Skip to content

Latest commit

 

History

History
36 lines (25 loc) · 1.2 KB

README.md

File metadata and controls

36 lines (25 loc) · 1.2 KB
Raw

named-entity-recognition-microservice

An example of an API which extracts named entities from a provided user query.

Some notes:

  • I have not load tested it at all

  • The model is being downloaded on the first client request (making it very slow). The model download should be moved to the Dockerfile (i.e. to the image build step)

Deploy it on a google Cloud Run service as follows:

GCP_PROJ_ID='your-gcp-proj-id'
GCP_REGION='europe-west2'
CREATE_ARTIFACT_REG_REPO_NAME='your-artifact-registry-repo-name'
CREATE_CLOUD_RUN_SERVICE_NAME='named-entity-recognition-test'

gcloud auth login
gcloud config set project $GCP_PROJ_ID # verify with `gcloud config get project`
gcloud config set run/region $GCP_REGION # verify with `gcloud config get run/region`

DOCKER_IMG_URI="${GCP_REGION}-docker.pkg.dev/${GCP_PROJ_ID}/${CREATE_ARTIFACT_REG_REPO_NAME}/${CREATE_CLOUD_RUN_SERVICE_NAME}"

make create_artifact_registry_repo \
  ARTIFACT_REG_REPO_NAME=$CREATE_ARTIFACT_REG_REPO_NAME \
  GCP_REGION=$GCP_REGION \
  CLOUD_RUN_SERVICE_NAME=$CREATE_CLOUD_RUN_SERVICE_NAME

make build \
  DOCKER_IMG_URI=$DOCKER_IMG_URI

make deploy \
  CLOUD_RUN_SERVICE_NAME=$CREATE_CLOUD_RUN_SERVICE_NAME \
  DOCKER_IMG_URI=$DOCKER_IMG_URI