Getting 'undefined' as a response

[MendelBak]

I'm getting undefined as a response when I attempt to use the extension.

I'm getting a 502 error, when inspecting the request in the console

[CakeCrusher]

@MendelBak
Hey thanks for bringing this up, I will fix it by the end of the week.

[MendelBak]

Thank you! Are you aware of the problem? If it's not too complex, I can give it a shot at trying to fix it, if you point me in the right direction.

[CakeCrusher]

@MendelBak Awesome please go ahead! This should be a relatively simple fix. I am aware of the issue, and it is related to the following background script on the client

tech-int-cheat/background.js

Lines 26 to 74 in f333dad

	if (request.type === "GET_RESPONSE") {
	const { meetId, startChat, endChat } = request.data;
	// get the slice of chat from startChat to endChat and make a deep copy of it
	const chat = chats[meetId];
	const chatSlice = JSON.parse(
	JSON.stringify(chat.slice(startChat, endChat + 1))
	);
	// preprocess the chatSlice so that any recurring roles are merged into one
	const preprocessedChatSlice = [];
	let i = 0;
	let currentChatInstance = null;
	while (i < chatSlice.length) {
	currentChatInstance = chatSlice[i];
	const nextChatInstance = chatSlice[i + 1];
	if (
	nextChatInstance &&
	currentChatInstance.role === nextChatInstance.role
	) {
	currentChatInstance.content += " " + nextChatInstance.content;
	chatSlice.splice(i + 1, 1);
	} else {
	delete currentChatInstance.ticId;
	preprocessedChatSlice.push(currentChatInstance);
	i++;
	}
	}
	// send the preprocessedChatSlice to the backend https://635f-68-234-232-23.ngrok.io/infer_response and pass the preprocessedChatSlice as the json body
	const response = await fetch(
	`https://t23m9fecmc.execute-api.us-east-2.amazonaws.com/Prod/infer_response`,
	{
	method: "POST",
	mode: "cors",
	headers: {
	"Content-Type": "application/json",
	},
	body: JSON.stringify(preprocessedChatSlice),
	}
	);
	const responseJson = await response.json();
	const generatedResponse = responseJson.response;
	// send the generatedResponse back to the popup
	chatGptResponses[meetId] = { generatedResponse, startChat, endChat };
	chrome.runtime.sendMessage({
	type: "CHATGPT_RESPONSE",
	data: chatGptResponses[meetId],
	});
	}

, and this endpoint on the backend https://github.com/CakeCrusher/tech-int-cheat-backend/blob/b9a348760f6c325fe77d78475eadb85013003866/flaskAPI/main.py#L17-L37 . The only disclaimer is that this runs a bit differently on production so if you can't reproduce it let me know.

[MendelBak]

I'm taking a look at it now. How do I run this locally to reproduce the issue? I'm assuming that I need to be running the backend at the same time in order to debug the backend error, correct?

Which backend service are you running? The flask or Node one?

[MendelBak]

Ok, so I figured out how to run it all locally but I don't have an API key, so, obviously, I am can't run it against the local python server.

When I try to debug the extension using your AWS creds, the network just returns a 502 error but I can't actually see any more information than that.

[CakeCrusher]

In response to your message here: #13 (comment)

(For the time being I implemented a catch for the instance where the response is undefined f2e9905)

For some reason, even with your advice to lose focus of the current popup, I was not able to reproduce the issue.

Although I can see how this could happen as a result of focusing away from the current popup, I believe a likelier cause is that there is an out of bounds index passed from the popup here

tech-int-cheat/popup/popup.js

Lines 59 to 68 in adf6e5b

	generateResponseBtn.addEventListener("click", function () {
	// send a message to the service worker of type "GENERATE_RESPONSE" with the startChat and endChat
	chrome.runtime.sendMessage({
	type: "GET_RESPONSE",
	data: { meetId, startChat, endChat },
	});
	startChat = null;
	endChat = null;
	reactToSelectedChats();
	});

to the background resulting in a invalid body: JSON.stringify(preprocessedChatSlice), leading to an error in the backend.

If you could make a video of how you managed to reproduce it that would be awesome. Cuz at the moment I'm bewildered.

[MendelBak]

GIF showing the error reproduction.

I'm pretty sure the error is occurring at the end of the closed captioning file since it repeats every second, which is the setInterval time for that function.

However, it likely isn't the most pressing issue. A more significant issue is the 502 error that is being returned from ChatGPT / the backend server. Like I said, I can't debug the server without an OpenAI key, which I don't have at the moment.

[CakeCrusher]

You can get an API key from here https://platform.openai.com/account/api-keys (you will need to sign in).
As for that error I just noticed its only on deployment 😖 (the web store version), as I don't get that issue in development. I will look into it tomorrow.

I created some basic instructions for setting up the api key here https://github.com/CakeCrusher/tech-int-cheat-backend/blob/master/README.md although you can use the deployed backend if you prefer. Nonetheless this error does not have anything to do with that.

[CakeCrusher]

I just learned why that is happening. You cannot have both the development and deployed "tech-int-cheat" activated simultaneously. You will find that if you turn off the development "tech-int-cheat" it will begin to pick up the messages.

Then were back to the old error, which at least happens when the server wakes. I think I will have to resolve that on my end (deployed server issues).

[CakeCrusher]

@MendelBak I created another issue here #14 to address that.

[CakeCrusher]

deployed a new server that is already configured as of 1db6d3c . It is currently pending review on the web store

[CakeCrusher]

@MendelBak the backend is now the same as the deployed Heroku dyno so you can test things there. I have yet to receive any errors with the new deployment.