SD1.5 inpainting is OOM-killed at the stock 6g compose limit on CPU hosts and completes at 8g. Read the cgroup limit and refuse hq up front with an actionable message; GPU hosts and unlimited containers are untouched. Fixes#670.
A fresh GPU deployment has working passthrough but no torch/ONNX runtime until a bundle installs, so the dispatcher reports gpu=false and the boot banner said no GPU was detected. Add the hardware-present state, driven by the same /dev/nvidia0 check the instance census uses. Fixes#673.
app_version was attached only to feedback events; every other allowlist stripped it, so the install base could not be segmented by release. One property on the once-per-boot census event covers it. Fixes#674.
trim-audio stream-copies, and a window below one frame ships a container with zero audio frames as a 200 success. Floor the window at 0.1s in the schema, and refuse after endS clamps to the real duration if the window drops under the floor. Fixes#671.
A release-readiness QA pass over the whole product. The commits split into
defects a user would hit and gates that were reporting green while measuring
nothing.
## Fixes that change behaviour
Rate limiting was bypassable on every install: TRUST_PROXY defaulted to true, so
request.ip came from a client-set header and a forged X-Forwarded-For got past
the login limiter. The default is now a private-network trust list.
A transient Postgres outage stranded in-flight jobs, leaving finished output on
disk with no row pointing at it. A reconciler now resolves those rows and adopts
the bytes rather than dropping the work.
A Redis connection that moved to a new address wedged every read-blocked
consumer, so completions stopped signalling while health still answered 200.
Socket timeouts plus subscriber pings recover it.
Installing more than one AI bundle left the shared venv multi-versioned and
silently broke three tools. The installer now reconciles distributions to one
version each.
Converting an image to JXL at quality 1 through 4 returned a 500, because
libjxl 0.7 rejects the distance those values compute. The quality is floored at
what the encoder honours. A missing ffmpeg was also reported to the user as a
corrupt upload; it now says the engine is unavailable.
RAW uploads reached an unpatched LibRaw on arm64, so it is built from source at
0.22.2, and the release scan was split so it can fail on an unfixed critical
instead of hiding it behind ignore-unfixed.
## Gates that could not fail
Two mutation lanes ran zero mutants because Stryker crawled the gitignored docs
build; coverage discarded its whole report on any failing test; the lint gate
skipped root tests, scripts, and two workspaces; and several generated matrices
counted a host missing ffmpeg as a passing tool. Each now measures what it
claims.
Full evidence and the outstanding release items are tracked locally and are not
part of this branch.
Three defects from #645, all of which let the server report something
that was not true.
Tool access was enforced per route, so it drifted. createToolRoute calls
requireToolAccess and the factory tools were fine, but all 45
hand-written routes had to remember the same call and none of them did.
A role without tools:use could run image-to-pdf, svg-to-raster,
erase-object, favicon, qr-generate, upscale, sign-pdf and the rest. The
issue described this as affecting two routes; it was every one of them.
The check now lives in a single preHandler keyed off the tool the router
matched, so it covers sub-paths (/batch, /info, /preview, /analyze,
/inspect) and any route added later without that route opting in. Ids no
tool claims stay unresolved, which keeps an unknown or misfiled tool a
404 rather than telling an unauthorized caller which ids exist.
Resolution reads request.routeOptions.url, the pattern the app itself
registered, rather than parsing request.url a second time. find-my-way
decodes before matching, so an independent parse disagrees with the
router and the router wins: `/api/v1/tools/image/%66avicon` ran favicon
while the gate saw no tool at all. Absolute-form request targets slipped
it the same way. Taking the router's own answer removes the disagreement.
A ZIP stream that failed after the 200 headers were out called
reply.raw.end(). On a chunked response that is indistinguishable from
success, so a client kept an archive with no central directory believing
it whole. Worse, a source stream that errored had no listener: the
request hung until it timed out and the error surfaced as unhandled. A
poisoned-storage probe reproduced both. The socket is destroyed instead,
and every source stream is listened to. svg-to-raster additionally ran
its append loop past the hijack with no try/catch, where a throw leaves
Fastify logging and walking away with the socket neither ended nor
destroyed. pdf-to-image is fixed alongside the other two: it shipped in
#643 with the destroy half but not the listener, so it hung the same way.
A zero-byte upload was dropped during parsing. The client pairs results
with its own file list by index, so every later result shifted onto the
wrong file: one document's output was presented as another's, under
another's name, while the file that actually converted was marked "not
found in batch results". Empty parts now keep their slot and fail in
place with a reason.
Two tests in adversarial-extended.test.ts asserted the old zero-byte
behavior, including a comment that batch "silently skips zero-byte
parts". They now pin the replacement: still rejected, nothing processed,
but the caller is told which files were empty instead of being told it
sent none.
A guard walks the whole catalog and fails if any of the 241 tools answers
anything but 403 for a role without tools:use, so a tool cannot escape
the gate by being registered in a shape nobody thought to sample.
Fixes#645
validateImageBuffer() never listed heif in CLI_DECODED_FORMATS, so real iPhone HEIC uploads hit Sharp's own metadata probe (its bundled libheif only supports AV1/AVIF) and got rejected before reaching the working heif-convert/heif-dec decode path already wired up downstream. Adds heif to that set, same as raw/psd/tga/bmp/etc.
Also fixes the same gap on erase-object's mask input, which validates through the same function but had no matching decode step, so a HEIC mask reached an unguarded sharp() call and came back as a misclassified server error instead of a clean 422.
Fixes#622
Coverage 83.6 to 87.36% lines, 81.63 to 84.14% branches. Mutation testing across five packages: image-engine 85, media-engine 92, doc-engine 87, shared+enterprise 86, apps/api security and jobs slice. Runs all five lanes weekly. Fixes the silently-broken mutation CI (babel pin), a redact-pdf envelope-shape test bug, an untested enterprise license valid-signature path, and an audit test that only exercised a hand-copied reproduction. Test and config only, no product code changes beyond the babel pin and one test-only oidc export. Full suite: 16,712 pass, 0 fail.
