2050 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Paul Nothaft	bbce3cd2a2	feat(faces): let the photographer choose which photo represents a person (#1119) ... Phase 1 of #1096. Clustering picks the cover, and its idea of a good one and a human's do not always agree. A cluster whose avatar is turned away or softer than the rest stays that way in the guest-facing people strip too, and nothing in the UI could change it. A picker reachable from each person row, reusing the face list the split dialog already loads — same query, same grid, different action on a click. Making the choice actually stick took four changes --------------------------------------------------------------------------- event_people.cover_face_id has existed since migration 177 and the PATCH already accepted it, so the first version of this was frontend-only. It was also a no-op: - facePeopleService.listPeople SELECTED cover_face_id and then discarded it, recomputing the cover as the best-scoring VISIBLE face on every read. The picker saved, said so, and the avatar reverted immediately. It now prefers the stored pick whenever this audience can see it, and falls back to the score-ordered choice otherwise — so visibility scoping still wins, and a guest is never handed a crop of a photo they cannot open. - recomputeCentroid overwrote cover_face_id unconditionally. It runs on rescan and on photo replacement, so any reprocessing silently undid a deliberate choice. It now keeps the chosen face while it is still a member of the cluster. - The face list is cached per person, and split/merge move faces between people. Until now the only reader closed itself after acting, so nobody saw the stale copy; the picker is a second reader of the same key. - cover_face_id meant two things. assignFaces seeded it with whichever face opened the cluster and recomputeCentroid overwrote it with the highest scoring one, so an automatic guess was indistinguishable from a deliberate choice — and honouring it would have pinned every UNCURATED person to that guess, which is worse than the fallback it replaced (the fallback is computed per audience and skips photos a guest cannot open). Both writers are gone, migration 179 clears the stored guesses, and the column now means one thing. That also removes the need to defend the choice against rescans: nothing overwrites it, and a dangling id self-heals to the derived cover. Clearing existing values is safe rather than destructive: no install has ever been able to SET a cover, so every stored value is an automatic guess by construction. Also fixes a PostgreSQL-only 500 --------------------------------------------------------------------------- GET /admin/events/:id/people/:personId/faces joined `photos` but did not table-qualify its WHERE, and photo_faces and photos BOTH have an event_id: column reference "event_id" is ambiguous Postgres refuses it, so the endpoint 500s and the Split dialog — its only consumer until now — has been broken on every PostgreSQL install since the join was added. SQLite resolves the ambiguity silently, which is why the suite stayed green. Reproduced against a real Postgres before and after. The query is now a named builder the route calls and the test imports, rather than a copy: an earlier version of that test re-declared the query, so the route could regress to the bare form while the assertions kept passing. Merge and recluster preserve the choice as well. Both already carried labels and privacy flags across; the chosen cover is human state of the same kind, so it now rides along — through a merge when the target has none, and through a recluster by following its FACE into whichever cluster ends up holding it, rather than the majority-descendant rule the label uses. The picker and the endpoint disagree past 500 faces, so the picker now says when it is showing a capped list rather than presenting it as exhaustive. Frontend suite 178 passing, backend 23 across the touched suites, build clean, no new type errors. Mutation-checked twice: dropping the cover preference fails the new listPeople test while the visibility-scoping test still passes, and restoring the auto-seed in assignFaces fails it too.	2026-08-21 19:26:52 +02:00
Paul Nothaft	887bdbe6e5	feat(gallery): responsive grid thumbnails (#1095) (#1109) ... * feat(gallery): responsive grid thumbnails (#1095) The half of #1095 that #1099 deliberately left out. Grid tiles are ~175 CSS px at the mobile 2-column default — about 530 device px on a DPR-3 phone — so the 300px thumbnail is upscaled ~1.8x and faces visibly mush. Backend mirrors the preview tiers exactly: ?w= on the gallery thumbnail route, whitelisted to 300/600/900, cached by width in storage, never written to photos.thumbnail_path, and keyed by photo id for every source type — basenames are not unique across events and a tier is served from a cache hit without re-reading the source, which is how the preview tiers nearly leaked one gallery's photo into another. The tier is in the ETag, or a client holding the 300px file gets a 304 for its 600px request. Cleanup and regenerate invalidation are wired the same way. generateThumbnail now takes width/height overrides; it keeps the configured `fit`, because the grid renders with object-cover and tiers that were framed differently would visibly jump as the viewport changes. The srcset only advertises tiers the SOURCE can fill. Thumbnails are generated withoutEnlargement, so a 400px original asked for 900 comes back at 400 — advertising "900w" would have the browser pick that candidate and upscale it, which is the reported softness made worse. That exact trap is why this was held back from #1099; the photo's own dimensions are now the guard, measured on the SHORT edge because thumbnails are square and a 4000x600 panorama can still only fill a 600 tile. A source that clears only one tier gets no srcset at all rather than a single pointless candidate. Two things this surfaced, both worth knowing separately: `npx tsc --noEmit` type-checks NOTHING in this project — the root tsconfig is `files: []` with project references, so the real command is `tsc -b`, which is what build:check runs. Under tsc -b the repo has 43 files with pre-existing type errors; this branch adds none, and the one error in a file I touched (PeopleManagerModal:91) is on main already and unrelated to the line I changed. * fix(gallery): wire grid tiers into the component that actually renders The srcSet landed in PhotoGrid.tsx, which nothing imports — GalleryView renders PhotoGridWithLayouts, and every grid layout funnels its tile through the shared PhotoCard. The frontend half of #1095 shipped nothing. Moved to PhotoCard, and switched from srcSet to a single sized URL, the same shape PhotoLightbox already uses for preview tiers. AuthenticatedImage fetches its src with the gallery bearer token and renders the blob; an <img> carrying a w-descriptor srcSet ignores src entirely, so that fetch would have been discarded and the browser would have issued its own — unauthenticated, and resolved against the page origin rather than the configured API host. One URL keeps the auth path and halves the requests. The tier comes from the tile's measured width via the IntersectionObserver entry, read on the same render that reveals the image so nothing is fetched twice. Column counts differ per layout and shift again with thumbnailScale, so the breakpoint table is only a fallback. Also closes what the tier cache leaked or served stale: - ensureThumbnailAtWidth short-circuits videos. Their thumbnail is a poster frame, so the tier path handed the video file to Sharp — after downloading it in full on S3, uncached, once per request. - The ETag names the tier actually served, not the one requested. A fallback to the canonical thumbnail was caching a 300px image under a 900px key. - Tier height scales from the configured aspect ratio instead of forcing a square; with fit:'cover' a 300x200 canonical and a 600x600 tier are two different crops and the photo reframed between tiers. - The canonical short-circuit compares against the configured thumbnail_width, not the 300 default, so a 600px install stops generating duplicate tiers. - Tier invalidation on /admin/thumbnails/regenerate, above the local-file check that skips S3 and external rows. - Tier cleanup in replacePhoto and deleteEventCascade. Both derive keys from the photo row, so the rows have to be read before they change or vanish. Preview tiers had the same two holes and are swept alongside. The clamp no longer drops a tier when the source falls between them: a 400px short edge asked for 600 returns all 400 pixels, where clamping to 300 threw 100 of them away. Backend 18 tier tests, frontend 22. Full suites green: 293 backend across the touched areas, 185 frontend, build clean, no new type errors. * fix(gallery): measure the tile, and stop regenerating the w300 tier Follow-up to the review of #1095. Closes the three items left open there, plus a defect the previous commit introduced. **The w300 tier regenerated on every request.** Decoupling the canonical short-circuit from the hardcoded 300 left generateThumbnail still tagging against DEFAULT_THUMBNAIL_WIDTH. On an install with thumbnail_width=600 a w=300 request wrote `thumb_<name>` while the caller probed for `thumb_w300_<name>`: the cache never hit, so every request re-downloaded the original and ran Sharp, and the file it left behind was in no cleanup list. The tag now follows the configured width, and thumbnailTierKeys lists all three widths — which one is canonical is a setting, so excluding 300 stranded exactly the file a 600-configured install generates. **The tier is chosen from the tile's measured width.** The observer entry only exists for `lazy` cards, and Mosaic, Masonry and Timeline don't pass it — Mosaic is 1-up on mobile where Grid is 2-up, so they are the layouts a breakpoint guess gets most wrong. Measured in a layout effect and gated: the image is not rendered until the width is known, so AuthenticatedImage never mounts with a src it has to replace. Attaching the observer ref unconditionally instead refetches every tile, since React flushes passive effects before the sync re-render a layout effect triggers — removing the gate makes the new single-request test fail, which is how that was confirmed rather than assumed. **Gallery Premium has its own card** and never reached the shared one, so its tiles kept pulling the canonical thumbnail. MasonryPhotoAlbum already hands the laid-out width to the render prop, so it needed no measurement. **Event rename orphaned tiers.** The key embeds the basename, so the DB update is the point past which the old keys cannot be derived. Dropped inside the filename-changed branch, not the loop body: unconditional would fire four storage deletes per photo on every rename, 20k calls against S3 for a 5,000-photo event that merely had its slug adjusted. Preview tiers had the same hole and are swept alongside. Carousel is the seventh layout and deliberately gets no tiering: its filmstrip thumbs are 80 CSS px, under the canonical 300 even at DPR 3. Tests: first PhotoCard suite (6), backend tier suite 21. Both new behaviours mutation-checked — reverting the width tag, the render gate, the measurement, or the rename sweep each fails a test. Full suites green: 298 backend across the touched areas, 191 frontend, build clean, no new type or lint findings. * fix(gallery): mount masonry cards once, into a measured layout Found while capturing screenshots for this PR, by attributing every thumbnail request to a photo id rather than eyeballing the grid. Masonry columns mode starts at 3 columns and runs its greedy distribution off a hardcoded 300px estimate until the container has been measured. Cards mounted into that guess are torn down when it settles — photos move to a different parent column, so React unmounts them — and since #1095 each mount picks its tier from its own width, the two mounts request two DIFFERENT urls. Measured on a 1440px desktop, production build, 62 photos: before 45 photos fetched at canonical AND w600, 17 stuck on w600 107 requests after 62 photos, canonical only, 62 requests Mobile was already landing on one tier either way, so both mounts produced the same url and the second was a cache hit — which is why it looked clean and the desktop case did not. The fix is the gate the rows/justified mode in this same file already applies for the same reason (line 346): hold the cards back until containerWidth is known. Only columns mode was missing it. Grid and Justified take their column counts from CSS breakpoints, so they have no transient measured value to discard and are unaffected. Worth noting this was NOT visible on main: without tiering both mounts request the same url, so the browser cache absorbs the duplicate. Tiering is what turns a harmless remount into a second download — the regression is this PR's, which is why it is fixed here rather than deferred. Frontend suite 194 passed (3 new). Mutation-checked: removing the gate fails the mount-once and placeholder tests. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-21 19:26:21 +02:00
Paul Nothaft	00776234fd	fix(security): let cors() own Access-Control-Allow-Origin on protected images (#1118) ... Closes #1116. secureImageMiddleware set its own Access-Control-Allow-Origin, overwriting the one cors(corsOptions) had already computed. server.js:247 mounts cors() on all of /api with credentials:true, so by the time the route handler ran the correct header was already there — and the local assignment replaced it with a worse answer in BOTH directions: unresolved -> '*'. Combined with the credentials:true that cors() sets, that is an invalid pair browsers reject outright. Unreachable on Docker until #1104 stopped compose injecting FRONTEND_URL; reachable on a fresh install from then until the wizard stores general_site_url. resolved -> the frontend origin, even when the request legitimately came from the allowlisted ADMIN_URL. A split admin host got a header naming the wrong origin and the browser rejected a request cors() had allowed. Deleting the line fixes both. cors() already validates the request Origin against the allowlist, sets Vary: Origin, omits the header entirely for a disallowed or absent Origin, and pairs correctly with credentials. Methods, Headers and Max-Age stay here: they are route-specific and cors() does not contradict them. Observed against a running instance before and after: allowlisted Origin ACAO: <that origin> + Vary: Origin + credentials:true disallowed Origin no ACAO no Origin header no ACAO Six tests, mounted on a real Express app with server.js's middleware order. Deliberately NOT a unit test against a response double: the first version of this fix was a guarded assignment that looked correct in isolation and still overwrote cors() whenever an origin resolved. A double cannot see middleware composition, which is exactly how that slipped through. Mutation-checked both ways — restoring the original `|| '*'` fails 5 of 6, and restoring the guarded assignment fails 3 of 6 including the admin-origin case.	2026-08-21 19:25:56 +02:00
Paul Nothaft	d241919604	fix(ui): stop iOS Safari zooming in on 14px form fields (#1113) ... Closes #1105. iOS Safari zooms the whole page in when a focused form control computes to under 16px, and it does not zoom back out. Unlocking a gallery is a client-side transition rather than a document navigation, so the zoom the password field triggers carries straight into the gallery: the layout pans horizontally and the header actions sit off-screen until the visitor pinch-zooms out by hand. A real page load would have reset it. `.input` and `.input-themed` are `text-sm`, so the field is 14px on every phone, and GalleryPage inverts the breakpoint on top of that (`text-sm sm:text-base` — 14px below 640px, where iOS zooms, and 16px above it, where it never does). Keyed to the POINTER, not a width. The zoom depends on the computed font size and a touch device, never on how wide the viewport is — and a phone in landscape is 667-956 CSS px, above any width you could call "phone". Measured on the admin login page, which has no `sm:` override: main portrait 390x844 14px zooms main landscape 844x390 14px zooms main iPad 820x1180 14px zooms fixed all three 16px fixed desktop (mouse) 14px unchanged, no zoom off touch One media query rather than flipping each call site. `.input` alone backs 334 `<Input>` usages, but there are also ~440 raw inputs, selects and textareas carrying their own `text-sm`, and Tailwind utilities sit in a later layer than @layer components — so a fix at the component definition misses most controls and any new `text-sm` silently reintroduces the bug. The `:not()` on each selector is load-bearing, not decoration: it buys the specificity to beat a utility class. Measured in a browser — input.text-sm 16px (0,2,1 beats .text-sm) select.text-sm 14px (0,0,1 loses) textarea.text-sm 14px (0,0,1 loses) 24 selects and textareas in the tree carry `text-sm`, so the bare form would have left them zooming. Checkbox and radio stay excluded so font-size never sizes their box. max(16px, 1em, 1rem) is a FLOOR, not a size. A flat 16px would make controls that are already bigger smaller: Typography -> Large sets --font-size-base to 18px on body, so anything inheriting it would be clamped down and the setting quietly ignored. Each term covers a case the others miss: normal (body 16) 16px Large theme (body 18) 18px Small theme (body 14) 16px browser default 20px 20px The viewport meta is deliberately left alone: `maximum-scale=1` would suppress the zoom by disabling pinch-to-zoom for everyone.	2026-08-21 19:25:47 +02:00
Paul Nothaft	69edee36fc	ci: give the backend job headroom over its observed tail (#1108) ... The backend job normally finishes in about 3 minutes — the last eight runs on main were 2.6 to 3.4 — but it is the only one that boots Postgres and runs the full integration suite, so it is the only one exposed to runner contention. The observed spread has reached 9.2 minutes against a 10-minute cap, and release PR #1088 was cancelled at 10.3 with every test in the log passing and jest still running. That failure mode is expensive out of proportion to how often it happens: a cancelled job is a red X on a branch that is actually green, so it costs a diagnosis and a re-run each time, and it lands on release PRs because those are the ones that run when everything else does. The cap is a runaway guard rather than a performance budget, so 20 buys real headroom over the worst run seen while still killing a genuinely hung suite well inside the hour GitHub would otherwise allow. frontend and ml keep 10: they finish in seconds and have never been close. Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-21 19:25:23 +02:00
Paul Nothaft	ba221ba2a0	chore(main): release 3.109.0-beta.0 (#1115) v3.109.0-beta.0	2026-08-21 10:58:27 +00:00
Luca	9431b9f094	feat(setup): configure the public address and SMTP in the wizard, not .env (#1104) ... * feat(setup): configure the public address and SMTP in the wizard, not .env A fresh install could not configure its own public address. `general_site_url` and the `email_configs` row already existed as admin settings, but nothing could reach them: - docker-compose.yml injected FRONTEND_URL=${FRONTEND_URL:-http://localhost:3000} and Dockerfile.aio baked in ENV FRONTEND_URL=http://localhost:3000, so getFrontendBaseUrl() returned on its first branch every time and the setting was never read. .env.example shipped the same value as an uncommented placeholder for FRONTEND_URL / ADMIN_URL / API_URL. - the wizard never asked for the address at all, and skipped its whole config step unless a CRM-ish feature was selected — so a gallery-only install was also never offered SMTP, despite gallery links, guest invites and expiry warnings all going out through email_configs. - eleven call sites read process.env.FRONTEND_URL directly rather than the resolver, three of them defaulting to placeholder hosts that reached real recipients: https://app.example.com in payment-reminder emails, localhost:3005 in admin invitation emails, https://app.example.com in dev template previews. Stop injecting a default anywhere, and resolve the origin instead: FRONTEND_URL -> general_site_url -> the origin the request arrived on -> whichever exists -> ''. A loopback candidate is treated as unconfigured so the installs that already have http://localhost:3000 baked into their environment self-heal; the same guard previously lived inline in routes/gallery.js for the slideshow QR (#848) and is now shared. The empty return is preserved because shareLinkService and the SSO redirects in routes/auth rely on it to emit relative urls — callers needing an absolute url use getAbsoluteFrontendUrl(), which still ends at http://localhost:3000. The wizard now persists window.location.origin right after the admin account is created, so an install that skips the rest still has a usable origin for background jobs that have no request to derive one from, and offers it as an editable "Public address" field. Settings -> General shows the field read-only when FRONTEND_URL pins it, instead of silently ignoring edits. Also drop the `|| 'mailhog'` fallback when seeding email_configs: that host only exists in the dev compose profile (which does not even start by default), so a fresh install came up with a live config pointing nowhere while the wizard showed empty SMTP fields. With no row, blank fields are the truth and emailProcessor logs "No email configuration found". Developers set SMTP_HOST=mailhog explicitly. backend/src/services/emailService.js is deleted: nothing in backend/ references it, and it was the only consumer of the SMTP_* variables, which misrepresented how mail is configured. Refs #705 * fix(setup): keep FRONTEND_URL ahead of ADMIN_URL/APP_URL when resolving links The previous commit routed two call sites through the resolver but put the site-specific variable FIRST, silently reversing precedence: userManagementService was: FRONTEND_URL || ADMIN_URL || localhost:3005 became: ADMIN_URL || resolver adminEvents/crud was: FRONTEND_URL || APP_URL || '' became: APP_URL || resolver An install with both variables set would have flipped which one won. Call the resolver first instead — it starts with FRONTEND_URL, so the original relative order is preserved and only the final fallback changes: localhost:3005 (not even the frontend's port) and '' (a relative link inside an email) both become the resolved origin. Refs #705 * fix(setup): unpin loopback FRONTEND_URL, keep ADMIN_URL/APP_URL reachable Review feedback on #1104. isEnvPinned() reported ANY FRONTEND_URL as authoritative, including the loopback values getFrontendBaseUrl() deliberately demotes. An install upgrading with the old compose default FRONTEND_URL=http://localhost:3000 therefore resolved its origin from general_site_url correctly, but got the Site URL field rendered read-only in Settings and skipped by the wizard's seeding - locking the exact operators this change exists to unblock out of configuring a public address anywhere. The predicate now mirrors the resolver, and the derived general_site_url_effective the General tab reads comes from the same helper instead of re-normalising process.env inline. APP_URL and ADMIN_URL had become dead code: getFrontendBaseUrl() only returns falsy when NOTHING is configured, so `|| process.env.ADMIN_URL` after it never ran once a site URL existed - which after this PR is the normal case. A split-origin install pointing ADMIN_URL at a separate admin host got invite links on the public gallery origin instead. They are now passed as an explicit `override` that resolves directly below FRONTEND_URL, preserving the historic FRONTEND_URL-before-ADMIN_URL order while beating the database- and request-derived fallbacks. general_site_url now feeds the CORS allowlist and the Access-Control-Allow-Origin header, not just email links, so a schemeless value is an allowlist entry no browser origin can match. Validate it server-side in PUT /general (isURL with require_protocol, require_tld off so LAN/NAS installs on http://nas:3000 still work) and client-side in both surfaces that write it - type="url" never fires in either, since neither input sits inside a form. Two more wizard fixes: the General tab no longer reposts general_site_url while it is env-pinned, because the field then holds the effective env value rather than the stored one and the round-trip read as a change to a protected key, 403ing a settings.edit-without-settings.domains admin on an unrelated save. And SetupConfigStep validates the From address before posting - /admin/email/config rejects a blank one, which used to surface as a generic warning while the wizard advanced from its finally block anyway, discarding every SMTP value the user had typed, password included. A failed save now keeps them on the step. * fix(setup): surface a rejected public address instead of swallowing it Review round 2 follow-up on #1104, pushed onto the branch. saveSiteUrl() caught and discarded every error. That was defensible before round 2 added a server-side URL check, but PUT /general can now answer 400 — and the two validators disagreed: http://my_nas.local client: accepted server: rejected http://foo_bar:3000 client: accepted server: rejected validate() let those through, the 400 was swallowed, `failed` stayed false and onDone() ran. The operator finished the wizard believing the public address was stored when nothing had been. That is the silent misconfiguration this whole change exists to remove, landing on the LAN and NAS installs it targets. Three parts: - saveSiteUrl() throws. finish() resolves it before anything else is posted and puts the message on the address field rather than the generic "some settings could not be saved" warning. Skip for now still always leaves, by contract, but warns instead of dropping the value in silence. - allow_underscores on the server check, for the same reason require_tld is off: browsers resolve http://my_nas.local and the client accepts it, so rejecting it server-side only produced the mismatch above. Both validators now agree across the LAN/NAS, IDN, bare-IP and scheme-less cases. - LOOPBACK_BASE_RE anchors its host token. Bare prefix matching also demoted https://localhost-nas.example.com, and now that this predicate gates the whole resolver rather than just the slideshow QR, being demoted means a configured address is silently ignored. 