⚡│ask-community

[primal acorn]

Welcome to the use case feedback #quick-questions channel!

We’ve set up this space for you to share any feedback, thoughts and ideas related to this particular use case.

Some examples for queries that would apply to this channel would be:

• Asking for product suggestions when shopping
• Any type of conversion
• Fun facts

We also offer a form that enables you to provide a comprehensive and precise report, which assists us in maintaining a record of your valuable feedback. This information allows us to address your concerns and potentially follow up with you in the future.

Feedback form

https://perplexity.typeform.com/to/v61rN7oF

[fluid lark]

I have a question regarding context length. OpenAI recently increased its limits via API quite significantly, see https://arstechnica.com/information-technology/2023/06/openai-rolls-out-big-chatbot-api-upgrades-for-developers/ for more info. Will we benefit from this as well, as it is likely that you use OpenAI API for Perplexity?

[hidden elk]

yes, we are going to integrate this

[edgy patrol]

Same concern. Being forced to keep ChatGPT open in another window because my contexts get too long for Perplexity

[thick lichen]

Am I doing something wrong?
https://www.engadget.com/the-morning-after-openai-and-microsoft-arent-happy-111544921.html?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAACQXOoXhejumKEkQ9YeotOWVPtSPW-qmaOtduUr9I7uwXSysentxlMOGOtCC0kkgcd59Fm4xOJUBughlku8VYnMajkOZ27v1lVQ3tYDkn5itVsTGnX8sSm76PPanCFCMpOcpooEwMW1owFEDJ1Skbs6vaP6CPONXTlXZyWWZq0yo
It gets the article way off

Using the 3.5 was more accurate

[glad imp]

@thick lichen, I think you posted the wrong link, it's not a Perplexity query 🙂

[thick lichen]

My bad
here is the link
https://www.perplexity.ai/search/936d03df-fcd1-44a3-8109-fd850b6cdf63?s=c

[rich marsh]

Its probably because it's using the search request instead of querying the actual site. Sadly Perplexity doesn't seem to be able to "visit" a site directly but relies on search results

[thick lichen]

It does read the website if your query is specific.
I had it read the documentation of an API once

[fluid lark]

what does it mean to "visit" a site?

could it have been a cached page?

who am i?

🤔

[thick lichen]

Like web scrap the body of the site

[burnt bronze]

Could you please tell me what is the function of this switch? What are the advantages and differences when conducting searches while it's turned on or off?

[rich marsh]

https://www.perplexity.ai/search/57ae9acc-1308-4a91-bad7-c0780470dd26?s=u

[haughty brook]

#📘│faq .14 explains the copilot function

[fluid lark]

kind of, not very well, FAQs could be updated i think

[fluid lark]

anybody know the character/token limit for queries in Writing Mode? (GPT-4, no copilot)

[fluid lark]

how to add bot for context to thread for perplexity ? when i ask something and i later ask a new question in the thread should it at least remmeber last response ?

[slim night]

Hey @fluid lark it should, as on the web version. When you are in the #🔎│ask channel, you can start a new thread and the bot will keep the context. You can check it proceeding to your browser and see those questions in a thread.

[patent yoke]

Does perplexity support summarizing youtube videos?

[velvet tangle]

PerplexityAI does not (yet) support the summarisation of YouTube videos. Up to now, you can only focus on YouTube videos in a search.

[wet karma]

I have Perplexity bot on my server, the lenght of the questions etc is the same than in the web?

[velvet tangle]

I think you have the same possibilities as when you use PerplexityAI anonymously.

[real tinsel]

Does only the copilot read your ai profile or also the normal AI?

[haughty brook]

it will always use your ai profile, when enabled

[real tinsel]

Good

Thanks

[real tinsel]

I created a thread as guest when I discovered the website, then I made a account. I can currently continue that chat, but its not in my threads, is it possible to add it to my thread somehow?

Currently I use a bookmark

[haughty brook]

Login into your account then go the thread. You'll see an icon like this. That means if you ask a new question it will fork the thread to your threads. that way you can continue the thread and save it to your account.

[real tinsel]

Available in web? I don't see it

[haughty brook]

yes it's on the web

[haughty brook]

are you currenly logged in there?

[real tinsel]

Yes

It say Anonymous

From my guest writing

[haughty brook]

@real tinsel could you open this link. and tell me what icon you see

https://www.perplexity.ai/search/5145b562-e219-41cb-9d8c-281dd83392a1?s=u

[real tinsel]

Why not on my own? 😭

[real tinsel]

Your link icon

[haughty brook]

If you ask a follow up on my thread do you see your profile picture in the top left?

[real tinsel]

I first have to click that blue button right?

[haughty brook]

yes

[real tinsel]

Because in my chat is only a arrow

[haughty brook]

yes just ask a random follow up on my thread I just send

[real tinsel]

Yes it makes it with my avatar

[haughty brook]

you see your own name and profile there and not Anonymous?

[real tinsel]

Yes

[haughty brook]

Hmmm weird could you go back to that one Anonymous thread and refresh and see if the icon changes

[real tinsel]

I use my old thread since 2 days it doesn't appear

I can continue it but not convert it to my own

I can also see all messages

[haughty brook]

would you mind sharing the thread with so I can better check the problem? If you're comfortable with sharing the thread link, you can dm if you want to.

[real tinsel]

Its too private.

There are hours of work involved xD I spent today like 5-6 hours in the thread 😁

[haughty brook]

Alright no problem I understand. The only thing that maybe would help is a refresh or try to recreate the thread on your account. Or maybe try to login in private browser tab

[real tinsel]

Well I refresh the thread always

Logged in via chrome

It appears but it doesn't work....

