Dropbox is exposing parts of Riviera, its internal content-processing platform, through public APIs that can prepare files for search and AI applications. Developers can submit asynchronous jobs to convert documents into Markdown, transcribe audio or video and extract structured metadata, then poll for completion.
Riviera began as a preview generator and has expanded into a universal pipeline supporting more than 300 file formats and over 100 transformations. Instead of building a separate processor for every output, Dropbox composes reusable steps—for example, converting a presentation to PDF and then rendering pages as images. A central orchestrator validates jobs, builds pipelines, dispatches them to specialized workers and caches intermediate results.
That cache lets multiple products reuse expensive work. Dropbox previously said Riviera handled about 2.5 billion daily requests and nearly an exabyte of data while supplying text and embeddings for summaries and question answering. The newer Dash product uses the same infrastructure to normalize content from Dropbox and connected services before indexing. External projects can now incorporate selected capabilities into retrieval-augmented generation and workflow systems. The APIs do not provide every internal transformation, and jobs are asynchronous rather than instant, but they give developers access to a mature file-normalization layer without maintaining separate parsers for hundreds of formats.