Skip to content

Azure OpenAI Stored Completions

Azure OpenAI Stored Completions is a feature that allows you to store the completions generated by the OpenAI API. This can be useful for various purposes, such as evaluating the performance of the model, training new models, or simply keeping a record of the generated text.

When PromptPex uploads the test results, it attaches metadata to allow filtering of the tests:

  • runId: a unique identifier for the run.
  • testId: a unique identifier for the test.
  • scenario: the scenario name.
  • dataset: training or test to allow for different sets to be used
  • prompt: the name of the prompt
  • storedCompletions: enables the stored completion feature.
  • storeModel: the model to use for storing completions. This is a string that specifies the model name. The default value is the store model alias which resolves to large alias.