How our universal content processing platform Riviera evolved for AI and beyond

Summary

Dropbox developed Riviera, a centralized content processing platform that evolved from an internal file preview generator into a shared infrastructure supporting over 300 file formats. By decomposing complex conversion tasks into modular, reusable transformations, the system now powers search indexing, media transcoding, and AI context extraction across Dropbox products.

Takeaways

1. 1

   Decomposing monolithic file workflows into small, reusable transformation steps (e.g., PowerPoint-to-PDF followed by PDF-to-image) avoids redundant logic and simplifies supporting new formats.

2. 2

   A plugin architecture that separates request dispatching and caching from backend execution allows product teams to add capabilities without modifying core infrastructure.

3. 3

   AI-powered applications like Dropbox Dash rely heavily on upstream content transformation to normalize, extract, and structure documents before passing context to LLMs.