Carter Offline Non-PyTorch

Carter Offline is an intent classifier which is based on Janex: Python Edition, and utilises LazyLyrics' Carter-Py library.

This particular implementation of C-O is designed to run without being dependent on PyTorch, which sacrifices some level of accuracy, but can be run on hosting providers that don't support PyTorch freely.

It compares the input you send to your carter agent with a list of patterns in your chosen intents file, and then remembers the class. Once your Carter agent creates a response, the program saves the response to your intents, recording the interaction and improving your intents dataset.

This is based on an older version of Janex (release 0.0.5.beta, specifically), and so it may not include the latest features from there unless explicitly added in.

See also:

https://github.com/LazyLyrics/carter-py
https://github.com/Cipher58/Janex

How to use

Adding to your project

Firstly, you'll need to install the library using the Python pip package manager.

python3 -m pip install Carter-Offline-NonPT

Secondly, you need to import the library into your Python script.

from CarterOfflineNonPT import *

Using Carter-Offline in your code

Define Variables

To use this program, you will need to define your API key, name and intent file path.

intents_file_path, thesaurus_file_path, UIName, CarterAPI = "intents.json", "thesaurus.json", "YOUR CARTER AGENT'S NAME", "YOUR CARTER AGENT'S API"

input_string = "Hello!" # for example

User = "YOUR NAME"

# Create an instance of the CarterOffline class

CarterOffline = CarterOfflineNonPT(intents_file_path, thesaurus_file_path, UIName, CarterAPI)

Speak to your agent

Here is the simple bit of code that sends the message to your Carter agent using the Carter-Py library, and then uses the Janex code to classify it, and save it to your intents file.

ResponseOutput = CarterOffline.SendToCarter(input_string, User)

print(ResponseOutput)