Follow-up to a full re-audit of the 2.0 tree. Most prior findings were already
fixed; this closes the ones that were not:
- SAML assertion replay: validateInResponseTo ifPresent plus a Redis-backed
CacheProvider, so a captured signed assertion cannot be replayed. ifPresent
keeps IdP-initiated SSO working.
- MFA login challenge burned after 5 wrong TOTP codes.
- api_keys.key_prefix indexed; the per-request lookup was a full table scan.
- MAX_AI_JOBS_PER_USER caps a user's in-flight single-file AI jobs (the AI pool
runs at concurrency 1). Batch and pipeline AI stay uncapped.
- MAX_WORKSPACE_SIZE_GB enforced instead of being dead config.
- SUBPROCESS_MEMORY_LIMIT_MB (default off) for the native media and doc engines;
not applied to the AI sidecar.
- SVG sanitizer closes unquoted and whitespace-prefixed javascript: hrefs and
the animateTransform/animateMotion/handler/mpath elements.
- Windows-style paths stripped from error output to match the Sentry scrubber.
- Postgres and Redis compose services get cap_drop plus pids_limit and cpus.
- .env.example ships MAX_SVG_SIZE_MB=50 (0 disabled the cap).
Adds security-focused unit and integration tests. typecheck, biome, and the
full unit and integration suites pass.
Defers the onboarding usage survey to the instance's first successful processing (the worker writes a one-time onboarding.firstProcessedAt marker and the overlay gates on it), so it reaches engaged users instead of first-landing visitors.
Replaces the two questions telemetry already answers (modality preference from tool_used, install method from instance_started) with what it can't infer: prior tool, self-host motivation, and discovery source.
Adds feedback_prompt_shown and feedback_prompt_dismissed on all five feedback surfaces (usage survey, per-job prompt, admin install card, global nav dialog, search-miss) so skip and completion rates are measurable, not just submissions. New survey strings translated into all 20 non-English locales.
Five fixes to the PostHog event stream, from an audit of what we actually collect versus what's flowing in. Each one is test-first.
## What changed
**Silenced the `$opt_in` noise.** `initAnalytics` called `opt_in_capturing()` on every page load to clear a stale opt-out flag, and posthog-js emits an `$opt_in` event on every call. That was 10k+ events a month (up to 55 per user) carrying no signal: analytics is on by default with an admin opt-out, so there is no per-user consent to record. Both call sites now pass `captureEventName: false`.
**Split the onboarding survey out of `feedback_submitted`.** The onboarding usage survey rode the same event as real feedback, so about 93% of "feedback" was actually onboarding profiling. It now emits `onboarding_survey_submitted`, so feedback metrics mean feedback again.
**Set `pipeline_executed.file_count`.** It was declared in the properties interface but never populated. A pure `pipelineExecutedProps` helper now derives it (batch size for a batch run, else 1) and is shared by the success and failure paths, which also drops a duplicated payload.
**Tracked OIDC login failures.** All six OIDC callback failure branches bumped the Prometheus counter and wrote an audit log but never emitted `auth_login_failed`. A `recordOidcFailure` helper mirrors the password path.
**Added `TELEMETRY.md`.** A contributor-facing event dictionary: every event, its properties, where it fires, and the privacy invariants, with the allowlists as source of truth. A drift test fails if any `ANALYTICS_EVENTS` value goes undocumented.
I left the published telemetry guide (`apps/docs/guide/telemetry.md`) alone. It is high-level and still accurate, and editing it would pull in the 21-locale stale-gate for no gain.
## Verification
- Unit (63 tests): `analytics-events`, `telemetry-doc-drift`, `api/analytics`, `web/analytics`, `worker.behavior`
- Integration (41 tests): `oidc-auth`, `feedback`
- Full typecheck across all 9 workspaces
- Biome clean on the changed files
All green locally.
AI sidecar failures reached Sentry as 'Error: Error': the scrubber type-onlys plain Errors and the tool wrappers threw them from result.error. The bridge now exports toSidecarError(), wrapping the sidecar reason in a SafeError (memory-allocation text classifies as operational, the rest as bug); all 14 wrappers use it, plus the dispatcher crash/stdin/spawn rejection paths and parseStdoutJson. toBgRemovalError from #535 delegates to the shared helper.
On the web side, DOMExceptions report their specific name via err.name, so the NATIVE_ERRORS allowlist dropped the whole family's browser-authored messages. It now carries the full WebIDL DOMException name table; messages still pass through url/path redaction.
Bridge-mocking test files switched to importOriginal passthrough mocks.
Stackless uncaught errors reached Sentry as a bare Error with no frames and collapsed into one ungroupable issue. beforeSend now fingerprints frameless events by safe identity (name, code, one-way hash of the message) so distinct crashes separate without leaking PII. Only frameless events are touched; an upstream fingerprint is never overridden.
BullMQ raises UnrecoverableError when a job loses its lock (a stall), e.g. a heavy upscale under memory pressure. We never throw it ourselves, so classifyError now treats it as operational (one warning per hour) instead of a bug. ReplyError stays a bug.
When a PDF has no text layer (scanned or image-only), pdf-to-text now
returns a 422 that points at the OCR tool instead of a silent empty file,
and text downloads carry charset=utf-8 so UTF-8 Arabic renders correctly
when the .txt is viewed inline.
Fixes#589
The library file-storage helpers joined FILES_STORAGE_PATH with a database stored_name and never checked containment, so a crafted name could read or delete files outside the storage root after a malicious 1.x SQLite import (which copies stored_name verbatim). Add assertSafeStoredName() and apply it in every helper that resolves a stored name to a path, matching the containment guard object-storage already uses.
Reported by Alpesh Bhagwatkar.
Replace the job timeout message that hardcoded "background-removal" for
every tool with a tool-agnostic one that sets the CPU-vs-GPU expectation,
the usual reason heavy AI times out on modest hardware. The client-side
SSE stall message gets the same treatment. Both stay under friendlyError's
280-char limit so the guidance reaches the user instead of collapsing to a
generic "Processing failed".
