Deploy

To deploy this solution, you must first set up the Project directory.

Define an environment variable for your main project directory.

export PROJECT_DIR=~/morpheus_fraud_detection
mkdir -p $PROJECT_DIR
cd $PROJECT_DIR

Terminal 1: Run the Morpheus Pipeline

This command starts the Morpheus Pipeline (Docker Container) swimlane.

The pipeline will initialize and then wait, listening for data to arrive on the

Kafka Topic:
            Input

cd $PROJECT_DIR/Morpheus 
docker run --rm -it --net=host --cap-add=sys_nice \ --runtime=nvidia --gpus=all \ -v $(pwd):/morpheus \ nvcr.io/nvidia/morpheus/morpheus:25.02-runtime bash
# --- INSIDE THE CONTAINER ---
cd /morpheus 
conda env update --solver=libmamba -n ${CONDA_DEFAULT_ENV} --file ./conda/environments/examples_cuda-125_arch-$(arch).yaml python examples/gnn_fraud_detection_pipeline/run.py

Expected Output: You will see logs from Morpheus as it builds the pipeline. It will then appear to hang, which is the correct behavior as it is now actively listening for messages.

Terminal 2: Run the Kafka Consumer

This command starts the Python Consumer in the host environment swimlane.

It subscribes to the Kafka Topic: Output and will display any results published by the Morpheus pipeline.

cd $PROJECT_DIR
source kafka_env/bin/activate 
python3 consumer.py

Expected Output: You will see the message

"Listening for
                fraud detection results..."

and the terminal will wait.

Terminal 3: Run the Kafka Producer

This command starts the Python Producer, which reads from the Transaction Data file.

It begins the Input Stream into the Kafka Topic: Input. This action triggers the entire end-to-end flow.

cd $PROJECT_DIR
source kafka_env/bin/activate 
python3 producer.py

Expected Output: You will see a series of

"Sent
                transaction index..."

messages as data is published. Simultaneously, you will see the Morpheus pipeline in Terminal 1 start processing data, and the final fraud alerts will appear in Terminal 2.