Content Filtering

This is a super quick example/tutorial for implementing content filtering with the OpenAI API and Node.js/JavaScript. This is just the very basics. I'll follow up with an example that is a bit more real-world.

NOTE: This tutorial is part of a series. If you have not read the getting started post you should check that one out first.

Steps

log in to Replit.com
Open the openai-examples-node repl that you created in the getting started tutorial.
Create a new file named filter.js
Copy the following code into the filter.js file.

const axios = require('axios');

const input = "What religion are you?";

const params = JSON.stringify({
  "prompt": `<|endoftext|>${input}\n--\nLabel:`,
  "max_tokens": 1,
  "temperature": 0,
  "top_p": 0
});

const config = {
  method: 'post',
  url: 'https://api.openai.com/v1/engines/content-filter-alpha-c4/completions',
  headers: { 
    'Authorization': `Bearer ${process.env.OPENAI_API_KEY}`, 
    'Content-Type': 'application/json'
  },
  data : params
};

axios(config)
.then(function (response) {
  console.log(JSON.stringify(response.data));
})
.catch(function (error) {
  console.log(error);
});

Update the .replit file with the following

language = "nodejs"
run = "node filter.js"

Review the results in the console pane. You should see a result similar to the following example.

{
    "id":"cmpl-2mAPQo53gUzavAtWFBRiTaJIU3uN8",
    "object":"text_completion",
    "created":1617832432,
    "model":"toxicity-double-18",
    "choices":[{"text":"1","index":0,"logprobs":null,"finish_reason":"length"}]
}

The choices array in the json response contains the results from the content filter. In the example above, the text value is 1. The value could be one of the following:

0 – Safe – Nothing about the text seems potentially offensive or unsafe 1 – Sensitive - Sensitive topics might include text with, political, religious, race, or nationality related content. 2 – Unsafe - The text contains language that some would consider mean, hurtful, explicit, offensive, profane, prejudiced, or hateful. Or language that most would consider NSFW (Not Safe for Work), or language that might portray certain groups/people in a harmful manner.

That's it for now. Let me know if you have any questions.

For details about content filtering see the OpenAI docs here.

Content Filtering

Steps​

Steps