Challenge
The product needed to store, organise and serve large volumes of media assets. Users expected fast, on-demand rendering of thumbnails and derivatives, and flexible search across metadata — without the cost of re-processing every asset up front.
Approach
- On-demand rendering with two paths. Built a Go (Gin) service that generates thumbnails and derivatives when they are first requested: a synchronous path for light media, and an asynchronous fallback for heavy jobs so requests never block on them.
- Pluggable media processors. Processors are injected through dependency injection, so new asset types plug in without touching the core service.
- Flexible, queryable metadata. Modelled metadata with PostgreSQL JSONB (via GORM), supporting filtering and sorting on arbitrary fields without a schema migration for every new attribute.
- Containerised from day one. Packaged with Docker and deployed on Kubernetes, with assets in object storage.
Outcome
A complete platform — API, processing pipeline, metadata search and deployment — built end to end. Derivatives are produced only when needed, heavy jobs don’t block the request path, and new media types and metadata fields can be added without changes to the core.
