Register custom BiRefNet-matting ONNX session in install_feature.py so
rembg.new_session("birefnet-matting") no longer raises ValueError during
on-demand installs. The session was already registered in remove_bg.py
(runtime) and download_models.py (build-time) but was missed in the
install path, causing background-removal bundle installs to always fail.
Send JSON body on install/uninstall POST requests to avoid Fastify 5's
strict content-type parser rejecting body-less POSTs with 415.
Fix error message extraction to preserve structured {"error": ...} JSON
from the Python script and filter out pthread_setaffinity_np noise.
- Update OCR engine expected name from "paddleocr" to "paddleocr-v5"
to match actual PaddleOCR PP-OCRv5 engine (eliminates spurious
fallback warning in logs)
- Show "waiting for AI sidecar startup" instead of misleading
"No GPU detected" when Python dispatcher hasn't reported yet
- Fix playwright.docker.config.ts testDir to ./tests/e2e-docker
and align auth storage state path with auth.setup.ts
Closes#17, #18, #19, #31, #32, #33, #34
Format preservation (#17, #18, #19):
- Add resolveOutputFormat to rotate, resize, text-overlay, watermark-text,
border, replace-color, blur-faces, upscale, erase-object, restore-photo
- Alpha-aware fallback: border with corner radius/shadow and replace-color
with makeTransparent fall back to PNG for non-alpha formats (JPEG)
- Python sidecar tools (blur-faces, upscale, erase-object) now convert
PNG output back to input format, matching restore-photo/colorize pattern
- Upscale and erase-object default to "auto" format detection instead of PNG
Dispatcher stability (#31, #32):
- Add gc.collect() and torch.cuda.empty_cache() after each dispatcher request
- Add configurable max_requests (default 50) for periodic dispatcher restart
- Add exponential backoff to dispatcher crash recovery in bridge.ts
- Circuit breaker: 5 crashes within 60s permanently disables dispatcher
- Reset crash counter on successful dispatcher startup
Health & security (#33, #34):
- Export getDispatcherStatus() from @snapotter/ai with running/ready/failed/
gpu/pid/consecutiveCrashes fields
- Admin health endpoint now includes full dispatcher status
- Add pip-audit job to CI workflow for Python dependency scanning
Add djxl (libjxl-tools) as primary JXL decoder with ImageMagick
fallback — fixes JXL format failures on Ubuntu where stock ImageMagick
lacks a JXL delegate. Also make Playwright Docker config respect
BASE_URL env var for testing against remote containers.
Three disconnected systems caused the theme to never apply from server
settings: the DEFAULT_THEME env var was parsed but never seeded to the
database, the settings store ignored defaultTheme from the API, and the
settings dialog wrote to the DB without updating the active theme store.
- Seed DEFAULT_THEME and DEFAULT_LOCALE env vars into the settings table
on first startup (ensureDefaultSettings in index.ts)
- Add applyServerDefault() to theme store that applies the server's
default theme only when the user hasn't made an explicit choice
- Extract defaultTheme from the settings API response and apply it on
fresh sessions (no localStorage preference)
- Apply theme immediately when admin saves settings
- Allow "system" as a valid DEFAULT_THEME env var value
- Unit: 1,353 tests (42 files) — +256 new tests covering AI bridge
modules, image-engine sharpen/optimize-for-web, Zustand stores, and
icon-map validation
- Integration: 1,640 tests (57 files) — +826 new tests across all
tool routes, pipeline/progress/batch infrastructure, user-files,
edit-metadata, and a 321-test cross-format matrix
- E2E-Docker: 389 passing (20 spec files) — 6 new spec files for
batch processing, format conversion, layout, optimization,
watermark/overlay, and pipeline chains. Tests verified against fresh
Docker container with all 6 AI bundles installed.
Bug fixes discovered during testing:
- fix(compress): SVG/BMP/exotic formats crashed Sharp encoder — added
format-safety fallback to PNG
- fix(rate-limit): increase default login attempt limit from 10 to 500
per minute — previous value caused false test failures and is too
restrictive for a self-hosted app
- fix(auth.setup): wait for consent button visibility before clicking
to prevent flaky E2E-Docker auth setup
- Add Cloudflare Pages deployment for landing page (snapotter.com) and
docs (docs.snapotter.com)
- Create deploy-landing.yml and update deploy-docs.yml workflows
- Update CI to ignore apps/landing/** paths
- Fix logo transparency (remove white background) across all apps
- Recreate social-preview.png with SnapOtter branding
- Update all docs URLs from GitHub Pages to docs.snapotter.com
- Update VitePress config: light theme default, fix llms.txt paths
- Add .vitepress/cache/ and .env.* to gitignore
PostHog and Sentry were silently disabled in production Docker builds.
The app runs as ESM ("type": "module") via tsx, where require() is not
defined. The catch blocks swallowed the ReferenceError, leaving both
clients as null. Switch to await import() and store the Sentry module
reference for use in the error handler.
Pipeline steps and batch size are now unlimited by default. The old
"20 steps" and "200 images" figures had no basis in the actual code
(MAX_BATCH_SIZE already defaulted to 0/unlimited). Both remain
configurable via MAX_PIPELINE_STEPS and MAX_BATCH_SIZE env vars.