127. stays a bare prefix on purpose: all of 127.0.0.0/8 is loopback. Resolver suite 31 passing, up from 26. Mutation-checked: restoring the unanchored regex fails the three new host-boundary cases. * fix(settings): don't lock the General tab on a site URL nobody typed Review follow-up on #1104, pushed onto the branch. general_site_url was free-text until this PR added a server-side check, so an upgraded install can hold something schemeless that predates it. The tab flagged that on load, and `disabled={!!siteUrlError}` then killed Save for EVERY General setting. An admin holding settings.edit but not settings.domains could not clear it either: correcting the address is a change to a protected key and 403s. The tab has no permission gating, so that role was simply locked out of the tab with no self-service way back. That is the same role adminSettings.js:85-95 documents the no-op round-trip allowance for. The allowance only helps if the request is made, and this blocked it in the browser first. Validation now waits until the field is actually edited, and an unchanged value is dropped from the payload rather than reposted — matching what the env-pinned case already does one line above, and for the same reason. stored value invalid, untouched Save works, key not sent edited to something unusable Save blocked edited to a usable absolute url saved Four tests, first coverage for this feature. Mutation-checked: removing the dirty gate fails the untouched-value case. --------- Co-authored-by: Paul Nothaft <53005142+the-luap@users.noreply.github.com>	2026-08-21 12:54:14 +02:00
Paul Nothaft	7583b4f6b0	chore(main): release 3.108.1-beta.0 (#1103) v3.108.1-beta.0	2026-08-20 12:22:34 +00:00
Paul Nothaft	b3a7ab27ea	fix(faces): face avatars were cropped against a cropped rendition (#1100) ... * fix(faces): face avatars were cropped against a cropped rendition Found while triaging #1096, which reported the People manager showing unusable cluster covers — a bare shoulder, the back of a head, a patch of background — and asked for more sample faces to compensate. Most of that is not a detector problem and not a UI limitation. It is a bug. faceCropStyle positions an avatar by scaling the WHOLE frame and offsetting so the face lands centre. That holds only while the rendition shown is the entire image at a uniform scale. Thumbnails are not: imageProcessor.js:93 DEFAULT_THUMBNAIL_FIT = 'inside' migration 040:6 thumbnail_fit seeded to 'cover' imageProcessor.js:229 fit: settings.fit The 'inside' constant is only a fallback for a missing settings row, and the row is seeded on every install — so thumbnails are centre-cropped essentially everywhere, and every face avatar rendered against one is silently offset on any non-square photo. The reporter read the setting as safe because of that constant, and the code comment at :87-92 says the same thing; all three places disagree with what is actually stored. It presents as a bad detector, which is why it survived: the boxes are right, the frame they are drawn against is not. All three surfaces — the admin manager and the guest-facing strip and sheet — now read a preview, which uses fit: 'inside' and is therefore the whole frame. At w=640: plenty for a 64px avatar at DPR 3, and small enough that a strip of a dozen people does not pull a dozen 1920px renditions. Face scanning already calls ensurePreviewImage for anything it scans, so a preview exists for every photo that has a face. Adds the admin preview route the manager needed; the gallery already had one. Both whitelist ?w= the same way. The first version of the call-site test passed with every surface still reading thumbnail_url, because an import alone satisfied it. It now matches inside the src={...} expression, and each of the three surfaces was individually reverted to confirm the test fails. * fix(faces): size the face tier by bbox, and keep admin_preview auth The face half of the external review; the tier-key and long-edge fixes live on the #1099 branch this is stacked on. Face avatars used one fixed 640 tier. In a 6000px group shot a 200px face is ~21px there, and faceCropStyle then blows that up ~9x to fill a 64px avatar at DPR 3 — mush, and indistinguishable from the mis-positioning bug this PR exists to fix. The tier is now derived from the bbox's share of the frame, so a face across a hall gets 1920 and a close-up still gets 640. The synthesized face URL also dropped admin_preview. verifyGalleryAccess only accepts the admin cookie when admin_preview=1 is on the request (middleware/gallery.js:28), and the preview flow deliberately mints no gallery JWT — so every avatar 401'd in exactly the mode an admin uses to check a gallery before sending it to a client. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-20 14:17:25 +02:00
Paul Nothaft	56034af0f0	chore(main): release 3.108.0-beta.0 (#1102) v3.108.0-beta.0	2026-08-20 12:14:09 +00:00
Paul Nothaft	011f6ae7ec	feat(gallery): sized preview tiers so phones stop pulling 1920px (#1095) (#1099) ... * feat(gallery): sized preview tiers so phones stop pulling 1920px (#1095) A phone can display ~1170px at most, but the preview tier is a single 1920px JPEG with no size parameter — so every lightbox swipe ships roughly twice the bytes it can use, and the slide track preloads neighbours, which multiplies it. On the reporter's all-external install a null preview_url falls back to the untouched NAS original, which makes it worse again. Backend: ?w= on the gallery preview route, whitelisted to 640/1280/1920. A whitelist rather than a free-form width because every distinct value is a permanent rendition on disk — an open parameter is an invitation to fill the volume. Unrecognised or absent values fall through to the canonical 1920 preview, so old clients and hand-typed URLs behave exactly as today. Extra tiers are cache, not state: ensurePreviewImageAtWidth keys them by width, looks them up in storage and generates on miss, and never writes photos.preview_path. That column owns the canonical rendition, and threading a width through it would mean the last size anyone requested silently becomes "the" preview. Requesting 1920 resolves to the existing preview rather than a w1920 duplicate, so no install grows a second copy of every preview it already has. The tier is part of the ETag. Without it a client holding the 1920 rendition gets a 304 for its 640 request and renders the wrong size, which is this feature inverted. Frontend: the lightbox picks a tier from innerWidth x devicePixelRatio, capped at DPR 3 — uncapped, a DPR-10 device asks for 3900px and lands straight back on the desktop rendition. At the top tier the URL is left byte-identical so existing caches and ETags stay valid and desktop sees no change at all. saveData and a 2g/3g effectiveType drop one tier; both are Chromium-only, so they are a bonus rather than the mechanism. Grid thumbnails are NOT tiered here, deliberately. generateThumbnail resolves its width from admin settings rather than an argument, so tiering it is a separate change — and shipping a srcset whose candidates the server ignores would be worse than shipping none: the browser would take the "600w" candidate, receive the 300px image and upscale it, which is the reported softness made slightly worse. That half of #1095 lands separately. * fix(gallery): scope tier keys per photo, size by long edge, clean up tiers External review. Three findings against the tier work, one a cross-gallery leak. The tier cache key was the photo's BASENAME. Managed uploads keep camera basenames, so two events can each hold an IMG_0001.jpg — and a tier is served straight from a cache hit without re-reading the source, so the second gallery gets the first gallery's photo. Keys are now scoped by photo id for every source type. The RAW branch passed proc.outputBasename, which would have dropped that scoping again; it now passes the scoped name. Tier selection used viewport WIDTH, but ?w= bounds the LONG edge (fit:'inside'). On a 390x844 phone at DPR 3 a 2:3 portrait is bound by height and renders ~1755 device px, so width-only picked 1280 and made portraits softer than today; landscape on the same phone needs ~1170. It now computes the rendered long edge from the photo's own dimensions and falls back to the top tier — today's behaviour — when they are unknown. Tiers live outside photos.preview_path, so nothing else knew they existed: delete, bulk-delete and archive left them orphaned in previews/ forever, and regenerate-previews refreshed only the canonical rendition while phones kept the stale copy. previewTierKeys derives them from the same deterministic scheme and all four paths clean up. Deliberately outside the preview_path guard — a tier can exist when the canonical rendition never did, so keying cleanup off preview_path would strand precisely the photos only ever viewed on a phone. The existing tier tests encoded the old width-only semantics and were updated rather than kept; that is a behaviour change, not a test fix. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-20 14:08:07 +02:00
Paul Nothaft	0b886ed942	fix(faces): defer on unreachable storage, and commit the import path first (#1097) ... * fix(faces): defer on unreachable storage, and commit the import path first The two follow-ups left open by #1091, both consequences of external photos becoming scannable at all. **A dropped mount no longer burns the gallery.** ensurePreviewImage returns null for "this JPEG is corrupt" and "the NFS share is gone" alike, and faceProcessor marked both 'failed'. Nothing re-queues a failure automatically and the queue only ever claims 'pending', so a mount that blinked mid-scan cost the whole event a manual Re-scan — on external libraries, where network storage drops far more often than local disk, that is the common case rather than the corner one. faceProcessor now probes the containing DIRECTORY before failing, and throws TransientSourceError when it cannot be reached; faceQueue treats that exactly like SidecarUnavailableError — release to pending, back off, retry — with the warning rate-limited to one per five minutes, since an outage hits every photo in the event. The directory rather than the file is the whole point: a missing file inside a healthy directory is a broken photo and should still fail, and it still does. Anything that goes wrong deciding which case it is falls through to 'failed', because guessing 'transient' on an unknown condition would retry forever. **The import commits its path before inserting rows.** enqueueEvent accepts processing_status NULL (faceProcessor.js:243-246), which these inserts leave unset, so an admin hitting the toggle or Re-scan during a long import could queue partial rows while the event still resolved against the old directory — and burn them to 'failed'. Moving events.external_path ahead of the loop closes that, and fixes a pre-existing bug on the same line: an import that died at photo 500 of 1000 used to leave those 500 rows pointing into the new tree while the event still resolved against the old one, making every one of them unreadable. Safe to do first because the path is already validated above, and existing photos are unaffected — photo.source_origin takes precedence over event.source_mode in both resolvers and is NOT NULL defaulting to 'managed'. The #1090 test that asserted a missing source fails needed its setup corrected rather than its intent: it created no directory at all, which is now (correctly) a dropped mount. It now creates one, so it tests the case it always meant to — healthy storage, dead photo. Verified both fixes discriminate: removing the probe fails the defer test, and moving the event update back after the loop fails the ordering test. Face suite 59 passed across 8 suites; the full backend suite fails the same 10 pre-existing suites as unmodified main, no more. * fix(faces): stop a dead mount stalling the whole queue External review, and the first finding is one my own change created. Deferring by returning the row to 'pending' was a trap: claimNextPhoto orders by id ascending and the queue defaults to a single worker, so the same unreachable row becomes the oldest pending one after every backoff and the worker never reaches a higher id. One dead mount would have stalled face scanning for the entire install — unrelated events, fresh uploads, everything. Strictly worse than the permanent 'failed' this set out to replace. The row is now left parked in 'processing' with face_started_at intact. It is not claimable, so the worker moves straight on; the janitor that already exists returns it to 'pending' past STUCK_TIMEOUT_MS, which is the retry. No new column and no new timer. The sidecar branch still releases, because a down sidecar blocks every photo anyway — there is no other work to get on with. Second: probing existence was not enough. Unmounting an NFS or SMB share usually leaves the mountpoint behind as an ordinary empty directory, so fs.access succeeded on storage that was entirely gone and the photo was failed anyway — the exact case this was written for. An empty directory where the photo should live now counts as unreachable. The trade is deliberate and documented: a directory an admin genuinely emptied is retried rather than failed, which now costs one attempt per janitor sweep and nothing else. Third: a comment in adminExternalMedia claimed source_origin isolates existing photos from the early external_path update. That is true of managed rows and false of external ones — resolveExternalPath prefixes every external row with event.external_path, so importing folder B into an event referencing folder A rebases the A rows. Pre-existing rather than introduced here (the update always did this, just later), but the comment asserted otherwise, so it now says what actually happens and names the underlying single-base-path limitation. The deferral test initially passed against the blocking version too — database state alone cannot tell the fix from the bug. It now inspects the branch directly, the way the #596 contract tests do, and fails when releaseToPending is put back or the two branches are merged. * fix(faces): back off per event, and stop clobbering concurrent scans Round two of external review. Parking a row in 'processing' fixed the head-of-line block but not the cost: every janitor sweep handed the whole dead gallery back, and the worker walked all of it again — one stat per photo against storage that may be hard-mounted and slow to time out — before reaching any healthy event. Every one of those attempts also went through generatePreviewImage first, which logs an error per photo, so a down mount produced a recurring flood that the rate-limited warning did nothing about. So the backoff is now per EVENT and separate from the janitor: TransientSourceError carries the event id, the queue records a cooldown, and claimNextPhoto excludes those events while it lasts. The janitor keeps doing its own job, which is rescuing rows a crashed worker abandoned. Cooldown is in memory on purpose — a restart is usually what follows fixing a mount, so it should retry at once. Second: committing the event path before the loop means a toggle or Re-scan firing mid-import can now genuinely queue and finish some of those rows. The final bulk update was unconditional, so it dragged 'done' rows back to 'pending' for a duplicate sidecar scan and knocked 'processing' rows out from under the worker. It is now whereNull — only rows nothing has touched are ours to queue. Also corrected a comment of mine that had gone stale in the same file: it still described the enqueue as happening after the event path was written "below", which stopped being true when that update moved above the loop. * fix(faces): judge the mount, the path and the file separately Round three of external review. The probe was too coarse in both directions. It read any ENOENT on the photo's own directory as a mount-wide outage, so a deleted or renamed subfolder — individual/ gone while collages/ is healthy — deferred the entire event and starved every sibling folder, renewing the cooldown on each retry. It now judges the EVENT ROOT for that verdict: root missing, or present-but-empty, is an outage; anything below a populated root is a broken path and fails. And it read a listable directory as proof the photo was at fault, so EACCES on a reconnected share, EIO, or the classic NFS ESTALE handle were burnt as permanent failures. Only ENOENT now means genuinely gone; any other error opening the file defers. The event-wide backoff was also too broad. A reference event can hold managed uploads alongside imported external ones, and those live in local storage that is fine — excluding the whole event id left them unscanned for as long as external rows kept renewing the cooldown, which during a real outage is indefinitely. The exclusion is now scoped to external and reference rows. One of my own tests had modelled the unmount wrongly: it emptied the photo's subdirectory rather than the event root, which under the corrected logic is a populated mount with a missing folder — a failure, not an outage. It now empties the root, which is what an unmount actually leaves behind. Dropped the path require the first version of this probe needed; the event-root form does not. All three fixes mutation-checked: reverting each one fails the test written for it. Face suite 69 passed across 9 suites; full backend suite fails the same 10 pre-existing suites as main, no more. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-20 14:06:21 +02:00
Paul Nothaft	b7f04f6992	chore(main): release 3.107.5-beta.0 (#1101) v3.107.5-beta.0	2026-08-20 11:18:12 +00:00
Luca	2a84efef71	Merge pull request #1098 from Luca-Timo/docs/aio-readme-existing-tag ... docs(readme): point the single-container install at a tag that exists	2026-08-20 14:13:12 +03:00
Luca	e47c103c2a	docs(readme): point the single-container install at a tag that exists ... The all-in-one quickstart tells people to pull `ghcr.io/picpeak/picpeak/aio:stable`, which has never been published, so the documented one-liner fails for anyone who copies it: docker: Error response from daemon: failed to resolve reference "ghcr.io/picpeak/picpeak/aio:stable": not found `merge-aio` does gate `:stable`/`:latest` on `refs/heads/stable` or a non-prerelease `v*` tag, same as backend/frontend — but `Dockerfile.aio` only landed on `main` in 0874a30a (#1068, 2026-08-18), and the current `stable` head (3.46.1) does not contain it. So the aio image has only ever built off `main`/beta refs, and its full tag list on GHCR and Docker Hub is `main`, `beta`, and `3.10x.y-beta.0`. backend and frontend both have `stable` and `latest`; aio is the only image that does not. Point the quickstart at `:main`, which exists today, and note when `:stable` will start working so this can flip back after the next stable promotion.	2026-08-20 11:51:36 +02:00
Paul Nothaft	731df9dc6e	chore(main): release 3.107.4-beta.0 (#1094) v3.107.4-beta.0	2026-08-20 06:39:49 +00:00
Luca	375bc5303d	Merge pull request #1093 from Luca-Timo/ci/aio-dockerhub-mirror ... ci(docker): publish the all-in-one image to Docker Hub, and give aio + ml a Hub page	2026-08-20 09:31:56 +03:00
Luca	899c9b3407	docs(docker): Hub pages for aio + ml, and the image table in the README ... picpeak/ml has an empty Hub overview and picpeak/aio has none at all, while backend and frontend carry hand-written ones — so the two newest images are the two with nothing on their registry page. Adds .github/dockerhub/{aio,ml}.md as the source of those pages and a dockerhub-descriptions job that pushes them on every main merge, so the page cannot drift from the release it describes. backend/frontend stay hand-maintained for now: capturing their current Hub text into files is a prerequisite, not a side effect of this change. README gains a registry table for all four images (both registries share digests and tags), the org-move callout lists the full set, and the feature list finally mentions People in this gallery, which shipped in #1074 without a README line.	2026-08-20 08:24:19 +02:00
Luca	2df455784c	ci(docker): mirror the all-in-one image to Docker Hub ... The aio image (#1042) shipped GHCR-only with a TODO to wire the Docker Hub mirror once the Hub repo existed. backend, frontend and the ml sidecar all publish to docker.io/picpeak/*; aio was the only image a Docker Hub user could not pull. merge-aio now follows merge-backend/merge-ml verbatim: DOCKERHUB_ENABLED computed from the repository slug (so forks stay GHCR-only), a gated Docker Hub login, docker.io/picpeak/aio added to the metadata images list, and a Docker Hub manifest inspect. Tag scheme is untouched — the same beta/main/stable/latest/semver tags land in both registries. The build summary drops the "Docker Hub mirror pending" note and lists the aio (and ml) Hub images when the mirror is active.	2026-08-20 07:49:53 +02:00
