Skip to content

FHIR Client for Retrieving and Posting FHIR Resources#

Authors: Holoscan Team (NVIDIA)
Supported platforms: x86_64, aarch64
Language: Python
Last modified: May 13, 2025
Latest version: 1.0.0
Minimum Holoscan SDK version: 2.5.0
Tested Holoscan SDK versions: 2.5.0, 2.7.0
Contribution metric: Level 1 - Highly Reliable

This is an application to interface with a FHIR server to retrieve or post FHIR resources.

It requires that the FHIR Server endpoint URL is provided on the command line as well as client authentication credentials if required. Currently, authentication and authorization is limited to OAuth2.0 server-to-server workflow. When authorization is required by the server, its OAuth2.0 token service URL along with client ID and secret must be provided to the application.

This application also uses ZeroMQ to communicate with its own clients, listening on a well-known port on localhost for messages to retrieve resources of a patient, as well as publishing the retrieved resources on another well-known port. For simplicity, the listening port is defined in the code to be 5600, and the publishing port 5601. Messaging security, at transport or message level, is not implemented in this example.

The message schema is simple, with a well-known topic string and topic-specific content schema in JSON format.

The default set of FHIR resource types to retrieve are listed below, which can be overridden by the request message:

  • Observation
  • ImagingStudy
  • FamilyMemberHistory
  • Condition
  • DiagnosticReport
  • DocumentReference

Requirements#

Run Instructions#

There are several ways to build and run this application and package it as a Holoscan Application Package, an Open Container Initiative compliant image. The following sections describe each in detail.

It is further expected that you have read the HoloHub README, have cloned the HoloHub repository to your local system, and the current working directory is the HoloHub root, holohub.

Note:
The application listens for request messages to start retrieving resources from the server and then publishes the results, so another application is needed to drive this workflow (e.g., the LLM application). To help with simple testing, a Python script is provided as part of this application, and its usage is described below in this section.

Quick Start Using Holohub Container#

This is the simplest and fastest way to start the application in a Holohub dev container and get it ready to listen to request messages.

Note:
Please use your own FHIR server endpoint, as well as the OAuth2.0 authorization endpoint and client credentials as needed.

./holohub run fhir --run-args "--fhir_url <f_url> --auth_url <a_url> --uid <id> --secret <token>"

Run the Application in Holohub Dev Container#

Launch the container:

./holohub run-container fhir

This command builds the holohub:fhir container based on the application-specific Dockerfile.

Build and run the application:

Now in the container, build and run the application:

~$ pwd
/workspace/holohub

~$ ./holohub clear-cache
~$ ./holohub run fhir --run-args "--fhir_url <f_url> --auth_url <a_url> --uid <id> --secret <token>"

Once done, exit the container.

Run the Application in the Host Dev Environment (Bare Metal)#

First, create and activate a Python virtual environment, followed by installing the dependencies:

python3 -m venv .testenv
source .testenv/bin/activate
pip install -r applications/ehr_query_llm/fhir/requirements.txt

Then, Set up the Holohub environment:

sudo ./holohub setup

Note: Although this application is implemented entirely in Python and relies on standard PyPI packages, you still may want to set up Holohub environment and use ./holohub commandline .

Next, build and install the application with ./holohub:

./holohub install fhir --local

Now, run the application which is installed in the install folder, with server URLs and credentials of your own:

python install/bin/fhir/python/fhir_client.py --fhir_url <f_url> --auth_url <a_url> --uid <id> --secret <token>

Test the Running Application#

Once the FHIR application has been started with one of the above methods, a test application can be used to request and receive FHIR resources, namely applications/ehr_query_llm/fhir/test_fhir_client.py.

The test application contains hard-coded patient name, patient FHIR resource ID, etc., corresponding to a specific test dataset, though it can be easily modified for another dataset.

It is strongly recommended to run this test application in a Python virtual environment, which can be the same as that used for running the FHIR application. The following describes running it in its own environment:

echo "Assuming venv already created with \`python3 -m venv .testenv\`"
source .testenv/bin/activate
pip install -r applications/ehr_query_llm/fhir/requirements.txt
export PYTHONPATH=${PWD}
python applications/ehr_query_llm/fhir/test_fhir_client.py

From the menu, pick one of the choices for the resources of interest.

Packaging the Application for Distribution and Deployment#

With Holoscan CLI, applications built with Holoscan SDK can be packaged into a Holoscan Application Package (HAP), which is an Open Container Initiative compliant image. An HAP is well suited to be distributed for deployment on hosting platforms, be it Docker Compose, Kubernetes, or otherwise. Please refer to Packaging Holoscan Applications in the User Guide for more information.

This example application provides all the necessary contents for HAP packaging. It is required to perform the packaging in a Python virtual environment, with the application's dependencies installed, before running the following script to reveal specific packaging commands.

applications/ehr_query_llm/fhir/packageHAP.sh

Once the HAP is created, it can then be saved and restored on the target deployment host, and run with the docker run command, shown below with user-specific parameters to be substituted.

docker run -it --rm --net host holohub-fhir-x64-workstation-dgpu-linux-amd64:1.0 \
--fhir_url <f_url> \
--auth_url <a_url> \
--uid <id> \
--secret <token>

Note: Packaging this application requires holoscan-cli, which can be installed using pip. If you are using the same Python environment for packaging as your development environment, there may be a version conflict for the pydantic package, as it is required by both holoscan-cli and fhir.resources. To ensure your development environment can still run the application after packaging, reinstall fhir.resources:

pip install fhir.resources