The AI Chat activity lets you connect to the open AI api to make use of Chat GTP in your codolex project.
Uses System.Net.HttpClientComponent.TNetHTTPClient
Activity properties
begin ... try NetClient := THTTPClient.Create; NetClient.CustHeaders.Add('Authorization', 'Bearer ' + 'YourApiKey'); NetClient.ContentType := 'application/json';
var Response := NetClient.Post('https://api.openai.com/v1/chat/completions', JsonStream);
var IsSuccessful := (Response.StatusCode > 199) and (Response.StatusCode < 300); if not IsSuccessful then raise ENetException.CreateFmt('%d %s%s%s', [Response.StatusCode, Response.StatusText, sLineBreak, Response.ContentAsString]);
JsonResponse := TJSONValue.ParseJSONValue(Response.ContentAsString);
var ChoiceJson := JsonResponse.GetValue<TJSONArray>('choices'); var MessageJson := ChoiceJson[0].GetValue<TJsonObject>('message');
ChatMessage := JsonAdapter.CreateEntity; JsonAdapter.MapToEntity(MessageJson, ChatMessage); ... end; |
Part of resulting Code
More information about the API:
The model can be set to one of the 5 options, Choosing the right model for your needs and token usage is recommended.
•gpt-4o (default)
•gpt-4o-mini
•gpt-4-turbo
•gpt-4
•gpt-3.5-turbo
The role can also be selected, depending on the type of conversation needed.
more about roles: https://platform.openai.com/docs/guides/chat-completions/overview
Provide a list of messages if you want to keep a conversation instead of just one question. this needs to be a list of the plugin datasource entity 'ChatMessage'.
Keeping track of the conversation is important for asking follow-up questions, so the AI knows what you are talking about.
The question and response messages are directly added in the activity.
'Chat message' and 'Authorization key' are string fields that can be filled with the expression editor.
