- Create 53 per-tool VitePress documentation pages with accurate
parameters from Zod schemas, example requests, and response formats
- Add root llms.txt for LLM-friendly repo browsing
- Fix OpenAPI spec: add auth and 422 error schemas to
edit-metadata/inspect and strip-metadata/inspect sub-routes
- Fix tool count inconsistency (52 -> 53) across landing site,
e2e tests, and local docs
- Rename color-adjustments.ts to adjust-colors.ts to match tool ID
- Update VitePress sidebar with all 8 tool categories and top nav
* feat(a11y): add i18n keys for ARIA labels and screen reader text
* fix(security): harden API against pentest findings
- Default TRUST_PROXY=false to prevent XFF rate limit bypass (PT-01)
- Return 400 instead of 500 on malformed JSON input (PT-03)
- Default MAX_PIPELINE_STEPS=20 to prevent DoS (PT-04)
- Validate clientJobId length (max 128) across all routes (PT-06)
- Add security headers to all reply.hijack() streaming responses (PT-07)
- Sanitize usernames in audit log to prevent stored XSS (PT-08)
- Block TRACE method with 405 response (PT-10)
- Add 429 RateLimited response to OpenAPI spec (PT-12)
- Default MAX_SVG_SIZE_MB=50 to limit SVGZ decompression (PT-13)
- Pin Dockerfile base images by digest
- Sanitize OIDC IdP error and sub claim in audit log
- Sync Docker compose/Dockerfile defaults with env.ts
* feat(a11y): convert all hardcoded aria-labels to i18n keys
Replace 49 hardcoded aria-label="..." strings across 25 files with
their corresponding t.a11y.* and t.common.* i18n references. Add
useTranslation import and hook call to 15 components that lacked it.
Zero hardcoded aria-labels remain in the codebase.
* feat(a11y): add aria-labels to icon-only buttons, aria-hidden on decorative icons, sr-only status text
* feat(a11y): add aria-live regions for processing status announcements
* feat(a11y): add skip-nav link, route announcer, main content landmark, and page h1 elements
* feat(a11y): add prefers-reduced-motion support, preserve functional spinners
* feat(a11y): add useFocusTrap hook for modal focus management
* feat(a11y): add focus trapping and dialog roles to all modals
* feat(a11y): add toggle switch roles, form labels, and error association
* fix(a11y): fix contrast failures, touch targets, and add nav landmark to sidebar
* fix(a11y): add role=switch to remaining toggle buttons found in verification sweep
- Fix selective metadata stripping (P1): use Sharp's keepExif()/keepIccProfile()
instead of broken withMetadata({}) that preserved everything
- Fix meme font mapping (P1): add ArchivoBlack and ComicNeue fonts, map
arial-black and comic-sans to correct TTF files instead of Anton
- Fix meme contentType (P2): detect actual output format from Sharp metadata
instead of hardcoding image/png
- Fix info/text-overlay/color-palette i18n (P2): wire up existing translation
keys that were imported but never used
- Fix info and color-palette displayMode (P2): change from before-after to
no-comparison since neither tool produces a processed image
- Add missing i18n keys across all 21 locales
- Update displayMode test assertions
* fix(security): harden SVG sanitizer, rate limiting, and analytics defaults
- SVG: add control-char stripping in href values to block whitespace/null-byte
obfuscated javascript: URIs; block <feImage> with external href (SSRF via
SVG filter primitives); expand test suite to 32 inline bypass payloads
- Rate limiting: add per-route limits on tool endpoints (60/min) and batch
(20/min); fix compose files defaulting RATE_LIMIT_PER_MIN to 0 which mapped
to 50,000 in code; simplify rate limit registration to use env.ts default
- Analytics: default ANALYTICS_ENABLED to false so self-hosters do not
unknowingly send telemetry
- Docker: add --max-time 5 and -s flags to compose healthcheck curl commands
* fix: remove stale login limit bypass, reduce error log noise, clean up fixtures
- Fix getLoginAttemptLimit() ignoring LOGIN_ATTEMPT_LIMIT when global rate
limit exceeded 1000/min, which let the global limit override the stricter
per-route login brute-force protection
- Downgrade rate limit 429 responses from error to warn level in the global
error handler to avoid log noise and unnecessary Sentry reports
- Log 4xx client errors at warn level instead of error level
- Remove 11 orphaned SVG attack fixture files replaced by inline test payloads
- Apply biome formatting fixes to web app components
- Add required S3 credentials to loadEnv test when STORAGE_MODE=s3
- Update bento-grid test tool counts from 52 to 53 for html-to-image
Support reading secrets from mounted files instead of plain-text
environment variables, following the standard Docker/Kubernetes
convention used by MariaDB, Postgres, and Stirling-PDF.