Paul Nothaft	54b68fe6e8	chore(main): release 3.107.3-beta.0 (#1092) v3.107.3-beta.0	2026-08-20 05:23:01 +00:00
Paul Nothaft	576924fa57	fix(faces): scan external/reference photos instead of skipping them (#1090) (#1091) ... * fix(faces): scan external/reference photos instead of skipping them (#1090) faceProcessor short-circuited every photo with source_origin 'external' or 'reference' straight to 'skipped', before the sidecar was ever contacted. On an external-media install that is the entire library — the reporter's gallery sat at 0/3230 with every row skipped and no error, and a rescan changed nothing. The guard was correct when written: resolvePhotoStorageKey returns null for anything outside managed storage, so ensurePreviewImage could not build a preview and there was nothing to send. #1078 removed that limitation one release earlier — ensurePreviewImage now reads externals straight off the mount via resolvePhotoFilePath and writes the preview into managed storage, so the key faceProcessor already fetches through getStorage() is readable like any other. The guard outlived its reason. Photos whose source is genuinely gone still return a null preview key and land in the existing 'failed' branch, which is the honest outcome: that is a broken photo, not an unsupported one. The blanket skip was absorbing those too. No migration or manual reset needed — enqueueEvent already re-queues rows with face_status in (NULL, 'failed', 'skipped'), so previously skipped photos get picked up on the next scan. * fix(faces): queue external imports for scanning (#1090) The other half of the same bug, found by external review — and my first counter-argument against it was wrong. Managed uploads are enqueued by photoProcessor, which writes face_status 'pending' once a photo is processed (photoProcessor.js:573, commented as "the only correct place to enqueue"). External media never goes through photoProcessor at all: adminExternalMedia inserts rows directly, leaving face_status NULL. faceQueue.claimNextPhoto only claims 'pending' (faceQueue.js:64), so an import into an already-enabled event produced nothing until someone pressed Re-scan. Lifting the skip guard alone made external photos scannable but still not scanned — which looks like a complete fix right up until you import a photo. Resolved once per import rather than per file, since it is a per-event setting and the loop can run to a thousand files, and guarded on both the global flag and the per-event toggle exactly as photoProcessor guards it, so installs without the feature still never write a face_status. A failure to read the setting logs and imports anyway — the photos are the point. No video guard: walkDir only collects jpg/jpeg/png/webp, so nothing faceProcessor would skip as video can arrive through this route. * fix(faces): enqueue imports only after the event path is written External review caught a race I introduced in the previous commit. Marking rows 'pending' as they were inserted published claimable work while events.external_path still held the old value — or none at all, on a first import, since the route only writes it after the entire thumbnail loop. The face worker polls continuously, so on any import long enough to matter (the loop is ~100-300ms per photo, and the reporter's library is 6500+) it would claim those rows, resolve them against the wrong directory and mark them permanently 'failed' — a state only an explicit Re-scan clears. That is strictly worse than the unscanned photos this set out to fix. Ids are now collected during the loop and marked pending in one pass after the event path is written, chunked at 500 because SQLite caps a statement at 999 bound parameters. The test now drives the real route instead of re-implementing its logic, and observes the mid-loop state from inside the per-photo thumbnail call — the only hook that can see the window the race lived in. Verified it discriminates: deleting the enqueue fails two tests, and moving it back onto the insert fails the ordering test specifically. * fix(faces): read the face setting after the import, not before Third external-review round. The setting was captured before a loop that runs for many minutes on a large library, so an admin who enabled detection during an import left every photo imported after that moment at NULL forever — the toggle endpoint only queues rows that already existed when it fired. Ids are now collected unconditionally and the setting is evaluated immediately before the queue update, off a freshly read event row. The guard is unchanged in substance: both the global flag and the per-event toggle, so installs without the feature still never write a face_status. Test flips the toggle from inside the mocked per-photo thumbnail call, which is the same mid-loop hook the ordering test uses. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-20 07:18:45 +02:00
Paul Nothaft	198df02d2a	test(ml): add an embedding-space fingerprint tool (#1084) (#1089) ... * test(ml): add an embedding-space fingerprint tool (#1084) requirements.txt is pinned exactly so rebuilds produce byte-identical embeddings, but nothing verified that. The API tests stub FacePipeline, so a decode/resize/kernel change could move every stored cluster without failing anything. This prints a hash per stage — decode, resize, cvtColor, YuNet detect, FaceNet forward pass — so the old and new image can be diffed on the same host before any base-image or dependency bump lands. Used it to answer the open question in #1084: Debian/Python 3.12 and Wolfi/Python 3.14 produce identical hashes at every stage, so a base swap would not invalidate stored clusters. The input is generated rather than a fixture, and the hashes are deliberately not compared across architectures — OpenCV and onnxruntime dispatch different SIMD kernels on x86 and aarch64, so this answers "did this change move the numbers", not "is every platform identical". * test(ml): fingerprint the production path, not a parallel one External review found the first cut was largely theatre. The documented command could not run: tools/ is in .dockerignore and the Dockerfile copies only app/, so the script is never inside the image. It has to be mounted — which is what I actually did when producing the numbers, while documenting something else. Three stages were fingerprinting the wrong thing: - The detector recorded "none" plus a return status, because synthetic input has no face to find. It would have stayed green through any change to YuNet or its kernels. Now the ONNX graph is driven directly, so all twelve output heads always produce numbers, and the reported thresholds are the service's (0.6/0.3) rather than FaceDetectorYN's 0.9 default. - The embedding used a hand-rolled tensor, bypassing everything that actually places a face in the embedding space: umeyama + warpAffine, BGR->RGB, per-image standardization, layout, and the L2 normalization the backend's cosine similarity depends on. It now calls _align and _embed directly. Private, deliberately — reimplementing the maths here would drift from pipeline.py and fingerprint a path nothing runs. - Decode exercised PNG, but the worker only ever receives the preview rendition, which imageProcessor.js writes as JPEG. Now a fixed JPEG, embedded as bytes so the input cannot depend on the encoder version being held still. Verified SOI/EOI-clean; the first attempt at this produced "Corrupt JPEG data: 22 extraneous bytes". Sensitivity checked rather than assumed: a one-pixel landmark nudge moves align_warp and embed and leaves decode and the detector heads alone, which is exactly the dependency structure expected. Debian/Python 3.12 vs Wolfi/Python 3.14 remain identical across all sixteen stages, so the #1084 parity conclusion still holds under the stronger check. * test(ml): measure the image's own pipeline, and the detector OpenCV runs Two more from external review, both of which let matching hashes mean less than they claimed. The documented bind mount put the checkout's app/ ahead of the image's /app/app, so comparing two images built from different revisions would have executed the same pipeline source twice and reported a match no matter how the images differed. /app now wins whenever it exists, so the tool measures the image under test however it is invoked, and the loaded path is printed as _app_source so that is auditable rather than assumed. The detector was fingerprinted through onnxruntime, but production runs cv2.FaceDetectorYN — OpenCV's own preprocessing, DNN engine and NMS/landmark decode, none of which ORT touches. An OpenCV upgrade could therefore move real landmarks, and with them alignment and embeddings, while every detector hash held still. It now runs the OpenCV path too, with the score threshold at the floor so synthetic input still yields candidates (594 here) instead of the empty result the production 0.6 gives on an image with no face. The ORT pass is kept alongside it to separate a model change from an OpenCV change. Parity across debian/3.12 and wolfi/3.14 still holds across all 19 stages, and a one-pixel landmark nudge still moves align_warp and embed and nothing else. * test(ml): cover the orchestration and progressive decode too Round three of external review found two more ways the hashes could match while production moved. The isolated stages never fed the detector's output into alignment — _align got fixed landmarks — so INPUT_LONG_EDGE resizing and the row -> landmark scaling in _one_face were invisible. process() now runs end to end on the fixture, with the pipeline's own detector threshold dropped so a faceless frame still yields rows to carry through (26 faces here). A first attempt at that still missed the resize: the embedded fixture is 48px, so `long_edge > INPUT_LONG_EDGE` never fired and changing 1920 to 960 moved nothing. It now runs a second pass with the threshold lowered under the fixture, which executes the same downscale and inverse landmark scaling without carrying a 1920px image in the source. Verified sensitive: moving that bound 32 -> 24 changes both the face count and the embedding. The fixture was also a baseline JPEG, while generatePreview writes progressive (imageProcessor.js:236/480/617) — a different path through libjpeg. Swapped for a progressive fixture, SOF2 confirmed present and SOF0 absent. 24 stages now. Debian/3.12 and Wolfi/3.14 remain identical across all of them. * test(ml): close three more false-negative paths in the fingerprint Round four of external review. All three let hashes match while production moved. INPUT_LONG_EDGE was used but never printed. The fixture is too small to trip the resize in either image, and the forced pass overrides the value in both, so a production change from 1920 to 960 moved no hash at all. It is now emitted alongside the other thresholds, where a reviewer sees it in the diff. The fixture was square, so a width/height swap in setInputSize or the resize produced identical dimensions and identical hashes. It is now 64x48. The forced-downscale pass hashed only an embedding, which is derived from separately scaled landmarks — a regression in the inverse scaling of row[0:4] would have shown up nowhere, because the normal pass runs at scale 1. That bbox is now hashed too; a wrong one is what breaks avatar crops and area calculations. Changing the fixture to 64x48 also broke the forced pass: at the old bound of 32 the downscaled frame is 32x24 and YuNet returns nothing, so the stage pinned nothing. The NO-DETECTIONS-STAGE-VACUOUS marker added last round caught it immediately rather than printing a reassuring hash of an empty result. Bound moved to 48, which still triggers the resize and still yields rows. 25 stages, no vacuous markers. Debian/3.12 and Wolfi/3.14 identical across all of them. * test(ml): hash every detection, not just the first Round five of external review. Both end-to-end passes hashed only candidate 0, so a change that moved candidates 1..n — or merely reordered them — matched as long as the count and the first candidate held. With the threshold at the floor those passes return 24 and 27 candidates, so that was most of the evidence being thrown away. Both now stack every returned face, in order, via a shared _hash_all. Stacking preserves order, so a reshuffle is caught too. Verified against the exact case: reversing candidates 1..n while leaving the count and candidate 0 untouched now moves process_embedding and process_bbox. Before this it moved nothing. * test(ml): hash every persisted field, and emit the model version Round six of external review, plus the adjacent gaps it implied. Two findings: MODEL_VERSION was never emitted, and _hash_all discarded score. Both matter to the backend rather than to the numbers — a model_version change makes faceClustering.js:190 refuse to compare new faces against existing people, forcing a rescan, and det_score decides via meetsQualityFloor (faceClustering.js:96-100) whether a face joins clustering at all. Either could change while every hash held still. Rather than fix only the two named, I checked what faceProcessor.js actually stores per face (:157-167) and covered all of it: bbox, score, yaw, pitch, blur, embedding. yaw/pitch/blur were heading for the same finding next round. One hash per field, so a diff says which thing moved rather than only that something did. model_version is emitted as a compatibility key alongside the thresholds, not hashed — it is a string, and its job is to be read. Verified: scaling score alone by 0.999 now moves process_score and nothing else. 33 stages, no vacuous markers, debian/3.12 and wolfi/3.14 still identical. * test(ml): split verdict from diagnostic, and stop masking the threshold Round seven of external review. The ORT detector hashes were being read as part of the compatibility verdict, but production never runs YuNet through onnxruntime. An ORT change touching a YuNet operator would have moved them while real behaviour was untouched, and the docstring said any difference means re-scan — so the tool could have ordered a full-gallery rescan for nothing. They are now diag_-prefixed, and the docstring states which keys carry a verdict, which are diagnostic, and which are metadata a reviewer has to read rather than diff. setScoreThreshold(1e-6) also overwrote the detector's real threshold before anything recorded it, and _thresholds.det_score only echoes config. If FacePipeline ever stopped applying DET_SCORE_THRESHOLD — falling back to OpenCV's 0.9 default — production would detect a different face set while every hash matched. The constructed value is now read first and emitted as _effective_det_score; simulating the regression makes it read 0.9 instead of 0.6. MAX_FACES is emitted for the same reason INPUT_LONG_EDGE is: the fixture never reaches the pipeline.py:138 slice, so 64 -> 128 would move no hash while real group photos persisted a different face set. 21 verdict keys, 12 diagnostic, no vacuous markers, debian/3.12 and wolfi/3.14 still identical across both sets. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-19 22:43:58 +02:00
Paul Nothaft	3c5bedc2cf	chore(main): release 3.107.2-beta.0 (#1088) v3.107.2-beta.0	2026-08-19 15:10:22 +00:00
Paul Nothaft	37a15e3d49	fix(faces): restore the :beta image tag and surface sidecar health (#1087) ... * fix(faces): restore the :beta image tag and surface sidecar health Both halves of what a user hit on discussions/1069: the People card sat at "Scanning… 0 of 227" for 30 minutes with no explanation, because the sidecar container could never have started. docker-build.yml — republish `:beta`. It used to come for free via `type=ref,event=branch` when the active development branch was literally named `beta`; the rename to `main` silently retired it. backend:beta has been frozen at 2026-06-29 (448da950) ever since while :main moved on, so PICPEAK_CHANNEL=beta has been serving a seven-week-old build across every image. The ml sidecar was added after the rename and so never had a `:beta` at all, which left docker-compose.production.yml:158 unable to resolve ghcr.io/picpeak/picpeak/ml for any documented channel — the image simply does not exist as :beta or :stable, only as :main and pinned versions. Tag added to all four merge jobs, gated on main. `:stable` stays absent for ml on purpose: it is gated on refs/heads/stable and the sidecar does not exist there. stable's docker-compose.production.yml carries no picpeak-ml service, so nothing can reference the missing tag. FaceRecognitionCard — show when the sidecar is unreachable. An unreachable sidecar is not an error by design: faceQueue.js:132-136 releases the photo back to `pending` and retries forever so a restart does not burn the queue. The cost was that a stopped container looked exactly like a slow scan, indefinitely, and the only signal was a backend log line rate-limited to once per five minutes. /admin/events/faces/health already existed and nothing in the frontend called it. It is now polled while a scan is in progress, and a failing check replaces the spinner with the sidecar URL, the underlying error (which distinguishes a stopped container from a token mismatch) and the command to start it. Health is only polled while a scan is running — an idle card has no reason to care whether the sidecar is up. * fix(faces): tell the three sidecar failure modes apart Follow-up to the health surface in this branch, from an external review pass. The original warning was right about "the sidecar is not working" and wrong about almost everything after that. faceClient.checkHealth now returns a `reason` rather than only a message, because the caller has to know whether photos survive: - 'unauthorized' (401) and 'rejected' (any other 4xx) both become SidecarRejectedError in classify(), which workerLoop does NOT retry — every claimed photo is marked 'failed'. Telling the admin the scan resumes on its own was simply untrue there; both now say to fix the cause and Re-scan. - 'unreachable' (refused/DNS/timeout/5xx) is the retryable one. The card also no longer cries wolf. /faces runs inference synchronously inside an `async def`, so one slow photo blocks the event loop and stalls /info past its 5s timeout — a healthy sidecar can fail a probe. Verified with an isolated uvicorn repro: a blocking call in an async handler stalled the sync /info endpoint to 5.01s. The warning now needs three consecutive failures AND no drop in `pending`. Three because a single /faces call may legitimately run to FACE_ML_TIMEOUT_MS (30s) and two probes 15s apart both fit inside that window; `pending` rather than `scanned` because scanned counts only 'done', so a run producing skipped/failed photos is progress that counter misses. A 4xx burns the queue with no backoff, so it can empty before anyone opens the card — in_progress goes false and only "227 failed" is left. The probe therefore also runs when a finished scan has failures, and the notice renders under the counts instead of replacing them. It is worded as present-tense service state, not as a claim about those specific failures: a live probe cannot know whether they came from this misconfiguration or from corrupt images earlier. Attributing them exactly needs stored face_error rows, which is a bigger change than this. Also adds the missing-token case to the unreachable text: FACE_ML_TOKEN has no default and the container refuses to start without it, so the most likely first run fails as a plain connection refusal that "just start it" does not fix. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-19 16:49:47 +02:00
Paul Nothaft	74a8f9bf24	chore(security): shrink the ML image's CVE surface, override deepmerge-ts (#1083) ... Three Trivy cleanups off the code-scanning tab. ml/Dockerfile — install no runtime apt packages at all. Neither libgl1 nor libglib2.0-0 is needed: opencv-python-headless 4.14 bundles what it needs and `ldd .../cv2/cv2*.so` resolves fully on a bare slim base. The old comment claimed the headless wheel still links libGL, which was true of much older wheels. libgl1 was dragging in 36 transitive packages (mesa, LLVM, X11) for a service that never opens a display. Measured with `trivy image` on locally built variants: before: 165 findings — 88 low / 49 med / 19 high / 6 crit without libgl1: 133 findings — 58 low / 48 med / 19 high / 5 crit without either: 123 findings — 57 low / 46 med / 13 high / 4 crit 42 findings gone, image 1.05GB -> 774MB. Not one of the 165 had an upstream fix available, so not installing the packages is the only lever there is. docker-build.yml — set ignore-unfixed on all four Trivy steps. All 123 remaining ML findings are unfixed base-OS CVEs; Debian has them resolved in sid and pending backport to trixie, and apt-get upgrade -y behind CACHEBUST picks each one up automatically. Reporting them buries anything actionable, and suppressing them is the precondition for ever setting exit-code: 1. backend — deepmerge-ts <8.0.0 has a stack-exhaustion advisory (CVE-2026-40345, high) reached via mailparser -> html-to-text, which pins ^7.1.5 so npm cannot get there alone. Not reachable in our code: html-to-text only feeds deepmerge-ts its options object (html-to-text.mjs:1468, :1442), never parsed email content. npm audit goes 3 high -> 0. The lockfile also picks up the version field release-please had left at 3.103.1-beta.0, plus some "peer": true metadata npm 11.6 recomputes. Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-19 13:56:01 +02:00
Paul Nothaft	9f825be01e	chore(main): release 3.107.1-beta.0 (#1081) v3.107.1-beta.0	2026-08-19 08:23:14 +00:00
Paul Nothaft	af7970b069	fix(preview): generate lightbox previews for external/reference photos (#1078) (#1079) ... * fix(preview): generate lightbox previews for external/reference photos (#1078) ensurePreviewImage() resolved its source only via resolvePhotoStorageKey(), which returns null for external/reference photos by design — those live on a media mount outside the managed storage tree. The null went straight into withLocalCopy(), which throws ("LocalFsStorage: invalid relative path: null"), so the preview route fell back to redirecting at the full-size original. A gallery whose photos are all external got no benefit from the preview tier (#492) at all: guests paid 5-12 MB on every lightbox open, with nothing surfaced in the admin UI. Add the external branch ensureThumbnail() has had since #423: resolve via resolvePhotoFilePath() and feed the mount path to generatePreviewImage() directly, with an ext<id>_ output basename so two events referencing the same NAS filename can't clobber each other's preview. Also close the adjacent hole that made the failure a throw rather than the documented null: a row with no source_origin in a reference-mode event takes its mode from the event, so resolvePhotoStorageKey returns null for it too. Return null instead of handing that to withLocalCopy. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(preview): select the columns the external branch needs on bulk regenerate POST /api/admin/thumbnails/regenerate-previews selected only id, event_id, path, media_type, mime_type and preview_path, so photo.source_origin was undefined by the time ensurePreviewImage branched on it. Every external row in a reference gallery took the managed path, resolvePhotoStorageKey returned null for it, and the endpoint reported success while generating nothing. Add source_origin, external_relpath and filename to the select, plus a source-inspection test pinning the caller contract and a service-level test showing a column-starved row is indistinguishable from a managed one. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * style(test): single-quote the source-inspection needles Matches the repo eslint quotes rule (no avoidEscape) by dropping the nested quotes from the search strings rather than escaping them. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-19 10:17:44 +02:00
