Multi-Document Support (Technical Guide)

Deckly's multi-document engine extends beyond PDFs to support Microsoft Office (PPTX, DOCX, XLSX) and CSV files. This document outlines the technical implementation of the processing pipeline, display modes, and security gating.

---

Processing Pipeline

1. File Upload & Identification

When a user selects a file in ManageDeck.tsx, the system identifies the file type via extension:

• Supported: .pdf, .pptx, .docx, .doc, .xlsx, .csv.
• Logic: File selection happens in ManageDeck.tsx, while processing/conversion orchestration now lives in useManageDeckWorkflow.ts and workflows/deckProcessing.ts.

2. PDF "Interactive" Processing (Client-Side)

For PDFs, high-fidelity slide viewing and link preservation are handled entirely in the browser:

• Technology: Uses pdfjs-dist (PDF.js) for rendering and metadata extraction.
• Link Preservation:
  • The extractPdfLinkHotspots() utility scans the document's annotation layer.
  • It identifies Link subtypes and extracts their URLs and bounding boxes.
  • Coordinates are mapped to 0-1 percentages relative to the page viewport.
• Workflow:
  1. Load: PDF is loaded into browser memory.
  2. Extract: Links are extracted and stored in the SlidePage.links array.
  3. Rasterize: Each page is rendered to a canvas and converted to a WebP blob.
  4. Upload: Images and metadata are uploaded through the shared deck processing/storage flow.

3. Office "Interactive" Conversion (Edge Function)

For Office files (PPTX, DOCX), conversion is handled by a Supabase Edge Function (document-processor):

• Provider: Uses ConvertAPI for server-side image generation.
• Current Limitation: Link preservation is currently restricted to native PDF uploads. Office-to-Image conversion produces static rasterized slides without hyperlink metadata.
• Workflow:
  1. Download: The function downloads the raw file from Supabase Storage.
  2. Convert: ConvertAPI transforms the file into high-quality JPG images.
  3. Upload: The resulting images are stored in a unique folder.
  4. Metadata Update: The pages array in the decks table is updated with the image URLs.

---

Display Modes

1. Interactive Mode (display_mode: 'interactive')

Renders the document as a slide-by-slide web experience using ImageDeckViewer.tsx.

• Data Shape: Deck.pages contains an array of image URLs.
• Analytics: Supports slide-level engagement heartbeat tracking.
• Requirement: Restricted to PRO users.

2. "Raw" View Mode (display_mode: 'raw')

Integrates the Microsoft Office Online viewer for native-fidelity rendering.

• Implementation: Renders an <iframe> pointing to https://view.officeapps.live.com/op/embed.aspx.
• Best For: Multi-sheet Excels or content-heavy Word docs where web-native navigation (PDF-style) isn't the priority.

---

Tier Enforcement & Gating

The tiered access logic is centralized in src/constants/tiers.ts.

• Feature Flag: canUseOfficeSupport and canUseInteractiveMode.
• UI Gating:
  • Upload: ManageDeck.tsx prevents uploading Office files for FREE users.
  • Interactive Toggle: Displays a "PRO" badge; clicking triggers the TierUpsellModal.
  • Server-Side: The Edge Function verifies the user's tier before processing, preventing API-level bypasses.

---

Security Hardening

To protect sensitive pitch data, the processing pipeline includes multiple security layers:

1. JWT Verification: Processors require a valid Supabase access_token in the Authorization header.
2. Ownership Check: The Edge Function queries the database to ensure auth.uid() === deck.user_id before touching any data.
3. Error Shielding: Errors from ConvertAPI or Storage are caught and logged server-side. The client receives only a generic Internal Server Error to prevent leaking infrastructure details.
4. Data Integrity: Conversion results are stored in isolated, user-specific paths (userId/deck-id/converted/).