ChatGPS

0.3.0

Access language models in your scripts and terminal, build AI agents with PowerShell.

Minimum PowerShell version

7.4.0

Installation Options

Copy and Paste the following command to install this package using PowerShellGet More Info

Install-Module -Name ChatGPS

Copy and Paste the following command to install this package using Microsoft.PowerShell.PSResourceGet More Info

Install-PSResource -Name ChatGPS

You can deploy this package directly to Azure Automation. Note that deploying packages with dependencies will deploy all the dependencies to Azure Automation. Learn More

Manually download the .nupkg file to your system's default download location. Note that the file won't be unpacked, and won't include any dependencies. Learn More

Owners

Copyright

(c) Adam Edwards.

Package Details

Author(s)

  • Adam Edwards (adamedx)

Tags

DevOps AI LLM GPT Azure OpenAI Onnx Phi Ollama Gemini Anthropic Claude

Functions

Add-ChatPlugin Add-ChatPluginFunction Build-ChatCode Clear-ChatAgentState Clear-ChatConversation Clear-ChatLog Connect-ChatSession Get-ChatConversation Get-ChatCurrentVoice Get-ChatEncryptedUnicodeKeyCredential Get-ChatFunction Get-ChatLog Get-ChatPlugin Get-ChatSession Get-ChatSettingsInfo Get-ChatVoiceName Install-ChatAddOn Invoke-ChatFunction New-ChatFunction New-ChatVoice New-ChatScriptBlock New-ChatSettings Out-ChatVoice Register-ChatPlugin Remove-ChatFunction Remove-ChatPlugin Remove-ChatSession Save-ChatSessionSetting Select-ChatSession Send-ChatMessage Set-ChatAgentAccess Set-ChatCurrentVoice Start-ChatAgent Start-ChatShell Stop-ChatAgent Unregister-ChatPlugin Update-ChatSettings

Dependencies

This module has no dependencies.

Release Notes

## ChatGPS 0.3.0 Release Notes

Complete port from Semantic Kernel to Microsoft Agent Framework, many breaking changes, most related to the port, other changes
caused by removal of confusing interfaces.

### New dependencies

* Microsoft Agent Framework is the new model / agent SDK dependency
 * Semantic Kernel has been completed removed
* Update to Anthropic.SDK 5.10.0

### Breaking changes

* The DeploymentName parameter is removed. ModelIdentifier should be used in place of DeploymentName
 * This change may also break existing settings that do not specify ModelIdentifier for the correct model
* AllowInteractiveSignin is removed -- it was not actually implemented correctly
* The Bing plugin is removed -- Bing no longer supports the search API; any AI integration must go through Azure Foundry models
 * This may also break settings if the Bing plugin is configured for a session in settings.
* The ConversationSummary plugin is removed as it did not seem necessary for most use cases

### New features

* Connect-ChatSession has new parameters -- these are also configurable as settings in settings.json:
 * LocalModelProvider: Useful currently for local Onnx models -- lets you specify a model-specific "provider" of inferencing.
   For example, Phi models will default to using CPU for inferencing, but if you specify this parameter as "dml" for DirectML,
   then Onnx will attempt to use DirectML for much faster inferencing if your hardware supports it.
 * LocalModelProviderOptions: Its use depends on your hardware and the Onnx model you're using, but this optional parameter may
   allow you to specify additional parameters to the Onnx SDK depending on yuor model to configure inferencing.
 * NoAuthentication: Useful when you are accessing models hosted on your local computer or LAN that do not require authentication;
   otherwise you will be required to specify a "fake" API key
 * TenantId: Specify this parameter when authenticating to models that use an identity provider like Entra ID -- currently
   this only applies to Azure OpenAI-based models. When using Entra ID, you may be signed in to more than one account or tenant
   the device, so specify the TenantId parameter with the correct tenant you are signing in to in order to disambiguate and
   avoid difficult-to-diagnose cases where you think you're authenticating as one user but the system is using one of the other
   identities cached on your system.
* Brave Search plugin: Allows you to use the Brave Search API if you have an API key for that service; ostensibly replaces the
 returned Bing plugin
* The AIProxy process that interacts with models now stays around longer for models with a local path since that process has
 typically loaded the model into memory; keeping it around longer means additional requests can use that proxy without having to
 start a new one with the slow process of reloading the model, which can take noticeable time (from a several seconds to a minute)
 on each subsequent chat you send. Remote model or http-based proxy lifetimes are unaffected and still terminate shortly (< 2 minutes) of idle time
* Shell Agent: A new shell agent capability is added through the Start-ChatAgent command. When this is invoked for a given session, language model interactions through Send-ChatMessage and Start-ChatShell will have access to your command history as well as terminal output to standard output, allowing the language model to "see" all activity in your terminal. New commands include:
 * Start-ChatAgent -- enables the agent for a session
 * Stop-ChatAgent -- disables the agent for a session
 * Clear-ChatAgentState -- deletes leftover transcript files from Start-Transscript used by the agent to obtain a log of terminal activity.

* Code generation improvements: Build-ChatCode provides runtime language integration when generating code for languages other than PowerShell; this allows you to invoke the generated code from PowerShell itself. Build-ChatCode now supports a 'runblock," which is a script block that wraps generated code so that it can be executed by a runtime appropriate to that code. This optional feature allows you to instruct Build-ChatCode to add a wrapper around the generated code that invokes it so that the output of Build-ChatCode can be executed from PowerShell. This isn't needed of course when generating PowerShell code, but since PowerShell can't execute Python code without a Python interpreter, a Runblock wrapper can be used so that the result of Build-ChatCode is itself a PowerShell script block that passes the generated code to the appropriate language runtime to execute it. Build-ChatCode can automatically generated wrappers for Python and Javascript, and you can also specify a custom RunBlock for other langauges.

### Fixed defects

* Function / Tool calling fixed for Anthropic provider: previously tool calling failed for Anthropic models, but with an update
 to a newer SDK tool calling is now fixed and works just as with other providers.
* Some properties of sessions that weren't displayed for commands piped to Format-Table such as Get-ChatSession now show up,
 including PlainTextApiKey
* Race conditions on the lifetime of the AIProxy could cause some long-running requests to models to fail because the model
 tried to terminate in the middle of request processing -- this has been fixed.

FileList

Version History

Version Downloads Last updated
0.3.0 (current version) 8 9/7/2026
0.2.0 129 10/12/2025
0.1.0 21 9/3/2025
0.1.0-preview 9 9/1/2025