[haughty brook]

oh.. raised my hands too quick

[real tinsel]

Doesn't show up in the threads or anything

[haughty brook]

what doesn't work exactly? You need to ask a new question to fork it, it can't be empty

[real tinsel]

No in your thread did I click the button and got a fresh thread

Tested it again and now it didn't worked in your thread xD i had to write something

[haughty brook]

yes that should always be the case

[real tinsel]

I really clicked it without a text and got the thread cloned with my name on top without a reply

[haughty brook]

On which browser was that?

[real tinsel]

Opera

This IS perplex xD

🙂

This is really crazy

I make a short video brb

@haughty brook

I can't continue the math chat but I can continue my own chat but i have to go through the follow up...

Well atleast I got a link now xD

[haughty brook]

ye it's a super weird bug haven't seen that before but atleast you were able to fork it now 🙌

[real tinsel]

I can now access the thread from my threads but its really weird, but tbh. I prefer that I can continue it and not make a follow up. It feels like the follow up would not check the ladt thread, did you had some tests with it? How much of the last conversation does the ai remember?

[wet karma]

Hi, I'm trying to do a discord bot for my own server that simply makes the prerplexity bot read the instructions at https://www.aiexperiments.org/instructions.txt and later give me a midjourney prompt with those rules, but first I would like to know if it violates you politics or is discouraged (then I will not continue) you can see the actual status on https://github.com/imigueldiaz/ai-experiments

[real tinsel]

Can I set it to default on writing?

And does Youtube mean it search for content on YouTube?

[velvet tangle]

Summarize YouTube videos is not supported yet, but it will search for videos that match your search.

[velvet tangle]

No, that is not possible.
But maybe you can suggest it.

[real tinsel]

I tried it. It gave me 6 videos with text, but video 2+3 didn't match the text. Like nr 3 should be tom sayer but it was some weird tiktoklike trash not even mentioning tom Sawyer

[real tinsel]

Yes, I don't like the standard system. Too much input.

[velvet tangle]

Yes this is because no videos can be summarized yet, only a normal search is performed.

[real tinsel]

But why was 4 of 6 100% what it described?

Why had 2 nothing to do with its summary of the video?

All other was correct

Atleast fitting to the yt title

[velvet tangle]

It's like googling normally, you usually find what you're looking for.

[real tinsel]

How can I tell the AI to give me a list like this:

1. Blabla
2. More bla
3. More

He create it as Uncopyable list that I can't grab the numbers.

I can't copy them -_-

Gave him now the sample "1.did" it seems like otherwise can he not avoid the list format

[devout canyon]

Can Perplexity AI access videos on Facebook and Tiktok in addition to YouTube?

[real tinsel]

Its from websites from April.

Thats funny... I didn't got any tiktok or FB video urls.
I would say no.
(FB is really not good anymore and tiktok is a place of trash and chaos)

[devout canyon]

I wish that Perplexity AI can also access videos from Facebook and Tiktok in addition to YouTube videos. How can Perplexity AI solve this?

[quaint mango]

use bing

[real tinsel]

How can I say write markdown readable?
I think I saw it used a slash "\" to make a * visible

I used here →`← signs

[haughty brook]

So you want it to give back unstyled markdown?

[real tinsel]

Yes, I can't copy a fat text it will copy just "fat"

And I need it for my preparing texts for my rpg

Like sound effects **BANG** and he closed loudly the door!

Now I can copy the bang with *

[haughty brook]

The copy button will copy it as unstyled markdown.

this will be copied to your clipboard when perplexity outputs bold text

[real tinsel]

No. Not in opera mobile working. I reported that.
Also I don't want all the text

I don't Roleplay with it, I use it as helper, so I don't want to read bold or italic. I want to see the formatting

Same for lists.

I can't copy.

1. Sun
2. Night
3. Dark

Because the numbers are a list element

[haughty brook]

You could add this to your ai profile

[real tinsel]

Uff can't read it

Can you put it in a txt file?

Or pastebin

https://privnote.com/

Quick and easy

[haughty brook]

https://pastebin.com/raw/FFzhGmvB

[real tinsel]

Thanks, I hope it does read it and that it helps.

[fluid lark]

does “deploying our own in house LLMs” mean that perplexity will eventually make it open source and public for anyone to use? That would be really cool

[oblique prairie]

how many tokens does llama perplexity

[real tinsel]

*can read 😅

Or use

[oblique prairie]

yeah

[real tinsel]

😭 this is absolutely not what I wanted

I showed you 8n the dm how they made it

Oh well... Atleast the copy works!

But its very bad to read from left to right only

[real tinsel]

It asked me the same thing twice.
I wanted that he make me a new rule and he wanted me to do it 😞

What do I wrong?

[real tinsel]

Could I explain it to use line breaks?

[real tinsel]

Does it not understand "un-/dress"? Because of the slash?

"un-/dress"

[wind hare]

what's Perplexity's contextual information size?

[wind hare]

So, on free service we have 5 copilot questions every 4h, and on pro plan we have 300 every day?

[haughty brook]

with pro you’ll have 600 gpt4 uses every day

[wind hare]

tnx, at what hour is it renewed?

but is it GPT-4 use or copilot use?

[haughty brook]

Its both gpt4 and copilot uses combined

[haughty brook]

and the uses work like this : #💬│general message

[wind hare]

I had tested and confirm, GPT-4 usage counts on Copilot counting too.

[honest otter]

Will there be an API for Perplexity? @haughty brook ty

[eternal holly]

#📘│faq message

[haughty brook]

Perplexity doesn’t have plans for an API atm. If this will change they will make sure to share this publicly

[timid robin]

Hi Can I ask if there is anything wrong with my version of perplexity - it have been struck at the "checking if the site connection is secure" most of the time when you guys change the interface

however, when try msft edge it work as soon as I open

[velvet tangle]

With which browser does the problem occur?

[timid robin]

it is normal chrome browser

[velvet tangle]

Are you using any extensions or anything that might block something?

