dataflow

Run using Socket

• In terminal 1, create the FIFO and start socat to forward FIFO input into the socket:

# Create the FIFO if it doesn't already exist (suppress error if it does)
mkfifo /tmp/drava_in 2>/dev/null || true

# Start socat to forward everything from the FIFO into the Unix domain socket
socat -u OPEN:/tmp/drava_in,rdonly,ignoreeof UNIX-LISTEN:/tmp/accel_2048.sock,fork

• In terminal 2, run the publisher script

cd examples/dataflow
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python publisher_socket.py

• In terminal 3, run the app. With no DRAVA_STAGE_CONFIG set, the runtime uses the socket transport by default, matching this data flow:

cd examples/dataflow
python app.py

Example run using Jetstream

• In terminal 1:

cd ~/nats_binary
./nats-server -js -sd ./jsdata
# Type any messages
# To end: ctrl + c

• In terminal 2 run the app. To use the JetStream transport, create a pipeline.yaml with transport.type: nats and a stage1 entry (see examples/ptychonn/pipeline.yaml), then point the runtime at it:

export DRAVA_STAGE_CONFIG=$PWD/pipeline.yaml
export DRAVA_STAGE_NAME=stage1
python app.py

• In terminal 3, run the the publisher script:

cd examples/dataflow
pip install -r requirements.txt
python publisher_jetstream.py