Supported vars: DEFAULT_PASSWORD, S3_ACCESS_KEY_ID,
S3_SECRET_ACCESS_KEY, OIDC_CLIENT_SECRET, COOKIE_SECRET,
SNAPOTTER_LICENSE_KEY.
The WHATWG URL parser canonicalizes dotted IPv4-mapped IPv6 literals
(e.g. ::ffff:127.0.0.1) to hex form (::ffff:7f00:1). The SSRF guard
only checked the dotted form, so hex literals bypassed the private-IP
classifier and allowed access to loopback, cloud metadata, and RFC1918
addresses.
Decode hex IPv4-mapped suffixes to dotted IPv4 before the private-range
check.
Reported-by: tonghuaroot
Add the enterprise package with Ed25519 license key validation and
feature gating. Enterprise code lives in the public repo under a
proprietary license (Cal.com/PostHog model), protected legally, not
by code hiding.
Implement S3-compatible storage backend as the first enterprise
feature. The file-storage module now delegates to either local
filesystem or S3 based on STORAGE_MODE env var. Works with AWS S3,
Cloudflare R2, DigitalOcean Spaces, MinIO, and any S3-compatible
provider. Workspace files remain local (ephemeral processing).
New env vars: STORAGE_MODE, S3_BUCKET, S3_REGION, S3_ENDPOINT,
S3_ACCESS_KEY_ID, S3_SECRET_ACCESS_KEY, S3_FORCE_PATH_STYLE,
S3_PREFIX, SNAPOTTER_LICENSE_KEY.
Tested against MinIO: 10 S3 integration tests + 82 existing tests
pass with zero regressions.
Phase 1 quick fixes:
- Add isInputFocused() guard to Cmd+A/D/T/J shortcuts (P1-7)
- Add Go Home button to tool-not-found page (P2-30)
- Fix hardcoded "Import from Library" string in file library modal (P2-28)
- Fix TeamEntry.id type from number to string to match API (P2-6)
- Add eye toggle to confirm password field (P2-10)
- Add Apply/Cancel buttons to Free Transform options bar (P2-14)
Phase 2 state fixes:
- Add sessionStorage persistence to pipeline store (P1-26)
- Fix Free Transform 0 dimensions by falling back to selection bounds (P1-6)
Phase 3 editor features:
- Constrain brush/eraser drawing within active selection bounds (P1-5)
- Add feather radius control to selection options (P2-21)
- Add flow control slider to brush options (P2-22)
- Add brush/block mode selector to eraser options (P2-23)
- Add estimated file size display to export dialog (P2-18)
Phase 4:
- Add Playwright e2e tests for key fixes (404 page, routing, pipeline persistence, export dialog)
- Expose birefnet-hr-matting in UI (People/Ultra) and fix model defaults
(People/Max now uses birefnet-matting for true alpha matting)
- Add output format selector (PNG/WebP/AVIF) with lossless alpha support
- Add edge smoothing post-processing (Off/Light/Medium/Strong) via
morphological mask refinement to reduce gray halo artifacts
- Add color decontamination to remove background color spill from
semi-transparent edge pixels
- Thread new settings through full stack: frontend -> API schema ->
Python sidecar -> Sharp effects pipeline
- Add i18n keys for all 21 locales
- Add unit tests for new option serialization (3 tests)
- Add integration tests for new settings validation (4 tests)
* fix: extend SSE result retention from 2 to 10 minutes for mobile recovery
* fix: check server health on tab visibility change for mobile recovery
* fix: recover feature install SSE on tab visibility change
* fix: reconnect SSE on tab visibility change in tool processor
* fix: reconnect SSE on tab visibility change in pipeline processor
* test: add visibility recovery tests for connection monitor
The Open File button in the Files section did nothing due to a race
condition where the home page reset the file store on mount before files
from handleOpenFile could render. Upload on the files page used fetch
with no timeout, progress, or retry, causing silent failures on mobile
and slow connections. SSE connections for job progress had no keepalive
pings, allowing reverse proxies to kill idle streams.
Add pre-built release archives (Linux amd64/arm64) to the release
workflow, published as GitHub Release assets. Each archive is a
self-contained tar.gz (~240MB) with built frontend, API source,
and production node_modules. Users extract and run without needing
pnpm build.
