Skip to content

Latest commit

 

History

History
131 lines (98 loc) · 5.33 KB

Migrate_ChatCompletion_to_Response.md

File metadata and controls

131 lines (98 loc) · 5.33 KB

Migrate from Request-ChatCompletion to Request-Response

This guide explains how to migrate your code from the traditional Request-ChatCompletion to the new Request-Response interface in PSOpenAI module.

Functional Differences

There are several important functional differences.
Request-Response is not a superset of Request-ChatCompletion. There are things that can only be done with Request-ChatCompletion.

Feature ChatCompletion Response Notes
Basic text chat ✅ ✅ Both support core chat functionality
Audio I/O ✅ ❌ Audio features only available in ChatCompletion
Built-in tools ❌ ✅ File search, Computer-Use, etc.
History management Manual Auto / Manual Response maintains state automatically

Usage Differences

The basic usage of Request-ChatCompletion and Request-Response is quite similar.

Request-ChatCompletion

$Output = Request-ChatCompletion -Message 'Hello' -Model 'gpt-4o-mini'
$Output | Format-List
id                 : chatcmpl-abc123
object             : chat.completion
model              : gpt-4o-mini-2024-07-18
choices            : {@{index=0; message=; logprobs=; finish_reason=stop}}
usage              : @{prompt_tokens=8; completion_tokens=10; total_tokens=18}
service_tier       : default
system_fingerprint : fp_abc123
created            : 2025/04/11 10:15:43
Message            : Hello
Answer             : {Hello! How can I assist you today?}
History            : {System.Collections.Specialized.OrderedDictionary}

Request-Response

$Output = Request-Response -Message 'Hello' -Model 'gpt-4o-mini'
$Output | Format-List
id                   : resp_abc123
object               : response
status               : completed
model                : gpt-4o-mini-2024-07-18
output               : {@{id=msg_abc123; type=message; status=completed; content=System.Object[]; role=assistant}}
previous_response_id : 
store                : True
truncation           : disabled
usage                : @{input_tokens=8; input_tokens_details=; output_tokens=10; output_tokens_details=; total_tokens=18}
created_at           : 2025/04/11 10:15:43
LastUserMessage      : Hello
output_text          : Hello! How can I assist you today?
History              : {@{role=user; content=System.Object[]}, @{role=assistant; content=System.Object[]}}

Here are several changes to consider when replacing:

Output Properties Have Changed

The property containing the last user message has changed to LastUserMessage, and the model output is now in the output_text property.

You might question why it's not OutputText, but this aligns with OpenAI's API reference documentation.

output_text is always a single string. (The Answer property in Request-ChatCompletion is always an array of strings.)

# Responses
$Output = Request-Response -Message 'Hello' -Model 'gpt-4o-mini'
echo "You say : $($Output.LastUserMessage)"
echo " AI say : $($Output.output_text)"

## (Chat Completions equivalent)
$Output = Request-ChatCompletion -Message 'Hello' -Model 'gpt-4o-mini'
echo "You say : $($Output.Message)"
echo " AI say : $($Output.Answer[0])"

Default Model Changed to gpt-4o-mini

When the -Model parameter is not specified in Request-ChatCompletion, gpt-3.5-turbo is used.
When the -Model parameter is not specified in Request-Response, gpt-4o-mini is used.

It's recommended to always specify the -Model parameter

Cannot Pass User Messages Through Pipeline

## You can do this
"Hello." | Request-ChatCompletion -Model 'gpt-4o'

## Cannot do this
"Hello." | Request-Response -Model 'gpt-4o'

Conversation history can be passed through the pipeline.

## Conversation state
$First = Request-Response -Message "Hello."
$Second = $First | Request-Response -Message "Can you assist me?"

History Property Type Has Changed

The History property added to maintain conversation history is an array of [OrderedDictionary] in Request-ChatCompletion, but is now an array of [PSCustomObject] in Request-Response.

-Store Is Now Boolean with True Default

The -Store parameter in Request-ChatCompletion is a switch parameter, defaulting to off ($false).
In Request-Response, -Store is a boolean and defaults to $true if not explicitly specified.

Request-Response -Message "Don't save this." -Store $false

Stream Not Output to Information Stream

When using Stream output in Request-ChatCompletion, it is output to both standard output and the Information stream.
In Request-Response, it only outputs to standard output. If you want to capture Stream to both the console host and a variable, you can achieve this using PowerShell's built-in Tee-Object function.

Request-Response -Message 'Hello' -Model 'gpt-4o' -Stream | Tee-Object -Variable StreamOut | Write-Host -NoNewLine
Write-Output $StreamOut