Analyses¶

Analyses run on Google Cloud Compute Engine.

High level overview of analyses process¶

Each analysis is built into a docker container
An instance template is created from each docker container
An auto-scaling instance group is created for each instance template
When audio is uploaded, each analysis job is pushed onto a pub/sub queue for the particular analysis, which is then picked up within the docker instances for that analysis
The instance group scales the analysis with the backlog of the queue.
Spot instances are used for lower cost, which may be shut down at any time and retried later

Bugg analysis lib is a python library that begins the work of splitting out the common boilerplate code needed to create a new analyser.

Currently it is only in full use by the analysis birdnetlib