Also includes updated hardware requirements and sidebar nav from
prior documentation audit.
When auth was disabled, users could log out, reach the login page,
and authenticate with the default admin/admin credentials to gain
full admin privileges — defeating the purpose of AUTH_ENABLED=false.
Defense-in-depth fix across five layers:
- Skip ensureDefaultAdmin() when auth is disabled (no admin user seeded)
- Return 403 from POST /api/auth/login when auth is disabled
- Return synthetic anonymous user from GET /api/auth/session when auth is disabled
- Hide logout button in settings when auth is disabled
- Redirect /login and /change-password to / via AuthGuard when auth is disabled
Closes#90
* feat: allow multi-file selection for automation pipeline
Add two ways to import server-stored files into the pipeline:
1. Files page: "Pipeline" bulk action button and "Open in Pipeline"
button in file details panel — navigates to /automate with selected
file IDs via React Router state.
2. Automate page: "Import from Library" button opens a modal with
thumbnails, search, and multi-select checkboxes to pick files from
the user's server-stored library.
Both paths download the selected files and load them into the existing
useFileStore, reusing the batch pipeline processing infrastructure.
Closes#35
* fix: resolve 8 pre-existing test failures across unit and integration suites
- file-validation.ts: Return valid:false when Sharp fails to read
metadata for standard formats (PNG, JPEG, BMP) instead of silently
accepting corrupt buffers. CLI-decoded formats already skip Sharp.
- pipeline.ts: Enforce hard cap of 20 steps via .max() instead of
relying on MAX_PIPELINE_STEPS env var (default 0 = unlimited).
Tighten name limit to 100 chars and description to 500 chars to
match test expectations.
- env.ts: Change MAX_LOGO_SIZE_KB default from 2048 to 500 to match
the branding upload size limit the tests verify.
- Convert all AI bridge inputs to PNG before writing to disk so PIL can
read AVIF/WebP/TIFF (7 bridge files; face-detection and OCR already
had this pattern)
- Add title/author aliases to edit-metadata schema so common field names
actually write EXIF tags instead of being silently stripped by Zod
- Port extend/pad crop logic from passport-photo single endpoint to the
batch pipeline so crop regions extending beyond the image get filled
with background color instead of producing all-white output
- Clamp quantized color channels to 255 in color-palette to prevent
Math.round(255/16)*16=256 from producing invalid hex like #100100100
- Compare OCR fallback warning against expected engine name per tier
instead of comparing engine name against tier name (always mismatch)
When auth was disabled, the backend middleware attached the first admin
user from the database to every request, and the frontend granted all 12
permissions. This gave every unauthenticated visitor full admin access
to user management, settings, teams, branding, and feature installation.
Now both layers use role "user" with user-level permissions so tools,
files, and pipelines still work without login while admin-only routes
correctly return 403.
Closes#72
Closes#73
AVIF was already supported in the core engine, convert, compress,
optimize-for-web, upscale, erase-object, svg-to-raster, and
pdf-to-image tools. This adds AVIF as an output format option to
the 6 tools that were missing it: split, collage, stitch,
image-to-base64, noise-removal, and red-eye-removal.
For each tool, both the frontend format selector (with quality
slider for AVIF's lossy encoding) and the backend Zod schema +
Sharp .avif() encoding were updated. AVIF defaults: quality from
the user slider, effort 4 (balanced encode speed).
Also fixes pre-existing Biome formatting violations in 5 files
that were blocking a clean lint pass.
1. split batch 404: register split tool in batch registry via
registerToolProcessFn() so /api/v1/tools/split/batch works
2. CodeFormer crash: inference_app() expects a file path, not a numpy
array. Save to temp file before calling, read result back.
3. OCR fallback chain: fix case-sensitive "Segmentation fault" match
that prevented PaddleOCR crash from triggering Tesseract fallback.
Also add "process crashed" check. Upgrade ARM paddlepaddle to >=3.2.1.
4. blur-faces large images: downscale to 1920px max before MediaPipe
detection, scale coordinates back. Also add rotation retry for
portrait-oriented images where BlazeFace misses faces. Applied to
detect_faces.py, enhance_faces.py, and restore.py.
5. color-adjustments tool ID: fix mismatch in index.ts registration
array (was "color-adjustments", should be "adjust-colors").
The info tool reads metadata directly via Sharp without going through
the format decoder pipeline. Added CLI format detection and decoding
before metadata read, matching the pattern used by all other tools.
