Releases: mkht/PSOpenAI
Releases · mkht/PSOpenAI
1.5.0
1.5.0
- Add
-MaxRetryCount
option for all functions.
Retries up to the maximum number of times specified if an API request fails with a429 (Rate limit reached)
or5xx (Server side errors)
error. The retry interval increases exponentially up to 128 seconds. - These obsolated functions are completely removed.
Request-CodeCompletion
Request-CodeEdit
- Minor fixes.
1.3.0
1.2.1
1.2.1
- OpenAI has announced that the Codex API will be discontinued on 2023-03-23, the following functions may no longer work in the future. In the future release, these functions will be completely removed from the module.
Request-CodeCompletion
Request-CodeEdit
- When a request specifying an AI model that has been or will be discontinued by OpenAI is executed, Warning messages will be output. (the request continues to be executed).
AI models that will output warning messages at this time:code-davinci-001
code-davinci-002
code-cushman-001
code-cushman-002
1.2.0
1.2.0
- Add
StopSequence
parameter forRequest-ChatGPT
,Request-TextCompletion
,Request-CodeCompletion
.
When specific words are output from the API, subsequent output is stopped.
e.g.)# This code generates only top 4 list. Request-TextCompletion -Prompt 'List of top 10 most populous countries' -StopSequence '5.'
1.1.2
1.1.0
1.1.0
- Improve error handling.
- Change the default value of the
MaxTokens
parameter ofRequest-TextCompletion
to2048
. The previous default value of16
was not useful for most use cases. - Add the
Request-CodeCompletion
function. This performs the same process asRequest-TextCompletion
, but is more suitable for generating program code since the AI model used by default iscode-davinci-002
. - Add Pester tests for all public functions.