kardolus / chatgpt-cli Goto Github PK

Great work on the CLI!
How do I configure it to use our azure openai deployment? Thanks!

Everytime i try a query, i get a 429.
nothing more is said

running on latest main commit (local build)

setting max_tokens to something low like 20 will still output long text

name: openai
api_key: <redacted>
model: gpt-4-turbo-preview
max_tokens: 20
role: You are a developer
temperature: 1
top_p: 1
frequency_penalty: 0
presence_penalty: 0
thread: default
omit_history: false
url: https://api.openai.com
completions_path: /v1/chat/completions
models_path: /v1/models
auth_header: Authorization
auth_token_prefix: 'Bearer '

I've tried the new version and I believe it's important to reinstate the -m option so that the model can be switched during runtime.

To balance the cost of the APIs, it is often useful to operate with GPT-3 and then pass the output to GPT-4, for example:

chatgpt -m gpt-3 "Do something" | chatgpt -m gpt-4 "Do something else" | ...

Using global variables or file configurations forces one to break the pipeline and save partial results in temporary files, which is not aligned with the rest of the utilities. This doesn't only apply to this scenario, but also when a user wants to change the model on the fly, or create a command alias.

Before this version, I had some very useful aliases:

alias gpt4="chatgpt -m gpt-4"
alias gpt3="chatgpt -m gpt-3-turbo"

which can no longer be created now.

In summary, I believe that:

• every option should have a default value
• every option present in the yaml file should have a corresponding argument
• the hierarchy should be: default, config, argument
• local variables should be used only for API_KEY

I hope these considerations may be of help!
Thank you!

The readline mode can be archived by rlwrap, but it's much better to have it built-in.

$ rlwrap chatgpt -i
[2024-04-01 00:00:00] Q1: hello world┃ # press Ctrl-A to move the cursor to the beginning
[2024-04-01 00:00:00] Q1: ┃hello world 

Thanks!

Hi there,

Since we are using cobra, any plan to use the completion to help with the flags?
I can do a PR if it's ok for you :)

Thanks

Is there a preferred way to provide escape characters especially in cases where we want to provide multi-line code snippets with the prompt?

Cheers and thanks!

This is a great project!

But, in my quest for looking for a way to use the same model that "chat.openapi.com" uses also from the CLI as the browser, I guess I need to keep looking.

I mean of course it's really Openapi's decisions not to offer the ChatGPT model via an API yet.

I think it would benefit this project, if it went to a little bit of effort to explain that in the Readme.

The fact that you can access many different OpenAI models via their APIs, but the ChatGPT model is not yet available.

It would also be worth mentioning the pricing difference.

• Installed the app (windows)
• Set the API key accordingly
• Ran the command
• Got 429 error (too many requests) despite only making a single query.
• Multiple attempts yield same result.

Hi @kardolus ,

Thanks for your work putting this CLI together. It's so nice to use ChatGPT in such a simple way. (And great to use in situations where browsers are disabled, like when using restricted airplane "message only wifi" packages !)

Having used it for a couple of days now, I'm finding the most UX friction to be in changing the threads. The simplest way I have been doing this from the command line is with sed on the config file.

Do you have any plans to add a --set-thread flag, and beyond that, a --list-threads flag?

I think these would make a really positive impact on the UX.

All the best,
Angus

a nice qol because I hand edit the yaml a lot and add comments

As more CLI apps are adopting the path conventions laid out by the XDG Base Directory Specification, it would be nice if config.yaml and other generated files like thread history are placed in these standard paths—or at least look for them in these paths before using ~/.chatgpt-cli/:

• config.yaml — use $XDG_CONFIG_HOME/chatgpt-cli/config.yaml (if XDG_CONFIG_HOME is NOT explicitly defined, use ~/.config/chatgpt-cli/config.yaml)
• history/default.json etc. — use $XDG_DATA_HOME/chatgpt-cli/history/default.json (if XDG_DATA_HOME is NOT explicitly defined, use ~/.local/share/chatgpt-cli/history/default.json)

How does one uninstall this?

Love this tool ❤️

❯ chatgpt -n
Error: Another chatgpt instance is running, chatgpt works not well with multiple instances, please close the other one first.
If you are sure there is no other chatgpt instance running, please delete the lock file: /tmp/chatgpt.lock
You can also try `chatgpt -d` to run in detach mode, this check will be skipped, but conversation will not be saved.

How (much effort) would it be to support multiple insances?
Would be amazing :)

Ready to throw a sunday at it (at some point this year 😅)

nice work, is there any tuto on how to using self-hosted LLM model with same openai API?

--set-<value> writes to file. I wouldn't want that.

Could be useful exclude history from context and/or have a way to start a new conversation. Something like: chatgpt --new-chat.

I just learned in the hard way how can be painful kept a large context in a loop.

Thx

Looking at their github stars progression, FOSS servers are seeing a fast grow. Supporting them will broaden the usage of this tool.
Thank you,
Piero

I just downloaded the program with
curl -L -o chatgpt https://github.com/kardolus/chatgpt-cli/releases/download/v1.3.2/chatgpt-linux-amd64
set up my API key, and run it with a first query. I get the error message:
http error: 429

System: Linux Mint 21.2 Cinnamon
I am using the free tier of chatgpt (3.5), if that is relevant.

so i just installed this that ms copilot told me to do so i can have chatgpt and it says: missing environment variable: OPENAI_API_KEY how am i gonna solve this?

I pay for ChatGPT premium and wondering if it's possible to use my session token so I don't need to pay for api usage?

Hello. I can't change from chatgpt-3 to 4. I have checked using openAI python tool that I have access to gpt-4.

`local:.chatgpt-cli yasin$ chatgpt -l
Available models:

• gpt-3.5-turbo-0613
• gpt-4-0314
• gpt-4-0613

• gpt-4 (current)

• gpt-3.5-turbo-instruct-0914
• gpt-3.5-turbo-instruct
• gpt-3.5-turbo-0301
• gpt-3.5-turbo-16k
• gpt-3.5-turbo
• gpt-3.5-turbo-16k-0613
  local:.chatgpt which model are you
  As an AI developed by OpenAI, I don't have a specific model number. I'm based on the GPT-3 (Generative Pretrained Transformer 3) architecture. My primary function is to assist and facilitate efficient communication.`

Not sure how I should go about debugging this.