Also includes AI install manifest fixes for Proxmox/bare-metal users:
- Pin setuptools<75 for Python 3.13 basicsr compatibility
- Pre-install basicsr with --no-build-isolation before realesrgan
- Loosen mediapipe pins from == to >= for Python 3.13 wheels
- Add retry logic to HuggingFace model downloads
AVIF encoding is slow in CI without hardware acceleration, causing
SVG->AVIF and WebP->AVIF tests to timeout at 30s. Set 120s timeout
for all AVIF output tests, matching the existing exotic format timeout.
- Fix installAll tests timing out by returning installed status from
refreshBundles mock (prevents infinite retry loop) and increasing
waitFor/test timeouts to accommodate the 2s inter-install delay
- Fix zustand-stores installAll test with same refreshBundles mock fix
- Icon-map test already fixed to derive from shared constants
- Separate unit tests (fast, no system deps) from integration test
shards to prevent vi.mock db leakage across test types
- Fix hero test: update expected subtitle to match current copy
- Fix FAQ test: add missing json-ld alias in vitest config and mock
- Integration tests run in 4 parallel shards (30min timeout each)
The sample.heif was 8736x5856 (2.5MB), causing each processing
operation to take 17-38s in CI. Resized to 1432x960 (224KB) and added
a 15-minute timeout to the test job.
When AUTH_ENABLED=false, seed an "anonymous" user row in the users
table so API keys, pipelines, and user files don't fail with FK
constraint violations. Previously, the synthetic anonymous user only
existed in memory (attached by the middleware), but any DB operation
referencing userId "anonymous" would violate foreign key constraints.
Also adds 25 new tests covering:
- Integration: ensureAnonymousUser, FK constraints, settings save,
API key and pipeline operations for anonymous mode
- Frontend: useAuth hook anonymous happy path (role, permissions,
hasPermission, session endpoint bypass)
- Frontend: settings dialog nav filtering (authRequired hides
security/people/teams/roles when auth disabled)
- Backend: session endpoint returns admin role when auth disabled
When AUTH_ENABLED=false, the anonymous user was assigned the "user" role
which lacks settings:write permission, making all settings saves return
403. Since no admin exists when auth is disabled, settings were
permanently read-only. Promote the anonymous user to "admin" so the
single user has full control of the instance.
Also adds DEFAULT_TOOL_VIEW env var (sidebar|fullscreen) following the
existing DEFAULT_THEME pattern, seeded via ensureDefaultSettings() on
first boot.
Closes#135
ai-canvas-expand was added to constants.ts and route files but never
added to the landing page bento grid, causing all hardcoded counts
to remain at 51. This updates all references across source, docs,
i18n, and tests to reflect the correct count of 52 tools.
- fix(db): migration 0012 column order mismatch causing NOT NULL
constraint failure on existing databases; use explicit column
mapping instead of SELECT *
- fix(db): disable FK checks during migrations to allow SQLite
table-recreation pattern (DROP + RENAME)
- fix(security): filter cookie_secret and instance_id from settings
API response for non-admin users
- fix(lint): resolve all 7 API lint warnings (noParameterAssign,
noImplicitAnyLet) in compose, image-enhancement, and workspace
- fix(docs): correct permission count from 16 to 14 in CLAUDE.md
- fix(e2e): resolve 44 Playwright test failures across 8 spec files
including locator specificity, compress mode defaults, format count,
restore-photo UI drift, stitch image count, GIF animated fixtures,
submit button timing, and processing timeouts
Add complete i18n infrastructure with 21 supported languages:
English, Simplified Chinese, Traditional Chinese, Japanese, Korean,
Spanish, French, Italian, Brazilian Portuguese, German, Dutch, Swedish,
Russian, Polish, Ukrainian, Arabic (RTL), Turkish, Hindi, Vietnamese,
Indonesian, and Thai.