Refs #591
Editing a file from the library used to silently supersede it: the worker auto-saved every result as a new version and the leaf-only listing hid the original, which read as a destructive overwrite. Tool pages now show a per-edit choice for library-sourced files. The default saves the result as an independent new file and keeps the original; picking overwrite keeps the old superseding-version behavior.
The client sends a saveMode multipart field next to fileId, validated with a 400 on unknown values, and autoSaveToLibrary branches on it. Every hand-written route that honors fileId parses the field the same way as the factory. The review panel shows where an auto-saved result went instead of offering a second, duplicate save. Tools whose route or submitter ignores fileId keep the selector hidden via a shared unsupported-tools set, and the choice resets to the non-destructive default whenever a new file is staged.
Closes#495
Comprehensive telemetry quality improvements across Sentry and PostHog, grounded in an audit of the live data plus current best-practice research.
Sentry: job_id/instance_id tags, operational fingerprinting, PII-safe settings context on bug events, web tag population + extension-noise filtering, an early-crash buffer, http status/method kept on breadcrumbs, and a gated-off-by-default performance-tracing re-enable (tracesSampler that zeroes db/redis/queue-poll root spans + drops the Redis integration) with worker job spans and canonical-host cron monitors.
PostHog: history_change SPA pageviews, instance_id super property for fleet rollups, enriched tool_used (formats, byte sizes, is_batch, execution_hint, real error_kind taxonomy), the previously-dead result_saved/batch_processed/ai_bundle_prompted events fired, search click-through, editor + Automate authoring + auth instrumentation, a before_send PII boundary, and minimal opt-in landing-site pageviews.
Every worker tool error now carries an input_format tag (file extension only, never the filename) for triage, derived once at the worker error path and added to the scrubber allowlist.
Postgres auth (28xxx), permission (42501), resource (class 53), and operator-intervention (class 57) failures now classify as operational via a cause-chain walk, not bug. pg query bugs (e.g. 42601) stay bugs.
Wrap convert and gif-tools process functions so a Sharp .toBuffer() failure carries an authored SafeError title (and the original as cause) rather than a scrubbed "Error: Error".
Fixes#533, found while working on #529/#531.
OIDC and SAML logins hard-blocked on the MFA policy with zero check of whether the user actually enrolled TOTP, and no challenge step at all. Once an admin turned on an MFA-required policy, every SSO user was permanently locked out regardless of enrollment status.
- Extract the post-auth MFA decision (challenge / enrollment-required / proceed) into a shared, unit-tested function so OIDC and SAML can't independently diverge again
- An already-enrolled user now gets a real challenge (reusing the existing, auth-method-agnostic MFA completion flow) instead of being blocked
- An unenrolled user under a required policy gets a distinct, correctly mapped error instead of the old generic one
- Fix a real fail-open regression caught in review: a transient DB error during the enrollment-status check could have silently skipped MFA entirely for an enrolled user; now it fails closed and logs
- Strip the one-time challenge token from the URL after consuming it
Target-size compression had only a coarse DPI lever, so it undershot badly (a 350KB target could land at 216KB) and silently missed unreachable targets. Adds JPEG quality as a second lever (forced re-encode so it bites on JPEG scans), folds both into one monotonic quality axis that target-size binary-searches, reports targetMet honestly in the panel across 21 locales, and flips the tool to async for the extra passes. Quality-mode output sizes shift intentionally (slider now drives JPEG quality at full resolution in its top half).
Add a once-per-boot instance_started event (arch, os, deploy_mode,
gpu_present) so the fleet architecture mix is measurable. It reuses the
existing per-instance instance_id and is exempt from the volume sample
rate, since a census that fires once per boot must not be thinned.
Restore useful capture depth now that the sponsored plan removes the
quota pressure behind the earlier hardening:
- PostHog sample rate 0.1 to 1.0 (full analytics when enabled); the
property allowlist still blocks file data.
- Sentry per-instance ceiling 20 to 500/hr, breadcrumb trail restored
(sanitized: urls/paths redacted, data payloads dropped), full stack
paths kept; local vars, request bodies, and PII still dropped. Both
api and web.
Honor ANALYTICS_ENABLED=false as an opt-out alias: it was documented on
the Docker Hub README but never wired in 2.x, so anyone who set it was
still tracked.
All capture stays behind the analytics opt-out gate.
Adds a dedicated remove-gif-background AI tool: removes the background from an animated GIF, WebP, or APNG frame by frame and reassembles a transparent (or composited) animation in WebP, APNG, or GIF, with full per-frame effects. Reuses the background-removal bundle. Verified end-to-end with the real rembg model.
Closes#496.
Follow-ups to the v2.1.0 Sentry telemetry overhaul, found by reviewing live release:2.1.0 events:
- error_code tag was empty because reportError read only the top-level err.code; add extractErrorCode() to walk the cause chain (pg SQLSTATE, node E-code, else first short code).
- InputValidationError from a tool's processV2 in the worker was logged as error_class=bug; classify it as expected for any source. Worker-side ZodError stays a bug (schema drift).
- AI dispatcher timeouts rejected with a bare Error, which the sanitizer scrubbed to a message-less "Error: Error"; reject with an operational SafeError (code "timeout") at both timeout sites.
Each fix written failing-test-first; affected and adjacent unit suites green plus full CI (integration + e2e).
Shared Claude Code translation pipeline (scripts/i18n, no API key) plus Astro/VitePress/Scalar i18n wiring. Landing and API reference translated into all 20 languages; docs i18n wiring + English source anchors. The translated docs markdown (apps/docs/<locale>/**, 3,620 files) follows in a companion PR because it exceeds GitHub's per-PR CI file limit.
Removes Sentry tracing entirely (BullMQ idle polling burned 4.8M transactions in 2 days at the baked 0.1 rate), decouples PostHog sampling, and replaces the type-only error scrub with a vetted-field sanitizer plus SafeError/ToolInputError contracts. One classified capture path with per-signature throttles and a per-process ceiling makes storms impossible (NODE-1E was 4,541 events from one 30s loop). Browser errors move to a dedicated web Sentry project with their own source maps. Adds the SNAPOTTER_TELEMETRY runtime kill switch and silences test fleets.