Paul Nothaft	98fe9c7699	chore(main): release 3.107.0-beta.0 (#1077) v3.107.0-beta.0	2026-08-18 20:41:12 +00:00
Paul Nothaft	b69dd134d0	feat(faces): People in this gallery — face recognition via an optional ML sidecar (#1074) (#1075) ... * feat(ml): optional face-detection sidecar, opt-in and inert by default (#1074) First of four PRs for "People in this gallery". This one ships only the sidecar, its wiring and its CI — no schema, no backend code, no UI. Nothing in PicPeak calls it yet. picpeak-ml is a single FastAPI + onnxruntime container: three endpoints (/health, /info, /faces), no database, no volumes, no egress, no model download at runtime. Clustering, person identity and every privacy decision stay in the backend where the data already lives. Models are YuNet (detection) + FaceNet-512 (embedding), both MIT, both pinned by URL and SHA-256 and verified at build time. The licence analysis is in ml/LICENSES.md: the more accurate InsightFace weights are non-commercial-only and PicPeak's users are working photographers, so they are never baked into an image we publish. Two things worth review attention: - Alignment uses a least-squares similarity transform (Umeyama), NOT cv2.estimateAffinePartial2D. RANSAC and LMEDS exist to reject outliers among many correspondences; given five landmarks and no outliers they fit a three-point subset exactly and let the rest drift. Measured on a real off-frontal portrait: eyes and nose pinned to 0.11px, mouth corners 11.8px out on a 160px crop. Umeyama distributes it (max 6.5px, rms 5.1 vs 7.4). The failure mode is silent — a bad warp still yields 512 confident floats — so tests/test_pipeline.py pins it numerically. - FACENET_ONNX_URL has no default and the build fails loudly without it. deepface distributes FaceNet-512 as Keras .h5 only, so the ONNX is produced once by tools/convert_facenet.py and published as a release asset. Converting inside the build would drag TensorFlow through both architecture legs of every build to produce a byte-identical file. The CI jobs are gated on the FACENET_ONNX_URL repository variable and skip cleanly until it is set. Off by default, twice over: the sidecar is behind the `faces` compose profile, and the backend will gate on a `faces` feature flag that defaults to false. FACE_ML_URL defaults to http://picpeak-ml:8000 so the standard deployment needs no configuration — nothing dials that host while the flag is off, which is why a non-resolving default is harmless. Verified: 27 pytest tests green; YuNet loads and detects against a real portrait with its landmark order matching the alignment template index-for-index; both compose files validate and the faces profile is correctly excluded from a default `up`; workflow YAML parses and the job graph resolves. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(ml): pin the converter toolchain, verify parity, drop a false reproducibility claim (#1074) Ran the FaceNet-512 conversion for real and corrected what the previous commit assumed about it. The conversion works: 23,497,424 parameters, 89.6 MB ONNX, and the converted graph matches the Keras original to 2.086e-06 absolute / cosine 1.0000000000. That check is now part of the script rather than something I did once by hand — a subtly wrong graph still returns 512 plausible floats, so it refuses to leave the file on disk if parity fails. Also ran the full pipeline against both real models end to end. The embedding is L2-normalized to 1.000000, and the same face survives being re-rendered: half scale 0.973, double scale 0.984, JPEG q40 0.987, rotated 8 degrees 0.984, brightness +40 0.988. Scale invariance in particular is evidence the alignment warp is doing its job. Corrected claim: the conversion is NOT byte-reproducible. Two runs with the same pinned versions on the same machine gave different SHA-256s. The graphs are functionally identical — same 336 nodes, same 271 initializers, every weight matching to 0.000e+00 — but a few initializer names differ because tf2onnx's traced-op naming is not deterministic (Keras layer naming is deterministic; I checked). The previous commit message and README both claimed byte-identical output. They were wrong, and it matters: anyone re-running the conversion gets a different hash, and without this note that reads like tampering. The build-time SHA-256 pins one published artifact so its URL cannot start serving different bytes; validating a fresh conversion is the parity check's job. requirements-convert.txt now pins the exact set that produced the artifact, including transitive keras/protobuf/numpy, and documents that the converter needs Python 3.11 while the image runs 3.12. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * feat(faces): schema, queue, clustering and API for People in this gallery (#1074) Backend half of the feature. Migration 177, a face-detection queue, the clustering engine, the gallery and admin APIs, and the privacy wiring. No UI yet; nothing is reachable until the `faces` feature flag is on, which defaults to false. The flag is the gate, not FACE_ML_URL. That variable now has a working default (the compose service name), so its presence proves nothing about intent — if it were the gate, every install would poll a hostname that does not resolve. faceQueue re-checks the flag every tick, so turning it off stops the workers without a restart. Visibility scoping is the part worth reviewing closely. Face rows have no concept of photo visibility, but guests are restricted to photos.visibility='visible'. A raw count leaks how many hidden photos someone appears in, and an unscoped cover face renders a crop of a photo the guest may not open — with the best-scoring face being the likeliest pick, so it would happen often rather than rarely. facePeopleService recomputes both per request against the caller's own scope, and event_people.face_count_total is named to be conspicuous in a guest path. Six tests cover it, including the case where a person's photos are ALL hidden and they must vanish entirely. Face data is excluded from backups and .picpeak exports, per the decision in the thread: it is derived, so a restore re-scans rather than carrying biometrics between operators. Three separate mechanisms, because the engines cannot be filtered alike — EXCLUDED_TABLES for export, --exclude-table-data (not --exclude-table; the CREATE TABLE must survive or restore breaks on the first query) for Postgres, and DELETE + VACUUM on the temp copy for SQLite, which has no way to exclude a table from a whole-file .backup. The VACUUM is not cosmetic: without it the pages stay in the file and the claim is false on disk. Archiving now purges face data explicitly. photo_faces cascades off photos, but archive deletes neither the photo rows nor the event, so without this an archived gallery kept its biometrics indefinitely. Other decisions: clustering keeps names across a re-cluster by majority inheritance (without it, one button click silently discards every name the photographer typed); consolidation refuses to merge two people who were named differently; assignment never compares across model_version, since embeddings from two pipelines are not comparable; low-quality faces are stored but left unassigned so they show in "this photo contains" without spawning junk people. Migration is 177, not 174 — 174/175/176 landed on main while this branch was open. 29 tests green: 7 migration (idempotency, down(), cascade, and that installing it enqueues NOTHING), 11 clustering, 11 privacy/visibility. Lint clean; the pre-existing error counts in databaseBackup.js and server.js are unchanged. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * feat(faces): People strip, face filter and admin controls (#1074) Frontend half. Renders nothing anywhere unless the `faces` feature flag is on AND the photographer enabled detection for the gallery — the whole guest surface hangs off one boolean, event.people_enabled, which the server computes from the flag, the per-event toggle and the show-to-guests toggle together. Guest side: a People strip between the filter bar and the grid, circular crops from each person's cover face, an active-filter chip row, and a "Show all" bottom sheet. The face filter composes with category, search, media type and the liked/saved/rated filters in the same useMemo rather than replacing them, so "photos of Anna that I liked" works. Two people selected means AND by default — that is what picking a second face almost always asks for — with a toggle to OR that appears only once a second person is picked. Unnamed people show a photo count and never "Person 7". A number is honest about what the system knows; an invented name is not. There is a test asserting we don't do it. The strip renders nothing below two people, collapses to one line when dismissed (persisted per slug, so dismissing one gallery says nothing about the next), and appears mid-backfill with a progress line rather than blocking the gallery behind a spinner. Avatar crops are computed in ratios of the source dimensions so they survive whatever rendition the browser gets; without width/height they fall back to an uncropped thumbnail, since a wrongly-offset crop is worse than no crop. No new download endpoint: "download these N" rides the existing photoIds path, which already enforces access level and per-category permissions server-side. Adding a person_id selector would have been a second thing to authorize for no gain. Guest-facing copy never says "biometric" or "recognition" — those words describe our implementation, not the guest's experience. The sheet's footnote answers the first question every guest has (where does this go?) inline. The admin card, by contrast, is explicit: it states the controller obligation next to the toggle, and warns that scanning materializes the preview tier on galleries that never generated one, which is real CPU and disk an admin should know about before a 2,000-photo backfill. EN + DE translations. 140 frontend tests green (8 new), tsc and eslint clean. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(faces): measured match threshold, working build defaults, 89MB smaller image (#1074) Ran the Phase 0 spike that had been outstanding, published the model, and fixed what both turned up. THRESHOLD IS NOW MEASURED, NOT GUESSED. LFW's standard 1000-pair protocol run through this exact pipeline (YuNet -> Umeyama alignment -> FaceNet-512 ONNX), 100% detection on 2000 images: same person cosine 0.6958 +/- 0.1415 diff person cosine 0.0849 +/- 0.1674 separation 0.6109 peak accuracy 96.60% @ 0.405 So the pipeline separates people well — the thing I could not previously claim, since every earlier number was the same face re-rendered. Default moves 0.62 -> 0.50. The old value was a placeholder and a bad one: it gave 0% false merges but 22.4% false splits, i.e. roughly one in four same-person pairs failing to join, which fragments a gallery badly. 0.50 gives 1.0% false merge / 8.2% false split. Peak accuracy (0.405) is deliberately NOT chosen: for clustering the two errors do not cost the same. A false split is a duplicate row the photographer can merge away; a false merge puts a stranger into someone's "download my photos" — and until the Phase 2 merge/split UI ships, there is no way to undo one. So this sits on the conservative side of the optimum. The spike is committed as ml/tools/benchmark_threshold.py rather than thrown away, so "why 0.50?" has an answer in six months and a re-tune is one command. BUILD DEFAULTS. FACENET_ONNX_URL/_SHA256 now default to the published ml-models-v1 release asset, so `docker build ml/` and `docker compose --profile faces up` work with no arguments. Blanking either still fails loudly — a URL without a checksum is never acceptable, since the checksum is what makes the URL safe to trust. Found by running compose for real: it failed exactly as designed, which was correct behaviour and a bad out-of-box experience now that a canonical artifact exists. IMAGE SIZE. 389MB -> 300MB single-arch. `chown -R` after COPY rewrote every copied file into a fresh layer, duplicating the 90MB model for nothing; the user is now created before the copies and ownership set via COPY --chown. Also drops pip/setuptools from the runtime image. Measured RSS is 186MiB idle, and the container answers /faces end-to-end in well under the 80-150ms/photo the issue budgeted. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(faces): threshold 0.50 -> 0.60 from real clustering, theme-aware People strip (#1074) Both fixes come from running the feature on an actual gallery — 61 photos, 5 real identities — rather than reasoning about it. THRESHOLD. The LFW pairwise sweep in the previous commit said 0.50, and it was wrong. On a real gallery at 0.50, three of six visible clusters were contaminated: two different people merged into one strip entry, which is the exact failure that puts a stranger into someone's "download my photos". Pairwise error rates do not predict cluster purity. Greedy assignment compounds — one wrong face drags the centroid toward the midpoint between two identities, making the next wrong face likelier. A 1% pairwise false-merge rate is not a 1% chance of a clean gallery, and no amount of staring at an ROC curve would have shown that. Sweep against ground truth (5 identities): 0.50 -> 6 clusters, 3 contaminated 0.56 -> 6 clusters, 0 contaminated 0.60 -> 5 clusters, 0 contaminated <- exactly right 0.64 -> 5 clusters, 0 contaminated, fewer faces assigned 0.60 recovers the right number of people with no contamination; higher only loses coverage. Migration 177 carries the full reasoning so the next person to touch this knows why the obvious pairwise answer is the wrong one. THEME. The People strip hardcoded `text-neutral-800` for named people. On a dark gallery — which the screenshot immediately showed — that renders a named person's label almost invisibly, while UNNAMED people stayed legible. Exactly backwards. Labels, headings, the collapsed summary, the scan line and the filter chip row now read the gallery's own theme tokens (--color-text / --color-muted-text / --color-accent / --color-surface-border) like the rest of the gallery surface. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(faces): keep the mobile filter row inside the viewport (#1074) At 390px the photo count and Clear link were pushed against the right edge by ml-auto and clipped. Only apply it from the sm breakpoint up, where there is room; below that they flow after the chips. Found by screenshotting the real thing on an iPhone-sized viewport. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * feat(faces): complete Phase 1, add People management and auto-categories (#1074) Closes the two Phase 1 gaps, then builds Phase 2 and Phase 3. PHASE 1 GAPS. "Download these N" was specified, described as done in an earlier summary, and never actually built — I had verified the backend needed no new endpoint and let that stand as if the button existed. It now hands the filtered photo ids to the same path as a manual selection, so the server re-applies access level and per-category permissions on the way through. Photos in a downloads-disabled category are excluded client-side too, so the number on the button is the number the guest receives. Hidden entirely when downloads are off for the gallery. Lightbox person chips ("In this photo: Anna") are the second way into the face filter — a guest looking at a photo of themselves can act on it without scrolling back to the strip. Tapping one closes the lightbox and filters the grid behind it. PHASE 2. A People management modal over the endpoints that already existed and were already tested: rename inline, merge (multi-select, first pick is the target so the name a photographer typed survives), split via a face picker, hide, ignore. This matters more than it sounds — clustering deliberately errs toward splitting because a wrong merge puts a stranger into someone's download, and that trade only works if merging is easy. PHASE 3. Rule engine over face_count plus face-area ratio: 0 -> Details, 1 large -> Portraits, 2-5 -> Small groups, >5 -> Groups. The area ratio is what separates "a portrait of someone" from "someone is in this landscape". Three guarantees, all tested: it only ever fills an EMPTY category (enforced in the query AND re-checked in the UPDATE, so a photographer setting one mid-run still wins), everything it touches is marked auto_categorized so undo is exact, and it is a no-op unless separately enabled. Migration 178 adds the column — separate from 177, which has already run wherever this branch is deployed. Verified on the real gallery: 61 photos -> 48 portraits + 13 small groups, undo cleared exactly 61 and left the manual ones alone. Merge moved faces and removed the source. Both confirmed against the database, not just the UI. TWO BUGS THE BROWSER CAUGHT, both invisible to tsc: - The lightbox destructure never landed — my patch targeted a line that has a default value, matched nothing, and failed silently. `people` resolved to something else entirely and the chips would never have rendered. eslint's "outer scope value" warning is what surfaced it. - Admin face thumbnails 403'd because <AuthenticatedImage> attaches whatever gallery token is in session storage; an admin who has also opened one of their own galleries sends a type:"gallery" bearer to an admin route. Admin routes authenticate from the httpOnly cookie, which a plain same-origin <img> sends by itself. Worth noting AdminPhotoGrid has the same latent shape; not touched here. Also: the admin card now reports "N people (M shown to guests)" when those differ, so the settings page and the gallery stop disagreeing without explanation. 45 backend tests (8 new) and 140 frontend tests green; tsc and eslint clean. EN + DE for every new string. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * perf(faces): batch migration DDL and drop the face stack from server.js import (#1074) CI's backend job timed out at 10 minutes on the first run of this branch. Nothing failed — 132 of 182 suites passed and the wall clock ran out. Main does the same 182 in 124s, and where main has 12 suites slow enough for jest to print a duration, this branch had 77. Two changes, both worth making regardless of how much of the gap they close: - Migration 177 added its columns one ALTER TABLE at a time (four on photos, three on events, plus a separate index statement) and seeded settings with a SELECT and an INSERT per key. It now uses one alterTable per table and one SELECT plus one bulk INSERT. 178 folds its index into the same statement as its column. That chain replays in ~90 suites, so statement count there is multiplied by 90. - server.js required faceQueue at module scope, which pulls in axios and — through imageProcessor — sharp. Every supertest suite that imports server.js was paying for a module graph it never uses. Now required inside the startup block, next to the call that needs it. Honest about the evidence: locally the migration delta measures at zero (1.15s vs 1.13s for the same suite, three runs each), so batching alone does not explain an eight-minute regression. A fast local disk and many cores mask per-statement and per-import costs that a two-core runner with a shared disk does not. These are the two real costs this branch added to a path that runs in almost every suite; whether they are sufficient is a question for CI, not for another round of local speculation. 37 face tests still green after the change. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * i18n(faces): complete EN and DE coverage for the face feature (#1074) The admin card and the Features toggle were rendering entirely from inline English `defaultValue` fallbacks — 22 keys existed in no locale file at all, so a German admin saw an English consent notice, English toggles and English buttons. The gallery side was already translated; the admin side was not, and nothing in the toolchain flags this because a `defaultValue` always renders something. Adds the missing `admin.faces.*` (19), `settings.features.faces.*` (2) and shared `common.clear/saved/saveFailed` in both languages. Existing keys are left alone (setdefault, not overwrite), so the shared `common` strings other features rely on are untouched. Committed the audit as frontend/scripts/i18n-faces-audit.py rather than throwing it away: it extracts every t() key the face components actually use and diffs it against each locale, and it also reports German values that are byte-identical to English, which is the usual shape of an untranslated copy-paste. Currently: 69 keys in use, EN complete, DE complete, no identical pairs. Verified in the browser, not just in the JSON — the German card reads "61 / 61 Fotos durchsucht · 16 Personen (5 für Gäste sichtbar)" end to end. Also checked the components for hardcoded user-facing text (JSX nodes, title/aria-label/placeholder attributes) outside t(); there is none. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(faces): 13 defects from external review — coordinates, counts, erasure, races (#1074) Codex reviewed the branch against main. Thirteen findings, nine P1. I checked every one against the code and could not dismiss a single one as a false positive, so all thirteen are fixed here. THE WORST ONE: bounding boxes were stored in the wrong coordinate system. The sidecar reports coordinates in the space of the image it was HANDED — which is the ≤1920px preview, not the original — while every consumer compares them against photos.width/height, the original dimensions. A 6000px photo therefore produced boxes ~3x too small and areas ~9x too small: avatar crops landed in the wrong place and the Portraits rule could never fire. It is invisible on any photo already under 1920px, which is exactly why the demo gallery and every screenshot looked correct. Now scaled once in faceProcessor so everything downstream can assume original-image coordinates. ERASURE. The FK cascade on photo_faces is decorative on SQLite: PicPeak never enables `PRAGMA foreign_keys`, so deleting a photo left its embeddings behind. I first enabled the pragma globally and reverted it — six unrelated suites immediately failed on pre-existing dangling references, and switching it on would start rejecting inserts on every existing install. That is a real change worth making, but it is its own PR, not a rider on this one. Instead deletion purges explicitly: purgePhotoFaces in the photo paths (single, bulk, service) and photo_faces/event_people in deleteEventCascade. Tests assert this with the pragma explicitly OFF, so they can only pass if the code does the work. COUNTS. A re-scan deleted the old face rows without undoing their contribution to event_people, so counts inflated on every re-scan and ghost people survived. Now the affected people are recomputed before the replacements are assigned. My own "must not double its faces" test only checked photo_faces rows, which is why it passed throughout. RACES. A worker that finished after an admin purged the event committed its rows anyway — erasure reported success and the data reappeared. The commit is now conditional on the row still being 'processing'. And assignFaces is read-modify-write over an event's people, so two workers lost each other's updates; it is now serialised per event with an in-process mutex plus a Postgres advisory lock for the multi-pod case the queue advertises. METADATA LOSS. Merging discarded the source's name and suppression flags, so a merge could erase a typed name or un-hide someone. Reclustering remembered only people with a label, so an unnamed-but-hidden bystander came back guest-visible after one "Re-group people" — and suppression now propagates to every descendant cluster, not just the majority one. Also: export reset face_status so a restored gallery re-scans instead of claiming to be scanned forever; manual category edits clear auto_categorized so "undo automatic" cannot delete a photographer's own choice; external photos are skipped rather than failed (resolvePhotoStorageKey returns null for them by design); the gallery refetches photo memberships as a scan progresses so filtering is not stale; a failed VACUUM now fails the backup rather than publishing one that may retain biometric pages; and the ML Dockerfile's `|| true` is scoped to the uninstall — as written it was `(install && uninstall) || true`, so a failed dependency install produced a green layer and an image with no onnxruntime. Four new regression tests. Full backend suite failure set verified identical to origin/main; frontend 140 green; tsc and eslint clean. