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.
RealESRGAN's enhance() and rembg's remove() run in one opaque call, so the
progress bar froze at 30% for the whole inference. Add a time-based
heartbeat that advances the bar in a background thread while the model runs
and stops when it returns, so the bar moves instead of freezing. Verified
end to end on a GPU box (forced CPU): a 55s upscale emitted 26 steady ticks
then completed; background removal too.
Fixes#591
Add a one-line notice to the upscale and background-removal settings that
heavy AI runs much slower without a GPU and a large image can take minutes,
so a CPU-only self-hoster isn't caught off guard by a slow run. Translated
into all 21 locales.
Refs #591
Adds a Rounded Crop image tool for logo, favicon, and app-icon work. It masks the framed square to a rounded rectangle (with a corner-radius control) or an iOS-style squircle, reusing circle-crop's zoom/offset framing, border ring, background fill, and output-size options. Includes translations across all 21 locales.
Closes#601
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
Pin the compose Postgres healthchecks to POSTGRES_DB (pg_isready was
defaulting to the username, silently reporting healthy while spamming
FATAL logs when USER and DB differ), and make docker/wait-for-postgres.mjs
log the target host and error code instead of a silent retry loop. Adds a
change-me note next to the default password in README and Docker Hub.
Refs #592
Two OCR-install fixes surfaced while verifying the accurate-OCR (v3 runtime) path end to end:
- Installer timeout must be a safe integer, not a performance.now() float. With the default INSTALL_MAX_MS this failed every accurate-OCR install via the app right after the download (masked by the unpublished runtime; CI drives install_runtime.py directly so it never surfaced). Fixed via remainingInstallerTimeoutMs().
- Classify an absent or forbidden runtime index (401/403/404/410) as OcrRuntimeNotPublishedError with a clear "Fast OCR still works" message instead of a raw HTTP 404, without retrying.
Refs #552
Closes#565. Wires the fileId/saveMode pair into the ocr, erase-object, remove-background, background-replace, and blur-background submitters so the library save-mode selector works for them; remove-background's two-phase effects route now auto-saves the final composite instead of the transparent intermediate.
Adds an opt-in High Quality mode to the Object Eraser, backed by a new inpaint-hq feature bundle (Stable Diffusion 1.5 inpainting via diffusers). The default fast LaMa path is unchanged. Both arch archives are published to deepsafe/feature-bundles and the manifest carries their real sha256/sizes.
Verified end to end: a fresh container pulls the bundle from HuggingFace, checksum-verifies it, extracts torch/diffusers plus the fp16 model, and the HQ sidecar erases a large object with a plausible fill.
Refs #141
Exports a pure buildRedirects() from generate-redirects.mjs and adds a unit test asserting the committed apps/landing/public/_redirects matches it, so tool additions can't silently leave the generated redirects stale (see #573).
Fixes#568. Adds a real --color-ring token (ink orange #A85518 light / #F0A766 dark) and sweeps all 57 focus-indicator occurrences onto it: soft opacity rings blended to 1.2-1.7:1, border-only indicators sat at 2.6-3.0:1, and four focus:ring-ring sites referenced a token that never existed. Sponsor button keeps pink via pink-700; range sliders and the file list gain their missing keyboard indicators; landing skip link and form borders hardened; the palette contrast guard pins the ring at 3:1 in both themes.
Fixes#557. Vivid fill, ink label: brand #E07832 stays on fills while primary-foreground flips to #1A1814 (5.83:1); new theme-aware ink tokens carry orange, destructive, and success text roles; opacity-modified text purged; landing, demo, and the docs fund button retuned. Guarded by a CSS-parsing unit contrast test, rebuilt axe baselines with zero contrast entries, a new landing axe smoke, and fully regenerated darwin visual baselines.
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
The Convert Audio tool promised configurable bitrate, sample rate, and channel count, but only format and bitrate were exposed. Adds an optional sampleRate setting (8000 to 96000 Hz, omitted = preserve source) wired through the Zod schema, the FFmpeg -ar flag, the standalone settings panel, and the pipeline builder controls.
Impossible combinations fail loudly instead of degrading silently: MP3 + 96000 Hz is rejected (libmp3lame caps at 48 kHz), and MP3 bitrates above the encoder ceiling at low rates (64 kbps at 8 kHz, 160 kbps at 16/22.05 kHz) are rejected rather than clamped. The UI offers only legal combinations and sanitizes stored pipeline settings on load.
Docs updated in English plus all 20 localized pages with refreshed i18n_source_hash stamps; two new UI strings added to all 21 locales.
Fixes#558
On mobile browsers 100vh is the tall viewport (URL bar retracted), so the
full-height shells (h-screen + overflow-hidden) rendered their bottom strip
below the visible area with no way to scroll to it. After an upload the tool
page's bottom control is the "Process" peek bar, so it landed off-screen on
every tool.
Switch the fixed-height shells from h-screen (100vh) to h-dvh (100dvh), which
tracks the visible viewport. AppLayout covers every tool page; the same sweep
covers the editor, login, change-password, 404, and the error/loading screens.
On desktop dvh equals vh, so desktop is unchanged. Adds a source-level guard
(tests/unit/web/app-shell-viewport.test.ts) since headless Chromium has no URL
bar and cannot reproduce the bug.
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.
A genuine pdf.js load failure (corrupt or password-protected file) was swallowed by the same catch that silences teardown rejections, leaving a blank canvas that looks like it is still loading. The two cases are now distinguished by the load effect's cancelled flag, and a real failure renders a clear message pointing at Unlock PDF for encrypted files. New loadFailed string in all 21 locales.
Item 4 of #478.