Crash fixes: remote 204/304 SSRF process kill (NODE-20), conversion-preset boot crash loop (NODE-21), Redis version preflight + unhandled subscribe rejection (NODE-1T), Sign PDF on plain-http origins (NODE-1K/1M), wavesurfer/pdf.js teardown rejections (NODE-1P/1N), bundle-import ZlibError to 400 (NODE-1Z), chart-maker input errors declassified (NODE-1H/1J), asset requests skip the session DB lookup (NODE-1D).
Make on-demand AI feature-bundle installs reliable and self-healing, closing
the failure modes behind most "some tool doesn't work" reports.
Multi-bundle installs: tools needing more than one bundle (Passport Photo,
Enhance Faces) install every required bundle from one action and stay
not-installed until all are present. Verified across all 19 AI tools.
Downloads: self-heal the accelerated Hugging Face (Xet) client so an upgraded
venv no longer silently falls back to slow urllib; restart instead of
corrupting a resumed partial when a proxy ignores Range and returns 200;
verify the completed size; fail fast on disk-full and HTTP 4xx; retry
transient errors five times; add hf_transfer fallback and document Xet egress.
Install integrity: crash-atomic venv writes so a killed or out-of-space
install can no longer tear the shared venv and break other tools; a boot
breadcrumb reseeds a torn venv to a clean state automatically; a post-install
smoke import test refuses to record a bundle whose libraries cannot load; an
install watchdog stops a wedged installer that would otherwise hold the venv
writer lock forever.
Adds unit and end-to-end tests for every failure mode above.
* fix(api): prevent a crash when an over-limit upload stream has no consumer yet
busboy's "limit" handler destroyed the file stream with an error but never
attached its own error listener, relying entirely on whatever consumes
part.file downstream to do so. On a fast enough connection (or a fully
buffered body, e.g. Fastify inject()), busboy can process enough bytes to
hit the size limit before the route handler's receiveUpload() call has
attached its own stream listener, leaving the resulting "error" event with
zero listeners -- which crashes the whole process by default in Node.
Surfaced by tonight's FULL_MATRIX+FUZZ integration run (880 uncaught
exceptions, all the same root cause). Reproduces deterministically in
isolation; unrelated to this release's actual code delta (file untouched
since PR #413, well before the baseline QA pass).
Fix: attach a baseline no-op error listener the moment the stream is
created, guaranteeing at least one listener always exists. EventEmitter
delivers "error" to every registered listener, so the real consumer's own
error handling is unaffected.
* fix(ai-bundles): rebuild upscale-enhance and photo-restoration to reconcile scipy ABI
upscale-enhance and photo-restoration both depend on codeformer-pip, whose
transitive closure (basicsr -> realesrgan -> gfpgan) pulls in an unpinned
scipy. Both bundles were last built ~June 18-19, before PR #437 added the
manifest's `constraints` array (numpy==1.26.4, scipy==1.12.0, etc.) to pin
exactly this kind of dependency during bundle builds. Only the ocr bundle
was rebuilt after that fix landed.
install_feature.py has no pip install step -- it's a raw tarfile extraction
with no cross-bundle conflict resolution, so installing OCR alongside either
stale bundle left three incompatible scipy versions' files mixed in the same
site-packages directory (a compiled _rotation.*.so from one release next to
Python files expecting a different release's API), breaking the `upscale`
tool and OCR's higher-quality tiers with an ImportError.
Rebuilt both bundles for amd64-gpu and arm64-cpu from the current manifest,
verified scipy/scikit-learn/scikit-image/pandas all resolve to the pinned
versions in the tarballs themselves, then verified end-to-end on real
hardware (Mac arm64 CPU and ubuntu_gpu .248 RTX 4070): installing all
affected bundles together now yields exactly one version of each constrained
package, `upscale` produces correct output, and OCR's balanced/best tiers
correctly use PaddleOCR-GPU instead of erroring out.
Published the rebuilt tarballs to the public deepsafe/feature-bundles
HuggingFace repo and updated this manifest's sha256/sizes to match.
Also adds verify-bundle-compatibility.sh: verify-bundle.sh checks each
bundle in isolation (a fresh venv per bundle), which is exactly why this
shipped twice -- nothing ever checked that bundles built at different times
agree once layered into the one shared venv real installs use. The new
script installs every bundle for an arch into one venv and asserts each
constrained package has exactly one, correct version.
Known follow-up (not fixed here, needs separate discussion): uninstalling a
bundle only removes its downloaded model weights, never the site-packages
it added, so existing installations that already hit this bug have no clean
self-service fix via uninstall+reinstall -- they need a full AI-venv wipe.
* fix(docker): bake a real rate limit default for the all-in-one one-liner
The documented single-container `docker run` install had RATE_LIMIT_PER_MIN=0
(effectively unlimited, ~50k/min) baked in, since only docker-compose.yml
carried a hardened override. A self-hoster following the one-liner path got
no meaningful throttling anywhere, including auth-adjacent routes with no
dedicated per-route limit. Bakes a generous-but-real 1000/min default into
the Dockerfile, raises both compose files' fallback to match so the two
documented install paths converge on the same posture, and updates the Zod
schema default plus docs that quoted the old value.
* fix(api): boot log undercounted tool routes by the conversion-preset total
The "Tool routes: N active" line logged before registerConversionPresets(app)
ran, so it only ever reported the base 158 tools, 83 short of the real
241-tool total. Presets have to register after the base loop (they delegate
to each base tool's own processV2), so the fix moves the log line to after
that call and has registerConversionPresets return its count instead of
reordering the dependency.
* fix(ai): forward {info}/{warning} stderr JSON instead of dropping it
The dispatcher stderr parser only recognized {ready} and {progress,stage}
shaped JSON lines; anything else that parsed as valid JSON (like ocr.py's
GPU-to-tesseract downgrade notice, an {"info": ...} line) matched neither
branch and fell through silently, never reaching docker logs. Adds explicit
{info}/{warning} handling that forwards to console.log/console.warn, same as
the existing [prefix]-tagged non-JSON path.