[eternal holly]

maybe vpn issue

[velvet tangle]

Only with chrome but not with edge? 🤔

[eternal holly]

Maybe not sure

[torpid zinc]

Its doing the same thing with iOS Safari for me

[haughty brook]

are you using a vpn on your iphone?

[torpid zinc]

Safari on both iOS and macOS undergo the verification process. Edge seems to be alright
Edit: sorry, didn't answer the question - no VPN, but I have an MDM with a toggleable VPN.

[haughty brook]

sorry for the late reply, could you maybe turn off private relay from apple and see if the problem still occurs?

[torpid zinc]

Np at all. It seems to be fixed. I’m no longer getting the verification screen when accessing the web via Safari on my iPhone.
I did not turn off private relay, for the record, but that’ll be my first go-to if it starts popping up again.

[torpid zinc]

Update:
Throughout the last 2 days, it’s been popping up. I thought maybe I’d wait and see if maybe there’s a pattern, but to no avail.
Before, it would verify twice before granting access. Now, it’s verifying once.

I turned off Private Relay and I am no longer encountering a verification screen.

[paper crag]

how long will my threads get stored on pplx?

[willow junco]

I am subscribed now to pro do I have access to plugins and if so how do I access them? Thanks!

[haughty brook]

Hi @willow junco, Perplexity doesn't have plugins. With the pro subscription you'll have access to 600 gpt4 uses every day. If you have more questions be sure to check the #📘│faq channel or feel free to send another message here if your question isn't listed there.

[granite hornet]

I am attempting to add the discount code to my account for the anniversary but I can't figure out how to do it. Does anyone know?

[velvet tangle]

This only applies to new users.
You then have to press pay as normal and somewhere in the Stripe window there should be a field for a promo code.

[torpid zinc]

Hey 👋, is the web version of Perplexity AI more “accurate” than the iOS app version? I’ve noticed the web platform uses a lot more sources compared to iOS.

[slim night]

Hey @paper crag! Sorry for the late response, if you are logged in, your threads are saved until you delete them.

[manic ether]

Hi. I'm trying the pro version copilot is great. I uploaded a book and started asking questions about it but then it starts searching the web for answers instead of interacting with the book. how should i prompt it so it follows up interrogating the book. thanks

[velvet tangle]

You can set the focus to "writing" (no internet access)

[manic ether]

great thanks

[shell verge]

Hi all,

I am not a tech guy but I would appreciate if someone could answer my question, I have a website and I was wondering if I can embed perplexity chat into that website and make it answer questions only about specific topic or sector as I would greatly benefit from the sources and updated information it provides.

Appreciate your help and guidance in this inquiry.

[quiet estuary]

At the moment you can't do that

[manic ether]

I did set the focus into writing, when I ask the first question about the uploaded text it works.
But the following one it asks me which text I'm talking about.
How can I prompt it so the follow up is only related to the uploaded text?
Tx 🙂

[shell verge]

Thanks, I really hope this will be in their plans.

[slim night]

Hey @manic ether! Did you try skipping that clarifying question? Usually it can still answer the question about the topic, keeping the context of the uploaded file. Please DM me the thread if it doesn't. We'll check

[quiet estuary]

Not the OP, but I have a thread: https://www.perplexity.ai/search/3f33bb0e-9aa6-40bd-9b24-f575967f4365
Copilot asked "Which games do you prefer?" suggesting several games. After I skipped it didn't list those games.

Never mind, bad example, it lists them in text. Sorry.

[runic charm]

hey is there a limit how many questions we can ask per day like openai?

[quiet estuary]

In quick search - no, even with gpt 4. Copilot is limited though.

[pale timber]

Does perplexity plan to release any APIs?

[velvet tangle]

There is no official information on this yet.

[slow pine]

Is there a way to avoid specific sources? For example i would like to exclude newyorktimes from my threads

[velvet tangle]

Hmm... I haven't tried that yet, but you could write it in your AI profile.
Maybe like this:
Do not use sources that start with https://www.nytimes.com/

[slow pine]

I’ll try that, thanks

[haughty brook]

You can use the -site: parameter

[slow pine]

In the prompt?

[haughty brook]

yes

example: What is newyorktimes?? -site:nytimes.com

[slow pine]

Nice, thanks

lol i need to block a whole list of sites. Would be great if it could be done in a global setting

[radiant lark]

question: why does the ai companion extension for chrome need permission to block content on any page?

[velvet tangle]

Please have a look @ #➕│feedback-extension message

[fast cliff]

if I pay for perplixity pro am I automatically getting gpt-4 access? There is a button in my settings that sort of visually indicates I need to turn it on but it is grayed out.

[haughty brook]

You already have access to gpt-4 in the free teir for 5 uses every 4 hours with copilot . But with pro you can also use gpt-4 without copilot and use it 600 times a day.

[fast cliff]

Thank you. I am very confused as to what that means. For context, I paid for a monthly subscription. What is copilot exactly? I thought I needed to use that to use gpt4 model. In order to use copilot I need to click the little button when I start a new thread, right? I assume that turns on GPT4? If thats not the case what does copilot do? Also, If I don't click the copilot button am I still using GPT4?

[haughty brook]

copilot is a helper that guides users through interactive follow-up questions to help the ai better understand and answer questions more accurately. this copilot uses gpt-4.

but when you have the pro subscription you have the option to always use gpt-4 even without copilot. to always use gpt-4 you'll have to toggle this button in the settings page (only available on pro subscription).

[haughty brook]

here is an example of copilot (see image). When your question isn't that detailed it will ask questions so the AI understands the questions better. Without copilot it will just search the results based on what you typed in.

[fast cliff]

Ok thank you. My initial attempt to click the little button there failed, and I assumed it wasn't working. Thank you.

[grim oriole]

Can I upload Figma files or image files in Pro version and ask to write code

[fluid lark]

Hello guys, regarding the pro version, how many messages/tokens do we have per day with GPT-4?
Is it possible to ask 2 or 3 questions simultaneously?

