code-yeongyu / aishell Goto Github PK

• The project relies on a module called revChatGPT, but it's using the leaked version of ChatGPT which is no longer functional.

• This issue has been discussed in more detail here: acheong08/ChatGPT#608. So now, AiShell isn't working either.

• I'm looking for a way to fix this.

  • I was thinking of either switching to GPT3 or finding a browser-based version of revChatGPT.
  • Any thoughts or suggestions would be greatly appreciated.

some approaches like yt-dlp has,

--cookies-from-browser

to get ChatGPT access key using browser's cookie

ref: https://github.com/yt-dlp/yt-dlp/blob/b38cae49e6f4849c8ee2a774bdc3c1c647ae5f0e/yt_dlp/cookies.py#L120

It's not required anymore

Context

Currently, the installation process involves a shell script and a Python script that installs Python modules globally, which isn't ideal.

Proposal

• Make AiShell installable via Homebrew or PyPI
  • For Homebrew, I'm considering to pyinstaller for executable binary

This issue contains the content of #52:

As a solution of this, I am thinking of:

1. Update config json like following
  ```json
{
    "language_model": "reverse_engineered_chatgpt",
    "openai_api_key": null,
    "browser": "brave" // firefox, chrome, safari or whatever
}

2. Replace session token retrieving code as access token retrieving code

• Using the browser's cookie, send request to https://chat.openai.com/api/auth/session to get the access token

3. Send request to ChatGPT using configured information from 2

It seems there’s an SSL handshake failure while attempting to establish a secure connection with the OpenAI API server.

When running the following command:

aishell 'print Hello World'

I'm getting this error:

SSLError: HTTPSConnectionPool(host='explorer.api.openai.com', port=443): Max retries
exceeded with url: /api/auth/session (Caused by SSLError(SSLError(1, '[SSL:
SSLV3_ALERT_HANDSHAKE_FAILURE] sslv3 alert handshake failure (_ssl.c:992)')))

Are there any specific Brave browser settings that could be causing this issue?

Is it possible to generate a git commit message based on a diff?
For instance, using pipes:

git diff | aishell "Generate git commit message for my changes"

It would be great if the system could implement some safety measures for commands that could potentially cause harm, such as:

• handling git
• removing files permanently

The idea is that, before executing any dangerous command, AiShell could ask the user for confirmation. This way, accidental execution of harmful commands can be prevented.

It supports loging into your chatgpt acoount with email and password and it works.

https://github.com/ayaka14732/ChatGPTAPIFree

What is YGK-a?

• https://github.com/code-yeongyu/YGK-a
• A brother project of AiShell, a unix pipeline supported ChatGPT client for terminal.

Why replace?

• Though AiShell is a older-brother of 'YGK-a', the codebase that does similar thing is more recent version in YGK-a
• Unifying similar codebase would decrease the pressure of maintaining two projects
  • The actual dependencies would be the same, but at least the dependencies that are should be maintained from AiShell would be decrease at the same time.

Now the project uses rich for the cli, which gives much better experiences.

Update README again to reflect such thing.

some kind of ~/.aishell_config.json file to set

• what model to use by default (reverse-engineered original chatgpt vs gpt3 vs gpt-3.5-turbo(chatgpt model))
• openai token or chatgpt access token

Same as: code-yeongyu/YGK-a#24

https://openai.com/blog/introducing-chatgpt-and-whisper-apis

and also revChatGPT supports it: https://github.com/acheong08/ChatGPT