* fix(api): fall back to a lower OCR tier when PaddleOCR itself is unusable
ocr.ts already retries lower quality tiers on a crashed dispatcher, but the
condition only matched crash-style messages (segfault, exited unexpectedly).
ocr.py's own ImportError/exception handlers already produce messages telling
the caller to use a lower tier (e.g. on the scipy ABI conflict class of bug),
but nothing ever acted on them, so a broken PaddleOCR hard-failed with 422
instead of degrading to Tesseract like ocr-pdf effectively does. Broadens the
retry condition to also catch PaddleOCR-engine-unusable messages.
Note: ocr-pdf's tesseract-only behavior turned out to be an unrelated,
pre-existing, deliberate design choice (PaddleOCR segfaults on rasterized PDF
pages on arm64), not a graceful-fallback mechanism to copy -- the two tools
weren't actually solving the same problem, so this fixes ocr.ts's own gap
rather than trying to mirror ocr-pdf.
* fix: remove all automatic third-party egress (OSM tiles, Scalar fonts, editor Google Fonts, AI model download fallbacks)
Phone-home audit follow-up. The product no longer makes any automatic
third-party request; user-initiated click-outs stay, and production now
fails closed on missing AI models.
1. GPS leak via OSM tiles: the strip-metadata panel auto-loaded
tile.openstreetmap.org tiles encoding the photo's GPS position. The
Leaflet mini-map is gone; coordinates render as text plus an explicit
View on map link (openstreetmap.org, opens on click only). Removed
tile.openstreetmap.org from the CSP img-src, dropped the leaflet
dependency, added the viewOnMap i18n key to all 21 locales.
2. Scalar docs fonts: /api/docs loaded Inter and JetBrains Mono from
fonts.scalar.com. Scalar now renders with withDefaultFonts: false and
both --scalar-font and --scalar-font-code pinned to system stacks;
fonts.scalar.com removed from the docs CSP font-src. Verified by
injecting GET /api/docs/: config carries withDefaultFonts false and
the served page has no fonts.scalar.com reference.
3. Editor Google Fonts: the editor font picker built
fonts.googleapis.com stylesheet URLs for 25 web fonts the served CSP
already blocked. The remote loading path is deleted; the picker now
offers system fonts only, with a SELF_HOSTED_FONTS seam (FontFace API,
same origin) for bundling fonts later. Unknown families saved in old
documents fall back to the browser default.
4. Python sidecar fails closed on model downloads: new
packages/ai/python/offline_guard.py gates every runtime download
fallback (inpaint, outpaint, restore, noise_removal, detect_faces,
enhance_faces, face_landmarks, red_eye_removal, remove_bg, ocr,
transcribe, upscale) behind SNAPOTTER_ALLOW_MODEL_DOWNLOAD=1 with an
actionable error. Bundled models keep working untouched.
5. OCR and transcription library-internal downloads: unbundled PaddleOCR
language and detection fallbacks now raise the guard error naming the
language instead of resolving models over the network; faster-whisper
gets local_files_only when downloads are off.
6. GFPGAN and CodeFormer cwd-relative weights: facexlib and
codeformer-pip resolve helper weights relative to the process cwd and
fetch them from GitHub when absent. They are now symlinked from the
installed bundle files under MODELS_PATH/gfpgan/facelib before the
libraries load, failing closed when unresolvable.
Defense in depth: HF_HUB_OFFLINE=1 and TRANSFORMERS_OFFLINE=1 are set in
the runtime image and in the sidecar spawn env; install_feature.py lifts
them for user-initiated bundle installs and restores them afterwards
(it can run in-process inside the dispatcher). SNAPOTTER_ALLOW_MODEL_DOWNLOAD
is documented in .env.example, default off.
Validation: typecheck 9/9 workspaces, Biome clean on touched files,
5178 unit tests pass, py_compile on all touched scripts, guard behavior
exercised in both dispatcher exec and per-request import modes, zero
remaining runtime references to the three hosts. Docker build and live
AI inference need post-merge verification on the GPU host.
Claude-Session: https://claude.ai/code/session_01XGB4pGvTvb7sUX4JN745U7
* fix: allow AI model downloads by default, make strict offline mode opt-in
Product call: ease of use first. The download gating from the previous
commit inverts its default: runtime model fetches (public model weights
only, never user data) are allowed out of the box so AI tools self-heal,
and SNAPOTTER_ALLOW_MODEL_DOWNLOAD=0 becomes the explicit strict offline
mode for airgapped deployments, where every fallback raises the
actionable error instead of fetching.
Changes: offline_guard blocks only on an explicit 0/false; the
unconditional HF_HUB_OFFLINE/TRANSFORMERS_OFFLINE image ENV is removed
and bridge.ts sets those flags for the sidecar only in strict mode;
.env.example documents the new default; install_feature's lift/restore
stays. All bundled-path preferences, pre-existence checks, and symlink
pre-placement remain, so installed bundles never trigger a download.
The OSM, Scalar font, and editor font fixes are unchanged.
Validation rerun: typecheck 9/9, Biome clean on touched files, 5178
unit tests pass, py_compile on touched scripts, guard behavior verified
for unset/1 (allowed) and 0/false (blocked with the new message).
Claude-Session: https://claude.ai/code/session_01XGB4pGvTvb7sUX4JN745U7
Fixes found by manually testing a fresh install end to end:
- auth: the must-change-password gate returned 403 on public routes
including /api/v1/health, so every fresh install showed a false
"Reconnecting to server" banner on the forced password change
screen. Public routes are now exempt (they need no session at all).
Adds the gate's first direct tests.
- multipart: @fastify/multipart's parts() iterator (9.4.0 and 10.0.0)
ends on the request stream's "close", which on a reused keep-alive
connection fires while an earlier part is still streaming to storage,
silently dropping the parts behind it. The object eraser lost its
mask file on every second POST per connection. Replaced with a
busboy-driven iterator (lib/multipart-parts.ts) that ends on busboy's
own "finish", installed for all routes via a preValidation hook;
the tool-factory field-recovery workaround for the same bug is now
unnecessary and removed.
