python-operator-dataflow

Python Dataflow Example

This examples shows how to create and connect dora operators and custom nodes in Python.

Overview

The dataflow.yml defines a simple dataflow graph with the following three nodes:

• a webcam node, that connects to your webcam and feed the dataflow with webcam frame as jpeg compressed bytearray.
• an object detection node, that apply Yolo v5 on the webcam image. The model is imported from Pytorch Hub. The output is the bounding box of each object detected, the confidence and the class. You can have more info here: https://pytorch.org/hub/ultralytics_yolov5/
• a window plotting node, that will retrieve the webcam image and the Yolov5 bounding box and join the two together.

Getting started

pip install -r requirements.txt
cargo run --example python-operator-dataflow

Installation

conda create -n example_env python=3.11
conda activate test_env
pip install -r requirements.txt
pip install -r requirements_llm.txt

Run the dataflow

• Start the object detection dataflow alone:

dora start dataflow.yml

• Start the llm dataflow (Only works on Windows and Linux):

dora start dataflow_llm.yml

Within the window you can ask question such as:

ask how are you
change bounding box plot to red
change confidence value to percentage
change object detection to only detect person
send 200 200 200 400 to topic line
record

The keyboard, microphone, whisper node, works in a very similar fashion as the object detection dataflow and I'll let you check it out by yourself.

The code modification flow works by first comparing an instruction with a vectordb of operators source code and then feeding the most similar operator to an llm with the instruction for code modification.

The end result is then saved using a file saver.