[velvet tangle]

Hi, information on Copilot/GPT-4 can be found in the FAQ: #📘│faq message

[velvet tangle]

Yes, you can ask several questions at once. For example, you can ask a question and then ask another question while the answer is still being generated.

[fluid lark]

Hello, I mean if I only use GPT-4, how many messages can I make per day?
2. Can I make simultaneous questions from 2 different computers using the same account?

[velvet tangle]

The number of copilot uses corresponds to the number of GPT-4 uses.
Yes, simultaneous use also works with several devices.

[empty nexus]

Hello, how can I enable or use writing mode? I don't see it in the FAQ

[fluid lark]

Hi @empty nexus , to use writing mode, choose this option and then select writing mode from the list shown to you. Thanks

[real tinsel]

Quick question (QQ): when is it useful to use the Wolfram alpha mode?
What is that usage, AI never heard from it. I asked weeks ago about it, but forgot it 😓

Pls ping.

[velvet tangle]

https://www.perplexity.ai/search/4d9dfe6f-4687-41a6-817b-6debb60b81ac?s=u

[real tinsel]

I never understood one thing: is it also a AI? I saw their website and it looked completely confusing,like entering some experts zone.

So, it's more a service for science people and experts in high skills things? Not for the usual roleplayer, artist, gamer, like me?

[velvet tangle]

Yes Wolfram Alpha was developed primarily for students.

[real tinsel]

Alright, thanks. So it's for me a casual derp,not relevant

[real tinsel]

https://www.wolframalpha.com/input?input=if+you+are+looking+for+a+term+that+summarizes+individuals+who+are+responsible+for+the+well-being+of+children%2C+"caregiver"+or+"childcare+provider"+would+be+more+appropriate+than+"supervisor."+"caretaker"+is+also+a+possible+term

That is what perplexity gives me as source xD

It looks like it simply just throw in it's own text to wolfram

But the website looks not helpful...

[fluid lark]

I think only for processing natural language

[fluid lark]

yes, the focus - as you've found out - is quite... ineffective

hopefully, they'll implement a better integration similar to the chatgpt plugin

[real tinsel]

Okay

[fluid lark]

Is there any way to add the query to Perplexity URL? Trying to query a search from Alfred App

[mossy iron]

Question, so, Perplexity can display LaTeX and some Markdown elements, but for the latter, I see it only uses headings, subheadings, and bold text. Can it show more Markdown elements besides those?

[dusty otter]

@Prerplexity Team Is there somewhere I can read more about the model? I'm curious how it is citing sources when explaining answers.

[haughty brook]

https://www.perplexity.ai/search?q=PROMPT_HERE&focus=MODE&copilot=TRUE/FALSE

[fluid lark]

Thank you. What does focus=MODE mean?

I assume copilot is a fancy feature that is not necessary for most queries?

[frail plume]

Not necessary for most queries unless you really want some help when prompting. Usually you'll only use copilot if you are not too sure about the topic you are asking about, so it can understand the question and provide a more accurate answer.

[verbal bison]

Does anyone use perplexity as their default search engine (say, instead of Google, DuckDuckGo, and whatnot)? Do you find it practical and fast enough for pedestrian queries that are not in form of a question?

[frail plume]

I do, and yes.

Setting your AI profile to your native language is the most useful thing you can do. This way, you can prompt queries as you would on Google Search, and the AI will search for results in that language and answer with information in your native one.

[frail plume]

Example. Asking in English, search results are in English too, but the answer is in Spanish.

[real tinsel]

I use it, when my google search didn't satisfied me.
But I also use sometimes just the Aria AI inside my opera Browser.

[torpid zinc]

Does the Copilot feature take into account more sources compared to non-Copilot/GPT-4?

I guess what I’m asking is, does Copilot search differently? I see some others that have asked, but I can’t seem to find a solid answer.

Is there something I can put into my AI Profile that will force Copilot to use more sources?

[frail plume]

Yes, up to 25 sources. It searches more in depth to provide a better answer.

[frail plume]

But it can sometimes prompt you more questions. When that happens it will search through more sources and then take all the necessary information to provide a response.

[torpid zinc]

I noticed before one of the big updates that it was searching beyond 25 sources. Was this a change mentioned anywhere? Is answering the follow-up questions/prompts the only way to bypass the 25?

In my case, the more sources, generally the better, as fewer sources can accentuate some unintended bias through a lack of cross-referencing information.

[velvet tangle]

Threads are only viewable with the link for your thread.

[grim matrix]

Would it show up in the discover tab?

[velvet tangle]

No

[haughty brook]

Only auto generated threads made by the perplexity bot are showing up there

[grim matrix]

Ah, got it. So if I asked a question, and someone else has previously asked the same question with copilot, it would not reuse the answer from my thread?

[haughty brook]

Indeed, it wouldn't

[grim matrix]

Thanks

[whole sand]

Hi everyone,
Can someone help me how to upload documents on perplexityt to train the model on personalised data and to set the context and learn from pdfs

[velvet tangle]

You can´t train Perplexity with your data.

[whole sand]

ohh! Train might be wrong term- sorry for that.
I want to unlpoad a pdf and get answers from it.

[velvet tangle]

Click the + under the prompt to upload a file

[whole sand]

yeah- I am getting that options in a start of new thread - but I want to do that same with my current thread.

[velvet tangle]

Not possible atm

[whole sand]

do we have any extension which can do that ?
or aipdf kind of service?

[velvet tangle]

No, there are no extensions, you have to wait until the devs add this function.

[whole sand]

okay! it's disappointing - I had a thought that perplexity ai is more UX friendly than chatGPT(by open ai)

[dusky jay]

Main advantage over ChatGPT is (super)fast real-time search and summarization atm, the devs spent a lot of resources on that. Other things like extensions are being developed but will take time since they're a small team.