- eraser: the mask canvas backing store is natural resolution, but
"absolute inset-0" does not stretch replaced elements, so the
canvas rendered at intrinsic size and the brush ring, strokes, and
exported mask were all misscaled on photos larger than the viewport.
The canvas now gets an explicit CSS box at the fitted size.
- compare slider: solid white divider with a dark halo so it stays
visible over light images; still initialised at the painted region.
- tool page: the AI bundle install prompt now centers in the content
area instead of hugging the top.
- api docs: disabled Scalar's cloud features (Ask AI, Generate MCP,
Open API Client, dev toolbar), hid the "Powered by Scalar" footer
link, and set the page title to "SnapOtter API Reference". The docs
CSP blocks those cloud calls by design, so the buttons were dead UI.
- docker: embedded Redis comes from packages.redis.io pinned to the
8.x major (was Debian's 7.0.15), matching the Compose stack and the
documented claim. Build fails fast if the major ever drifts.
- docs: DOCKERHUB.md quick start now leads with the one-command docker
run (matching the README) with Compose as the production path;
README says embedded Postgres 17 + Redis 8.
Claude-Session: https://claude.ai/code/session_01XGB4pGvTvb7sUX4JN745U7
* fix(api): correct format/filename/container handling across tool routes
Found during a comprehensive QA sweep exercising every tool against its
full accepted-format matrix:
- watermark-image, compose: preserve the requested output format and a
matching download filename/extension instead of always emitting the
source format
- compose: crop oversized overlays to the visible base area instead of
crashing Sharp's composite, and reject only overlays fully outside the
base image instead of any oversized one
- compare, vectorize: switch to the shared image input handler so
filenames and formats like .svgz/.tga/RAW survive validation instead
of being rejected pre-processing
- tool-factory, images-to-video: normalize frames through Sharp before
handing them to FFmpeg, fixing GIF/AVIF/RAW image-to-video jobs that
previously failed or hung
- media-tool, replace-audio, embed-subtitles: fix legacy container
MIME/codec handling for MPEG sources and subtitle remux cases
- files: expand download MIME mapping for text/data/document/video/audio
outputs that were falling back to a generic content type
- convert-document/presentation/spreadsheet: same-format conversions now
return the original validated file instead of erroring or producing
corrupt tiny output
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(web): dropzone a11y, stale localStorage getter, dead code
- dropzone: stop making the whole drop-zone section clickable/focusable.
A section acting as an interactive element around a real upload button
is a nested-interactive-element anti-pattern that confuses screen
readers; drag-and-drop doesn't need focus semantics, only the button
fallback does. Keeps that button semantic and keyboard-reachable.
Updates the two e2e call sites that clicked the section directly.
- api, use-auth: read through window.localStorage via the existing API
storage helper instead of the bare global, which resolves to Node's
experimental localStorage getter under Vitest and threw
- find-duplicates-settings, info-settings, login-page: remove dead code
(unused zip-download handler, a stale mount-only effect dependency
that left cached info stuck at reused indices, an unused response
variable)
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(i18n): pt-BR, zh-CN, zh-TW were silently falling back to English
The locale loader looked up dynamic-import exports by the raw locale
code (mod["pt-BR"], mod["zh-CN"], mod["zh-TW"]), but those three modules
export camelCased bindings (ptBR, zhCN, zhTW) since identifiers can't
contain hyphens. The lookup returned undefined and every consumer
silently fell back to English for these three locales. Replaces the
generic lookup with explicit per-locale loaders so the mapping can't
drift out of sync again.
Also updates the dropzone helper copy across all 21 locales to match
the drag-only dropzone wording from the previous commit.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(docs): clear build warnings in the VitePress site
- config.mts: add an onwarn handler for the @vueuse INVALID_ANNOTATION
warnings emitted during the docs build
- deployment.md: the caddyfile code fence language isn't a shiki grammar
VitePress ships with, so it warned on every build; use txt instead
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* test(qa): update QA harness for the drag-only dropzone and regen metadata
- api-sweep, qa-helpers, verify-ai: add JSON-body tools, multi-input
secondary fixtures, async polling for slow valid jobs, 501
FEATURE_NOT_INSTALLED skip handling, and safer per-tool settings
- input-preview, pipeline-ui specs: update upload flow for the
drag-only dropzone surface
- add tests/fixtures/data/valid/chart.json, a valid chart fixture the
updated helpers route to
- regenerate tools-meta.json against current TOOLS[]
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(security): close a login timing side-channel, harden zip-slip tests
Found during a black-box security sweep of the real auth-enabled
production container: a nonexistent username returned 401 in ~3-10ms,
while a wrong password for a real user took ~35-42ms, because scrypt
verification only ran when a user row existed. That timing gap lets an
attacker enumerate valid usernames without ever guessing a password.
Now runs verification against a cached dummy hash on the unknown-user
path too, so both cases cost the same regardless of outcome.
extract-zip already had a relative-traversal regression test
(../evil.txt), but its absolute-path rejection branches
(name.startsWith("/") / startsWith("\\")) had none. Added the three
missing cases: deep relative traversal, absolute Unix path, and
Windows-style absolute path.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* test(qa): add UI-driven AI bundle install scripts
QA_PROMPT.md's Phase 2 requires installing AI models the way a user
does -- through the UI, on demand from HuggingFace -- and treats the
curl-based admin install endpoint as fallback-only. Nothing in the
harness actually drove that flow; tests/qa/seed-ai-models.sh installs
via docker exec + pip, which is further from a real user than even the
API fallback.
install-ai-bundles-ui.mts logs in, opens Settings > AI Features,
screenshots the pre-install state, clicks Install All, and screenshots
progress -- then exits, since installs continue server-side once
triggered. verify-ai-install-complete.mts polls bundle status,
screenshots the completed state, and runs one real tool per installed
bundle to prove the freshly-downloaded model actually executes.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(qa): correct the apiToolPath import in the AI verify script
Dynamic import of the package name failed under tsx's module resolution
from apps/api's node_modules context; use the same relative-path import
api-sweep.mts already uses successfully.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(web): correct AI bundle size estimates shown before install
Measured real downloads during GPU-node QA verification: photo-restoration
pulls ~4.4GB (was advertised as 800MB-1GB, off by 4-5x) and ocr pulls
~5.5GB (was advertised as 3-4GB). Both estimates only accounted for model
weights, not the pip dependencies (torch/paddle) that come down with them.