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(faces): 12 more defects from review round 2 — cross-event purge, leaks, lifecycle (#1074) Second Codex round on the same diff, now including round 1's fixes. Twelve findings, seven P1. Again none were false positives. SECURITY, AND MINE FROM ROUND 1: the bulk-delete face purge iterated the raw `photoIds` from the request instead of the event-scoped `photos` rows the handler had already validated. purgePhotoFaces has no event scope of its own, so an editor could pass another gallery's photo id and delete its face data — even though the photo deletion right below it was correctly scoped. Fixing one thing and introducing another is exactly why the second round was worth running. ANOTHER VISIBILITY LEAK, same class as the one round 1 fixed: /people returns scan progress, and getScanStatus counted every photo with a face_status — including hidden ones. Guests could read the hidden-photo count off the progress bar while the people list and covers beside it were properly scoped. Now scoped by the same predicate, with the caller passing its audience. RECLUSTER, ROUND 1'S FIX WAS INCOMPLETE. I made suppression follow every descendant but still copied the flags from the majority ANCESTOR. When reclustering merges a visible named person with a hidden one, the majority ancestor is often the visible one — republishing the hidden person's photos. Suppression is now OR-ed across every ancestor contributing faces. The name also now goes to the genuine largest descendant; the previous code took whichever cluster came first in map order, which the comment already claimed it did not. LIFECYCLE. Face data is excluded from backups and exports, but photos. face_status came across intact, so a restored install claimed every photo was scanned while holding no faces — and the worker only claims 'pending', so it stayed that way forever. Now: the SQLite backup requeues in the dump, restore requeues after the pool reinit (the Postgres path cannot rewrite rows inside pg_dump), the portable importer purges LOCAL face tables (they were excluded from the replace list, so another instance's embeddings survived an import with FK checks suspended) and requeues, and archiving disables detection so a restored archive is honestly off rather than enabled-and-empty. WRITE PATHS. Only processPhoto enqueued. The synchronous upload path (chunked-upload completion, watch-folder) left photos unscanned, and replacePhoto kept the OLD image's faces on a row now pointing at a different picture — stale identities shown on the new photo. FRONTEND. PeopleSheet and the admin manager rendered centred thumbnails and ignored the bbox, so on group photos the avatar showed whoever stood in the middle and two people from one photo were indistinguishable — in the manager whose entire job is telling faces apart. The crop maths is now one shared helper (faceCrop.ts) so the three surfaces cannot drift again. Full-page layouts (gallery-premium, gallery-story) render their own lightbox and never received the people props. Backend failure set verified identical to origin/main; frontend 140 green; tsc and eslint clean. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(faces): round 3 — five of round 2's fixes were wrong or no-ops (#1074) Third and final Codex round. Eight findings, four P1 — and the important part is that FIVE of them are defects in round 2's fixes, not in the original code. - The sync-upload enqueue I added was a silent no-op. It queried through `trx` after the transaction had already been committed, which throws "Transaction query already complete" straight into the catch I had wrapped it in. Chunked uploads and watch-folder imports were still never scanned, and the code read as though they were. Uses `db` now. - The post-restore requeue ran BEFORE the files were restored, in both the portable importer and the native restore. The face worker is live during a restore, so it could claim those rows and scan the previous instance's files, or fail them for originals not yet on disk — with nothing to requeue them afterwards. Both now run after file restoration; the native one is extracted into requeueFaceScans() and called from the full and database-only paths. - The admin face crop mixed coordinate spaces: an original-pixel bbox scaled against the THUMBNAIL's natural size. The API now returns the source dimensions alongside the box, so there is one space to reason about. - Forwarding people props through layoutProps did not make them work — the full-page layouts never destructured them. GalleryStoryLayout now threads them to its own lightbox. Genuinely new findings, all in the same class as ones already fixed: - releaseToPending updated unconditionally, so a photo purged while its sidecar request was in flight came back as 'pending' and was rescanned — biometric rows reappearing after the purge reported success. Round 2 fixed exactly this on the COMMIT path and I did not carry it to the retry path. Now guarded on 'processing'. - purgePhotoFaces left face_status alone, so a worker mid-scan still satisfied its commit guard and could write fresh faces into a photo being deleted — orphans, since the FK cascade is inert on SQLite. It now clears the claim as part of the purge. - Phase 3 was unreachable: the migration seeds face_auto_categorize_enabled false and nothing could ever write it, so the rule engine and its undo endpoint returned "disabled" in every real flow. Added GET/PUT and a toggle on the admin card, EN + DE. NOT fixed, deliberately: GalleryPremiumLayout uses yet-another-react-lightbox rather than the shared PhotoLightbox, so person chips there are a real port rather than a prop forward. Recorded as open rather than bodged. Backend failure set identical to origin/main; 41 face tests and 140 frontend tests green; i18n audit reports EN and DE complete at 71 keys. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * feat(faces): block face recognition on the all-in-one image (#1074, #1042) The single-container image cannot run this feature, so it is refused there rather than left to degrade. WHY, since the reason is not obvious from the code: the AIO image runs the backend, the frontend, SQLite and every background worker inside one container aimed at "one photographer plus guests browsing". It has no Redis, SQLite gives it a single writer, and it contains no ML sidecar to talk to. Face detection would add a second image-processing pipeline competing with Sharp for the same CPU and memory. That failure is not loud — the install just becomes slow and looks broken, which is the worst possible shape for a deployment whose whole promise is one container and no decisions. Gated on an explicit PICPEAK_SINGLE_CONTAINER marker, NOT inferred from SERVE_FRONTEND or a SQLite path: plenty of legitimate multi-container setups serve the frontend from the backend or run SQLite, and none of them should lose the feature by accident. Three layers, because the first is the only one that enforces: - faceSettings.isFeatureEnabled() returns false before consulting the flag, so a database restored from a full deployment with `faces` enabled still cannot switch it on here. - The feature-flag API forces `faces: false` in both directions, so the admin UI reflects reality instead of offering a switch that refuses to stay on. - The Features tab renders the card disabled with a plain-language reason, read from a new `single_container` field on /admin/system/version (an endpoint the admin UI already calls). Documented in ml/README.md and .env.example. Three tests pin the behaviour, including that the marker only accepts explicit truthy values. NOTE FOR PR #1068: this expects `Dockerfile.aio` to set `ENV PICPEAK_SINGLE_CONTAINER=true`. That one line lives on that branch and is not in this commit — until it lands, an AIO build would still offer the feature. Worth adding alongside the `Limits` section of docs/single-container.md. 44 face tests green; EN + DE complete at 72 keys. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * test(faces): pin the bbox coordinate space with a real scale factor (#1074) The coordinate-space bug — boxes stored in preview space while every consumer reads them as original-image pixels — had no test, and could not have been caught by the ones that existed: every photo in the demo gallery is 750px, so the scale factor was always exactly 1.0 and the correction never executed. Verified by hand first, on a real 4000x3000 upload with the face placed off-centre so a wrong crop would be unmistakable. Before the fix the stored box was 1493,204 (preview space, face actually at x≈2850-3618); after, 3110,426 — a factor of 2.083, exactly 4000/1920, landing inside the face. The admin crop then resolved to left=-395px/top=-46px on a 64px window, which is the face centred. That verification is now a test rather than a memory. Three cases: a 4000px photo must scale by 4000/1920, a 1920px photo must NOT change (the case that hid the bug), and a row with no width must fall back to unscaled rather than storing NaN. Note for anyone extending these: jest hoists mock factories above the file, so anything they close over has to be `mock`-prefixed. Getting that wrong fails at transform time with a message that does not name the variable. 47 face tests green. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-18 22:37:28 +02:00
Paul Nothaft	6ebdd13dde	chore(main): release 3.106.0-beta.0 (#1076) v3.106.0-beta.0	2026-08-18 20:21:44 +00:00
Luca	0874a30ac9	feat(docker): all-in-one image (#1042) — my version of #1067 (#1068) ... * feat(docker): add all-in-one image — backend + frontend in one container (#1042) One container, one Node process, SQLite by default: `docker run` with no compose file, no nginx, no supervisor, no bundled Postgres/Redis. - Dockerfile.aio (repo-root context): frontend build stage + backend deps stage + a runtime stage mirroring backend/Dockerfile's production stage, with the built SPA copied to /app/frontend/dist and SERVE_FRONTEND=true. DATABASE_CLIENT=sqlite3 and STORAGE_PATH=/app/storage are pinned explicitly — the storage fallback resolves to container-root /storage, which EACCESes after the su-exec drop. - server.js: the SERVE_FRONTEND block now does what the nginx image did — renders ${BRAND_TITLE}/${BRAND_DESCRIPTION} into index.html once at boot, serves that rendered shell on /index.html and every SPA route, caches hashed /assets/* immutably while the shell revalidates, and gzips the bundle via compression() mounted after all /api routers. express.static now runs with index:false so `/` keeps flowing to handlePublicSiteRequest — its default index option was shadowing the landing page on native installs. - wait-for-db.sh: skip the Postgres readiness wait when DATABASE_CLIENT is sqlite3. The engine resolver still runs, still logs, and still refuses the populated-both conflict (#1038). - .dockerignore: **/node_modules, so the root-context build can't pick up host deps from backend/ or frontend/. - docker-build.yml: build-aio / merge-aio follow the same per-arch build → digest-merge → per-version tag scheme as backend/frontend (GHCR only for now; the Docker Hub mirror is wired once the Hub repo exists), plus a smoke-aio job that boots the image on every PR and asserts /health, the SPA shell, the rendered brand title, immutable asset caching and the SQLite engine resolution. Pointing DB_HOST/DB_USER/DB_PASSWORD + DATABASE_CLIENT=pg at an external Postgres works exactly like the backend image. * fix(ci): correct three smoke-aio assertions that would fail a green image (#1042) Found by running the smoke job locally against a real build — the image passed every behavioral check, but three assertions were wrong: - `/` asserts 200, but handlePublicSiteRequest 302s to /admin/login while the public landing site is disabled, which is the state of the fresh install the smoke container always is. Assert the redirect target instead — that still proves express.static's index option is not shadowing the handler, which is the thing the check exists for. - The placeholder-leak grep matched index.html's explanatory comment, which mentions BRAND_TITLE in prose and survives into the built shell. Match the literal ${BRAND_TITLE}/${BRAND_DESCRIPTION} tokens with -F, and cover the description token too. - Add a gzip assertion, probing with GET: the compression middleware skips bodyless responses, so a HEAD probe reports no Content-Encoding even when compression is active. Verified locally on linux/arm64: image builds clean, boots to healthy in ~8s on the SQLite default, and 25/25 checks pass (SPA shell, rendered brand title, immutable+gzipped assets, no-store shell, SPA fallbacks, npm removed, su-exec drop to nodejs, no errors in the boot log). The DATABASE_CLIENT=pg override was exercised against a real Postgres too — the readiness wait still runs and the engine resolves to postgres. * fix(server): serve the SPA for every client route, not just /admin and /gallery (#1042) nginx did `try_files $uri $uri/ /index.html`, so behind compose every client-side route survived a direct hit or a refresh and the short `['/admin', '/admin/*', '/gallery/*']` list was never exercised. Without nginx that list is the whole contract, and everything outside it 404'd: /setup /customer /impressum /datenschutz /payment-check /quote/:token /contract/:token /invite/:token /transfer/:token /transfer-upload/:token /setup is the first URL a new install visits, so the all-in-one image was unusable from a cold start. The catch-all is registered after `app.use('/api', notFoundHandler)`, so an unknown /api route still answers JSON instead of being handed the HTML shell, and after the /s/:shortSlug resolver, so a typo'd short URL still 404s (#699). It is GET-only — a stray POST keeps 404ing rather than getting a 200 page back. The handler is hoisted out of the SERVE_FRONTEND block via `spaCatchAll` because that block runs before the API 404 handler is registered. Verified on the built image: all ten routes above now 200, /api/nope still returns JSON 404, /s/nonexistent still returns 404, / still 302s to /admin/login, and the smoke suite is 25/25. Both boundaries are now asserted in the smoke-aio job. * docs(readme): document the single-container install (#1042) The README had no mention of the all-in-one image, so the only way to discover it was reading the workflow file. Adds a Quick Start subsection with the one-line `docker run` and the `docker exec … cat SETUP_TOKEN` step, plus a row in the documentation table. Deliberately does not sell it as the default: the note says the compose stack is still the right choice for anything busier, gives the reason (SQLite takes one writer at a time), and points at the `.picpeak` restore as the way out, so nobody picks it and then finds themselves stuck. Full details live at docs.picpeak.app/deployment/single-container (PicPeak/docs#8). * feat(docker): fold #1067's items into the all-in-one image (#1042) Consolidating the two parallel AIO branches into this one. This PR's approach is kept wherever the two differed on design — in particular the in-process brand render, `index: false` (which fixes express.static shadowing handlePublicSiteRequest, a bug #1067 had), the compression middleware, and the smoke-aio job. What follows is what #1067 had that this branch did not. Layout — the issue asks for a single mountable root, and this moves to one: /data/db picpeak.db (+ -wal/-shm) and SETUP_TOKEN /data/storage originals, thumbnails, archives /data/logs application logs /data/backup built-in backup output; /backup symlinks here `-v picpeak:/data` and nothing else to remember. README and the smoke job's database-path assertion follow the new layout. Correctness items: - sqlite CLI. DatabaseBackupService SPAWNS `sqlite3` for `.backup` and PRAGMA integrity_check; the npm module does not ship that binary. backend/Dockerfile omits it because compose always runs Postgres — this image defaults to SQLite, so every database backup failed with ENOENT. - /backup wired in. Migrations 029 + 030 seed /backup/picpeak and /backup/database as the backup destinations; nothing created or mounted them, so backups had nowhere to write and anything written would die with the container. Symlinked into the volume, subdirectories created at startup (a bind mount hides the tree baked into the image), and adopted only when BACKUP_DIR is set so it never gates boot for compose deployments that do not mount it. - logger.js honours LOG_DIR. It hard-coded <backend>/logs, so logs could not leave the container. Unset keeps the old path for every existing install. - wait-for-db.sh derives its writable roots from STORAGE_PATH / DATA_DIR / LOG_DIR instead of hard-coded /app paths, and mkdir -p's them before chown — a bind-mounted /data hides the image's tree, and chown against a missing path reports "the filesystem rejects chown", which is both wrong and a dead end. - .dockerignore excludes backend/-prefixed runtime data. Docker reads only the root file, so the unprefixed data/*.db, logs/* and storage/* rules missed backend/data, backend/logs and backend/storage entirely; a checkout used to run PicPeak would bake its database, photos, logs and SETUP_TOKEN into a published layer. - HEALTHCHECK follows $PORT rather than a hard-coded 3000. - --max-http-header-size=32768 matches nginx's large_client_header_buffers 4 32k; Node's 16 KiB default would reject a guest carrying several per-gallery JWT cookies. docs/single-container.md is added as the in-repo reference the README links to. The smoke job gains four assertions for the above: the one-volume layout and writable backup destinations, the sqlite3 CLI, logs landing on the volume, and the image carrying no runtime data from the build context. Verified on a built image — named volume, bind mount and PORT=8080 all healthy; every existing smoke assertion still passes, including / -> 302 /admin/login, the rendered BRAND_TITLE, immutable assets, gzip and /s/<unknown> -> 404. Co-authored-by: Luca-Timo <102960244+Luca-Timo@users.noreply.github.com> * fix(docker): restore the SPA-fallback exclusions and close the build-context leak (#1042) Both found by external review of the consolidated branch. - The SPA catch-all had no backend-owned exclusions. This was a regression I introduced while merging: #1067 carried a BACKEND_OWNED prefix list, and taking this branch's server.js wholesale (correctly — its index:false and in-process brand render are the better design) dropped it. /photos, /thumbnails, /uploads and /fonts are static mounts whose middleware calls next() on a miss, so the catch-all was answering 200 text/html under image and font URLs instead of 404. nginx gave each of those its own location block, so try_files never applied to them. - backend/data is now excluded wholesale rather than by suffix. The suffix list (*.db, *.db-wal, *.db-shm, SETUP_TOKEN) let real secrets through: a used checkout carries ADMIN_CREDENTIALS.txt next to the database, plus -journal files and any DATABASE_PATH not ending in .db. Since Dockerfile.aio builds from the repository root and COPYs backend/ wholesale, any of those would be baked into a published layer. The directory holds only runtime state and is already gitignored in full. smoke-aio gains an assertion that the backend static routes still 404, so the exclusion cannot be dropped again silently. Verified on a built image: /photos, /thumbnails, /fonts and /uploads misses all 404; /setup, /impressum, /gallery/x, /admin/login still 200; / still 302s to /admin/login; /api/nope still answers JSON; /s/<unknown> still 404s; and the image carries no *.db, ADMIN_CREDENTIALS.txt, logs or storage from the context. * fix(aio): three failures that only surface outside a dev laptop (#1042) Backups aborted on SQLite. getTableChecksums() built its digest with `CAST(t.* AS TEXT)`, which is Postgres row-to-text syntax; SQLite parses `*` there as a syntax error, so every backup threw before reaching the .backup call. Since the all-in-one image ships SQLite by default, that is every AIO install. Enumerate the columns via columnInfo() and sum their lengths instead. The shared /data mount root was never adopted. wait-for-db.sh chowned the children it creates but not the mount point itself, so a host directory arriving as 0700 with a foreign owner stayed untraversable by UID 1001 after the su-exec drop. Docker Desktop's permissive bind mounts hide this completely, which is why local testing passed; a NAS share does not. DATA_ROOT is now adopted first. Maintenance mode locked the admin out of the box. The middleware runs at server.js:493, long before the static block at 891, and exempted the auth endpoints but not the page that calls them. With the backend serving the frontend, /admin/login and /assets/* returned 503 JSON, so an admin who enabled maintenance mode could never load the UI to turn it off. nginx serves those paths in the compose stack, which is why it never surfaced there. Guest and API surfaces stay gated. Verified on a built image: checksums compute across all 95 tables; a bind mount created 0700/4000:4000 boots healthy and ends up 1001:1001; with general_maintenance_mode=true, /admin/login, /admin and /assets/* return 200 while /gallery/* and /api/gallery/* return 503 — and 503 across all three once the exemption is removed again. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(aio): stop leaking .env into the image, fix the broken checksum test (#1042) The Jest suite was red: mocking db.raw is no longer enough now that the SQLite checksum branch asks the query builder for its column list, so db(table) came back undefined and getTableChecksums failed on every PR. The production code is right; the fixture needed to know about the call. backend/.env was landing in the published layer. The root ignore file's `.env`, `.env.*` and `data/*.db` rules read as unanchored but Docker matches them from the context root, so they catch ./.env and never backend/.env — and `COPY backend/ .