Phase 1 — Docker Artifact Optimization:
- Replace broad `COPY . .` with targeted frontend source copies (API/Python
changes no longer bust the frontend build cache)
- Replace build-essential with gcc/g++ (leaner runtime)
- Fix LOG_LEVEL=debug → info for production
- Harden .dockerignore (exclude worktrees, IDE, CI, test artifacts)
Phase 2 — State & Persistence:
- Add PUID/PGID support in entrypoint.sh for bind mount compatibility
- Guard against PUID=0/PGID=0 to prevent accidental root execution
- Evict conflicting system users (e.g. node:1000) before UID remap
Phase 3 — Security:
- Always register @fastify/rate-limit so login brute-force protection
works even when global rate limit is disabled (RATE_LIMIT_PER_MIN=0)
- Add trustProxy support (TRUST_PROXY env var, default true) so rate
limiting and audit logs use real client IPs behind reverse proxies
- Strip stack traces from 500 error responses in production
- Fix FSTDEP022 deprecation: maxParamLength → routerOptions
- Add multi-file guard on single-file tool endpoint with clear error
message pointing to the /batch endpoint
Phase 4 — Graceful Degradation:
- Add consolidated hardware detection startup banner (GPU, rate limit,
upload limit, proxy status)
- Add ConnectionMonitor component with health polling and reconnecting
overlay that auto-dismisses when the server comes back
Phase 5 — Deployment Docs:
- Rewrite deployment.md with copy-paste CPU and GPU compose templates
- Add hardware requirements table (minimum, recommended, heavy workloads)
- Add PUID/PGID bind mount documentation
- Add complete env var reference table
- Add reverse proxy guides for Nginx, Nginx Proxy Manager, Traefik,
and Cloudflare Tunnels
dcraw_emu is not included in libraw-dev (only libraries). ImageMagick
already delegates to LibRaw for all major RAW formats (DNG, CR2, NEF,
ARW, ORF, RW2). Moved ImageMagick helpers above RAW decoder to fix
declaration order.
Extends the platform to handle 7 new image format families alongside
the existing AVIF support gap-fill. Uses the established HEIC decoder
pattern (CLI decode → PNG → Sharp) for formats Sharp can't handle
natively: Camera RAW via dcraw_emu/LibRaw, PSD/TGA/EXR/HDR via
ImageMagick. JXL and ICO are Sharp-native. Adds server-side preview
for non-browser-displayable formats and JXL as a new convert output
target. All 27 validateImageBuffer callers updated with filename for
extension-based format detection.
Fastify's find-my-way router defaults maxParamLength to 100 chars.
Files with long names (e.g., stock photo filenames >100 chars) would
return 404 on download because the :filename param exceeded the limit.
- Add 8 new E2E specs for AI tools (upscale, enhance-faces, colorize,
restore-photo, erase-object, smart-crop, passport-photo, red-eye-removal)
closing all HIGH/MEDIUM coverage gaps from the test matrix audit
- Fix ensureAiDirs() crash on non-Docker environments by gating on
isDockerEnvironment() — prevents ENOENT when /data doesn't exist
- Bump torch 2.6.0→2.7.0 and torchvision 0.21.0→0.22.0 in feature
manifest for broader Python version compatibility
- Add Python 3.14 version guard warning in install_feature.py
- Remove duplicate torchvision shims from upscale.py and enhance_faces.py
(dispatcher.py already handles this at startup)
- Remove orphaned tools.batch i18n key and dead pipeline-builder filter
- Regenerate 4 visual regression baselines for current UI state
- Add data-testid to passport-photo generate button for E2E testability
The torchvision compatibility shim for basicsr 1.4.2 was missing the
parent-package binding and only proxied a single attribute, causing
upscale and enhance-faces to fail at import time. The fix adds a
__getattr__ proxy for all attributes, binds the shim to the parent
package, and installs it in the dispatcher at startup for defense-in-depth.
Also removes unused anyInstalling variable, redundant `as any` cast,
and applies Biome formatting fixes across the codebase.
- Fix SKIP_MUST_CHANGE_PASSWORD not affecting login/session API responses,
causing frontend redirect even when the env var was set after user creation
- Increase Docker Playwright timeouts (test: 600s, expect: 60s, AI processing: 300s)
to support CPU-only self-hosted environments
- Increase default rate limit from 100 to 50000 req/min for self-hosted deployments
- Fix OCR tests: use filechooser pattern (Dropzone has no static file input),
correct enhance checkbox default, rewrite for actual fixture behavior
- Fix remove-bg tests: update quality labels (Balanced→HD, Best→Max)
- Fix noise-removal skip guard: use waitFor() instead of instant isVisible()
- Fix automate pipeline save test: clean up stale E2E pipelines before assertion
- Add tool-specific suffix to output filenames so downloads don't overwrite originals (batch & single-tool routes)
- Skip deleting shared models when uninstalling a bundle that shares models with another installed bundle
- Auto-detect NVIDIA GPU and swap GPU-only pip packages (onnxruntime-gpu, paddlepaddle-gpu) for CPU equivalents
- Refactor docker-compose with YAML anchors and explicit cpu/gpu profiles
- Add libheif-plugin-x265 to Dockerfile
- Fix install-all queue logic to handle concurrent individual installs and clear stale errors
- Unify playwright docker config to use same test dir with API_URL env var
- Fix flaky e2e selectors, rename Strip Metadata → Remove Metadata, handle collage custom dropzone, improve fallback test image generation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds objectFit property to CellTransform (cover/contain). When set to
"contain", the entire image is shown within the cell with background
color fill. Toggle button in the cell controls toolbar switches between
modes. Server-side rendering handles both modes via Sharp.
SnapOtter