mkht / psopenai

PowerShell module for OpenAI API.

PSOpenAI

PowerShell module for OpenAI and Azure OpenAI Service.
You can use OpenAI functions such as ChatGPT, Speech-to-Text, Text-to-Image from PowerShell.

This is a community-based project and is not an official offering of OpenAI.

• About OpenAI API
  https://platform.openai.com/docs

• About Azure OpenAI Service
  https://learn.microsoft.com/en-us/azure/ai-services/openai/overview

日本語版のREADMEはこちら

---

Supported Platforms

• Windows PowerShell 5.1
• PowerShell 7 or higher
• Windows, macOS or Linux

You need to sign-up OpenAI account and generates API key for authentication.
https://platform.openai.com/account/api-keys

---

Installation

You can install PSOpenAI from PowerShell Gallery.

Install-Module -Name PSOpenAI

---

Functions

Common

• ConvertFrom-Token
• ConvertTo-Token
• Get-CosineSimilarity

OpenAI

Chat

• Enter-ChatGPT
• Request-ChatCompletion
• New-ChatCompletionFunction

Assistants

Guide: How to use Assistants

• Get-Assistant
• New-Assistant
• Set-Assistant
• Remove-Assistant
• Get-Thread
• New-Thread
• Set-Thread
• Remove-Thread
• Get-ThreadMessage
• Add-ThreadMessage
• Get-ThreadRun
• Start-ThreadRun
• Stop-ThreadRun
• Wait-ThreadRun
• Receive-ThreadRun
• Get-ThreadRunStep

Images

• Request-ImageEdit
• Request-ImageGeneration
• Request-ImageVariation

Audio

• Request-AudioSpeech
• Request-AudioTranscription
• Request-AudioTranslation

Files

• Get-OpenAIFile
• Register-OpenAIFile
• Remove-OpenAIFile
• Get-OpenAIFileContent

Others

• Get-OpenAIModels
• Request-Embeddings
• Request-Moderation
• Request-TextCompletion

Azure OpenAI Service

• Get-AzureOpenAIModels
• Request-AudioSpeech
• Request-AzureAudioTranscription
• Request-AzureAudioTranslation
• Request-AzureChatCompletion
• Request-AzureChatGPT
• Request-AzureEmbeddings
• Request-AzureImageGeneration
• Request-AzureTextCompletion

Assistants for Azure OpenAI Service

• Get-AzureAssistant
• New-AzureAssistant
• Set-AzureAssistant
• Remove-AzureAssistant
• Get-AzureThread
• New-AzureThread
• Set-AzureThread
• Remove-AzureThread
• Get-AzureThreadMessage
• Add-AzureThreadMessage
• Get-AzureThreadRun
• Start-AzureThreadRun
• Stop-AzureThreadRun
• Wait-AzureThreadRun
• Receive-AzureThreadRun
• Get-AzureThreadRunStep

---

Usages

See Docs and Guides for more detailed and complex scenario descriptions.

ChatGPT (Interactive)

Communicate with ChatGPT interactively on the console.

$global:OPENAI_API_KEY = '<Put your API key here.>'
Enter-ChatGPT

ChatGPT (Scripting)

You can ask questions to ChatGPT.

$global:OPENAI_API_KEY = '<Put your API key here.>'
$Result = Request-ChatCompletion -Message "Who are you?"
Write-Output $Result.Answer

This code outputs answer from ChatGPT

I am an AI language model created by OpenAI, designed to assist with ...

Tip

The default model used is GPT-3.5.
If you can and want to use GPT-4, you can specifies model explicitly like this.

Request-ChatCompletion -Message "Who are you?" -Model "gpt-4"

Audio Speech (Text-to-Speech)

Generates audio from the input text.

$global:OPENAI_API_KEY = '<Put your API key here.>'
Request-AudioSpeech -text 'Do something fun to play.' -OutFile 'C:\Output\text2speech.mp3' -Voice Onyx

You can combine with ChatGPT.

Request-ChatCompletion -Message "Who are you?" | Request-AudioSpeech -OutFile 'C:\Output\ChatAnswer.mp3' -Voice Nova

Audio transcription (Speech-to-Text)

Transcribes audio into the input language.

$global:OPENAI_API_KEY = '<Put your API key here.>'
Request-AudioTranscription -File 'C:\SampleData\audio.mp3' -Format text

This code transcribes voice in C:\SampleData\audio.mp3. Like this.

Perhaps he made up to the party afterwards and took her and ...

Image generation (Text-to-Image)

Creating images from scratch based on a text prompt.

$global:OPENAI_API_KEY = '<Put your API key here.>'
Request-ImageGeneration -Prompt 'A cute baby lion' -Model 'dall-e-2' -Size 256x256 -OutFile 'C:\output\babylion.png'

This sample code saves image to C:\output\babylion.png. The saved image like this.

Multiple conversations with ChatGPT while keeping context.

Request-ChatCompletion accepts past dialogs from pipeline. Additional questions can be asked while maintaining context.