[velvet tangle]

You mean because there are extensions from users?
Perplexity is much more secure this way, because there are no vulnerabilities due to external extensions.

[autumn pendant]

Hey Pro users!

I'm planning to subscribe but I want to ask if Perplexity also has cap of 25 msgs / 3 hrs using GPT-4?

[velvet tangle]

600 uses/24h (the count restores dynamically)

[whole sand]

No!! Its for 600msgs I guess.

[autumn pendant]

What if I send more messages then they will go to GPT 3.5?

[dusky jay]

Yeah you're right, but since Perplexity's mission afaik is facilitating easy acquisition and manipulation of knowledge, I think it would be in their best interests to develop an in-house plugin that integrates well with WolframAlpha say, currently a major issue with LLMs is their lack of mathematical accuracy and effective reasoning and if Perp can rectify this by using tools they integrate with, then that's a genuine unique value prop for them.

[velvet tangle]

You can´t send more, you have to switch to GPT-3.5. But you won´t reach 600 msgs!

[velvet tangle]

The devs are constantly working on improvements, soon they will have a solution for this too!

[autumn pendant]

Yeah, makes sense 600 is a lot for one day.

I want to use Pro version for my coding work which will also involve explaining research papers for me. Is Pro version good for this?

[dusky jay]

Yes, absolutely

[velvet tangle]

I can only recommend it. For complicated code tasks, Copilot is great!

[autumn pendant]

Okay!

[autumn pendant]

Awesome and interesting too! I never used Copilot (not referring to GH Copilot) for coding.

I really gotta try it!

Have anyone tried OpenAI's website one GPT 4? Does it do things well as Perplexity?

Sorry for so many questions. I just don't want to spend anywhere and regret.

[dusky jay]

Nope, nowhere close, if you want to get info on the latest code libraries/repos and build something custom based on integrating that with a new paper on aarxiv say, then Perp's the way to go

[autumn pendant]

Just perfect! I'll go for it!

Thanks all for answering me! 😊

[eternal holly]

Btw ChatGPT Plus limit is 50/3 hrs ( which was updated long time back )

[autumn pendant]

Okay. I subscribed to Perplexity Pro!!!

[eternal holly]

Nice ! Welcome I use both so thats why Clarified that wrong info

[autumn pendant]

But I took that screenshot right now. It shows 25/3

[eternal holly]

Its wrong 😦

[autumn pendant]

I’m in India. Could it be because of different regions?

[eternal holly]

I am also in India

[autumn pendant]

I can a video of it.

[dusky jay]

Hmm... that's odd. OpenAI did increase the limit across all ChatGPT plus users...strange.

[eternal holly]

official Source-https://help.openai.com/en/articles/6825453-chatgpt-release-notes

[eternal holly]

I dont need a proof ! Chill

[autumn pendant]

That’s partiality with me. Lol.

Nvm. I use Perplexity since I discovered it & never touched OpenAI Chat since. Perplexity also rocks with free version.

[eternal holly]

Yes just cross checked with my other Free Account they forgot to update the 25 msgs to 50 msgs there lmfao

[autumn pendant]

Hahahha. Okay.

[dusky jay]

Huh, OpenAI web UI team is getting sloppy

[eternal holly]

They're sloppy from the very start

[autumn pendant]

Wait 50 x 12 = 600 msgs. Per day should be higher for perplexity then.

[eternal holly]

.

wait

[dusky jay]

38 = 24, so 508 = 400 msgs on ChatGPT4 vs 600 on Perp

3*8, sorry

[eternal holly]

discord now attaches * directly ?

[dusky jay]

and 50*8

[eternal holly]

i was type typing same but for some reason 50*8

[dusky jay]

meaning?

[eternal holly]

was not getting typed initially

[eternal holly]

ignore probably a glitch

[autumn pendant]

Ok.

What is 3, 8?

[dusky jay]

So, you get 600 GPT4 msgs on Perp per day vs 400 on ChatGPT

[dusky jay]

3 multiplied by 8 yields 24 and 50 messeages per 3 hours implies, 400 messages per day on ChatGPT Plus

[autumn pendant]

Oh, I got confused.

But nice I subscribed to Perplexity then.

🤣 Now I don't know what to ask Perplexity.

Looks nice. It explain the SD model pretty well and in details. GPT 4 writes shorter answer.

GPT 3.5 gives long answer.

GPT 4 writes short answer.

[autumn pendant]

The coding capabilities are insane!

[frail plume]

It shows like that when you are not suscribed. It's a bug on their site.

[real tinsel]

Where does that feedback get sent? Does it really helps the team? @primal acorn

[primal acorn]

Hi! yes it does, it gets sent to the AI team directly. So it's good to report inaccurate responses through this medium 👍🏼

I would do any other feedback through here or support email

[pastel vigil]

Greetings to all members! 👋🏻 I have a question. After much research and attempts to work with various AI tools, I often find myself only halfway there. I'm looking for an AI that can help me create cool marketing videos. I'd like to make a short video (sort of like a company advertisement) – PowerPoint is okay, but it's quite boring. My presentation should be presentable and succinct, designed to spark interest. I have my text ready, but now comes the part where I need to visualize this presentation or advertisement. Animation is what interests me, and I believe it's becoming a necessity. Can someone suggest AI tools that I may not have tried yet? With so much research, I'm losing a lot of time. It would be great if you or someone could provide me with an idea ☺️."

[dusky jay]

AI video is still in its infancy, its gonna take a couple of months to get up to scratch to where you likely want it to be, this might be a good site to help you find what you're looking for.
https://www.futuretools.io/

[pastel vigil]

Thank you very much for the tip. I will take a close look at the websites. Thanks!

[brittle pagoda]

For a Thread, why do I usually see more Sources then embedded Sources referenced?

[velvet tangle]

The AI evaluates all sources, the most suitable sources according to the AI's opinion are displayed in the answer.

