Asynchronous Call

If you do not want to wait for the OpenAPI response when making a request to its API, you need to use asynchronous code.

To make asynchronous calls, you must provide a callback 4D.Function(Formula) in the OpenAIParameters object parameter to receive the result. For streaming chat completion see OpenAIChatCompletionsParameters.

The callback function will receive the same result object type (one of OpenAIResult child classes) that would be returned by the function in synchronous code. See examples below.

Process Considerations

The asynchronous method is based on 4D.HTTPRequest, so the response will be received within the current process.

⚠️ If your process ends at the conclusion of the current method (e.g., using New process, or playing in the method editor), the callback formula might not be called asynchronously. In such cases, consider using CALL WORKER or CALL FORM.

Examples of Usage

model list

$client.models.list({formula: Formula(MyReceiveMethod($1))})

$1 will be an instance of OpenAIModelListResult, so MyReceiveMethod method could be:

#DECLARE($result: cs.AIKit.OpenAIModelListResult)

If($result.success)

   Form.models:=$result.models

Else

  Alert($result.errors.formula(Formula(JSON Stringify($1))).join("\n"))

End if

chat completions

var $messages:=[{role: "system"; content: "You are a helpful assistant."}]
$messages.push({role: "user"; content: "Could you explain me why 42 is a special number"})

$client.chat.completions.create($messages; { onResponse: Formula(MyChatCompletionsReceiveMethod($1))})

$1 will be an instance of OpenAIChatCompletionsResult, so MyChatCompletionsReceiveMethod method could be:

#DECLARE($result: cs.AIKit.OpenAIChatCompletionsResult)

ASSERT($result.success) // We use onResponse here, callback receive only if success
Form.assistantMessage:=$result.choices[0].text

chat completions with streaming

When you want to receive the response progressively as it's being generated (streaming), you can use the stream parameter along with an onData callback:

var $messages:=[{role: "system"; content: "You are a helpful assistant."}]
$messages.push({role: "user"; content: "Could you explain me why 42 is a special number"})

// Enable streaming and provide onData callback
$client.chat.completions.create($messages; { \
    stream: True; \
    onData: Formula(MyStreamDataReceiveMethod($1)); \
    onTerminate: Formula(MyStreamTerminateMethod($1)) \
})

The onData callback will be called multiple times as data chunks arrive. $1 will be an instance of OpenAIChatCompletionsStreamResult:

// MyStreamDataReceiveMethod
#DECLARE($streamResult: cs.AIKit.OpenAIChatCompletionsStreamResult)

If ($streamResult.success)
    // Check if we have content in the delta
    If ($streamResult.choices.length>0)
        var $choice: Object
        $choice:=$streamResult.choices[0]
        
        If ($choice.delta#Null) && ($choice.delta.content#Null)
            // Append the new content chunk to the existing message
            Form.assistantMessage:=Form.assistantMessage+$choice.delta.content
        End if
    End if
Else
    // Handle streaming error
    ALERT("Streaming error: "+$streamResult.error.message)
End if

The onTerminate callback will be called once when the stream is complete:

// MyStreamTerminateMethod
#DECLARE($result: cs.AIKit.OpenAIChatCompletionsResult)

If ($result.success)
    // Stream completed successfully
Else
    // Handle final error
    ALERT("Stream terminated with error: "+$result.errors.formula(Formula(JSON Stringify($1))).join("\n"))
End if