OCR using GCS and Cloud Functions

Sample code to use google functions with vision API and google cloud storage. The result is dumped into elasticsearch which is running on K8s cluster exposed via IAP.

Pipeline

Pre setup

• Create pub/sub topic with name image-file-type
• Create pub/sub topic with name pdf-file-type
• Create pub/sub topic with name es-upload
• Copy env.yaml.sample to env.yaml and update required values

WARNING: The env.yaml file has service account credentials which should be stored as a secret but it is out of scope for this PoC to do so. To learn more about it see [this](https://cloud.google.com/solutions/secrets-management/).

Deployment

gcloud functions deploy DispatchWithFileType \
  --runtime go111 \
  --trigger-resource <target_bucket_name> \
  --trigger-event google.storage.object.finalize

gcloud functions deploy DetectTextsFromImage \
  --runtime go111 \
  --trigger-topic image-file-types

gcloud functions deploy UploadTextsToES \
  --runtime go111 \
  --trigger-topic es-upload \
  --env-vars-file env.yaml