[brittle pagoda]

Good to know. Thanks!

[velvet tangle]

@knotty plover
What data exactly are you concerned about?

[knotty plover]

My company data, if I provide my company data Mr Ice Man, then what do you do with that data sir?

[velvet tangle]

Sharing of Information
We do not sell, trade, or otherwise share your personal information with third parties, except as described in this Policy. We may share your personal information with the following types of third parties:

Service providers. We may share personal information with service providers that perform certain services on our behalf. For example, we may use service providers to send email or other communications, process payments, or provide customer support.
Analytics service providers. We may share anonymized usage information with analytics service providers that assist us in the improvement and optimization of our website. For example, Google Analytics: https://policies.google.com/technologies/partner-sites.

[lofty eagle]

I'm Canadian. Would Perplexity Pro give me access to Claude 2? Sorry if this is the wrong channel for this.

[velvet tangle]

Yes

[gentle charm]

I live in Russian Federation, will Perplexity Pro give me access to Chat GPT 4.0 ?

[haughty brook]

Yes, there are no restrictions

[sterile hatch]

Claude 2: When uploading a file is that taken into the context as a whole like when using https://claude.ai/chats ? (At least up to the 100.000 tokens.)

[delicate flume]

not really, it uses Claude 2, but u dont have access to the functionality. I bought PRO and wanted to paste large text like in claude 2, and its not possible you have to convert every text in .pdf and upload it. Thats frustrating... i also thought i can use perplexity for Claude 2, but this way is much to complicated, because everytime make a pdf document is a waste of time

and in addition GPT4 on OpenAI gives a much better answer, than perplexity with GPT4 activated. It seems they have access to API, but its only a reduced version with reduced functionality

[uneven shadow]

so it won't take text as advertised somewhere?

[velvet tangle]

Text files, yes, but you can't paste a 20k tokens text in the text input field.

[uneven shadow]

Ahh, I have not tried that in perplexity though I have done it at Anthropic. it should take about 25k token no?

[velvet tangle]

#💬│general message

[uneven shadow]

or is it paste you are talking about and not size. Yes I noticed that. Anthropic does a nice thing that converts a paste to a text file for you. I use that all the time for chapters from a book instead of a whole book. Very nice to handle that instead of making us convert it ourselves.

[velvet tangle]

Adopting Anthropic's system has been suggested a few times, maybe it will be added in the future, but I can't comment on that.

[dense briar]

Hello, I live in Thailand, I am wondering when will the quota for co-pilot and/or chatGPT-4 will be renewal, thank you.

[primal acorn]

We have answers to these type questions on #📘│faq but I'll happy to answer: It's 600 uses every 24 hours for Copilot, GPT4 and Claude 2 queries. You can check how many you've used by hovering on the Copilot toggle.

[dense briar]

Thank you, I've checked it after 24 hours but the amount is slowly increasing from 498 to 502, that is why I'm wondering.

[primal acorn]

After 24 hours it will get back to 600 but it also slowly "recharges" during that time period

[dense briar]

I see, thank you for your reply.

[night pebble]

Dear Perplexity team, I hope you all have a good day.

This is Jun Koda, the Marketing Officer of fill_media, a Japanese web media dedicated to AI businesses.
*Please check my bio.

We would like to write articles about Perplexity, so that our Japanese readers can learn and know more about it.

Please note that this is not a paid publication offer.
We're offering this as a free publication.

May I know who to contact for further communication?
Thank you for your attention.

[velvet umbra]

Is it possible to upload another file after the first one when you start a new conversation? I was having a conversation about error logs and it asked to provide them again and I had initially given them in a file upload. The upload file button was gone so i attempted to cut and paste the log into the chat and got the error -7331 characters too long. So is the upload file button gone after the conversation starts? Perplexity shouldn't' query me to do things I can't do lol

[wet karma]

You can only upload your file on the first message by now, but there are multiple suggestions of let it be on any message ☺️

[cedar wren]

Hello, why does Perplexity AI get stuck and not answer when I ask a question after not using the computer for a long time and then operating it again? Any solutions for this issue?

[velvet tangle]

Hey, @strong quail, you get 600 Copilot/GPT-4/Claude uses/24h with a Pro subscription!
Please use @pastel ivy only in the commands category.

[alpine kernel]

So how do we add properly the PerplexityAI bot to our own server. I mean properly (any roles, permissions tweaks to do after inviting the bot) ? I invited it by its profile option, but the bot doesn't answers anything, it's silent.

[lavish topaz]

Hey @alpine kernel! You can use the bot in the #🔎│ask channel. Tag @ PerplexityAI and ask your questions.

[alpine kernel]

is there any setup guidelines ?

[velvet tangle]

Can the bot read messages, give reactions, send embed links? Does the bot have access to the corresponding channel?

[alpine kernel]

well yes i did setup a specific bots role wich i enabled all of these perms for. then assigned the role to PeplexityAI. i'm gonna check it's also enabled on its own role as it was all grayed at first. It's a small server i'm alone in, just for testing purposes.

yea can confirm all the perms related to viewing channels, posting messages, embeds, reactions are enabled

[velvet tangle]

You can also just give the bot Admin. Please check again if the bot has all permissions from this list:

• View Channels
• Send Messages
• Send Messages in Threads
• Embed links
• Add reactions
  If the bot still doesn't work you can also send me a server invite link and I can take a closer look.

[alpine kernel]

it works after allowing admins perms. although all the perms you mentionned above were allowed before, so that must be tied to another one wich was not, probably but can't say wich one :o

[alpine kernel]

=> that was probably because i needed to add the "use application commands" permission. i removed admin, added that, and it's still working

[timber basin]

Is it possible to use all three ai-models as pro in paralell to check what is the best result to fit my needs?

[velvet tangle]

Asking the same question in parallel with different models is not possible, you can only do it one after the other.

