Thumbnails: update the actual thumbnail endpoint (ConversionController.js
thumbnailFromBuild) to quality=90 and width=794. The previous fix targeted
ConversionManager.js which handles preview mode, not the thumbnail route
called by ThumbnailManager.mjs.
Mobile layout: move the isMobile guard before the stored-preference check
in getInitialLayout(). The autoSave race fix (build 274) stopped future
bad writes, but a stale 'flat' in localStorage was still being read on
every load, blocking the mobile check. Mobile now always starts in
verticalSplit regardless of any stored value.
CI: add node --check on all server-side .mjs files in the Dockerfile,
after source copy and before webpack compile, so syntax errors like the
escaped-backtick incident fail the build immediately.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- clsi-nginx: allow hyphens in project-id regex — conversion IDs are UUIDs
which nginx was rejecting, causing 500 on file download after conversion
- CLSI ConversionController/Manager: add 'latex' export type (typst→latex via pandoc)
- Web: add 'latex' to SUPPORTED_CONVERSION_TYPES
- Frontend: add Export as LaTeX button (visible only for typst projects)
- Fix visibility logic: export-as-latex shows for typst, export-as-typst shows for latex
- Add export_as_latex translation key (en + fr)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- ConversionController.js: add typst to CONVERSION_CONFIGS (missing entry caused 400→500 chain)
- export-project-with-conversion-button: hide button for non-LaTeX projects (typst/quarto) via compiler check
- project-list-lumiere.tsx + scss: revert lumiere-card-actions back inside .lumiere-card (put them back like they were)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
thumbnailFromBuild() now tries output.pdf → output-slides.pdf → decktape
on output.html (slide 1 only). The web service's ThumbnailManager already
calls this endpoint fire-and-forget on every successful compile, so RevealJS
project cards will show the first slide thumbnail automatically.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Increase pdftocairo output from 190px/q50 to 380px/q82 — 2× resolution
for crisp rendering on retina displays, higher quality to eliminate
visible compression artefacts.
Inset the thumbnail image 6px from the tile edges (inset: 6px) with a
4px border-radius so the card's colour gradient is visible as a frame
around the document preview.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The previous implementation delegated to ConversionManager which uses
the Docker-based CommandRunner and is gated behind enablePdfConversions
(ENABLE_PDF_CONVERSIONS env var). Neither is configured in the Verso
deployment, so every thumbnail request 404'd before doing any work.
poppler-utils (which provides pdftocairo) is already installed directly
in the CLSI base image via install_deps.sh. Rewrite thumbnailFromBuild
to call pdftocairo via execFile instead — no feature flag, no Docker
image, no ConversionManager indirection.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After a successful compile, web service calls a new CLSI endpoint
(GET /project/:id/user/:uid/build/:bid/thumbnail) which runs pdftocairo
page-1 to a 190px-wide JPEG using the existing thumbnail preset. The
JPEG is stored in Redis (90-day TTL, overwritten on next compile) by
the new ThumbnailManager.
GET /project/:Project_id/thumbnail serves the cached JPEG to authenticated
users, returning 404 when no thumbnail exists. Project cards in the
Lumière grid show the image overlaying the coloured gradient tile; if
the image 404s (project never compiled or cache expired) the onerror
handler hides it and the gradient + initial letter shows through.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* [monorepo] consolidate clsi-lb host/ip env-vars
Target env-var is CLSI_LB_HOST. Keep CLSI_LB_IP populated for a week.
* [clsi] initial version of /convert/pdf-to-jpeg
* [rails] use fake-secrets in CI and Codespaces
* [rails] adapt tests for using clsi to convert PDFs to image
* [rails] add rake task for comparing clsi conversion with transloadit
* [clsi] double check that output.jpg is a regular file
Co-authored-by: Brian Gough <brian.gough@overleaf.com>
* [clsi] fix composing basename
* [monorepo] fix clsi-lb host env-var post merge
* [monorepo] sort dev-environment.env hosts
* [rails] use local pdf file rather than downloading it again
Download from the old renderer code path still. It's dead code.
* [terraform] clsi: enable pdf to jpg conversion
---------
Co-authored-by: Brian Gough <brian.gough@overleaf.com>
GitOrigin-RevId: 5ecaa8559d299486340bb3961f06b29f7c4dfcca
* [clsi] add request flag for isCompileFromHistory
* [clsi] derive cacheKey for history snapshot from compile dir
* [clsi] migrate convert project to document to compile from history
* [clsi] address review feedback
* [web] determine root doc at the time of converting the project
* [web] wait for flush before starting document conversion
* [saas-e2e] add tests for root doc override when converting project
GitOrigin-RevId: 71c578030949b89f3a74e7f7ab882dfa9c98c17a
* [clsi] Forward pandoc errors to web
* [clsi] Remove unused import
* [clsi] Align warning logs
* [clsi] Update HTTP response for errors
* [clsi] Update acceptance test with 422
* [clsi] Always return json body on 422
* [clsi] Include stderr in logs for non user facing errors
GitOrigin-RevId: 4284c8d4e8b7b45eac4997cd9e52ca4894b20412
* using CLSI logic for fetching the project contents and skip the .zip export
* Use unique conversion directory for project-to-docx export to avoid corrupting the shared compile
directory when a compile runs concurrently
* Remove X-Accel-Buffering header — not needed as CLSI does not run behind nginx
* moving log before sending the data
* Return CLSI stream directly instead of buffering to disk on web
Previously convertProjectToDocx wrote the CLSI response to a temp file
on disk, then the controller read it back to stream to the client.
Now the stream is returned directly and piped to the response,
avoiding unnecessary disk I/O on the web server.
* Use href redirect for docx export instead of fetching blob into memory
* making functions and files more generic so they can be used in future for other documents exports as well
* adding export-docx split test
* adding unit tests
* adding cypress E2E test
* format:fix
* renaming the route to download from convert
* adding new icon for export docx button
* format:fix
* remove unused showExportDocumentErrorToast export and adding guard against invalid Content-Length header from CLSI
* format:fix
* refactor(clsi): move promisify(parse) into RequestParser
* refactor: generic conversion endpoint with type as route
param
* refactor: use type→extension map for validated conversion types
* refactor(clsi): remove --standalone flag and fix rejection test
* fixing the href in cypress test
* renaming function
* adding type to Metrics.inc
* fix: rename exportProjectDocument, add WithLock wrapper and metrics type label
* format:fix
* fix: hide docx export from anonymous users and add WithLock wrapper
* format fix
* remove redundant Content-Length validation from DocumentConversionManager
* format:fix
* removing trailing icon
GitOrigin-RevId: e9764fefac2c4b625d23be9e942ea4a8b283c70d
