About

This guide will illustrate how to host the entire BioMedICUS pipeline as a RESTful service.

Pre-requisites

You will need to have installed biomedicus. You will need the MTAP API gateway, available as a binary from the MTAP releases page or installable using the command: go install github.com/nlpie/mtap/go/mtap-gateway/mtap-gateway.go. This tutorial also uses the jq utility to format json.

Deploying and using the pipeline

First deploy the BioMedICUS default pipeline:

b9 deploy --rtf > deployment.log &
tail -f deployment.log

Wait for the deployment to finish (“Done deploying all servers.”) and then Ctrl+C out of tail.

Next deploy the pipeline service.

b9 serve-pipeline --rtf --include-label-text -p 55000 > serve-pipeline.log &

Next we need to create a configuration file for the API Gateway. Save the following contents to a file named “gateway.yml”.

gateway:
  port: 8080
  events: localhost:50100
  processors:
  pipelines:
    - Identifier: biomedicus-default-pipeline
      Endpoint: 127.0.0.1:55000

After saving this file, we can start the gateway using the following command:

MTAP_CONFIG=gateway.yml ./mtap-gateway-<version> -v=3 -logtostderr &> gateway.log &

The ./mtap-gateway-<version> is the binary downloade earlier. If you installed the gateway using go replace ./mtap-gateway-version with mtap-gateway.

Now the biomedicus servers are running, the pipeline and api gateway are hosted, and we can send documents to the pipeline to process. You can use either this file or one of your own:

BODY=$(jq --null-input --arg doc "$(base64 "97_204.txt")"  \
'{ "event": { "event_id": "97_204.txt", "binaries": { "rtf": $doc }}, "params": { "document_name": "plaintext" }}')
curl -X POST http://127.0.0.1:8080/v1/pipeline/biomedicus-default-pipeline/process \
-H 'Content-Type: application/json' \
-d "${BODY}" | python -m json.tool

This functionality will work with either .txt files or .rtf files using the same command.

Deploy using the docker image

You can also deploy the end-to-end REST pipeline using the biomedicus docker image:

 docker run -it -d --rm -p 8080:8080 --name b9 --entrypoint "./rest_e2e.sh" ghcr.io/nlpie/biomedicus:latest

You can follow the deployment logs using this command:

docker logs -f b9

Once you see the following message the server is ready to use on the 8080 port.

Starting new pipeline gateway for service: biomedicus-default-pipeline with address: 127.0.0.1:55000

RTF to Text Only Processing

The above docker image also contains functionality for RTF-to-Text conversion without the BioMedICUS pipeline. You can test the functionality using either this file or one of your own:

BODY=$(jq --null-input --arg doc "$(base64 "97_204.rtf")"  \
'{ "event": { "event_id": "97_204.txt", "binaries": { "rtf": $doc }}, "params": { "document_name": "plaintext" }}')
curl -X POST http://127.0.0.1:8080/v1/pipeline/biomedicus-rtf-to-text/process \
-H 'Content-Type: application/json' \
-d "${BODY}" | python -m json.tool

Conclusion

Additional information on the mtap-gateway as well as API specifications can be found in the MTAP Documentation.