maxijonson / gpt-turbo Goto Github PK

Scope

Library (gpt-turbo)

Type

New feature

Description

I tried implementing some tests, but because GPT Turbo is ESM only with TypeScript, it's surprisingly very hard to even get started with Jest. Unfortunately, I have very small experience with testing with Jest and it's usually in CommonJS front-end environments, so this setup is totally new to me. I've tried for about a week a bunch of solutions but couldn't get it started...

I'd really appreciate it if someone could just get the tests started. I'm not even asking to write them, I just want the following test (or something like it) to work fine when I npm test. If this works, then I don't mind picking up from there.

The only requirement I have is that tests are in a test folder, separate from the src folder.

// packages/library/test/classes/Conversation.test.ts
import { describe, expect, test } from "@jest/globals";
import { Conversation } from "../../src/classes/Conversation.js"; // notice ESM .js extension
import { DEFAULT_CONTEXT } from "../../src/config/constants.js"; // notice ESM .js extension

describe("Conversation", () => {
    test("creates a Conversation with default parameters", () => {
        const conversation = new Conversation();

        expect(conversation.getMessages()).toHaveLength(0);
        expect(conversation.getMessages(true)).toHaveLength(1);

        const firstMessage = conversation.getMessages(true)[0];
        expect(firstMessage.role).toBe("system");
        expect(firstMessage.content).toBe(DEFAULT_CONTEXT);
    });
});

Notice the use .js extensions to import .ts modules. This is the syntax used across the library, since it's the required import syntax when working with ES Modules. The library simply won't compile without it. Removing the .js extension fixes the issue where Jest is unable to find the module (with .js, since files are actually .ts), but then crashes because these modules are actually importing other modules with .js.

Benefit

Adding tests to the library is crucial as it helps to reinforce the robustness of GPT Turbo. It can also be seen as a good indicator for potential devs willing to work with the library and I believe that it conveys a certain degree of maturity for the project as well.

I rarely add tests to my projects as they are often left unfinished, so I don't want to spend too much time testing something that won't go far enough to be worth the time. In the case of GPT Turbo, I feel like it is in a stable enough state that it can be used in a variety of projects and fulfill its purpose, so it feels right that it gets tested properly.

Participation

Yes

Type

New feature

Description

It would be great to have a command that would save the current chat history to a timestamped file. As an extension of this, it would be very useful to have a flag that would do this automatically by creating a file on launch, and then append requests and responses to it.

Benefit

One of the main drawbacks of gpt-turbo-cli is that I can't save my chat history, which is a critical feature. It's much nicer to use than other CLIs, so it would be great if this was possible in any form.

Participation

No

Since this is provided as-is, do you maybe want to add an open-source license so others can pick it up and extend it as needed, as well as make contributions back to your source?

Scope

Web (gpt-turbo-web)

Type

New feature

Description

I'm not sure this is in the scope of this application but would be very nice to be able to save custom prompts and start conversations with them. I have some prompts that I use a lot and ended up storing them in a markdown file and copy-pasting them when needed.

The main points would be:

• Users should be able to view a list with custom prompts (maybe in a modal?)
• Users should be able to add/delete/edit custom prompts that they created
• Users should be able to select a custom prompt from a select input when starting a new conversation
• Users should not be able to see other users prompts or have any kind of cloud like functionality
• User prompts should be stored in the user browser (maybe localstorage or something like that) and would be lost when clearing cache (just to avoid all the work of syncing it)

It is a somewhat complex feature so I would definitely understand if you think it shouldn't be added.

Benefit

Users would have easy access to prompts they use a lot when creating a new conversation, making it easy to do repetitive tasks in different conversations (like generating code or translating text)

Participation

Yes

Type

New feature

Description

gpt service is not available in my country, so I have to use a proxy to reach it

Benefit

as above

Participation

None