option) => (
<option key={option} value={option}>
{option}
</option>
))}
</select>
</Field>
</div>
<p className="text-[11px] leading-relaxed text-ctp-overlay0">
e621 posts per page drives the Online browser and how many pool posts
load per request; e621 caps a request at 320.
</p>
<label className="flex flex-col gap-1.5">
<span className="text-xs font-medium uppercase tracking-wide text-ctp-overlay1">
Thumbnail size
@@ -0,0 +1,23 @@
import { Link } from "react-router-dom";
import { linkButtonClass } from "@/components/ui";
export function TokensCard() {
return (
<section className="rounded-lg border border-ctp-surface0 bg-ctp-base p-5">
<h2 className="text-sm font-semibold text-ctp-subtext1">
Shell tokens
</h2>
<p className="mt-1 text-xs leading-relaxed text-ctp-overlay0">
Long-lived tokens that only work with the random-image endpoint — for
shell greetings and small scripts. They cannot read the library,
upload or change your account.
</p>
<div className="mt-3">
<Link to="/tokens" className={linkButtonClass}>
Manage API tokens
</Link>
</div>
</section>
);
}
+5 -2
View File
@@ -14,11 +14,14 @@ export default function DetailPage() {
if (J_ID_RE.test(itemId)) {
if (!hasBackend()) return <BackendNeeded />;
const normalized = `J-${itemId.slice(2)}`;
return <LibraryDetail jId={normalized} />;
// Key per item: remount so local state (delete confirm, optimizer, task
// view) cannot leak from the previously viewed item.
return <LibraryDetail key={normalized} jId={normalized} />;
}
if (/^\d+$/.test(itemId)) {
return <OnlineDetail postId={Number(itemId)} />;
// Key per post: the download task view must not survive a post change.
return <OnlineDetail key={itemId} postId={Number(itemId)} />;
}
return (
+56 -39
View File
@@ -1,5 +1,5 @@
import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
import { Eye, StarOff } from "lucide-react";
import { ChevronDown, ChevronRight, Eye, StarOff } from "lucide-react";
import { useState } from "react";
import { Link } from "react-router-dom";
@@ -13,7 +13,6 @@ import type {
FollowedPool,
FollowedTag,
FollowFeed,
Paginated,
} from "@/lib/types";
const ratingPill: Record<string, string> = {
@@ -149,48 +148,65 @@ function BlacklistCloudPanel() {
q.state.data?.status === "building" ? 2000 : false,
});
const cloud = query.data;
const [expanded, setExpanded] = useState(false);
return (
<section className="rounded-lg border border-ctp-surface0 bg-ctp-base p-4">
<div className="flex flex-wrap items-center justify-between gap-2">
<button
type="button"
onClick={() => setExpanded((value) => !value)}
title={expanded ? "Hide the tags" : "Show the tags"}
className="flex w-full items-center justify-between gap-2 text-left"
>
<h2 className="text-sm font-semibold text-ctp-subtext1">
Blacklisted tags
</h2>
<p className="text-xs text-ctp-overlay0">
{cloud?.source === "user"
? "From your e621 blacklist"
: "From e621's anonymous default blacklist"}
{cloud ? ` · ${cloud.blacklist_count} entries` : ""}
{cloud ? ` · ${cloud.posts} feed post(s) scanned` : ""}
{cloud?.computed_at ? ` · computed ${formatDate(cloud.computed_at)}` : ""}
<span className="flex items-center gap-1.5 font-mono text-[11px] text-ctp-overlay0">
{cloud ? `${cloud.blacklist_count} entries` : "…"}
{cloud?.status === "building" ? (
<span className="ml-2 inline-flex items-center gap-1.5">
<Spinner className="h-3 w-3" /> building…
</span>
<Spinner className="h-3 w-3" />
) : null}
</p>
</div>
{expanded ? (
<ChevronDown className="h-3.5 w-3.5" />
) : (
<ChevronRight className="h-3.5 w-3.5" />
)}
</span>
</button>
{query.isPending ? (
<div className="mt-3 flex justify-center py-4">
<Spinner className="h-4 w-4" />
</div>
) : cloud && cloud.tags.length > 0 ? (
<div className="mt-3 flex flex-wrap gap-1.5">
{cloud.tags.map(([tag, count]) => (
<span
key={tag}
className="rounded border border-ctp-red/30 bg-ctp-red/10 px-1.5 py-0.5 font-mono text-[11px] text-ctp-red"
>
{tag} ({count})
</span>
))}
</div>
) : (
<p className="mt-3 text-xs text-ctp-overlay0">
No blacklisted tags in your feeds.