PS C:\> $FirstQA = Request-ChatCompletion -Message "What is the population of the United States?"
PS C:\> Write-Output $FirstQA.Answer

As of September 2021, the estimated population of the United States is around 331.4 million people.

PS C:\> $SecondQA = $FirstQA | Request-ChatCompletion -Message "Translate the previous answer into French."
PS C:\> Write-Output $SecondQA.Answer

En septembre 2021, la population estimée des États-Unis est d'environ 331,4 millions de personnes.

PS C:\> $ThirdQA = $SecondQA | Request-ChatCompletion -Message 'Please make it shorter.'
PS C:\> Write-Output $ThirdQA.Answer

La population des États-Unis est estimée à environ 331,4 millions de personnes.

Stream completion outputs

By default, results are output all at once after all OpenAI responses are complete, so it may take some time before results are available.

To get responses sooner, you can use the -Stream option for Request-ChatCompletion and Request-TextCompletion. The results will be returned as a "stream". (similar to how ChatGPT WebUI displays)

Request-ChatCompletion 'Describe ChatGPT in 100 charactors.' -Stream | Write-Host -NoNewline

Restore masked images

Request-ImageEdit -Image 'C:\sunflower_masked.png' -Prompt 'sunflower' -OutFile 'C:\sunflower_restored.png' -Size 256x256

Masked image is restored to full images by AI models.

Source	Generated

Moderation

Test whether text complies with OpenAI's content policies.

The moderation endpoint is free to use when monitoring the inputs and outputs of OpenAI APIs.

PS C:\> $Result = Request-Moderation -Text "I want to kill them."
PS C:\> $Result.results.categories

# True means it violates that category.
sexual           : False
hate             : False
violence         : True
self-harm        : False
sexual/minors    : False
hate/threatening : False
violence/graphic : False

---

About API key

Almost all functions require an API key for authentication.
You need to sign-up OpenAI account and generates API key from here.
https://platform.openai.com/account/api-keys

There are three ways to give an API key to functions.

Method 1: Set an environment variable named OPENAI_API_KEY. (RECOMMENDED)

Set the API key to the environment variable named OPENAI_API_KEY.
This method is best suited when running on a trusted host or CI/CD pipeline.

PS C:> $env:OPENAI_API_KEY = '<Put your API key here.>'
PS C:> Request-ChatCompletion -Message "Who are you?"

Method 2: Set a global variable named OPENAI_API_KEY.

Set the API key to the $global:OPENAI_API_KEY variable. The variable is implicitly used whenever a function is called within the session.

PS C:> $global:OPENAI_API_KEY = '<Put your API key here.>'
PS C:> Request-ChatCompletion -Message "Who are you?"

Method 3: Supply as named parameter.

Specify the API key explicitly in the ApiKey parameter. It must be specified each time the function is called.
This is best used when the function is called only once or with few calls, such as when executing manually from the console.

PS C:> Request-ChatCompletion -Message "Who are you?" -ApiKey '<Put your API key here.>'

Azure OpenAI Service

If you want to use Azure OpenAI Service instead of OpenAI. You should create Azure OpenAI resource to your Azure tenant, and get API key and endpoint url. See guides for more details.

• Guide: How to use with Azure OpenAI Service

Sample code for Azure

$global:OPENAI_API_KEY = '<Put your api key here>'
$global:OPENAI_API_BASE  = 'https://<resource-name>.openai.azure.com/'

Request-AzureChatCompletion `
  -Message 'Hello Azure OpenAI Service.' `
  -Deployment 'gpt-35-turbo' `

---

Changelog

CHANGELOG.md

---

Plans and TODOs.

If you have a feature request or bug report, please tell us in Issue.

• More docs, samples.
• Performance improvements.
• Add an option for change output types / formats.

---

Licenses

MIT License

psopenai's Issues

Press Enter to confirm and for automatic line wrapping in PowerShell ISE

I usually use PowerShell ISE for code editing, and I have encountered two issues:

1. After inputting a question, I need to press Enter three times to confirm. Can I use Ctrl+Enter to confirm directly?
2. When the text length of ChatGPT's answer is too long, a scrollbar appears, displaying it all in one line, which is not very convenient for reading. Can it automatically wrap at the width of the window?
   demo Gif:
   

add -Proxy parameter

my System: windows 10 Pro 21H2, Powershell 5.1

I'm a novice in PowerShell. The network I'm using cannot directly access the ChatGPT API. I need to use a proxy for access. However, when using the following code, I don't receive any response.

$env:http_proxy = "http://127.0.0.1:1081"
$env:https_proxy = "http://127.0.0.1:1081"
$global:OPENAI_API_KEY = "sk-XXXXX"
$Result = Request-ChatCompletion -Message "Who are you?"
Write-Output $Result.Answer

Therefore, I hope to add a -Proxy parameter, just like below, so that I can successfully use the following code to install the module

Install-Module -Name PSOpenAI -Force -Proxy "http://127.0.0.1:1081"