Features

PezzoOpenAI Class

A new class named "PezzoOpenAI" used to execute OpenAI chat completion requests. Mirroring the native OpenAI.createChatCompletion method (plus a 2nd argument for variables and custom properties).

Usage example - with prompt management

Assuming the Pezzo client has been initiated:

const prompt = await pezzo.getPrompt("GenerateTasks");

const result = await openai.createChatCompletion(prompt, {
  variables: {
    goal,
    numTasks,
  },
  properties: {
    traceId: "SomeTraceId123",
  },
});

Usage without prompt management

const result = await openai.createChatCompletion({
  model: "gpt-3.5-turbo",
  messages: []
}, {
  variables: {
    goal,
    numTasks,
  },
  properties: {
    traceId: "SomeTraceId123",
  },
});

Paginated Requests Table

Filters

Basic filtering is in place. Still need to be extended to support custom properties, date picking etcetera.

"Unmanaged Prompt Warning"

If requests are reported but prompts are not managed by Pezzo, a warning will be displayed, with a tooltip on hover. In the future we will allow auto-importing of prompts.

Request Inspector