[작성중] wd analyze api workflow

haproxy > ibm-nginx > wd-discovery-gateway > wd-discovery-sateless-api-rest-proxy > wd-discovery-stateless-api-model-runtime > wd-discovery-wksml

1. wd-discovery-gateway POD
analyze API 요청을 수신하고 해당 백엔드 구성요소로 전달하는 역할을 합니다.
API 레이어 역할을 하는 Liverty 서버에서 실행되고 scala 애플리케이션을 제공합니다.
nginx 컨테이너에 들어오는 트래픽을 관리하고 서비스에 배포하기 위해 리버스 프록시 및 로드 밸런서로 사용 됩니다.

The Watson gateway service is our "front-end". It is in charge of receiving all the Watson Discovery REST API requests and forwarding them to the corresponding backend components. The api container serves a scala application running on Liberty server which plays the role of API layer for Watson Discovery; the nginx container is used as a reverse proxy and load balancer to manage incoming traffic and distribute it downstream or database services.

2. wd-discovery-sateless-api-rest-proxy POD

3. wd-discovery-stateless-api-model-runtime

This set of pods implement the Analyze API feature, an endpoint for processing text documents through the enrichment pipeline of the Discovery service without storing the source documents. The Proxy pods watch collection configuration and registers new models when there are changes. The Model Runtime pods uses model-mesh (a "Train and Serve" (TAS) runtime service) to load/unload models and processes the analyze operation on both Java enrichment server and Python enrichment server.

4. wd-discovery-wksml

WKSML is a stand-alone service that loads and processes document with WKSML models. Two components are included in this service: WLP Process (define REST APIs, download model from Minio, install downloaded model to local filesystem, invoke SIRE Runtime process) and SIRE Runtime(process analysis of input text).
WKS model service receives requests from: