Skip to main content
Let’s follow the complete flow of a request through Polyglot:

​
Request Processing

  1. Application creates an Inference object
  2. Application calls create() with parameters
  3. Inference creates an InferenceRequest.
  4. Inference creates a PendingInference object with the instances of request, driver and event dispatcher.
  5. Inference returns a PendingInference object to the application.

​
Response Processing

  1. Application accesses the PendingInference object content, e.g. via response() method.
  2. PendingInference checks if HTTP request has been already executed.
    • If already sent, it returns the cached response.
  3. PendingInference dispatches the InferenceStarted event.
  4. PendingInference dispatches the InferenceAttemptStarted event.
  5. Driver dispatches the InferenceRequested event and sends the HTTP request.
  6. Driver parses provider response into InferenceResponse.
  7. Driver dispatches the InferenceResponseCreated event.
  8. PendingInference dispatches InferenceAttemptSucceeded, InferenceUsageReported, and InferenceCompleted.
  9. Result InferenceResponse object is returned to the application.