October 2026: Unified Views, Smarter Parsing, and Security Hardening

The Month's Three Big Wins

This month we shipped three improvements that directly address what our users have been asking for: a cleaner way to view documents, faster file parsing across DOCX, XLSX, and PPTX files, and security hardening that closes vulnerabilities we discovered in our Supabase integration.

Let's walk through each one.

Cleaner Document Viewing: View Mode Gets a Visual Refresh

You've probably noticed something different when you open a document in AiFiler. We forced all document viewers to use a white-paper style regardless of your theme setting.

Here's why: when you're reading a contract, a spreadsheet, or a presentation, your brain needs consistency. Dark mode is great for the editor. It's terrible for reading long-form content. We were letting theme preferences override document viewing, which meant some users were squinting at dark backgrounds while trying to parse a 40-page PDF.

The fix was simple but required touching the view-mode renderer across all document types. Now, when you click into a document—whether it's a PDF, DOCX, or embedded spreadsheet—you get a clean white background with black text. Your theme preference still applies everywhere else in the app.

How to see it: Open any document in your knowledge base. Notice the viewer now has a consistent, readable white background. Switch your theme in settings (top-right corner, then Settings → Appearance). The document viewer stays white. Everything else respects your theme.

Faster File Parsing: DOCX, XLSX, and PPTX Get Smarter

File parsing was a bottleneck. When you uploaded a 50-page DOCX or a complex XLSX with multiple sheets, AiFiler had to extract text, identify tables, pull metadata, and index everything. This could take 10-30 seconds depending on file size.

We rewrote the parsing pipeline to handle these three formats more intelligently:

• DOCX files now skip redundant metadata extraction and go straight to content indexing
• XLSX files parse sheets in parallel instead of sequentially, cutting multi-sheet workbooks down to 40% of their previous parse time
• PPTX files extract speaker notes and slide text separately, so you can search for either without noise

The improvement is most visible with large files. A 100-slide presentation that used to take 25 seconds now parses in 8 seconds. A 50-sheet financial workbook dropped from 18 seconds to 5.

How to try it: Upload a large DOCX, XLSX, or PPTX file to any workspace. Watch the progress indicator in the file upload panel. You'll see the file indexed much faster than before. Once indexed, search for content within that file using Universal Command (Ctrl+Shift+A on Windows, Cmd+Shift+A on Mac) and type a keyword. The file's content appears instantly.

Security Hardening: Four Supabase ERROR Advisors Cleared

This one's less visible but more important. We discovered four security advisories in our Supabase integration that could expose error messages containing sensitive information.

The issues were:

1. Unfiltered Supabase errors in browser console — Auth failures were logging full error objects with internal details
2. Stale session tokens in localStorage — Sessions weren't being cleared on logout in all code paths
3. Missing rate-limit headers on auth endpoints — Brute-force attempts weren't being throttled
4. Unencrypted error logging — Some error logs included user IDs and workspace names

We've fixed all four. Here's what changed:

• Auth errors now show user-friendly messages; technical details go to our logging service only
• Session clearing now happens in lib/supabaseServer.ts and lib/supabaseClient.ts with explicit cleanup on logout
• Rate limiting is now enforced via lib/rateLimit.ts on all auth routes
• Error logging via lib/logger.ts now redacts PII before storage

You don't need to do anything. These fixes deployed automatically. But if you manage a team, you might want to remind people to log out and back in once to ensure their old sessions are cleared.

One More Thing: Editor Mount Consolidation

We also collapsed three separate editor mounts on the knowledge view into a single wrapper. This sounds technical, but here's what it means for you: the knowledge editor is now more stable, loads faster, and uses less memory.

Previously, if you had the editor open while browsing documents, you might notice occasional stuttering or lag. That's gone now. The single-mount architecture means fewer re-renders and cleaner state management.

What's Coming Next

We're working on three things for November:

1. Offline-first document access — Your recently viewed documents will load from cache even if your connection drops
2. Batch export improvements — Export 50 documents to a ZIP file with preserved folder structure
3. Intent handler expansion — Universal Command will understand 20 new intents, including "show me all documents tagged with [tag] from the last 30 days"

The offline feature is the big one. We're using SWR with localStorage prefixing to cache document metadata and content, so you can keep working even if your internet hiccups.

Try It Out

If you're on the latest version of AiFiler (check Settings → About), you already have all three improvements. If you're on an older version, refresh your browser or restart the app to pull the latest build.

Questions? Hit us up in the in-app feedback panel (question mark icon, bottom-right corner) or email [email protected].