- I18nProvider context with three-tier locale detection
(user preference > navigator.languages > instance default > English)
- ~1500 translation keys per locale with TypeScript-enforced completeness
- Dynamic code-splitting: only the active locale is loaded at runtime
- Language selectors in footer, login page, settings, and mobile sidebar
- Arabic RTL support with CSS logical properties across all components
- Tool names, descriptions, and categories translated via i18n helpers
- Public API endpoint GET /api/v1/config/locale for instance default
- Multi-script font stack (CJK, Arabic, Devanagari, Thai, Cyrillic)
- format() and plural() helpers for interpolation and pluralization
- API error translation mapping (translateApiError)
- 36 Playwright e2e tests verifying all 21 locales load correctly
- 25 unit tests for format, plural, locale detection, and completeness
- Updated translations.md docs and CLAUDE.md conventions
The GPU detection in gpu.py had two issues preventing GPU usage in
containers (especially rootless podman with CDI):
1. When torch was installed but torch.cuda.is_available() returned
False, the function returned immediately without trying the
ONNX Runtime + nvidia-smi fallback. This meant a CPU-only torch
build (installed before GPU was available) would block all GPU
detection, even for ONNX-based tools.
2. The failure logged a generic "torch loaded but CUDA not available"
with no diagnostic information, making it impossible to debug
whether the issue was a CPU-only build, missing libraries, or
device permissions.
The fix restructures gpu_available() into three detection tiers
(torch -> ONNX Runtime -> nvidia-smi) that always fall through on
failure. When torch CUDA fails, it now checks torch.version.cuda to
distinguish CPU-only builds from CUDA builds that can't access the
GPU, and logs LD_LIBRARY_PATH, torch.cuda.init() errors, and
nvidia-smi results.
Also fixes two env var passthrough bugs in buildMinimalEnv():
- SNAPOTTER_GPU was never passed to the Python subprocess, so the
user-facing GPU override env var had no effect
- MODELS_DIR was a dead entry (never set as env var); replaced with
MODELS_PATH which the Dockerfile sets and Python scripts read
Closes#134
The hardened SVG sanitizer may reject certain SVGZ content that was
previously accepted. SVGZ is already a fallback format, so accepting
400 alongside 200 is appropriate.
Corrupt image data now returns 400 (invalid image) instead of 422
(processing failure) because validation catches it earlier.
Long usernames now return 400 (Zod max length) instead of 401.
archiver v8 changed its default export, breaking all ZIP-producing
tools (pdf-to-image, split, batch, favicon, bulk-rename, svg-to-raster).
Reverted to v7 -- the lodash vulnerability via archiver is _.template
which is never called directly.
@fastify/static v9 has breaking changes incompatible with the current
static file serving setup. Reverted to v8 -- the path traversal CVEs
in v8 are mitigated by the existing path traversal guards in files.ts.
Updated edge-cases test to expect 400 for >64KB settings payloads
(new security limit).
When 88 integration test files run sequentially in a single-fork
Vitest process, the SQLite WAL file grows unbounded. Adding a
TRUNCATE checkpoint after each test app cleanup prevents the SHM
mapping from exceeding its size limit.
validateImageBuffer catches corrupt image data before processing
reaches the tool handler, so the correct status code is 400 (bad
request) rather than 422 (processing failure). Also fix SVGZ
watermark validation by returning early for compressed SVG (Sharp
cannot read gzip-compressed SVGZ directly) and passing the actual
watermark filename to validateImageBuffer for correct format
detection.
Auth: login rate limit 30/min (was 500), global rate limit 1000/min (was
unlimited), password/username max lengths on all Zod schemas, session
invalidation on role change, API key legacy scan bounded to 100 keys.
SVG: hardened regex sanitizer with CDATA stripping, XML entity decoding,
set/animate/iframe/embed blocking, comprehensive data: URI blocking,
use element external href blocking. 11 attack payload fixtures added.
SSRF: fixed DNS rebinding TOCTOU by pinning resolved IPs via custom
HTTP/HTTPS agents. Added 6to4 and NAT64 to blocked IPv6 ranges.
Docker: capability dropping (cap_drop ALL + minimal cap_add), resource
limits (4g/8g mem, 512/1024 pids), healthcheck timeout, password
removed from startup banner, default password warning comments.
Network: CSP and HSTS applied in all environments (not just production),
stack traces removed from all error responses, internal paths stripped
from error details, per-route rate limits on uploads (60/min) and URL
fetches (200/hour).
Files: exclusive temp file creation (O_EXCL), disk space circuit
breaker, per-user storage quotas, settings payload 64KB size guard.
Python sidecar: script name allowlist in dispatcher, minimal environment
for subprocess spawns.
Dependencies: fixed 6 production CVEs (drizzle-orm, fastify, fast-uri,
@fastify/static, next, archiver/lodash). Pinned all GitHub Actions to
SHA hashes.
114 security tests added. Full OWASP Top 10 penetration test matrix
verified against production Docker container (30/30 pass after
hardening).
SnapOtter