Connection Error, Pinged Proxy Successfully Though

[CMDMonaco]

Hey man, really amazing work with this. I can tell you're ahead of the wave that's coming. I'm having an issue, however. My goal here is to have the browser version working within my discord bot as an assistant for the server. I have GPT Plus subscription, and I'm hoping this will be able to take advantage of that.

I'll get to coding for the discord bot soon, but for now, I need to get this up and running. GPT is helping me fix the code. So far, we're stuck here:

⠸ Command Assist is typing...TypeError: fetch failed       
at Object.fetch (node:internal/deps/undici/undici:11413:11)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async create (file:///C:/Users/AJBEATX/Downloads/node-chatgpt-api-main/node_modules/@waylaidwanderer/fetch-event-source/lib/esm/fetch.js:52:34) {

cause: Error: read ECONNRESET
at TLSWrap.onStreamRead (node:internal/stream_base_commons:217:20) {
errno: -4077,
code: 'ECONNRESET',
syscall: 'read'
}
TypeError: fetch failed
at Object.fetch (node:internal/deps/undici/undici:11413:11)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async create (file:///C:/Users/AJBEATX/Downloads/node-chatgpt-api-main/node_modules/@waylaidwanderer/fetch cause: Error: read ECONNRESET
at TLSWrap.onStreamRead (node:internal/stream_base_commons:217:20) {
errno: -4077,
code: 'ECONNRESET',
syscall: 'read'
}
}

We've pinged the proxy, everything I can think of. I'm not a coder though, so if you can help that's amazing, if not no worries at all, I'm sure you're busy.
node-chatgpt-api-mainMonaco.zip

[waylaidwanderer]

You need to use the reverse proxy server for now, so pass it in as a parameter like so:

new ChatGPTBrowserClient({
  reverseProxyUrl: 'https://chatgpt.duti.tech/api/conversation',
  accessToken: 'access token',
});

Also, the file you linked has your access token in it. Generally, it's bad practice to commit any secrets or tokens to GitHub as automated bots scan for them and use them for malicious purposes. Logging out should revoke that access token, so you should do that.

[CMDMonaco]

Is it worth all the hassle I'm going through just to get ChatGPT rather than just be okay with GPT-3? If you want to help me out I'll pay you for your time in helping set this up.

[waylaidwanderer]

I'm not sure it's much more hassle using ChatGPT vs. GPT-3/text-davinci-003. It's just a matter of changing which client you're using, really.

IMO it's probably better to stick with ChatGPT in your use case (keeping in mind that it's restricted to writing one reply at a time), especially if you expect it to get a lot of use in your Discord server.

[CMDMonaco]

Thanks for the info, I appreciate it. A few more questions if you find time.

I want the best solution possible for hooking up GPT with a discord bot, in the role of a soldier, essentially. It's going to be an assistant for our DayZ Military Unit, helping us create missions from a template that we give it, great Situation Reports, even plan routes to and from our objectives. It will need to have a bunch of info stored about our group for it to be a great asset, so I'm wondering how prompts work for GPT and GPT-3. I know you can fit roughly up to 8000 words in GPT-3's memory (I think), but if we're using ChatGPT, how can I feed the bot this information, assuming it's roughly 3,000 words about our group, for instance, and another 1,500 that it will use each day, it would start forgetting things past a certain point after about a week.

I'm hoping to code it so that it has a permanent context of the group and our roles, what we do, what our structure is like, our templates, etc.

I also want it to have either rolling temporary context, so that while it knows our group and structure forever, it slowly accumulates real time data and maintains a cache of it over time that's fully accessible, or it to have rolling context, where it begins to forget certain things after a week, as they don't matter as much anymore.

Sorry for the essay, hoping you'll let me pay you and maybe you can help me write up this code.

[waylaidwanderer]

No need to send me the code. Make sure your settings.js file is like this if you're testing the CLI or API server. This is the bare minimum to use text-davinci-003:

export default {
    openaiApiKey: '<api key>',
    chatGptClient: {
        modelOptions: {
            model: 'text-davinci-003',
        },
    },
    cacheOptions: {},
    apiOptions: {
        port: 3000,
        host: 'localhost',
        clientToUse: 'chatgpt',
    },
    cliOptions: {
        clientToUse: 'chatgpt',
    },
    storageFilePath: './cache.json',
}

[CMDMonaco]

You fixed it.

[CMDMonaco]

I guess my last question is, how do I give it more prompts, or the best learning? Is there an easy way to just feed it blocks of text describing our military unit, like locations, weaponry, etc? Or is it best to let it learn slowly and just cache the responses forever. Wondering if there's a simple line of code to add. I'll ask GPT lol

[waylaidwanderer]

Refer to my comment above.

[CMDMonaco]

Last thing I need to do is get this to send messages to the discord bot. Hey, thank you so much for dedicating some time to help me. If you have any tips about getting the replies sent to discord, as well as reading the messages in there as prompts to the bot, let me know! I think I have the code for the most part.