` then puts a real JWT_SECRET at /app/.env. Matched at any depth instead, the way **/node_modules in the same file already is. Confirmed by building from a checkout carrying a planted secret: before, `cat /app/.env` printed it back. Business documents wrote outside the volume. quoteService, invoice sending/reminders and contract signatures build paths from process.cwd()/storage and never read STORAGE_PATH; compose hides it by setting STORAGE_PATH=/app/storage with WORKDIR /app so the two are the same directory. Here they are not, and /app is root-owned, so a quote or invoice PDF failed to write as UID 1001 — and would not survive the container if it had. Symlinked /app/storage into the volume, matching the /backup symlink beside it. Teaching those services STORAGE_PATH is the real fix and wants its own change. Two smaller ones: the mount root is now chowned shallow rather than recursively, since every child below it is already walked recursively and a NAS-sized photo library should not be traversed twice on each restart; and /assets/ joins the backend-owned prefixes, so a stale hashed chunk requested by a tab left open across an upgrade gets a 404 instead of index.html served with 200 under a .js URL. Verified on a built image: planted backend/.env and backend/probe.db are absent; /app/storage resolves to /data/storage and a business-doc write as UID 1001 appears on the host; a 0700 bind mount owned by 4000:4000 boots healthy; a missing /assets chunk 404s while the real bundle still serves 200 as application/javascript. The databaseBackup suite is green again, and the branch adds no failing suite that origin/main does not already fail on the same machine. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * test(aio): teach the leak assertion about the storage symlink (#1042) The previous check listed /app/storage/events and treated a hit as a leak. That was true while /app/storage was either absent or a copied directory; now it is a symlink into the volume, so the check followed it and found the empty tree the image itself creates — a false positive on its own design. Check the shape instead: /app/storage must be a symlink pointing at /data/storage, and the volume's photo tree must contain no files on a fresh install. A real directory there now fails loudly, which is the condition the assertion was always trying to catch. Also extended the path list to /app/.env and loose database files, matching the .dockerignore rules added alongside. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(aio): show the maintenance screen instead of raw JSON to guests (#1042) The previous commit exempted the admin shell so an admin could still reach the switch they had just flipped. Guests had the same problem for the same reason: with no nginx in front, /gallery/<slug> reaches this middleware long before the static block, so a visitor during maintenance got a 503 JSON body where every other deployment shows the branded maintenance screen the frontend already ships. Replaced the two path-specific exemptions with the rule they were both special cases of: a GET that is not an API call and not a backend-owned content mount is the SPA shell, and the shell is inert HTML — it boots, reads /api/public/settings (already exempt) and renders MaintenanceMode on its own. Everything that carries real data stays gated: /api/*, /photos/, /thumbnails/, /fonts/, and any non-GET. Compose is untouched by construction, since nginx answers those paths and they never arrive here. Verified on a built image with the flag on: /gallery/x, /customer/x, /admin and /admin/login return 200 text/html while /api/gallery/x/verify, /photos/x.jpg and /thumbnails/x.jpg return 503 and a POST to a public API still returns 503; with the flag off the same paths go back to 404. Added a middleware test over that exemption matrix — over-exemption is the real risk in this change, so it asserts the gated half too. It fails on five cases without the fix. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(aio): stop the shell exemption from un-gating /og and the public CMS (#1042) The previous commit exempted "any GET that is not an API call". That negative rule reads as safe and is not: /og/gallery/<slug> and its /cover render the event name and the hero thumbnail, /s/<code> renders short-link previews, and `/` is handed to the public CMS. All four are proxy_passed to the backend by nginx, so they were gated before this PR in every deployment — the rule un-gated them, and for compose too, not just the new image. A site switched to maintenance would have kept publishing gallery metadata. Replaced the guess with the split nginx already defines: exempt what the frontend container answers itself, gate what it proxies. That is the same rule the all-in-one image needs by definition, since its whole job is to be both halves of that stack, and it now matches compose in both directions rather than only in the direction the last commit tested. Verified on a built image with the flag on: /admin/login, /gallery/<slug> and /customer/* return 200, while /, /og/gallery/x, /og/gallery/x/cover, /s/abc, /robots.txt, /api/* and /photos/* return 503; with the flag off all of them behave normally again. The middleware test grew the gated cases — it now covers 21, most of them asserting what must NOT be exempt. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(aio): give the image a FRONTEND_URL default so share links are absolute (#1042) getFrontendBaseUrl() reads FRONTEND_URL, falls back to the general_site_url setting, and otherwise returns an empty string — which makes share_url come back as a bare "/gallery/<slug>/<token>". Compose defaults the variable to http://localhost:3000, but the documented one-liner for this image passes only JWT_SECRET, so every fresh single-container install handed out relative links in API responses, QR codes and emails. Defaulted to the same value compose uses; -e FRONTEND_URL=https://... overrides it, as does the site URL field in Settings. Found by pointing tests/e2e/local at a running AIO container: auth/06-api-tokens asserts share_url matches /^https?:\/\//, and it was the one spec that failed for a product reason rather than a harness one. It passes now, and the suite is 19/20 against the image — the remaining failure is smoke/02-auth-flow, whose seed helper shells out to a hard-coded `docker exec picpeak-backend`, so it cannot arrange its precondition against any other container. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * feat(aio): mark the image so face recognition stays off (#1042, #1074) Face recognition needs a separate ML container this image does not contain, and enabling it here would add a second image-processing pipeline competing with Sharp for the CPU and memory of a container sized for one photographer plus guests browsing. The failure mode would not be a clear error — just a slow install that looks broken. The backend gate for this lands in #1075 and keys on PICPEAK_SINGLE_CONTAINER. Without this line the guard never triggers on an actual all-in-one build, so the two changes have to arrive together: whichever merges second completes the pair. Verified against this file's exact value — isFeatureEnabled() returns false with it set. An explicit marker rather than inferring from SERVE_FRONTEND or the SQLite path, because legitimate multi-container deployments do both of those and should keep the feature. Also adds it to the Limits section of docs/single-container.md, next to the SQLite and Redis constraints, since that is where someone will look before choosing this image. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local> Co-authored-by: the-luap <paul-nothaft@hotmail.de>	2026-08-18 22:15:49 +02:00
Paul Nothaft	f22999aba6	fix(storage): write business documents under STORAGE_PATH, not the cwd (#1070) ... * fix(storage): write business documents under STORAGE_PATH, not the cwd persistDocPdf, the invoice sending and reminder writers and both contract signature writers built their target from `path.join(process.cwd(), 'storage', 'business-docs', ...)` and never consulted STORAGE_PATH. docker-compose.yml and docker-compose.production.yml both pin STORAGE_PATH=/app/storage and the image's WORKDIR is /app, so on a stock deployment the two expressions name the same directory and nothing looked wrong. Point STORAGE_PATH anywhere else and quotes, invoices, Mahnungen and contract PDFs land outside the configured storage root: missed by the backup walker, invisible to the storage accounting, and gone when the container is replaced. It also fails outright where the working directory is not writable by the runtime user. Routed all six writers through getStoragePath(), the resolver the rest of the app already uses. Two read-side sites of the same class came along: the custom PDF font lookup now checks the storage root before the legacy cwd path (a font under STORAGE_PATH/fonts was simply never found, and the document silently fell back to the built-in face), and the dev-test scratch directory follows the same root. Left alone deliberately: resolveLogoFile and adminBusinessProfile already try both roots, so their cwd reference is a legacy fallback rather than a miss. No migration needed — the persisted path is stored absolute, so rows written before this keep resolving to where those files actually are. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(storage): allow the configured contract root, and move signature images too Two holes in the previous commit, both found by review. Contract downloads would have broken. assertContractPdfPath() guards the admin unsigned/signed PDF routes and GET /api/public/contracts/:token/pdf, and it listed only <cwd>/storage/business-docs/contract. Moving the writers to STORAGE_PATH without moving that root meant every newly generated contract was refused with PATH_OUTSIDE_STORAGE — a worse failure than the bug being fixed, and only on the installs the fix was for. The configured root is now allowed alongside the cwd one, which stays for contracts written before the move; their absolute paths are in the database and still resolve. Note the sibling root on the next line already honoured STORAGE_PATH, so the helper was half-migrated already. persistSignatureImage() still wrote customer and admin signature PNGs under process.cwd(). It was missed because its path.join is spread over seven lines while the others are single-line — and the regression test compared against the single-line literal, so it reported green over a live bug. The test now collapses whitespace before matching, which is the only reason a formatting difference ever hid this. A sweep of the whole of src/ with the same normalisation confirms the remaining process.cwd()/storage references are all deliberate `STORAGE_PATH || cwd` fallbacks, not misses. Added a case that drives assertContractPdfPath against real files on disk — the guard realpaths both the file and its roots, so a test using imaginary paths proves nothing. It fails without the fix. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(storage): resolve the contract guard's root through the shared resolver The guard still built its own `STORAGE_PATH || <cwd>/storage`. That matches getStoragePath() only while STORAGE_PATH is set — with it unset the shared resolver falls back module-relative to <repo>/storage while this fell back to <cwd>/storage, and the backend is normally started from backend/, so the two name different directories. Writers and guard then disagreed about where contracts live and the download routes refused them, which is the same failure the previous commit fixed for the configured case, reappearing in the fallback case. One resolver on both sides now, which is the point of the whole change. Docblock updated to describe the three roots as they actually are. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc * fix(storage): make the fallback test safe, and align the backup diagnostics The test added in the previous commit was dangerous. To exercise the STORAGE_PATH-unset case it deleted process.env.STORAGE_PATH and then, in cleanup, recursively removed `<resolved root>/business-docs` — which with the variable unset resolves to the developer's real, gitignored <repo>/storage. Running `npm test` in a working checkout would have destroyed local business documents. This checkout has 65 MB there, including a populated business-docs tree. Rewritten to mock the shared resolver instead. That is both safe (every path stays in the tmpdir) and a sharper assertion: if the guard consumes getStoragePath() the mock moves its root, and if it went back to rolling its own expression the mock would have no effect and the test fails — which is exactly the regression being pinned. backupCoverageService and backupIntegrityService kept their own `STORAGE_PATH || cwd` roots. The backup walker itself already falls back module-relative, so with the variable unset the two diagnostics inspected a directory neither the walker nor the writers use and would report the business-docs tree as missing while it was in fact being backed up. Both now use the shared resolver. No regression: the same jest invocation over contract/quote/invoice/pdf/ backup suites gives an identical 11 failed, 24 passed before and after — those failures are a locally missing cron-parser dependency and reproduce on an unmodified tree. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-18 22:14:41 +02:00
Paul Nothaft	44adabca9e	test(e2e): read the admin JWT from the cookie, not the login body (#1071) ... Three specs acquire an admin token with `const body = await res.json(); return body.token`. The admin login has not returned a token in its body for some time — establishAdminSession() sets the JWT as the httpOnly `admin_token` cookie and responds with `res.json({ user })` — so the token was undefined and every one of them failed at the first assertion, before exercising anything they were written to cover. Server-side the cookie and an Authorization: Bearer header are interchangeable (see middleware/gallery.js, which reads the cookie first and accepts an admin-typed Bearer second), so the fix is to read the value back out of the context cookie jar and keep threading it as a Bearer. Every downstream call in these specs stays exactly as it was. Measured against a real stack, running only these three files: before 0 passed, 6 failed — all six at the token assertion after 3 passed, 3 failed The three that still fail no longer fail on auth: they get deep into the flow and then miss UI that has since changed (a settings label, a locator that no longer resolves). That is a separate and much larger staleness problem across this directory — a full run is 12 passed against roughly two dozen failures of that kind — and it is not addressed here. Worth knowing: no CI workflow runs tests/e2e at all, which is why this rotted silently while `npm run test:e2e` stayed documented in CLAUDE.md. Wiring it up is the obvious follow-up, but it has to wait until the suite is actually green, or it would just pin main red. Claude-Session: https://claude.ai/code/session_01Ra4hcsYiKuQLbbRsg6EjAc Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-18 22:14:00 +02:00
Paul Nothaft	d62e21c1ba	chore(main): release 3.105.1-beta.0 (#1066) ml-models-v1 v3.105.1-beta.0	2026-08-16 19:41:03 +00:00
Paul Nothaft	52db982661	fix(gallery): make per-event banner overrides actually work, both banners (#440, #932) (#1064) ... * fix(gallery): make per-event banner overrides actually work, both banners (#440, #932) The promo banner shipped with a per-event inherit/custom/off override that never reached a guest. GalleryView reads promo_mode from the /photos payload, and /photos never sent it — so every gallery resolved to 'inherit'. Setting a gallery's promo banner to "Off" did nothing; the global banner kept rendering. The info banner (#932) mirrored that shape and inherited the same gaps. Four places dropped the fields; all four now carry both banners: 1. GET /gallery/:slug/photos — send promo_mode/promo_markdown alongside the info fields. This is the fix that makes "Off" mean off. 2. POST /admin/events — the validators accepted both banners and the insert discarded them, so an API client could POST info_mode:'off', get 201, and find the row on 'inherit'. Markdown is stored only for 'custom', matching the PUT rule. 3. POST /admin/events/:id/duplicate — copy both from the source row. The dialog promises the copy "inherits the branding, behaviour, feedback, and category configuration"; a muted gallery un-muting on duplication is the opposite of that. 4. PUT /admin/events/:id — resolve the effective mode from the STORED row when a partial update sends only the markdown. Previously updates.promo_mode was undefined on such a request and the text was parked on an inherit/off gallery, then resurfaced when someone later switched it to 'custom'. The lookup is lazy: one extra query, only on that path. The two normalisation blocks are now one loop over both banners, so the pair can't drift apart again. Verified in a browser, both directions against the same global banner: promo_mode='off' -> not rendered; 'inherit' -> rendered. The /photos payload went from promo_mode ABSENT to carrying the value. * fix(gallery): thread promo into the reveal view, drop stale markdown on duplicate External review, round 1 on this PR. Two gaps in the plumbing it introduced: - The reveal-hidden branch copied only the info fields from /photos. Now that /photos carries promo too, a reveal-hidden gallery with promo_mode 'off' still fell back to 'inherit' and showed the global banner on the first load after login. Thread both banners there. - The duplicate copied markdown verbatim. A row written before the PUT normalisation landed can hold text while its mode is 'inherit'/'off', so the copy inherited hidden text that would resurface the moment someone switched it to 'custom' — violating the very invariant this PR establishes. Copy markdown only when the source mode is 'custom'. Test covers the stale-markdown source explicitly. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-16 21:37:17 +02:00
Paul Nothaft	180f19d70d	chore(main): release 3.105.0-beta.0 (#1065) v3.105.0-beta.0	2026-08-16 19:36:29 +00:00
Paul Nothaft	b48fa62eea	feat(gallery): info banner above the photo grid (#932) (#1063) ... * feat(gallery): info banner above the photo grid (#932) A short informational note rendered at the TOP of a gallery, above the photos. Distinct from the promotional banner (#440), which stays by the footer for marketing copy — the reporter's case is an onboarding hint ("use the menu button to filter"), which is useless below a gallery the guest has to scroll past first. Mirrors the promo feature's shape rather than inventing a second one: a global default in Settings → Branding (branding_info_markdown) plus a per-event inherit/custom/off override. Markdown via the existing MarkdownContent sanitiser — no raw HTML, no CSS injection. Empty global default means nothing renders, so upgrading changes nothing visible. Deliberately NOT included: an alignment knob (this is short helper copy, not marketing layout) and guest dismissal — the issue lists dismissal as a nice-to-have, and it needs per-guest persistence that is its own decision. Migration 176 is idempotent (hasColumn / existing-key guarded). Note on the payload plumbing: the per-event fields travel in the /photos response, not just /info. GalleryAuthContext seeds its cached event from the gallery LOGIN response — a small identity subset — so anything absent there is undefined right after a guest signs in. /photos is the payload that refreshes on every gallery load, which is why the fields were added there and why GalleryView reads them from `data.event`. Verified in a browser across all three modes; reading them from the context event instead silently collapsed every override back to 'inherit'. * fix(branding): map branding_info_markdown on read so saving can't wipe it (#932) External review caught this. BrandingSettings declared no info_markdown and formatBrandingSettings never mapped branding_info_markdown, so BrandingPage's hydration — setBrandingSettings(prev => ({ ...prev, ...formatted })) — kept the empty-string initializer instead of the persisted value. The form loaded blank and the next Save posted '' back, wiping a configured banner. Silently: the gallery keeps rendering the old copy until that save lands. This is the same bug the footer/promo fields hit in #441 + #440 / #460, which the read mapper still carries a comment about. Add the field to the interface and the mapper, and pin the round-trip for the whole editable branding set so the next field added is caught by a test rather than by a user losing copy. Verified: the new test fails 3/4 with the mapper line removed. * fix(gallery): honour the info-banner override in the reveal-hidden view (#932) External review, round 2. The hidden-until-reveal branch renders GalleryLayout with the context `event`, which is seeded from the gallery login response and carries no banner fields — so while a gallery was hidden, a per-event 'off' silently resolved to 'inherit' and the global banner appeared on a gallery the admin had muted. Resolve the fields there the same way the main render path does. The two full-page layouts (gallery-premium, gallery-story) are deliberately left alone: they return before GalleryLayout and render no header, footer or promo banner either — injecting a wrapper into layouts documented as having 'their own integrated UI' would be a design change, not a fix. --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-16 21:33:13 +02:00
Paul Nothaft	ab6ca6f82f	chore(main): release 3.104.1-beta.0 (#1061) v3.104.1-beta.0	2026-08-16 17:34:09 +00:00
Paul Nothaft	3a11e6ebb5	fix(pdf): RFC 6266-encode Content-Disposition on quote/invoice PDFs (#1024) (#1055) ... * fix(pdf): RFC 6266-encode Content-Disposition on quote/invoice PDFs (#1024) The six quote/invoice PDF endpoints interpolated buildPdfFilename()'s result straight into `inline; filename="${filename}"`. That result deliberately preserves non-ASCII (it doubles as the PDF's internal Title metadata), and HTTP header values are latin1 — so a customer label reaching the header directly failed in one of two ways: - U+0080-U+00FF (ä ö ü ß — every German umlaut): no throw. The raw byte goes out and the client reads back a mangled name. Silent corruption. - above U+00FF (Polish ł, Czech ř, Turkish ş, €, Cyrillic, CJK, emoji): Node's setHeader rejects it with ERR_INVALID_CHAR. The throw lands after the PDF buffer is already rendered, so the request 500s. Note this corrects the issue's diagnosis: it reported umlauts as the 500 case, but umlauts are inside latin1 and mangle rather than throw. Both symptoms share this root cause and both are fixed here. Route through buildContentDisposition() (utils/filenameSanitizer, already used by secureImages.js), which emits an ASCII fallback plus the RFC 5987 `filename*=UTF-8''…` form, so the unicode name survives in browsers and the header stays legal. Applied to all six sites: adminQuotes (persisted + preview), adminInvoices (persisted + preview), customer (quote + invoice). Also correct buildPdfFilename's docstring, which advertised the preserved non-ASCII as suitable for Content-Disposition — the exact misreading that produced these call sites. * test(pdf): pin the ASCII fallback for fully non-Latin customer names (#1024) A name written entirely in another script leaves the legacy filename= token with just the document number (Q-2026-0042_.pdf) — filename* carries the real name. That's the intended trade, but it's the token a client without RFC 5987 support actually saves, so assert it stays legal, non-empty and carries the document number rather than leaving it unpinned. * fix(pdf): don't split surrogate pairs when truncating the filename (#1024) Codex review caught this. sanitiseSegment caps each segment at 80 UTF-16 code units, so a cap landing inside an astral character (emoji, rarer CJK) left a dangling high surrogate. encodeURIComponent throws URIError: URI malformed on a lone surrogate, so buildContentDisposition — the helper this PR routes the six PDF endpoints through — 500'd for e.g. company_name = 'a'.repeat(79)+'🎉', well inside the 120-char validator limit. Same 500 the PR set out to remove, reached a different way. Drop the orphaned surrogate instead of widening the cap, so the byte budget the limit exists to protect is unchanged. Tests cover both boundary cases and assert the cap semantics; they fail against the previous slice(). --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-16 19:31:19 +02:00
