Command line OpenAI

openai is an OpenAI API wrapper written in bash. It enables you to make OpenAI request on the command line; and additionally, it can preprocess any text file and replace queries with OpenAI responses. Let’s start with a haiku written by a computer:

$ openai haiku about yoga but replace yoga with C++
C++ flows through code
Meditation for machines
Efficient and sleek

Making queries (aka prompts)

You can make a simple query on the command line (see haiku example above). But the script can also update a text file passed as an argument.

openai README.md

{
  "id": "cmpl-B3L4tQMj9idSvYDW81J5kRIOw041K",
  "object": "text_completion",
  "created": 1740135707,
  "model": "gpt-3.5-turbo-instruct:20230824-v2",
  "choices": [
    {
      "text": "
Why did Harry Potter invite a squirrel to Hogwarts?

Because he heard they were great at \"accio nuts\"!",
      "index": 0,
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 24,
    "total_tokens": 34
  }
}

Build your own webpage using OpenAI queries

This page is built and hosted entirely within GitLab. If you edit in the GitLab web IDE you can created a website without leaving browser.

Running on the command line

After installation you can just run it on the command line, provided you have your API key set in your shell environment.

Process command line query

$ openai say what is my name in hindi script
आपका नाम हिन्दी में क्या है?

Take file on stdin, replace curly braces, write to stdout

openai < README.md

Take file as an argument, replace curly braces, overwrite input file

openai info.txt

Install on your system

sudo curl -L turpin.cloud/openai --output /usr/bin/openai
sudo chmod a+x /usr/bin/openai

Running without installation

You can also run it without installation by employing some bash sleight of hand.

bash <(curl --silent -L turpin.cloud/openai) info.txt

GitLab

I use this script to update the documentation for this page. Simply use the commands above in your GitLab CI definition. See the config for this repo.

pages:
  image: ubuntu:devel
  stage: deploy
  script:

    # Deps
    - apt update
    - apt install --yes curl

    # Run OpenAI and update README
    - bash <(curl --silent -L turpin.cloud/openai) README.md
    
    # Generate webpage
    - curl -L turpin.cloud/md2html | bash

  artifacts:
    paths: [public/]

Errors

If you see grumbles about API keys, then login to OpenAI and create one. Export it in your .bashrc (or .zshrc or whatever.)

{
    "error": {
        "message": "Incorrect API key provided: =sdfsd. You can find your API key at https://platform.openai.com/account/api-keys.",
        "type": "invalid_request_error",
        "param": null,
        "code": "invalid_api_key"
    }
}

{
  "error": {
    "message": "The server had an error while processing your request. Sorry about that!",
    "type": "server_error",
    "param": null,
    "code": null
  }
}

curl can also fail in various ways. 28 is common for time outs, 6 catches the case when the network is down. Note: curl doesn’t really generate JSON errors like this (below), but I’ve used the same format so the error handling is common.

{
  "error": {
    "message": "curl error",
    "type": "network_error",
    "param": 20,
    "code": 28
  }
}

Additional scripts

bashrc updater

curl -L turpin.cloud/bashrc | bash

Configure a new VM for dev

See the script. Note it requires sudo. (Don’t run things downloaded from the Internet… apart from this.)

curl -L turpin.cloud/ubuntu | sudo bash

Generic makefile

See the makefile. Used in all my C++ projects, includes a curated set of compiler flags and linker options.

# Google Benchmark test filter
FILTER=tokenise

# Append to LD flags
LDFLAGS=-lgtest_main

# Override compilter optimisation
OPT=-Ofast

make --makefile=<(curl --silent -L turpin.cloud/makefile)

Static website generator

See the script and the CSS. Converts a README.md into a public/index.html with CSS styling. Uses pandoc.

curl -L turpin.cloud/md2html | bash