chatgpt_client_stream

Description:

chatgpt client stream

ChatGPT Client #
Access OpenAI ChatGPT Official API using Dart Language. Supports any Dart project and all Flutter target platforms (iOS, Android, Windows, Linux, Web)
Features #

Using OpenAI ChatGPT Official Completions API Endpoint.
Maintain Chat History List on the client (~4000 tokens) so new prompt is aware of the previous chat context.
Support any custom model as Parameter (GPT-3.5, GPT-4, etc)
Support Stream HTTP Response using native Dart Stream.
Support Standard HTTP Response.
Pass custom system prompt and temperature.

Getting started #
Register for API key from OpenAI.
Installation #
This package is hosted in Pub Dev Library
Flutter App #
flutter pub add chagpt_client
copied to clipboard
Dart App #
dart pub add chagpt_client
copied to clipboard
Usage #
Initialization #
Initialize with api key. Default model is gpt-3.5-turbo.
import 'package:chatgpt_client/chatgpt_client.dart';

const client = ChatGPTClient(apiKey: "API_KEY");
copied to clipboard
optionally, you can provide the system prompt, temperature, and model like so.
const client = ChatGPTClient(apiKey: "API_KEY",
model: "gpt-4",
systemPrompt: "You are a CS Professor",
temperature: 0.7);
copied to clipboard
There are 2 APIs: stream and normal
Stream Response #
The server will yield chunks of data until the stream completes or throws an error.
try {
var text = "";
final stream = client.sendMessageStream(prompt);
await for (final textChunk in stream) {
text += textChunk;
print(textChunk);
}
print(text);
} catch (exception) {
print(exception.toString());
}
copied to clipboard
Normal #
A normal HTTP request and response lifecycle. Server will send the complete text (it will take more time to response)
try {
final text = await client.sendMessage(prompt);
print(text);
} catch (exception) {
print(exception.toString());
}
copied to clipboard
History List #
The client stores the history list of the conversation that will be included in the new prompt so ChatGPT aware of the previous context of conversation. When sending new prompt, the client will make sure the token is not exceeding 4000 (using calculation of 1 token=4chars), in case it exceeded the token, some of previous conversations will be truncated
You can also delete the history list by invoking
client.clearHistoryList();
copied to clipboard