Paul Nothaft	959c18a864	chore(main): release 3.104.0-beta.0 (#1057) v3.104.0-beta.0	2026-08-16 17:04:35 +00:00
Luca	8809564aad	feat(backup): open sqlite → pg .picpeak restore as the supported upgrade direction (#1041) (#1043) ... Reshaped onto main after #1039 landed the coercion engine (typedColumnsFor / epochToIso / coerceForTargetEngine) — this PR is now only the policy delta on top of it: - validateManifest: replace the CLI-only allowEngineSwitch opt-in with a direction rule — sqlite → pg allowed (upload UI and CLI alike), pg → sqlite refused with a message naming the supported direction - importFromPicpeak: derive crossEngine from the manifest's engine (absent field = target engine, the exact pre-change behavior), log it, return it; route passes it through - scripts/migrate-sqlite-to-postgres.js: rely on the shared gate, drop the flag - restore card: direction stated in the intro, cross-engine notice after a converting restore; both strings in en.json + de.json; removed the orphaned settings.backup.picpeak locale node (unreferenced, stale copy) - picpeakCrossEngine.test.js: direction policy, epochToIso (ms, seconds, numeric strings), coerceForTargetEngine units, plus PICPEAK_PG_TEST_URL-gated real-Postgres stored-value assertions Co-authored-by: Paul Nothaft <53005142+the-luap@users.noreply.github.com>	2026-08-16 19:01:36 +02:00
Paul Nothaft	18b1e0f66e	ci(tests): run the gated real-Postgres .picpeak cases in the backend job (#1056) ... The .picpeak restore suites gate their Postgres cases behind PICPEAK_PG_TEST_URL and describe.skip themselves out when it is unset. That variable was set in no workflow, so those cases have never run in CI — the suites reported green while silently skipping the half that needs a real database: sequence resync, operator/role preservation across a cross-instance restore, and whether a coerced row lands with the right STORED VALUES rather than merely not throwing. Add a postgres:15-alpine service to the backend job (same shape schema-drift already uses) and point the variable at it. Everything else in the suite still runs on SQLite; this only un-gates the cases that were skipping. Verified against a real Postgres 15 before wiring: picpeakRestorePg 4/4 and picpeakCrossEngine 11/11 (8 of which were previously skipped across both). Matters now because #1043 opens sqlite -> pg restore to the upload UI, so the coercion layer's correctness stops being a CLI-only concern. Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-16 18:57:44 +02:00
Paul Nothaft	77cb65f5a2	chore(main): release 3.103.1-beta.0 (#1053) v3.103.1-beta.0	2026-08-16 13:17:54 +00:00
peipeimo	3600231d5f	fix(storage): add S3 client timeouts so a dropped connection can't wedge uploads (#1049) ... * fix(storage): add S3 client timeouts so a dropped connection can't wedge uploads * fix(storage): use socketTimeout, not requestTimeout, for the dead-connection guard * fix(storage): make S3 timeouts generous — short connectionTimeout breaks pooled reads --------- Co-authored-by: Peifu Mo <peipeimo@Peifus-MacBook-Pro.local>	2026-08-16 15:14:48 +02:00
Paul Nothaft	2728479d2a	chore(main): release 3.103.0-beta.0 (#1052) v3.103.0-beta.0	2026-08-16 13:02:55 +00:00
Luca	b118695474	feat(permissions): granular permission gating + role editor & presets (#747, phase 1 of #743) (#1045) ... * feat(permissions): granular permission gating + role editor & presets Make every admin feature permission-gateable so multi-user studios can split capability across roles (#747, and phase 1 of #743). - Split the catch-all settings.edit into dedicated dangerous-config perms (banking / domains / security / integrations / features): a team member can no longer change IBAN, domains, SSO, webhooks, API tokens or feature flags. Reads keep an OR with settings.view so existing roles keep visibility. The site-URL write inside /general is change-gated on settings.domains. - Add dedicated perms for admin surfaces miscategorised under settings.* (whatsapp, event_types, image_security, notifications, system) plus roles.manage and vat_codes.view; gate the previously-ungated VAT read. - Boot self-heal (_permissionsBoot.js): super_admin always holds every permission (tracks-all) so new perms never need a compensation migration; all other roles stay frozen (no silent escalation on upgrade). - Seed two presets: Solo Photographer (full operator) and Team Photographer (contributor — view events + manage photos + read-only CRM; no settings/users/billing edits, no events.edit). - Role editor: adminRoles CRUD (create/edit/clone/delete + permission matrix; system roles protected, super_admin immutable) and a Roles tab with a category-grouped matrix and preset cloning. - Settings page tabs are permission-gated with snap-back; i18n en/de. Migration 174. Backward-compatible: admin/editor/viewer unchanged. * feat(permissions): hide in-page action buttons a role can't use Wrap mutating controls on the surfaces restricted roles actually reach (Events list, Archives, gallery photo grid, event detail) in PermissionGate so they are HIDDEN when the user lacks the permission, rather than shown-then-403: - Events list: create / bulk archive / bulk delete / row archive / row delete / download-archive. - Archives: restore / download / delete. - Photo grid: single + bulk delete (photos.delete), per-photo download (photos.download), bulk move/hide/show (photos.edit). - Event detail: edit / rename / publish (events.edit), duplicate (events.create), archive (events.archive), create-invoice (bills.manage); the Actions card is hidden entirely for view-only roles. - Photos tab: upload / external import (photos.upload), export menu (photos.download). Backend already enforces these with 403; this is the matching UX so a Team Photographer never sees delete/settings controls. * fix(permissions): close settings-split bypass via generic settings writers Security review found the settings.edit split was bypassable: the generic settings writers (/general, /analytics, /seo, /security) upsert arbitrary setting_keys, so a role holding only settings.edit (or settings.security) could write keys owned by a narrower permission — repointing the public site URL (settings.domains), security policy (settings.security) or VAT/accounting config (settings.banking) via the wrong endpoint. Add stripUnauthorizedProtectedKeys(): before every generic upsert, drop any protected key the caller isn't permitted to write (general_site_url → settings.domains, security_* → settings.security, accounting_* → settings.banking). Dedicated routes still work because their caller holds the matching perm. Replaces the narrower in-handler site-URL guard. Also fix two tests affected by the RBAC changes: - authzPermissionGaps: API-token management moved to settings.integrations, so grant that (not settings.edit) to exercise the ownership 404. - AdminPhotoGrid.viewToggle: stub PermissionGate (its buttons are now gated and the test renders without a PermissionsProvider). * fix(permissions): address upstream review (#1045) - Renumber migration 174 -> 175 (174 now taken by 174_sqlite_nullable_event_dates from #1035; the collision made picpeakImportService's forward-only restore guard treat both as order 174 and accept a newer .picpeak onto an older schema). - Contain the roles.manage blast radius (delegation, not root escalation): a non-super_admin can no longer edit their own role, nor grant any permission their own role doesn't already hold (createRole + updateRole). - Protected-key denial now 403s (naming the keys + required perms) instead of silently stripping and reporting "saved" (adminSettings generic writers). - Reserve team_photographer so a custom role can't squat the preset name. - Boot self-heal: per-step try/catch so a role_permissions insert race on one replica doesn't skip preset seeding. - Forward-project the feature .manage perms that also replaced settings.edit gates (whatsapp/event_types/image_security/notifications/system), matching the settings.* split projection so the pattern is symmetric for phase-2. - Guard exports.down's roles/admin_users queries with hasTable. * fix(permissions): change-detection on protected-key 403 + commit guard tests (#1045) Round-2 review: - The protected-key 403 fired on key PRESENCE. The General tab re-posts general_site_url on every save, so a settings.edit-only role (the office manager this PR enables) got 403'd on every General save even when the URL was unchanged. Restore change-detection: compare the incoming value against the stored one and 403 only on an actual change; unchanged protected keys are dropped so the rest of the save proceeds. Only /general is affected. - Commit the self-amplification guard test (was run locally, never staged): adminRolesGuards.test.js — non-super can't grant perms it lacks, can't edit its own role, can't escalate another role; super_admin bypasses; team_photographer name reserved. - Add adminSettingsProtectedKeys.test.js pinning the change-detection: an unchanged general_site_url saves, an actual change 403s, super_admin changes it.	2026-08-16 14:59:52 +02:00
Paul Nothaft	9b976386f9	chore(main): release 3.102.2-beta.0 (#1046) v3.102.2-beta.0	2026-08-13 16:57:48 +00:00
Paul Nothaft	6de30e5bf1	fix(docker): default NODE_ENV=production so non-compose deploys don't fall back to SQLite (#1038) (#1039) ... * fix(docker): default NODE_ENV=production so non-compose deploys don't fall back to SQLite (#1038) knexfile.js selects its config block by NODE_ENV and the `development` block defaults to sqlite3. The image never set NODE_ENV, so every deployment that doesn't go through our compose files — Kubernetes, Helm, plain `docker run` — silently ran on SQLite and ignored DB_HOST/DB_USER/DB_PASSWORD. It stayed invisible because wait-for-db.sh is shell: it reads DB_HOST directly, connects to Postgres, creates the database and logs "PostgreSQL is up" in the same container where the Node process then writes to a SQLite file. Migrations go through src/database/db.js → the same knexfile, so they also ran against SQLite, leaving the provisioned Postgres database empty. Setting the default alone would be unsafe: an affected install would flip to Postgres on its next image pull and come up against an EMPTY database, which reads as total data loss. So this adds a guard that runs before migrations touch anything: - logs the resolved engine + target at boot (nothing did before, which is why this went unnoticed for so long) - refuses to start when pointed at a virgin Postgres while a populated SQLite file exists, naming the file and the .picpeak export path for moving the data, with PICPEAK_ALLOW_EMPTY_PG=true as the escape hatch - warns but boots when Postgres settings are present yet SQLite is in use Compose files already set NODE_ENV explicitly, so compose users are unaffected. The engine-selection tests resolve knexfile in a child process with a clean cwd — dotenv.config() would otherwise let a developer's backend/.env decide the answer instead of the knexfile defaults under test. Fake credentials in the describeEngine tests are built at runtime rather than written inline, so secret scanners don't flag a literal after `password:`. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): stay on SQLite instead of blocking, and add a migration path (#1038) Reworks the guard from the previous commit after walking through what an existing install actually experiences on its next image pull. Blocking was the wrong trade. An operator who had unknowingly been running on SQLite (because the image left NODE_ENV unset) would have pulled the fix and got a CrashLoopBackOff: data safe, galleries offline, for something they did not do. Now the boot RESOLVES the engine before migrations run and stays on whichever one holds the data: - Postgres configured but holding no galleries, while a populated SQLite file exists → keep serving from SQLite, print what happened and how to migrate. Nothing moves until the operator decides. - once Postgres holds the data, the next restart switches over on its own. - an explicit DATABASE_CLIENT is always honoured. The check is keyed on Postgres holding DATA, not on it having tables: a stray `run-migrations` against the empty database creates every table, which would otherwise blind the check and strand the operator on an empty install. wait-for-db.sh resolves the engine and exports DATABASE_CLIENT before the migration step, so the runner and the server always agree. Manual migration runs (no entrypoint, no exported client) now refuse rather than build a schema in the wrong database. Adds scripts/migrate-sqlite-to-postgres.js for moving the data across. It reuses the .picpeak export/import services rather than hand-rolling a cross-engine copy — they already handle FK suspension, JSON columns and Postgres sequence resync. Two things had to be added for the SQLite → Postgres direction, both opt-in and CLI-only so the upload/restore UI is untouched: - `allowEngineSwitch` relaxes the importer's same-engine guard - cross-engine row coercion: SQLite has no real date or boolean types, so its rows carry epoch numbers where Postgres wants a timestamp and 0/1 where it wants a boolean. Postgres rejects both outright ("date/time field value out of range: 1786548038763"). Coercion is driven by the TARGET schema, never guessed from the value. Verified end to end against a real PostgreSQL 15: a seeded SQLite install migrated across with booleans, timestamps and foreign keys intact, and the serial sequences correctly advanced (the next INSERT got id 2, not a primary-key collision). Photo files on disk are never touched and the SQLite file is left in place as a rollback. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close four review findings on the SQLite fallback + migration (#1038) External review (codex) found four issues, all confirmed against the code and fixed here. Two of them could have cost data. 1. The engine resolver was reachable only through wait-for-db.sh. A Kubernetes manifest that sets `command`/`args`, or a plain `docker run … node server.js`, bypasses the entrypoint — exactly the deployment styles this fix targets. With NODE_ENV now baked into the image, such an install would have resolved to Postgres and come up against an empty database while its SQLite data sat there unseen. server.js now resolves the engine itself, before anything requires knexfile, via the same script the entrypoint uses. Verified by running `node server.js` directly against an install with stranded SQLite data: it logs the banner and serves SQLite. 2. Cross-engine loads double-encoded JSON. SQLite has no json type, so its json columns are TEXT holding JSON; the export dumps that as a string and serialiseJsonColumns stringified it again, storing `true` as the scalar string "true". app_settings.setting_value is json on every install, so this reshaped every migrated setting. The text is decoded before serialisation now — verified against a real Postgres: json_typeof(setting_value) is `boolean`, matching a native install exactly. 3. The migration could silently miss concurrent writes. If the backend keeps serving, rows written after the export never reach Postgres and vanish from view once the engine switches. The script now fingerprints the SQLite tables whose loss would be noticed, checks for drift BEFORE loading Postgres (so a detected race leaves the target untouched) and again after, and refuses with the exact rows that moved. It also says plainly to stop the backend first. 4. The child phases shared stdout with winston. Outside production, and whenever LOG_TO_CONSOLE=true, createPicpeak's own log line was concatenated with the archive path and the migration failed on a bogus filename. Payloads travel through a result file now; verified with LOG_TO_CONSOLE=true. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 2 — six more data-safety findings (#1038) 1. The engine choice is now PINNED once the data is in Postgres. Previously the boot decided from "does Postgres hold galleries", so an operator who later deleted every gallery would be sent back to the stale pre-migration SQLite file while their settings, admins and CRM data stayed in Postgres. The migration writes a marker next to the database file (and retires the file itself by renaming it); the marker wins over any probe. 2. The migration refused to overwrite Postgres only when it held GALLERIES. A target with admins, customers, invoices or projects but no galleries was wiped without --force. Both the source and target checks now look for user data across the tables that are empty on a fresh install. 3. Same bug in the other direction: an install with no galleries but real admins/settings/customers was refused a migration it was entitled to. 4. Drift detection covered four tables and only count/max(id), so an in-place UPDATE (event edit, password change) or a write to any other table passed unnoticed. It now fingerprints every table the export carries, including max(updated_at). It still is not a substitute for stopping the backend, and the script says so rather than implying a guarantee. 5. probeSqliteData() treated an unreadable or corrupt file as "no data", which would have switched the install to an empty Postgres — the very failure this module exists to prevent. It fails closed now and stays on SQLite so the real error surfaces. 6. The "you are leaving SQLite data behind" warning was unreachable: setting DATABASE_CLIENT skipped the probes, so the branch that produces it never had the inputs. Postgres and SQLite are both probed whenever Postgres is the engine in play. Also: the final verification compares row counts for EVERY table rather than just galleries, and flags only a shortfall — the import legitimately adds an app_settings row (setSessionsValidAfter) that made the strict equality fail on a first real run. Verified against a real PostgreSQL 15 end to end, including: the marker keeps an install on Postgres after every gallery is deleted; removing the marker and restoring the file rolls back to SQLite as documented. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 3 — occupancy, bootstrap admin, secrets in /tmp (#1038) 1. Both engine probes judged occupancy by GALLERIES alone. An install whose galleries were all deleted, but which still has admins, customers or accounting records, was treated as empty: on the SQLite side that meant booting the empty Postgres and appearing to lose everything; on the Postgres side it meant diverting a live install to a stale SQLite file. Both now look across the tables that are empty on a fresh install, matching the migration script. 2. The migration ran migrate-schema BEFORE checking the target, and migration 001 seeds a bootstrap admin when ADMIN_PASSWORD is set (common on legacy installs). The occupancy check then saw that admin and refused, pushing the operator towards --force against a genuinely empty database. The target is read first now. 3. probeSqliteData()'s warning went through the app logger, which writes to STDOUT when LOG_TO_CONSOLE=true — and the resolver's stdout is the protocol channel wait-for-db.sh captures, so DATABASE_CLIENT could have been set to a JSON log line. Diagnostics take an injected sink (stderr in the resolver), and the shell now validates the value it captured instead of trusting it. 4. The .picpeak archive holds password hashes, SMTP credentials and API keys in plaintext, and was only removed on the fully-successful path — any drift or import failure left it in /tmp. Every exit path removes it now. 5. A database-only migration still hauled every business-doc and upload through /tmp and back into the same volume. createPicpeak takes includeFiles:false for this path; rows move, files stay where they already are. Verified against a real PostgreSQL 15: a gallery-less install with only an admin account now stays on SQLite and migrates successfully with ADMIN_PASSWORD set; the resolver emits exactly one token on stdout with LOG_TO_CONSOLE=true and a corrupt database; a drift failure leaves Postgres untouched and no archive behind. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): pin the boot to SQLite while a migration is unfinished (#1038) Review round 4. A migration that dies after touching Postgres leaves rows behind — schema creation alone seeds a bootstrap admin when ADMIN_PASSWORD is set, and a drift or row-count failure can leave a partial load. Since the occupancy probes were widened in round 3, those rows read as "Postgres is occupied", so the next restart would switch engines and hide the SQLite data that is still the database of record. The script now writes a pin file next to the database BEFORE its first Postgres write and clears it only on success (after the success marker exists, so no restart in between can pick the wrong engine). While the pin is present the resolver stays on SQLite and explains why. Verified against a real PostgreSQL 15 by reproducing the exact scenario: a migration failed mid-run with ADMIN_PASSWORD set, leaving one bootstrap admin in Postgres. With the pin the next boot resolves to sqlite3; with the pin removed it resolves to pg — the failure this closes. The subsequent successful re-run clears the pin and the boot moves to Postgres. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 5 — occupancy, path drift, retry, host default (#1038) 1. A seeded bootstrap admin counted as "Postgres is occupied". core/001_init.js inserts one whenever ADMIN_PASSWORD is set, so a Postgres that was initialised once and never used would have beaten a SQLite file full of real galleries — the exact failure the guard exists to prevent, reintroduced by widening the probe in round 3. The two sides are deliberately asymmetric now: the SQLite probe counts any user data (err towards keeping data visible), the Postgres probe ignores rows that schema creation seeds (err towards requiring proof of real use). 2. The guard resolved DATABASE_PATH with its own logic while knexfile trimmed whitespace and collapsed the legacy duplicated-backend form. A path either engine normalised differently meant probing a file nobody uses, concluding there was no SQLite data, and booting an empty Postgres. The resolution now lives in one module both require. 3. Re-running after a partial migration — the documented recovery — was refused unless the operator passed the destructive-sounding --force, because the half-written rows read as target data. An unfinished run of this same script is now recognised as a safe retry. 4. wait-for-db.sh verified readiness against its own default host (`postgres`) while knexfile's production block defaults to `db`. With NODE_ENV now baked in, a bare `docker run` without DB_HOST would have passed the readiness check against one host and then dialled another. The entrypoint exports the exact connection it verified. Compose sets DB_HOST explicitly and is unaffected. Verified: a Postgres holding only a seeded admin now loses to real SQLite data; a DATABASE_PATH with surrounding whitespace resolves to the identical file in both knexfile and the guard. