dtinth / threadgpt

Alternative frontend to ChatGPT (gpt-3.5-turbo / gpt-4) with a thread-based interface

ThreadGPT

An alternative frontend to ChatGPT (gpt-3.5-turbo) with a thread-based interface.

• History stored locally
• Bring your own OpenAI API key
• Explore the conversation tree!

threadgpt's Issues

add keybaord shortcut ` shift+enter` to reply

Some setup details for the readme

I just wanted to say that this is a really excellent program, but I wanted to share the steps I needed to take to set this up on my Windows machine:

npm i vite
npm install -g typescript
npm run build
npm run dev

An important note, I had an issue with running npm i typescript and instead found success with npm install -g typescript, it gave me an error that I couldn't resolve unless I installed typescript with the -g parameter. May be a noob mistake, but it might increase accessibility to include these instructions in the readme. Side note, when I clicked onto the documentation link, I got a 404.

Was able to resolve the issue thanks to this post:
https://stackoverflow.com/questions/35369501/tsc-is-not-recognized-as-internal-or-external-command

Thanks!

Edit: Also, thought I would share this, I asked on Reddit for a WebAPI where I can edit ChatGPT's answers and I was directed here.

Allow specifying a role for user-supplied message

This may help us model a good example for the ChatGPT model to follow.

not able to render output markdown

Today, I'm curious and test this against the normal https://chat.openai.com/chat interface.

Expected

Happened

[
  {
    "role": "user",
    "content": "n table format, could you compare HLS, HLS-LL, MSE, WebRTC, rtsp?"
  },
  {
    "role": "assistant",
    "content": "\n\n| Technology | Description | Pros | Cons |\n| --- | --- | --- | --- |\n| HLS | HTTP Live Streaming (HLS) is a adaptive streaming protocol developed by Apple for delivering media content over the internet. | Efficient video delivery, agile transcoding, inline privacy, and security | High latency, mixed browser support, playback errors |\n| HLS-LL | HLS Low Latency (HLS-LL) is an extension of the HLS protocol, which reduces the stream latency significantly to provide a better real-time user experience. | Low latency, compatibility with inexpensive infrastructure, less buffer waiting time | Limited ecosystem support, increased complexity for video delivery |\n| MSE | Media Source Extensions (MSE) is a web API that provides an interoperable mechanism to stream media content via JavaScript. | Better control over the user experience, custom player interface | Limited browser support, requires web developer integration |\n| WebRTC | Web Real-Time Communication (WebRTC) is a web API that enables peer-to-peer communication between browsers in real-time, with low latency and high-quality audio and video. | Real-time communication, no plugins or software download required, excellent user experience | More complex to integrate, requires a robust infrastructure |\n| RTSP | Real-Time Streaming Protocol (RTSP) is a network protocol used for delivering real-time streaming media content over the internet. | Low latency, high-quality video and audio, wide compatibility with media players | More complex to setup, potential firewall issues, security issues |\n\nIn summary, HLS and HLS-LL are adaptive streaming protocols that are widely used for delivering video content over the internet, while MSE and WebRTC are web APIs that provide more flexibility in terms of customizing the user experience. RTSP is a network protocol widely used in professional and enterprise-grade video streaming applications. All of the technologies have various advantages and disadvantages, making them suitable for different use cases and applications."
  }
]

allow changing model temperature

right now the temperature is defaulted to 0.7 and there is no UI to customize this temperature value

task: create a UI to allow changing the temperature (maybe top_p as well)

[object Object] error, improper handle 4xx response

Make the UI more mobile friendly

Right now when using ThreadGPT in mobile, each indentation level take up a lot of screen space.

Store number of tokens locally

would be good to store a number of tokens locally so that I can use them to compare pricing between using the API or subscribing to ChatGPT Plus

Let me know if it it good. I will contribut

Can I do code refactoring?

Seem App.tsx reach 800 lines already.

Tweak message

Sometimes the prompt that I send is not good enough for the model to produce a good result. I need to tweak my message.

Right now I have to do this:

1. Copy my message

   

2. Create a new thread

3. Paste the copied text

4. Submit and generate a reply

To make it easier:

• Create a new menu item "Tweak message" for user-supplied message

  

• When clicked on that, the compose form appears on the parent and the text is copied into the compose form