[timber basin]

and there is also no better option then jump into the settings every time to switch the model? --> could be a dropdown in the text-box-window ?

[velvet tangle]

No, but the devs are working on a faster model-switch.

[timber basin]

okay, nice to hear 🙂

[fair dagger]

Hello everyone,
Am I doing something wrong or is Perplexity not really good at remembering context within threads?
Latest example: I asked Perplexity to write a short paragraph, which it did. Then I followed up asking to translate it in another language, and Copilot asked me: "which text would you like to translate?" Ahem...

[haughty brook]

I think that might just be an issue with copilot. It sometimes ask unnecessary questions from time to time. But it should just work fine when you skip that question.

[fair dagger]

Ok thank you! I’ll give it a try!

[formal karma]

I would love to know a way to do this too. My attempts so far have just been a major PIA. Even if operating three threads in separate browser windows, it seems that changing the model in one changes the model in all of them so I can't even copypasta. Major current drawback for me atm.

[formal karma]

This has been my experience too, that hitting skip has been working just fine. It would be great if co-pilot would get better at realizing when to not ask those kinds of follow-up questions

[bright dove]

Hello everyone,
I was wondering how well I can use Perplexity AI for writing texts. I usually use the OpenAI site to write introductory texts and other short paragraphs that don't really require outside sources. Meanwhile I use Perplexity when I need to write paragraphs about something that requires more background knowledge. I was wondering if there is a way to get all my texts in one place, since that would save on some searching between the two.

So basically, is there a way to get more general texts without sources from perplexity AI instead of the answer-like structure it normally has?

[fluid lark]

Hi @bright dove , you can switch to "writing mode" for this purpose.

[bright dove]

Ahh that's great! Completely missed that somehow! Thank you!

[verbal bison]

are there plans to increase the allowed input size? 2k is a little bit low if I want to use Perplexity more like ChatGPT (for example: proofread this text). I may even stop paying OpenAI and start paying Perplexity if the input text got increased. Even increasing it for Pro users only would work.

[hearty quarry]

use file upload?

[verbal bison]

That is time consuming and it is not the same that using the whole context. I use file upload if I have a huge file to query

[hearty quarry]

iirc they limit the context for UX and speed

[abstract wing]

I have the same issue with copilot constantly.

[slim night]

Hi @abstract wing, @fair dagger, could you share a couple of examples with permalinks in the #1140622008086970420 channel and also mention if skipping those questions helped.

[abstract wing]

@slim night Done for a couple of them that are in my thread history: https://discord.com/channels/1047197230748151888/1148660688286400563

[real tinsel]

@oak yew Quick Question: Did you/your team used Perplexity to "code" the Survey? :3

I would like the idea, when you use your own AI for making surveys 😄

Okay no... you grabbed it from a template, sad sad 😉

[fair dagger]

Hi, thanks for your follow up! Unfortunately I deleted these threads, but skipping the question helped indeed!

[pseudo oar]

Hi, i have a question. in my thread the Question box shows this error: "too long -213". How can an empty box be too long?

[slim night]

Hey, @pseudo oar! Thanks for reporting this, could you also add the URL to this thread. You got this error in a follow-up question, are you still facing it? Which browser are you using? Thanks again, would be good if you could share those details in the #1140622008086970420 channel.

[eternal holly]

Hi Alex can you have a look here ? https://discord.com/channels/1047197230748151888/1149045584876028005

[left nest]

Hi! The annual plan gives a 7 day trial which I availed today with the SCHOOL25 promo, so I won't be charged for it until 7 days from now. The school25 promo ends tomorrow, so will it avail with the discount in 7 days? Thank you!

[late bolt]

Last week the references in the output was footnoted. Today, I still see the Quick Search links but the footnotes don't appear in the answer text. Why is that?

[slim night]

Hi @left nest! It should be applied when the trial ends. Email us at [e-mail], if you get any issues with that in 7 days.

[slim night]

Hi @late bolt! Which version are you referring to (the web or apps)? Please share a permalink to the thread here or in the #1140622008086970420 channel, so we could check.

[rigid quiver]

is it possible to change models in the Android app? seems like i can't

[formal karma]

What's the point of being able to favorite threads? Does it protect them from deletion or something? I wish I could just have the favorites appear at the top or be in a separate folder

[velvet tangle]

Other users can like your thread if you share it

[primal acorn]

it will serve a higher purpose down the line 😉

[cedar wren]

hi, why always block here?

block forever until I reenter the perplexity.ai

[velvet tangle]

Hey, @cedar wren!
You use the web version? On the computer or on the mobile phone?

[velvet tangle]

Do you still have this problem?

• Are you using any extensions or content blockers? If so, please deactivate them.
• In addition, you could clear the website data, cache and cookies and log in again.
• Are you having the problem with one or more browsers?
  If that doesn't help, please create a bug report in #1140622008086970420 with all the information as described in the pinned thread.

[cedar wren]

yes, web version, on the computer my os sometime is windows while another times is mac os

I only use the chrome browser

[velvet tangle]

Have you tried everything? #⚡│ask-community message

[cedar wren]

yes , When I don't interact for a long time, such as being away from the computer for a long period, and then come back and ask a question, I encounter a situation where it keeps getting stuck here, and it happens every time.

both on my windows pc or on my mac os m1 mac mini

In such cases, I have to reopen the PerplexAI website address, and only after passing the Cloudflare verification can I continue asking questions.

[fluid lark]

Why does perplexity voice recognition not use whisper?

[velvet tangle]

I think that only works with English and is also more expensive.

[unkempt pecan]

Loving Perplexity! As a web marketer, what would be the best way to create content that Perplexity would pick up?

[silk kite]

Are there any plans to save the history of edited individual queries in threads like you can in ChatGPT's interface?

[silent ridge]

Any way I can get the output in Mark Down?

[silent ridge]

Is there a way to have claude output a table?

[stable crow]

Is there a plan for API?