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 6 — explicit-client bypasses, retry scope, cleanup (#1038) 1. An explicit DATABASE_CLIENT bypassed the unfinished-migration pin, because decideBootEngine honoured it first. docker-compose sets DATABASE_CLIENT=pg, so a failed migration would have restarted on a half-written Postgres on exactly the deployments that pin it. Worse in the other direction: with DATABASE_CLIENT=sqlite3, a SUCCESSFUL migration renames the source file, so the next start created a NEW, empty SQLite database and served that. The pin now outranks explicit pg (clearing the marker is the override), explicit sqlite3 is left alone since it already points at the data, and the migration refuses up front when the deployment pins anything other than pg. 2. The retry allowance was bound to the SQLite file, not to the target. An operator who repointed DB_HOST/DB_NAME between attempts could have replaced an unrelated populated database without --force. The pin records the target and the allowance only applies when it matches. 3. The printed rollback did not roll back: with data on both sides and no marker, the resolver still selects Postgres. It now spells out all three steps, including DATABASE_CLIENT=sqlite3. 4. A failure inside createPicpeak left a partial archive — plaintext hashes and credentials — in the caller-supplied temp dir, which that service deliberately does not clean. The export phase removes it on error. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 7 — pin bypass on direct start, real admins (#1038) 1. server.js only ran the engine resolver when DATABASE_CLIENT was unset, so a deployment that both bypasses the entrypoint (Kubernetes `command:`) AND pins DATABASE_CLIENT=pg never consulted the migration pin — the round-6 fix was unreachable on exactly that path, and a failed migration would have served a half-populated Postgres. The resolver now also runs whenever a pin file exists. 2. Round 5 excluded admin_users from Postgres occupancy to stop a seeded bootstrap admin counting as real data. That over-corrected: an install that has completed first-run setup but has no galleries yet has exactly one user-created row — an admin — so Postgres looked empty and, with a stale SQLite file present, the boot would switch away and the admin's credentials and configuration would disappear. core/001_init.js seeds must_change_password=true; setupService writes false once a human completes setup. The FLAG, not the table, distinguishes them, and a legacy NULL counts as a real admin. Verified against a real PostgreSQL 15: a Postgres holding only the seeded row loses to real SQLite data, the same Postgres wins once setup is completed, and a server started directly with DATABASE_CLIENT=pg and a pin present comes up on SQLite with the warning. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 8 — reset admins, CLI config, JSON nulls (#1038) 1. must_change_password is mutable: resetAdminPassword() re-raises it on REAL accounts (userManagementService.js:474). Round 7's discriminator therefore read a gallery-less Postgres whose only admin had been reset as an untouched bootstrap seed — and with a stale SQLite file present, the boot would have switched away and hidden those live credentials. The rule is layered now: more than one admin, any admin that has logged in, or must_change_password false all count as use. Only core/001_init.js's exact leftovers — one admin, never logged in, still flagged — read as a seed. 2. The CLI read process.env directly but never loaded the configuration the child phases get through knexfile, so running it directly (or via `docker exec`, which does not inherit wait-for-db.sh's exports) failed the pre-flight checks even with valid settings in backend/.env or /run/secrets/db_password. Both sources are loaded up front now. 3. The migration's target check counted a seeded bootstrap admin as user data while probePgData classified the identical row as empty, so migrating into a previously-initialised-but-unused Postgres demanded --force. Same rule on both sides. 4. Cross-engine JSON handling is simpler and no longer lossy. SQLite keeps json columns as TEXT holding valid JSON and Postgres accepts JSON text directly, so the correct action is to pass them through untouched. Round 1 parsed then re-serialised them to undo a double-stringify; that round-tripped the JSON literal `null` into SQL NULL, changing data and breaking NOT NULL json columns. Not serialising at all fixes both. Verified against a real PostgreSQL 15: a migrated install now carries json_typeof = null for a JSON null, object for a nested object, and boolean for a boolean — matching a native install exactly. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): close review round 9 — probe error classes, marker ordering (#1038) 1. probePgData() answered every failure with "Postgres has data". That is right for an unreachable server — the app cannot run on it either way, and diverting a healthy pg install to a stale SQLite file over a transient blip would be worse — but wrong for a server that answers and then fails the query, which is what a half-built or damaged schema looks like. That is not evidence of data, and reporting it as such booted the empty Postgres and hid a populated SQLite file: the exact failure this guard exists to prevent. Reachability is now established with SELECT 1 first, so the two cases get opposite answers: unreachable → leave the configured engine alone; reachable-but-uninspectable → unproven, and the SQLite side wins if it actually holds data. 2. The success marker was written after the SQLite file was renamed away. A failure in between — a full disk — left the source retired with no marker: the next attempt reported "No SQLite database", the in-progress pin stayed, and the operator never saw the rollback path. The marker is written first and updated with the retired filename once the rename succeeds, so a failure at any point leaves everything recoverable. Verified against a real PostgreSQL 15: a reachable database whose admin_users table lacks the probed column now resolves to sqlite3 rather than hiding the data, while an unreachable host still resolves to pg. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): don't fail the migration on empty SQLite-only tables (#1038) Review round 10. The final verification flagged every source table missing from Postgres, regardless of whether it held rows — and SQLite-only tables do exist: initializeDatabase() creates an `events_new` scratch table and, when its legacy column copy throws, the catch swallows the error and leaves the empty table behind (db.js:236). The importer correctly skips tables Postgres does not have, so verification then reported a mismatch AFTER the data had already landed, exited 1, and left the install pinned to SQLite with no way to finish. An absent target table only matters if the source actually had rows. Empty ones are now listed and skipped. Reproduced both ways against a real PostgreSQL 15 with an events_new table present: without the fix the run ends in "ROW COUNTS DO NOT MATCH" and leaves the in-progress pin; with it, the table is reported as skipped, the migration completes and the pin is released. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): a completed migration overrides an implicit SQLite config (#1038) Review round 11. The migration allowed the one configuration it should have worried about most: DATABASE_CLIENT unset AND NODE_ENV not "production", which resolves to the development block — i.e. sqlite3. That is precisely the state the affected installs are in, since it is why they ended up on SQLite at all, so an operator can easily run the migration before fixing it. The script then renames the source database away, and the next start resolved to the implicit sqlite3, created a NEW empty database and served it — after reporting success. The success marker now overrides an IMPLICITLY resolved sqlite3 when Postgres settings are present, because the marker is durable proof of where the data actually went. An explicit DATABASE_CLIENT=sqlite3 still wins: that is the documented rollback. The script says something rather than refusing — refusing would block exactly the population this exists for. Reproduced with NODE_ENV and DATABASE_CLIENT both empty, against a real PostgreSQL 15: the migration completes, the source is renamed away, and the next boot resolves to pg with the data intact. Before this it resolved to sqlite3 and would have served an empty database. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * refactor(db): drop the dead reachability flag in probePgData (#1038) github-code-quality flagged `if (reachable)` as always true, and it is right: the unreachable branch returns, so everything below it runs only when the probe connected. The variable and the conditional were leftovers from a first draft that used a single catch for both failure classes. No behaviour change — the two error paths still return opposite answers. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): refuse to choose when both databases hold data (#1038) Review round 12. 1. An install that ran on PostgreSQL, lost NODE_ENV/DATABASE_CLIENT, and kept working on SQLite has REAL data on both sides: old rows in Postgres, newer ones in SQLite. The stranded-data rule only protected SQLite when Postgres was empty, so pulling this fix would have booted Postgres and hidden every gallery created since the switch — the exact failure this PR exists to prevent, in a variant I had not considered. A completed migration leaves a marker saying which side is current. Without one, two populated databases are a conflict: the boot stops and prints both targets, the two DATABASE_CLIENT values that resolve it, and the migration command that merges them. This is the only deliberate refusal in the change — guessing here would hide data AND split subsequent writes across two databases. 2. probePgData was handed knexConfig.connection even when knexfile had resolved to SQLite (a completed migration whose environment still says sqlite3), so node-postgres dialled its own localhost defaults instead of DB_HOST/DB_NAME — false "unreachable" diagnostics and a needless delay on every boot. The probe target is now built from the environment when the config is not pg. The conflict is honoured by all three entry points: the resolver exits 3 with an empty stdout, wait-for-db.sh stops the container, and server.js refuses to start. Two existing tests asserted that Postgres wins when both sides hold data. They encoded the pre-conflict assumption and described a state that cannot occur after a real migration (which always leaves a marker); both now pass the marker. Found while testing: the resolver's logger shim had no .error, so the conflict path threw, was swallowed by the fallback, and silently chose Postgres — the precise outcome this refuses to make. The shim is complete now. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): symmetric bootstrap rule, one resolved Postgres target (#1038) Review round 13. Both findings are consequences of earlier rounds. 1. The conflict rule added in round 12 counted an untouched SQLite bootstrap admin as data. core/001_init.js seeds one whenever ADMIN_PASSWORD is set — including into the accidental SQLite database — so a healthy Postgres install that had ever started once without NODE_ENV would have had a seeded-only SQLite file beside it, been declared a both-populated conflict, and REFUSED TO BOOT. The bootstrap discrimination is applied on both sides now; a setup-completed or logged-in admin still counts as real use on either. 2. The CLI's child phases inherited whichever knexfile block NODE_ENV selected. The development block defaults Postgres to localhost/postgres/photo_sharing, production to db/picpeak/picpeak — and this script is explicitly meant to run with NODE_ENV unset. With DB_USER/DB_NAME left to defaults it would therefore have migrated into `photo_sharing`, after which following the script's own advice to set NODE_ENV=production pointed the app at an empty `picpeak`. The target is resolved once, with production defaults, and passed explicitly to every phase — so the block knexfile happens to pick can no longer decide which database the data lands in. The pin and success marker record that same resolved identity. Verified against a real PostgreSQL 15: a live Postgres beside a seeded-only SQLite file now boots pg rather than refusing, flipping that admin to setup-completed restores the conflict, and a migration records localhost:7102/picpeak_r13b as its target rather than a defaulted guess. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): one Postgres identity everywhere; protect the credentials file (#1038) Review round 14. Three of the six findings were the same defect as round 13's, surfacing through paths that fix did not cover: the connection used to PROBE or MIGRATE could differ from the one the application then OPENS, because knexfile's development block points Postgres at localhost/postgres/photo_sharing while production uses db/picpeak/picpeak. 1. server.js exported only DATABASE_CLIENT=pg after the resolver decided, so knexfile filled in host/user/database from whichever block NODE_ENV selected. With SQLite already retired by a migration, that meant opening an empty database. The whole connection is pinned now. 2. Two defaults existed for DB_HOST: wait-for-db.sh resolves and exports `postgres`, knexfile's production block says `db`. Since the entrypoint exports its value, `postgres` is what a running container actually uses — so a `docker exec` migration, which inherits neither, has to agree with that, not with the default that is only reached when the entrypoint did not run. 3. The migration's Postgres phases inherited an unset NODE_ENV and therefore the development block, which ignores DB_SSL entirely — a managed Postgres requiring TLS could never be migrated into. The phases run with production semantics now. 4. core/001_init.js writes data/ADMIN_CREDENTIALS.txt, and that data directory belongs to the SOURCE install. Bootstrapping the Postgres schema replaced the operator's real credentials file with ones for a temporary admin the import immediately discards. The file is preserved across the phase, including when it fails. 5. The boot line described knexConfig, so an install redirected to Postgres by a migration marker still logged "Database engine: sqlite (...)", contradicting the warning printed one line earlier. 6. On a both-populated conflict resolveBootEngine returns client:null, and both migration runners told the operator their data was in "null" and to set DATABASE_CLIENT=null. They now present the two real choices. Verified against a real PostgreSQL 15: a migrated install started directly with NODE_ENV unset now logs `postgres (localhost:7102/picpeak_r14)` and opens it, where before it would have gone to the development block's photo_sharing. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * refactor(db): resolve the PostgreSQL target in exactly one place (#1038) Rounds 13 and 14 both traced back to the same thing, each time through a caller the previous fix had not covered: three different defaults existed for the same connection. knexfile development : localhost / postgres / photo_sharing knexfile production : db / picpeak / picpeak wait-for-db.sh : postgres / picpeak / picpeak (and it EXPORTS them) So a process that probed or migrated against one could hand over to a process that opened another. Patching each caller was not converging — the guard, then the CLI's child phases, then server.js — so this deletes the divergence instead. `src/utils/pgConnection.js` now owns the resolution and knexfile's development and production blocks both derive from it, as does the engine guard. Same shape as the earlier sqlitePath.js extraction, for the same reason. The database NAME is what made this dangerous: a wrong host or user fails loudly at connect time, while a wrong name connects fine and presents an empty installation. BEHAVIOUR CHANGE: with DATABASE_CLIENT=pg and no DB_* variables, a non-production environment now resolves to postgres/picpeak/picpeak instead of localhost/postgres/photo_sharing. Deployments are unaffected — compose sets these explicitly and wait-for-db.sh exports them — but a local machine running Postgres bare now needs DB_HOST=localhost DB_USER=postgres DB_NAME=photo_sharing (or DATABASE_CLIENT=sqlite3, which is what backend/.env already uses). The failure mode of getting this wrong is a refused connection, not a silently empty database. Side effect worth having: DB_SSL is now honoured whatever NODE_ENV says, so the managed-Postgres case is fixed at the root rather than by forcing production semantics onto the migration's child phases. The test block keeps its own photo_sharing_test default — isolation is the point there. Verified: every block plus the guard resolve identically from the same environment; explicit DB_* still wins; production's pool tuning is preserved; and a full SQLite → PostgreSQL migration with NODE_ENV unset lands in the right database with JSON shapes intact. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): two more components that guessed the database instead of asking (#1038) Both found while sweeping for copies of the connection defaults. Checked in detail first — one of my suspicions about them was wrong. scripts/set-admin-password.js hand-rolled its own knex config while all four sibling scripts (reset-admin-password, create-admin, show-admin-credentials, reset-admin-mfa) use the application's connection. Two consequences: - it read DB_CLIENT, a variable nothing else in this codebase sets, so it defaulted to Postgres and could not work on a SQLite install at all; - it defaulted to database `picpeak_dev`, a name no other component uses. It now uses `require('../src/database/db')` like its siblings, so it follows whatever engine the install actually runs on. Timestamps are written as ISO strings because it reaches SQLite now, where raw Date objects are the documented landmine. NOT changed: the script's "all existing sessions have been invalidated" notice is accurate — auth.js compares token iat against password_changed_at — and it deliberately leaves must_change_password alone, which is right for an operator choosing a password rather than being issued one. routes/adminSystem.js re-derived three things the live connection already knows, and each could disagree with it: - the engine, from DATABASE_CLIENT || 'sqlite3' — so a Postgres install without an explicit DATABASE_CLIENT took the SQLite branch; - the Postgres database, from DB_NAME || 'picpeak'; - the SQLite file, from a hardcoded ../../data/photo_sharing.db that ignored DATABASE_PATH entirely. All three now come from db.client.config, with pg_database_size(current_database()). Verified: set-admin-password works on SQLite (new hash verifies, old rejected) and still on PostgreSQL; and on a SQLite install with a custom DATABASE_PATH the size logic reports the real database (1,748,992 bytes) where the old code reported a different file entirely (1,851,392) — or 0 where that path does not exist. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM * fix(db): bind the migration marker to its target; fix a phantom table (#1038) Review round 15. 1. The marker records `host:port/database`, but only its EXISTENCE was checked. Repoint DB_NAME or DB_HOST at a different, empty PostgreSQL after migrating and the marker would vouch for that one too — booting it, presenting an empty installation, and suppressing the SQLite fallback while the real data sits in the recorded target and the renamed rollback copy. The marker is compared against the current connection now, and a mismatch stops the boot with both targets named and the two ways out. 2. `incoming_invoices` is not a table — supplier documents live in `inbound_documents` (core migration 124). Both occupancy lists skip tables that do not exist, so those records were silently not protecting anything: an install whose only remaining data was inbound documents could be switched away from, or overwritten without --force. Verified every other name in the lists against the live schema at the same time. Verified: a marker naming picpeak_original with picpeak_mk configured refuses with exit 3 and prints both; making them agree boots pg. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM --------- Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-13 18:50:56 +02:00
Paul Nothaft	89dc9623c1	fix(feedback): persist guest feedback settings, unshadow the guest route (#1030) (#1031) ... Enabling Guest Feedback on an event could silently do nothing. 1. `updateEventFeedbackSettings` spread the request body straight into the knex UPDATE. The admin event form posts its whole client-side state, including three keys that were never columns on event_feedback_settings (`enable_rate_limiting`, `rate_limit_window_minutes`, `rate_limit_max_requests`), so the write threw and the route answered 500. Writable columns are now whitelisted; identity columns and timestamps stay server-managed. 2. EventDetailsPage swallowed that 500 in a bare `catch {}` ("Error already handled by mutation" — it is a different request), so the admin was left looking at "Event updated successfully" while the toggle never persisted. The error is surfaced now and the settings query is invalidated on success. 3. gallery.js declared a duplicate `GET /:slug/feedback-settings`. server.js mounts galleryRoutes before galleryFeedback, so it shadowed the real handler and dropped the per-guest caps (#655) from the guest payload — the gallery could never render the favorite/like limits or their counters. Timestamps are written as ISO strings so they round-trip on both engines. Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-13 18:50:52 +02:00
Paul Nothaft	34ee31141b	fix(gallery): coerce SQLite 0/1 booleans in the guest surface (#1028) (#1034) ... SQLite stores booleans as 0/1, Postgres as true/false. The guest gallery compared strictly against `true`/`false`, so every flag read backwards on SQLite installs: allow_downloads: 0 !== false → true (header Download button shown with downloads disabled) allow_user_uploads: 1 === true → false (upload button hidden with uploads enabled) Worse, all five download guards used `allow_downloads === false`, which never fires against a stored 0 — so on SQLite "Allow photo downloads = off" was inert end to end: single photo, download-all, download-selected, download-jobs and the job-status poll all kept serving, as did the secure-images download route. Per-category blocking (#640) was ignored for the same reason, the protection toggles (right-click, devtools, canvas, watermark) reported false while enabled, overlay_protection was stuck on, and show_feedback_to_guests leaked feedback with the setting off. The /info endpoint was already correct — it checks 0/'0' explicitly. The two payloads had simply drifted. Everything now goes through parseBooleanInput (utils/parsers.js), which normalises both engines and takes a per-column default so legacy NULL rows keep their documented behaviour. Tests run on the SQLite harness, so they assert the real engine values. Every one of them fails on the unfixed code — the payload assertions return the inverted value, and the guard assertions never get their 403 (the request proceeds to serve instead). Claude-Session: https://claude.ai/code/session_0168gubtwYYacJv8weAjy8DM Co-authored-by: Paul Nothaft <paul@MacStudio-von-Paul.local>	2026-08-13 18:50:48 +02:00