Conceptual Model

What is an Asset?

We define the Asset as a single or collection of digital resources. Digital resources can be data, code or models etc. A good way to reason about this is basically anything that can be stored and retrieved.

Examples of Asset

How do we store an Asset

In AMA, assets are serialized to yaml files and stored in the gs bucket. Conceptually, an asset holds a list of objects and other metadata. 

top_hash: 7e23XXXXXXXXXXXXXXXe3d2c
class_name: trained_models
class_id: 88xxxxfa-bxxc-1xx0-1111-a6xxxxxxxxcf1
created_by: john.smith
created_at: 2023-06-15T14:27:33-PDT
objects:
- path: sequence_data/trained_models.json
  created_by: john.smith
  created_at: 2025-02-15T14:29:52-PDT
  content:
    id: gs:md5$+MRxxxxxxxx/hxxxxxxxh==
    type: application/json
- path: sequence_data/chr2_variants.txt
  created_by: john.smith
  created_at: 2025-02-15T14:31:08-PDT
  content:
    id: gs:md5$3KxxxxxxxxxxUnLcQ==
    type: text/plain

The underlying content is stored separately in a bucket or other data-stores i.e. database, ftp etc depending on the type and source of content.

Why do I need AMA?

AMA presents a shared hub to ensure that assets can be trusted, reused, and shared for greater efficiency and more accurate results over time, across collaborators, and in different machine environments
  • Trackability: Know and lineage of every asset
  • Availability: Run your experiments without worrying about how to plugin data. AMA will fetch the relevant dataset and make it available. Both locally and in cloud.
  • Shareability: human friendly name to communicate assets.
  • Searchability: Query and filter all data / metadata
  • Pipeline Integration: seamlessly integrated into pipeline