Skip to content

nitida

Hand it raw media. Get back every size, every format, every screen — crisp.

Upload that survives a phone

Multipart with resume across a reload, IndexedDB-backed, progress per file. A dropped 4G connection resumes where it stopped instead of starting over.

Compression in the browser

A worker and OffscreenCanvas — the main thread never blocks. HEIC decodes natively where the engine can (136 ms), and only falls back to WASM where it cannot, so the 341 KB decoder stays unfetched for most users.

Video, not just images

Poster, transcode, an adaptive HLS ladder, and a cheap proxy built for machine analysis rather than for eyes.

Multi-tenant from the first line

Keys per tenant, isolation by prefix, and usage you can actually read. Running in production across five tenants.

@aquienpz/asset-client builds URLs. No network, no key, nothing to configure beyond a tenant id. If all your app does is render media, this is the only thing you need.

@aquienpz/sdk adds the write half — upload, compress, register, poll — and re-exports every URL builder, so you never import both for the same job.

Every page under API reference is produced from the source on each build. Nothing there is transcribed by hand, which is the only way a reference stays true after a rename. The same content is published at /llms.txt and /llms-full.txt so an agent can read the whole surface in one fetch.