User Tips: Create a bot & have users chatting with ChatGPT in 30 seconds!

If you want your users to be able to chat 1-1 with a bot, this post is for you! This builds on the great post by Poorvik Palanikumar here that walked users through the process of creating a bot that could respond on a channel. If you followed that post and set everything up can can skip to "3." below and continue from there.

(The lazy instructions... Copy code below, create a bot, click "edit code" for the message handler & paste the code, insert your own token and save.)

To create & enable a bot to respond privately to user instead of on a channel follow the steps below.

Create a bot. (Click your profile pic>Bots & Tools > "Create Bot" (top right)
Name the Bot, add description and set it to "Organization" and Save
Click "Edit Code" from the "Message Handler" section and delete the existing code (presuming you don't use it!)
Paste in the code below and save. Make sure to add in your own token. Get one here

You're all set! Go and subscribe to your bot and ask it a question... you'll have to grant permissions from the popup message.

Message Handler Code:

(Bots & Tools > Edit Handlers (for your bot) > Go to Message Handler & click "Edit Code")

response = Map();

//

//Reassign the inbound message from the user, which we'll send to ChatGPT

question = message;

//

//

// Need to add your OWN openAI token below. Replace the xxxxxxx with your own token

token = "Bearer xxxxxxxxxxxxxxxx";

//

//

header = Map();

header.put("Authorization", token);

header.put("Content-Type", "application/json");

//

//

//Set the params which we'll send to chatGPT

//Increasing the temperature increases response creativity, decreasing has the opposite affect

params = { "model": "text-davinci-003", "prompt": question, "temperature": 0.4, "top_p": 1, 

"max_tokens": 256, "frequency_penalty": 0, "presence_penalty": 0, "stop": { " Human:", " AI:"} };

//

//

// Making post request to ChatGPT

fetchCompletions = invokeurl

[

    url : "https://api.openai.com/v1/completions"

    type: POST

    parameters: params.toString()

    headers: header

    detailed: true

];

//

//

info "Fetch completions: " + fetchCompletions;

//

//

//Check we got a good response from ChatGPT

if (fetchCompletions.get("responseCode") == 200) {

    // Grab the answer from Chat GPT

    answer = fetchCompletions.get("responseText").get("choices").getJSON("text");

//

//

    //Put the answer into the response map which is what we'll return to the user

    response.put("text", answer);

}

else if (fetchCompletions.get("responseCode") == 429) 

{

    response = { "text": "Oops! I can't help with this. Try asking something else :wink:" };

}

//

//

// Trigger the response back to the user.

return response;

27 Second Gif... use the extra 3 seconds to set your own OpenAI "token" before you paste the code (get your token here)

Some FAQ:

Will this allow users to subscribe to a bot and ask it questions and get answers?

Yes but it is not the current ChatGTP version that everyone is raving about. It's the ChatGPT-3 Da-Vinci model which is excellent but not on the same level as ChatGTP-4 which is still in research mode and does not yet have an API connection.

Will it remember my previous questions?

It generally depends on how you phrase the questions and give the context. It's not on the same level as ChatGPT-4.

I asked it a question but it does not give me a great answer, why?

How you enter your "prompt" (question) is the important part. Generally the more context and instruction you give the better the answer. Google or YouTube "ChatGPT Prompt Design" and you'll be an expert in no time.