Updated to reflect actual total download size, since that's what a user
deciding whether they have the disk/bandwidth actually needs to know.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(web): make desktop Settings reachable when auth is disabled
AvatarDropdown (the only desktop entry point to Settings) was gated
behind `!isMobile && authEnabled`. With AUTH_ENABLED=false the synthetic
anonymous admin user should have full Settings access per how auth.ts
documents this mode -- and the mobile bottom nav already worked this way,
showing Settings unconditionally. Desktop just had a stray extra gate the
component doesn't need: AvatarDropdown already resolves its own username
internally (falling back to "admin") and reads authEnabled itself where
it actually matters (hiding the Logout button). Removed the outer gate;
verified end-to-end against a fresh AUTH_ENABLED=false instance -- avatar
now renders, Settings opens, shows the anonymous/Admin identity correctly.
Also documents (not changes) a related finding in install_feature.py:
detect_arch() always resolves amd64 hosts to the GPU-bundled archive
variant regardless of actual GPU presence, since no CPU-only amd64
archive is published to the bundle repo yet. Left as a code comment
rather than a behavior change, since requesting an unpublished archive
key would hard-fail installs entirely -- worse than the current
oversized-but-working download. Full detail in the QA report.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(ai): stop logging expected dispatcher reloads as crashes
After each AI bundle install the Python dispatcher reloads because the
venv changed, and after every app shutdown it's SIGTERMed. Both took the
close handler's `code !== 0` branch (SIGTERM makes the exit code null),
so they were counted as crashes -- producing an alarming "crash" line in
the logs and a pointless ~1s recovery backoff after each of 7 installs.
A `stopping` flag set in shutdown() lets the close handler tell an
intentional stop apart from a real crash. The request-timeout kill path
deliberately does not set it, so a genuinely hung script still records a
crash and the 5-in-60s permanent-disable threshold is untouched.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(api): return a clean message when content-aware resize times out
Carving a very high-resolution image down to a tiny target could exceed
the caire subprocess timeout, and the raw error forwarded to the user was
caire's terminal output -- ANSI color codes and progress-spinner control
characters -- instead of anything actionable. Now: the timeout path
throws a clear "timed out; try a smaller image or larger target" message
(keeping the raw stderr as `cause` for server logs); friendlyError()
strips ANSI/control chars centrally so any subprocess dump surfaced
through the shared sanitizer is plain text; and the content-aware-resize
route (a custom route that bypassed the sanitizer) now routes its error
paths through friendlyError like every other tool.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(ai): stop bundle installs from exhausting host disk
Installing an AI bundle on a tight-disk host could push the root
filesystem to zero bytes free after the preflight check had already
passed. Two root causes:
- move_tree used copytree+rmtree, so during the move the extracted
payload existed in both staging and the venv at once -- a full
transient doubling on disk. Rewrote it to rename entries (a cheap
metadata op on the same filesystem, no copy), falling back to a copy
only across filesystems.
- the preflight budget used the manifest's extractedSize verbatim, which
is 0 for several archives, collapsing the estimate to just the
compressed size. Added a conservative fallback (3x compressed) so a
missing value can't under-reserve.
Also added a real-on-disk re-check immediately before the first
destructive venv write (measuring the actual extracted payload and
whether the move needs extra space for a cross-filesystem copy), which
also now covers the offline-import path that previously skipped the disk
check entirely; wrapped the moves so an out-of-space failure returns a
clean actionable error instead of a traceback; and made the disk check
resolve the nearest existing ancestor so it never throws on a
not-yet-created venv path.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* feat(web): show the real per-arch AI bundle download size
The bundle cards and install prompt showed a hardcoded, architecture-blind
estimatedSize string. That's misleading: amd64 hosts always pull the
CUDA-inclusive archive (there's no CPU-only amd64 variant published), so a
bundle labelled "1-2 GB" can actually download several times that, while
arm64 pulls a much smaller archive for the same label. The manifest
already carries the real per-arch compressedSize (and extractedSize where
measured), so surface those: a new optional downloadBytes/installedBytes
on FeatureBundleState, populated in getFeatureStates() for this host's
arch (resolver mirrors install_feature.py detect_arch), shown by the UI
when present with estimatedSize kept as the fallback label. Also nudged
upscale-enhance's fallback string (4-5 -> 5-6 GB) to match its real
compressed size, consistent with the earlier photo-restoration/ocr fixes.
Fields are optional so demo/mock and existing tests stay compiling; the
manifest's extractedSize is 0 for a few archives, which now surfaces as
null rather than a bogus 0.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(web): move the AI install queue to the server so it survives tab close
Installing multiple bundles could silently lose all but the first. The
server rejected a concurrent install with 409, so the client worked
around it by queueing the rest in browser-local state and only POSTing
each once it saw the previous finish. A single POSTed install is durable
(the installer child is detached from the request), but a queued one had
zero server footprint -- close the tab mid-queue and those installs
vanished with no error, while the UI still showed them "Queued". The
client "mutex" didn't even serialize: the queued bundles' local waits all
resolved at once and raced into concurrent POSTs that 409'd each other.
Now the queue lives on the server (a small in-memory FIFO leaf module).
