- We use GCP to create a container cluster called atforestry-cluster
- docker artifact repo atforestry-repo was created manually for Model and data
- container orchastration is done with Kubernetes
- prediction history data is managed by postgres
- model and data services configuration files are at ./yamls
To create the infrastructure and deploy, run:
$ make
or
$ make create-infrastructure
To deploy data, model, front-end, monitoring services, run:
$ make deploy-services
To clear predictions database, run:
$ make reset-db
To terminate all services, run:
$ make delete-services
To terminate the cluster, run:
$ make delete-cluster
To switch to working on atforestry-cluster, run:
$ make auth
-
batch-run daemon:
periodically (once a month) call fetch-data
-
fetch-data service:
get last month data from planet using planet-api and call model-predict
-
model-predict service:
classify each image to a cover-land class and push the result to predictions data base file
-
postgres service:
manage sql data base
-
grafana service:
monitor Kubernetes performance
-
user-query service:
front-end query to check if an area (identified by latitude&longitude) is predicted by our model to be suffering from deforestration