Introduction

A module for prototyping asyncflows actions that classify data.

This repo contains a classify action, which prompts the LLM to return a classification for a piece of data, given a provided list of labels.

To use this action in your own flows, simply:

pip install asyncflows-classify 

And include the action in your flow yaml file:

flow:
  sentiment_analysis:
    action: classify
    labels:
      - positive
      - negative
    data:
      var: data

Running the Example

The repo also includes an example of how to use the classify action in sentiment analysis:

• sentiment_analysis_example.yaml, a flow that classifies a piece of data as either funky, janky, or serious, and says hello world in that way
• sentiment_analysis_example.py, a script that runs the flow on a hardcoded piece of data

To run the example:

1. Set up Ollama or configure another language model

2. Clone the repository

git clone ssh://git@github.com/asynchronous-flows/asyncflows-classify

3. Change into the directory

cd asyncflows-classify

4. Create and activate your virtual environment (with, for example)

python3.11 -m venv .venv
source .venv/bin/activate

5. If not already installed, install poetry. Install dependencies with:

poetry install

6. Run the example

python sentiment_analysis_example.py