</p>
)}
{expanded ? (
<>
<p className="mt-2 text-xs text-ctp-overlay0">
{cloud?.source === "user"
? "From your e621 blacklist"
: "From e621's anonymous default blacklist"}
{cloud ? ` · ${cloud.posts} feed post(s) scanned` : ""}
{cloud?.computed_at
? ` · computed ${formatDate(cloud.computed_at)}`
: ""}
</p>
{query.isPending ? (
<div className="mt-3 flex justify-center py-4">
<Spinner className="h-4 w-4" />
</div>
) : cloud && cloud.tags.length > 0 ? (
<div className="mt-3 flex flex-wrap gap-1.5">
{cloud.tags.map(([tag, count]) => (
<span
key={tag}
className="rounded border border-ctp-red/30 bg-ctp-red/10 px-1.5 py-0.5 font-mono text-[11px] text-ctp-red"
>
{tag} ({count})
</span>
))}
</div>
) : (
<p className="mt-3 text-xs text-ctp-overlay0">
No blacklisted tags in your feeds.
</p>
)}
</>
) : null}
</section>
);
}
@@ -207,11 +223,12 @@ export default function FollowedPage() {
const tagsQuery = useQuery({
queryKey: ["follows-tags"],
queryFn: () => api<Paginated<FollowedTag>>("/api/follows/tags/"),
// Unpaginated: this is the complete follow list, shown as cards below.
queryFn: () => api<FollowedTag[]>("/api/follows/tags/"),
});
const poolsQuery = useQuery({
queryKey: ["follows-pools"],
queryFn: () => api<Paginated<FollowedPool>>("/api/follows/pools/"),
queryFn: () => api<FollowedPool[]>("/api/follows/pools/"),
});
const feedQuery = useQuery({
@@ -269,8 +286,8 @@ export default function FollowedPage() {
);
}
const tags = tagsQuery.data?.results ?? [];
const pools = poolsQuery.data?.results ?? [];
const tags = tagsQuery.data ?? [];
const pools = poolsQuery.data ?? [];
const feed = feedQuery.data;
return (
@@ -5,7 +5,7 @@ import { useState } from "react";
import { Button, Spinner } from "@/components/ui";
import { api, errorMessage } from "@/lib/api";
import { hasBackend } from "@/lib/backend";
import type { FollowedPool, Paginated } from "@/lib/types";
import type { FollowedPool } from "@/lib/types";
import { useAuth } from "@/store/auth";
import { toast } from "@/store/toasts";
@@ -23,12 +23,13 @@ export function PoolFollowButton({
const query = useQuery({
queryKey: ["follows-pools"],
queryFn: () => api<Paginated<FollowedPool>>("/api/follows/pools/"),
// Unpaginated on purpose: the button must see every follow, not page one.
queryFn: () => api<FollowedPool[]>("/api/follows/pools/"),
enabled: Boolean(user),
staleTime: 30_000,
});
const follow = query.data?.results.find((entry) => entry.pool_id === poolId);
const follow = query.data?.find((entry) => entry.pool_id === poolId);
const followed = Boolean(follow);
const mutation = useMutation({
@@ -4,7 +4,7 @@ import { useState, type MouseEvent } from "react";
import { api, errorMessage } from "@/lib/api";
import { cn } from "@/lib/cn";
import type { FollowedTag, Paginated } from "@/lib/types";
import type { FollowedTag } from "@/lib/types";
import { useAuth } from "@/store/auth";
import { toast } from "@/store/toasts";
@@ -16,13 +16,14 @@ export function TagFollowToggle({ tag }: { tag: string }) {
const query = useQuery({
queryKey: ["follows-tags"],
queryFn: () => api<Paginated<FollowedTag>>("/api/follows/tags/"),
// Unpaginated on purpose: the toggle must see every follow, not page one.
queryFn: () => api<FollowedTag[]>("/api/follows/tags/"),
enabled: Boolean(user),
staleTime: 30_000,
});
const normalized = tag.trim().toLowerCase();
const follow = query.data?.results.find((entry) => entry.tag === normalized);
const follow = query.data?.find((entry) => entry.tag === normalized);
const followed = Boolean(follow);
const mutation = useMutation({
+2 -2
View File
@@ -3,7 +3,7 @@ import { useState } from "react";
import { Link } from "react-router-dom";
import { Button, EmptyState, Spinner, inputClass } from "@/components/ui";
import { api, errorMessage } from "@/lib/api";
import { api, apiUrl, errorMessage } from "@/lib/api";
import { cn } from "@/lib/cn";
import { formatBytes } from "@/lib/format";
import type { MediaItem, Paginated, StorageInfo } from "@/lib/types";
@@ -245,7 +245,7 @@ export default function DeletePage() {
>
<div className="relative aspect-square overflow-hidden bg-ctp-mantle">
<img
src={item.thumbnail_url}
src={apiUrl(item.thumbnail_url)}
alt={item.filename}
loading="lazy"
className="h-full w-full object-cover"

Some files were not shown because too many files have changed in this diff Show More