Access ChatGPT API using Swift
License: MIT License
Is there a way to check if a stream is still going on or finished?
How to support azure OpenAI?
Is the it possible to show the usage data in the response?
'usage': {'prompt_tokens': 56, 'completion_tokens': 31, 'total_tokens': 87},
https://platform.openai.com/docs/guides/chat/response-format
And finish_reason
Can you please add support for stopping the generation of a response and regenerating a response, similar to how we have in ChatGPT?
I can't seem to open it when I run it on mymac I don't see the app I know I am a noob can someone please explain how to get it running on my Mac
Here https://github.com/alfianlosari/ChatGPTSwift/blob/main/Sources/ChatGPTSwift/ChatGPTAPI.swift#L58:
_ = historyList.dropFirst() just returns a new sequence without the first element and doesn't change historyList itself. Instead, you can use historyList.removeFirst().
PS Thanks for the package!
Hello 👋, I have an issue with AsyncThrowingStream. For some reason it is not working when app is switched to background. For now I am getting "The network connection was lost error" in my app. But the same result is with your test app too. My goal is to make streaming work in background, for now I am trying to save the AsyncThrowingStream<String, Error>.Iterator and continue it when app is returned to foreground but it is not working either. Can you please help me to solve this problem ?
How to cancel the stream request?
gpt 3.5 has a problem with only having data until 2021. As far as I know, the web has solved this with a plug-in. I wonder if this can't be done in the API. I am Korean, if I ask who the president is gpt will answer the 2021 president, not the current one.
Failed to resolve dependencies Dependencies could not be resolved because 'gptencoder' 1.0.0 contains incompatible tools version (5.8.0) and 'chatgptswift' depends on 'gptencoder' 1.0.0.
The request timed out in China.
Hi, thanks for the good work.
Just wondering can you add the ability to add the history back to the api?
Can you please add support for gpt-3.5-turbo-16k longer context model which support four times the context length of the 4k base model?
Looks like the image generation interface is missing in the ChatGPTAPI, is it intended? Are you able to add support for it? Thanks.
These are some missing response in stream, even in the demo app, output include missing stream.
Hey! Thank you for this amazing repository.
We are trying to handle a situation where we don't let the user send another message while the chat is responding to the user's input. Currently, as far as we can tell, there is no way to detect when the message stopped streaming (tho I do see something like): continuation.finish()
But it's not exposed. Thank you!
Is it possible to add content to the package, eg like we do in the web interface when we can add content or like we can do here:
https://betterprogramming.pub/how-to-build-your-own-custom-chatgpt-with-custom-knowledge-base-4e61ad82427e
where we add Custom Knowledge ?
I am not exactly sure what changed, but after updating the package from version 1.3.3 to 1.4.0 I noticed that while streaming the response, words were occasionally being skipped. Downgrading to 1.3.3 resolved this issue. Might be worth looking into.
I don't have a reproducible example but about every 100 words I notice a missing word.
It's much more noticeable while generating code since the syntax error becomes evident.
Hello, thank you for creating the ChatGPTSwift library. I noticed that the tokenizer currently used is the BPE tokenizer for ChatGPT-3, which is different from the Unigram language model tokenizer used by GPT-3.5.
Since we need to manually count the used tokens in stream mode, I was wondering if there is a plan to implement the GPT-3.5 tokenizer in the ChatGPTSwift library.
Thank you for your consideration.
Loks like
public static let defaultModel = "gpt-3.5-turbo"
is removed from ChatGPTAPI on line 18 should this be added back?
Was present in build 1.7.0 not present in 2.1.0
AnalyticsView - I call the function try await vm.analyseWeek()
AnswersAnalyticsVM:
private let api = ChatGPTAPI(apiKey: Constants.Keys.chatgpt)
private let gptModel = "gpt-3.5-turbo"
func analyseWeek() async throws {
do {
let weekAnswers = try await collectAllWeekAnswers() // Fetching the data to input ChatGPT
let analysisString = try await api.sendMessage(text: weekAnswers, model: gptModel)
if let userid = await Session.shared.user_id {
// Saving output to DB
} else {
throw CustomError.sessionError
}
} catch let err{
throw err
}
}
App crashes when I call api.sendMessage() with the message "Swift/RangeReplaceableCollection.swift:622: Fatal error: Can't remove first element from an empty collection".
XCode crash trickle-down:
public func sendMessage(text: String,
model: String = ChatGPTAPI.Constants.defaultModel,
systemText: String = ChatGPTAPI.Constants.defaultSystemText,
temperature: Double = ChatGPTAPI.Constants.defaultTemperature) async throws -> String {
var urlRequest = self.urlRequest
urlRequest.httpBody = try jsonBody(text: text, model: model, systemText: systemText, temperature: temperature, stream: false) // Here
...
private func json```
Body(text: String, model: String, systemText: String, temperature: Double, stream: Bool = true) throws -> Data {
let request = Request(model: model,
temperature: temperature,
messages: generateMessages(from: text, systemText: systemText),
stream: stream) // Here
return try JSONEncoder().encode(request)
}
private func generateMessages(fro```
m text: String, systemText: String) -> [Message] {
var messages = [systemMessage(content: systemText)] + historyList + [Message(role: "user", content: text)]
if gptEncoder.encode(text: messages.content).count > 4096 {
_ = historyList.removeFirst()
messages = generateMessages(from: text, systemText: systemText) // Here
}
return messages
}
Why does it happen? I think it's a bug.
Hello, thanks for your work, would you consider adding the ability to generate images.
Do you know why? I can see the init is there but Xcode won't let me build. I tried to uninstall and reinstall the package, cleaning the cache, cleaning the build, still no luck
Can you add an input box to set the API key?
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.