Infrastructure
How Dafado is hosted, cached, and processed, described at a high level.
Overview
Dafado runs as a WordPress plugin on managed hosting. The plugin handles recording, transcription, translation, publishing, and the place and profile pages. The pages a guest and a place owner see are all served by this one application.
The edge cache
Public pages are served through a global edge cache (Cloudflare). Review, place, and profile pages are cached at the edge so they load fast for anyone, anywhere. A guest who scans a QR code in a cafe gets a page from a nearby edge location, not from a single distant server.
One consequence is worth knowing. When a page changes, the fresh version does not appear everywhere at once. The edge keeps its cached copy for a while, so a newly published review or an edited detail can take time to update across all locations. This is normal and expected. The page catches up after the cached copy expires or is refreshed.
Audio processing
When a guest finishes recording, the audio is handed to a dedicated speech-to-text service. That service accepts the recording and returns a transcript. It can also return a rendered audio mix, which is the voice with an optional music bed under it. The original recording is never altered by this step.
Transcription runs on infrastructure Dafado operates, using open-source speech-to-text. It is not a third-party transcription company. Only the audio a guest chose to record is sent for transcription, and only to produce the transcript and the optional mix.
Storage
Recordings and images are kept in cloud object storage. When the browser uploads a recording or a photo, it uses a short-lived, pre-signed upload. This means the browser is granted permission to write one specific file for a short time, and the storage credentials never travel to the browser. Playback later reads the finished file back from storage through the edge cache.
Why this shape
Each part does one job. Hosting serves pages. The edge cache makes them fast. The speech-to-text service turns voice into text. Object storage holds the audio and images. Keeping these separate makes the system easier to reason about and keeps sensitive credentials off the guest's device.
What this means for you
If you publish or change something and do not see it immediately on a shared link, the edge cache is the usual reason. Give it a little time. The recording itself is safe in storage the moment it uploads, and the transcript follows once processing finishes.