The install endpoint enqueues instead of 409-ing and returns
202 {jobId, queued}; a pump starts the next bundle when the current one's
child exits (and after an offline import releases the lock), all behind
the existing venv + file locks, which are unchanged. The client just
POSTs every bundle immediately and reflects the server-reported
queued/installing status; Install All fires all POSTs and lets the server
serialize them, keeping the one-shot retry-on-failure. Adds "queued" to
FeatureStatus (the bundle card already rendered that state) and surfaces
it from getFeatureStates. In-memory is deliberate: it matches the
existing contract (survives a tab close, not a server restart, which
already clears the lock on boot).
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
* fix(qa): don't log env-derived credentials in the AI-install script
CodeQL flagged clear-text logging of sensitive information: the login
status line interpolated the QA base URL and username (both read from
the process environment) into a console.log. Replaced with a static
message. QA helper only, but it's a real hygiene issue and cleared the
high-severity code-scanning alert on the PR.
Claude-Session: https://claude.ai/code/session_019fpSXhLGLXWwfyZY2tWhLG
AI feature installs now keep copied Python venv metadata (bin/pip shebang,
bin/activate, pyvenv.cfg) pointed at /data/ai/venv, so scripts no longer
silently fall back to the baked, read-only /opt/venv after the venv is
bootstrapped into /data. Fixes#127 (AI tools incompatible with PUID/PGID).
The entrypoint repairs both fresh bootstraps and already-stamped runtime
venvs (self-heals existing deployments on next restart, no reinstall
needed), with regression coverage for literal path replacement and binary
file safety.
Independently reviewed and verified: traced chown/gosu ordering in
entrypoint.sh to confirm no permission regression, reproduced the exact
issue #127 scenario (custom PUID + manual venv activation) in a live
container both before and after the fix, and ran the PR's own test suite
locally (16/16 passing).
Co-authored-by: SyntaxSawdust
Adds a prefilled 'Request a tool' affordance to the home search empty state and beneath weak results. Opens the in-app feedback dialog with a new search_miss source and a structured search_query when analytics is on; links to a prefilled GitHub Discussions (Ideas) post when off, so a request is never silently dropped. Reuses the existing feedback pipe, dialog, and analytics gate; no new storage. i18n across all 21 locales.
* fix(landing): correct PDF tool count to 28 in alternatives copy
The pdf section has 28 tools (section.test.ts asserts bySection('pdf')=28). PR #363 corrected the docs breakdown but the alternatives pages still said 40 PDF tools (the document-modality count, not the pdf section) with 200 for the rest. Update to 28 PDF tools and 212 for the non-PDF remainder.
* chore: use "200+ tools" for the tool-count claim across public surfaces
Replaces the exact '240 tools' count (which drifts as tools are added) with the stable '200+ tools' on README, the Docker Hub overview, landing pages, the docs site (meta, homepage, search), the API self-description, the demo OG tag, llms.txt, package.json, the branding readme, and the en/nl app strings. Per-modality breakdown tables stay exact. Leaves the architecture doc's technical 'tool routes' figure, an internal vitest comment, and a QA report line unchanged. Updates the two tests that assert the docs strings.
Lands five integrated branches: pipeline templates (#355), analytics opt-out (#354), 83 conversion presets bringing the catalog to 240 tools (#356), self-hosted positioning (#353), and e2e modernization (#351).
Integration fixes: aligned stale web analytics tests with the opt-out/allow-list model, closed 3 CodeQL incomplete-sanitization alerts in the i18n generator, resolved settings/index/docs/format-matrix conflicts, and corrected tool counts to 240.
The entrypoint only fixed volume permissions when started as root (chown +
gosu-drop to snapotter). Launched under a non-root/foreign UID (TrueNAS app
user, Kubernetes runAsUser, OpenShift) it did no permission setup, so /data and
/tmp/workspace -- owned by uid 999 from the image -- were not writable by the
running user. Uploads and processing then failed with a cryptic EACCES
("workspace folder is not writable") and AI bundle installs failed the same way,
while health checks still reported the container healthy.
- entrypoint: source new entrypoint-lib.sh; verify writability up front when
non-root, and as snapotter after chown when root (catches root-squashed
mounts), failing fast with an actionable message (which dir, uid/gid, how to
fix) instead of a late, cryptic EACCES
- Dockerfile: own /data and /tmp/workspace as snapotter:0, group-writable with
setgid, so an arbitrary UID with the root supplementary group (OpenShift /
Kubernetes fsGroup) can write; keep /opt/venv world-readable for the AI venv
bootstrap under arbitrary UIDs
- api: assert storage writability at boot (lib/storage-writable.ts), failing
fast with the same guidance even when the entrypoint is bypassed
- docs: add a Storage permissions section (named volumes, bind mounts, TrueNAS,
Kubernetes/OpenShift) and cross-link it from the security guide
Fixes#230
RAW (DNG) processing crashed on ImageMagick's deprecated ufraw-batch
delegate, which fails on modern formats such as iPhone ProRAW DNG.
Root cause: the dcraw_emu (LibRaw) decode tier read the wrong output path.
dcraw_emu APPENDS the output extension (raw-in-X.dng -> raw-in-X.dng.tiff)
but the code looked for raw-in-X.tiff (replaced extension), so readFile threw
on every RAW, the tier silently fell through to ufraw, and the 24MB TIFF
leaked into the temp dir on each attempt.
- Repair the dcraw_emu output path; clean it up in finally (fixes the leak)
- Prefer LibRaw full decode over embedded-preview extraction so a
full-resolution RAW is never silently returned as a reduced-size preview
(sample DNG: was 1024x683 preview, now 3474x2314 full)
- Add RAW decode regression tests (DNG full-resolution + all 6 RAW formats);
these were absent, which let the bug ship
- Install libraw-bin on CI test runners so dcraw_emu is actually exercised
Replace textual prefix matching in isPrivateIPv6() with numeric bit-prefix matching via ipaddr.js. Closes reachable classifier bypasses for IPv4-compatible IPv6 (::a.b.c.d), site-local fec0::/10, and the full fe80::/10 link-local span, on top of the IPv4-mapped forms. Embedded IPv4 (mapped and compatible) is run through the full isPrivateIPv4 classifier; unparseable input fails closed.
Follow-up hardening to f64cbdda.
Reported-by: tonghuaroot