Both PyPI onnxruntime flavors unpack into the same site-packages directory, so a bundle carrying the CPU build (transcription, via faster-whisper) overwrote the GPU build's native libraries during install while the stale onnxruntime_gpu dist-info kept claiming otherwise. Every ONNX-backed tool then silently ran on CPU.
The installer now reconciles the flavor before the venv merge and the GPU build always wins, in both install orders; reinstalling any GPU bundle repairs a previously clobbered venv. gpu.py's warning now says exactly that. Build-side, build-bundle.sh gains the same reconcile and verify-bundle-compatibility.sh layers bundles through the real installer merge and asserts a single flavor.
Verified live on an RTX 4070 against the published bundles: reproduced the clobber with the stock installer, then confirmed both the prevention and repair paths with the patched one.
Fixes#490
The password policy toggles (uppercase, lowercase, digit, special) can all be switched off in Settings -> Security, but the minimum-length input clamped at 4, so homelab admins couldn't deliberately allow short passwords. The API never enforced a floor; only the UI did. Lower the input floor to 1 and pin it with a test.
Closes#136
Defer the overall deadline until the first fetch is dispatched (the deferred-deadline pattern the sibling 20ms test already uses), so a loaded CI runner cannot abort before fetch and randomly fail Unit Tests.
Briefly retry Postgres and Redis connectivity at boot (waitForService, DB_STARTUP_TIMEOUT_MS) so an ordered-but-not-yet-ready dependency recovers cleanly instead of crash-looping.
removeBackground failures wrap in a SafeError so the specific reason survives the Sentry scrubber; the OOM lighter-model fallback and bridge SafeError passthrough are preserved.
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.
Route both Python exit paths through pythonExitError so the reason survives the scrubber; OOM/segfault stay operational and keep "out of memory" for the lighter-model fallback.
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".
Route every doc_* helper JSON.parse through a guarded helper; non-JSON stdout now yields a diagnosable SafeError with the raw output in the cause instead of a bare SyntaxError.
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
Fixes#529 (opened investigating #515).
Setting MFA policy to "required"/"admins only" saved regardless of whether the mfa enterprise feature was licensed, and there was no enrollment UI at all, so any instance that flipped the toggle locked every unenrolled user out with no way back in. The login page and Settings save also both collapsed the resulting error into a generic message, hiding the real reason.
- Reject saving mfaPolicy to admins_only/required server-side unless mfa is licensed
- Surface the specific server error on login and on a failed settings save instead of a generic fallback
- Add a self-service two-factor authentication enrollment flow (QR code, manual entry, recovery codes, verify, disable) so a licensed admin can actually satisfy the policy before it's enforced
- Fix a pending-enrollment dead end, silent error swallowing in verify/disable, and a silent clipboard-copy failure on the recovery codes screen
- Add the integration test that actually proves the fix: a real login attempt returns 403 MFA_ENROLLMENT_REQUIRED
Add a proportion chip row (Free, Original, 1:1, 4:3, 3:2, 16:9, 3:4, 9:16) to the Resize tool's Custom tab. Picking a ratio locks width and height so editing one recomputes the other, and prefills the largest box of that ratio that fits the source so it never upscales. Free stays the default, preserving existing behavior. Replaces the previously non-functional lock-aspect button. Frontend only, no backend or schema change; adds strings to all 21 locales.
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.
Adds install method and hardest-setup-area as two optional questions to the first-login usage survey, so these fields are collected at onboarding scale instead of only via the low-traffic admin installer card. Both stay null until chosen and are omitted from the payload when unanswered. No API change and no new i18n; reuses the shared feedback schema and existing locale strings.
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.
Seed eight sample files across modalities in the demo Files library and mock the list (search + pagination), details with version history, bulk delete, and thumbnails. Thumbnails are generated colour-coded SVG tiles served through the interceptable fetch path, so the grid always renders with no real file bytes. Uploads/processing stay disabled.
The demo signs in as an admin on load (no login/change-password screen; /login and /change-password bounce to the dashboard), authEnabled stays true so the People/Teams/Roles/Security settings tabs remain available, the banner notes the admin data is sample data, and the mobile editor icon is an inline SVG so it always renders.
Rewrite the demo mock API around a seeded in-memory dataset with correct response shapes so the People/Teams/Roles/Audit/Usage/API-keys tabs stop crashing (the /auth/users vs /v1/users mismatch caused users.filter() on undefined) and show realistic sample data. In-memory CRUD makes the settings buttons work. Copy edit-image.png into the demo so the mobile editor icon renders. Adds unit shape guards and an e2e admin-settings walkthrough.
Uninstalling a bundle only deletes its downloaded model weights, never the
shared venv's site-packages, so self-hosters who already hit an AI bundle
conflict (e.g. the scipy ABI strand) have no clean self-service path via
uninstall+reinstall: reinstalling just overlays corrected files on top of
stale ones. Adds POST /api/v1/admin/features/reset, which wipes
/data/ai/{venv,models,pip-cache}, resets installed.json, and reseeds a real
working venv from the image's baked /opt/venv (extracted docker/reseed-ai-venv.sh,
now shared with entrypoint.sh's existing base-venv-upgrade bootstrap instead
of duplicating that logic) -- leaving an empty venv directory here would
make the very next install fail with "spawn .../python3 ENOENT", caught by
testing this live rather than assuming it. Ships with a matching Settings UI
section (inline confirm, same pattern as per-bundle uninstall) and strings
across all 21 locales.
Verified against a real snapotter/snapotter:1.17.2 image migrated to 2.0.0,
with real multi-GB bundles installed (background-removal + OCR): confirmed
the migrated instance's inherited python3.11 venv (2.0.0 itself uses 3.12)
still imports the fixed scipy/numpy/paddleocr correctly, then reset + real
reinstall + actual tool execution (remove-background, verified output image)
all worked end-to-end.
* 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.
