improvement(sandbox): expand document generation — style extraction, sandbox hardening, OOM errors, task guards

[waleedlatif1]

Summary

• Expand style extraction for DOCX/PPTX/PDF: OOXML font inheritance, bold detection (LibreOffice compat), heading ID case-insensitivity, compound font suffix stripping, PPTX slide count/aspect ratio/background color
• Fix auth order in style route (session before parseRequest per CLAUDE.md mandate) and add 100 MB size guard
• Fix contract schema drift: add pdf to format enum, make theme optional, add defaults/pageSize/fonts/slideCount/aspectRatio/background fields
• Add __docxDocOptions to docx sandbox bootstrap/finalize so chunked mode can set document-wide styles and numbering
• Add null guard to pptx finalize; align docx addImage validation with pdf/pptx (throw on missing dims)
• Surface friendly OOM error from isolated-vm worker instead of raw V8 message

Type of Change

• Bug fix
• Improvement

Testing

Tested manually

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		May 9, 2026 0:26am

[cursor]

PR Summary

Medium Risk
Medium risk: expands document parsing to PDFs and adds more OOXML style inference, which can affect performance/memory and edge-case handling for user-uploaded files. Also tweaks sandbox/worker error classification and generation guards, which could change failure modes in production.

Overview
Adds PDF support to workspace file style extraction (API + VFS), including a 100MB size limit, updated error messaging, and an expanded response schema (optional theme, plus PDF pageSize/fonts and PPTX slideCount/aspectRatio/background).

Improves OOXML extraction robustness by making DOCX themes optional (LibreOffice compatibility), resolving style inheritance/theme fonts, and enriching PPTX metadata; updates the sandbox worker to surface a friendlier OOM error before cancellation detection.

Enhances document generation tasks: DOCX chunked mode can now pass document-wide options via __docxDocOptions, and PPTX finalize now guards against globalThis.pptx being overwritten.

^{Reviewed by Cursor Bugbot for commit d251915. Configure here.}

[greptile-apps]

Greptile Summary

This PR expands document style extraction to cover PDF files (via pdf-lib) alongside the existing DOCX/PPTX support, adds PPTX metadata (slide count, aspect ratio, background color), and introduces OOXML font-inheritance resolution for DOCX styles. It also fixes a critical error-ordering bug in the isolated-vm worker where OOM exceptions were silently swallowed by the cancellation path, and hardens the sandbox with a 100 MB file-size guard and auth-order correction.

• document-style.ts: New extractPdfStyle function reads page dimensions and embedded font families from the first 10 pages; parseDocxStyles now walks basedOn chains and resolves w:asciiTheme font references; parsePptxPresentation and parseSlideMasterBackground add slide metadata.
• isolated-vm-worker.cjs: OOM check moved before the isDisposed guard in both executeCode and executeTask so that an OOM (which auto-disposes the isolate) is no longer misreported as a user cancellation.
• Route + contract: auth order fixed per CLAUDE.md mandate, 100 MB size guard added, and documentStyleSummarySchema updated to reflect all new optional fields.

Confidence Score: 5/5

Safe to merge — all logic paths are well-guarded and the OOM ordering fix corrects a genuine misclassification without introducing new failure modes.

The OOM reordering is the most load-bearing change and it is clearly correct: OOM auto-disposes the isolate, so the old code's isDisposed check would fire first and mask the real error. Moving OOM detection above the dispose check eliminates that ambiguity. The style-extraction additions are well-insulated behind try/catch at every layer (outer, per-page, per-font-entry), so a malformed or unusual document degrades gracefully to null. The contract schema changes are additive and backward-compatible. No existing call sites are broken.

No files require special attention.

Important Files Changed

Filename	Overview
apps/sim/lib/copilot/vfs/document-style.ts	Major expansion: adds PDF style extraction via pdf-lib, PPTX slide count/aspect ratio/background parsing, DOCX inheritance resolution and font theme mapping; well-guarded with try/catch; landscape orientation not detected as a named preset
apps/sim/lib/execution/isolated-vm-worker.cjs	OOM check moved before the isDisposed guard in both executeCode and executeTask; fixes a real ordering bug where OOM silently appeared as AbortError
apps/sim/app/api/workspaces/[id]/files/[fileId]/style/route.ts	Auth order fixed (session before parseRequest), 100 MB size guard added, PDF extension support added; all changes correct
apps/sim/lib/api/contracts/workspace-files.ts	Contract schema updated: pdf added to format enum, theme made optional, new fields for pageSize/fonts/slideCount/aspectRatio/background; aligns with runtime implementation
apps/sim/sandbox-tasks/docx-generate.ts	Adds __docxDocOptions global for chunked document-wide styles/numbering; spread correctly placed before sections so sections always wins
apps/sim/sandbox-tasks/pptx-generate.ts	Null guard added to finalize to catch overwritten pptx instance; error message is clear and accurate
apps/sim/lib/copilot/vfs/workspace-vfs.ts	Comment and type annotation updates to reflect pdf support; no logic changes

_{Reviews (2): Last reviewed commit: "chore(lint): suppress noTemplateCurlyInS..." | Re-trigger Greptile}

[waleedlatif1]

@cursor review

[waleedlatif1]

@greptile

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit d251915. Configure here.}