[haughty brook]

There are currently no plans for an API if this changes they will make sure to let the users know

[thorn gull]

Can anyone help on how to get perplexity pro to help with my writing. I have 8000 words and even after my prompt stating I want similar word count after editing for flow and headings - it keeps summerising to around 500!

[fair dagger]

I've also tried Whisper in French and Italian and it works perfectly.
In my experience, Perplexity's voice recognition only understands English.
But I get that using Whisper would probably cost more.

[muted saffron]

can I switch the engine to get an alternative take within a thread?

[urban lichen]

how to prevent internet search in follow up question? (or change focus)

[velvet tangle]

Not possible atm, but requested many times.

[urban lichen]

try very long questions

[velvet tangle]

What do you mean?

[urban lichen]

when i have very long instructed followup question, i often don't get search results

[velvet tangle]

Ah okay, but that's not intentional then, it's a bug!

[urban lichen]

is following instructions instead of search is also a bug?

it sometimes follow, sometimes not, and sometimes partially, and in another times it search entirely irrelevant data

i definitely need to combine writing and internet focus in one thread

[knotty hedge]

Hi all, I'm using Perplexity to help speed up research for the content I write for medical device and clinical tech clients. Every statement of fact and stat has to be cited and referenced, and those references have to be credible: scientific/medical journals, primarily, but industry publications are sometimes acceptable depending on the content. Here's my question -- which model is best for this kind of research? Claude or GPT4? Copilot or not Copilot? I've tried out both (on Perplexity and directly) and can't tell much difference. And I'm not sure what the benefit is of Copilot. I have caught errors in the answers, which is frustrating. Thank you! cc @primal acorn

[cedar wren]

Hi all, I am a pro subscriber of perplexity.ai, and I really love perplexity.ai. However, I would like to know if perplexity.ai can have a chat list on the left side, similar to chatting with different users, where I can select to chat with chatgpt4, 3.5, or claude2. Currently, it seems that there is no such feature, which is inconvenient. If I need to switch models, I have to go to the personal settings to switch, and I cannot use different models in different chat windows without switching every time.

just like this, will be more convenient

[velvet tangle]

Hey, @knotty hedge, errors can always occur, but you can always report them with the flag under the answer!
In your case, I would use GPT-4 in conjunction with Academic focus.
Copilot will ask questions back if your prompt is inaccurate to give you the best possible result.

[torpid zinc]

Does Perplexity Copilot always take the AI profile into account?

[velvet tangle]

Yes!

[solid parcel]

Is there a way to use the AI models in parallel?
Since they give different answers, and none are entirely reliable, it can be very beneficial to get the answer to a question from multiple models.
Currently I paste important questions manually into OpenAI-GPT, Bing-GPT, Perplexity-Claude2, and sometimes Bard, separately, but with your multiple models you could provide this out of the box and conveniently. I think that would be awesome.

[muted saffron]

@solid parcel yesterday I asked if it's possible to easily switch models within a thread and @velvet tangle said not yet but asked for many times; asking in parallel is another level of ask.:)

[velvet tangle]

You can contact [e-mail] for this problem!

[raven sundial]

Hello,

I would like to try out the Pro version of Perplexity (the best AI for my current needs). I want to make sure that I can use the 7-day free trial and have the option to cancel my subscription to avoid being charged $200.

Thank you.

[velvet tangle]

Hey, @raven sundial!
Within the seven days you can cancel your plan simply via Manage Subscription!

[raven sundial]

Yes, but I want to be sure that if we leave the subscription plan before the 7 days we are not billed on the 8th day to pay for a year.

[velvet tangle]

As I said, you can cancel the subscription within the seven days, then you don't have to pay the $200.

[raven sundial]

Thank you, I've already signed up for a "free trial" service in the past and even if we cancelled before the end of the "fee trial" we were still committed.

[raven sundial]

If i write my question in english Perplexity search ressource in english and if i write my question in french Perplexity search ressource in French, right?

[slim night]

Hi @raven sundial! There are several ways how people work with their language set as a default one.

1. You can set French in your AI Profile and get all of the answers in French even if they were asked in English.
2. You can prompt in English and add the output language of your choice, like: "your query, answer in French", or English, or other language you want.
3. You can try asking questions in different languages and receive answers in the language of your question, but if they switch to the default one, you can use ways mentioned in 1 or 2.

[raven sundial]

Awsome answer thank you very much !

[cedar wren]

My perplexity can't answer any questions! I just ask 5 question, but I am pro subscriber, why?
only today

why 0 left today?

I really like perplexity.ai, but this issue is affecting my usage. Could you please help me solve this bug? Thank you very much. I am in a hurry to write my paper.

[haughty brook]

hi @cedar wren could you reload the page and check if you still have this issue?

[cedar wren]

I just logout and re login, but seems the same result, while I can use chatgpt4 as usual...., so I think this might be a statistic bug??

[slim night]

Hey @cedar wren! Are you still facing the issue?

[haughty brook]

#💬│general message it has been resolved in #💬│general

[cedar wren]

yes ,I still facing the issue, today when I first login perplexity.ai, It shows to me "5 left"......

this is a huge different with normal show of 500+

@slim night

this is today's first login, no question ask yet...

as a pro subscirber, I can't make sure I have the pro's right to use about 500+ copilot, please help me!

perplexity.ai is the best GPT tool I have ever used, there is nothing better than it. So please help me solve the problem and let me use it more comfortably

@haughty brook

[velvet tangle]

It's best to contact support (in the settings https://www.perplexity.ai/settings) so that the devs can take a close look at what the problem is with your account.

[cedar wren]

already contacted

[velvet tangle]

Great, someone should be in touch with you in the next few days.

[cedar wren]

ok, just wait the fixing reply

[cedar wren]

no reply. My bugs existing for 4 days

this 4 days , I can't use gpt4 or claude2

as a pro user, Who can help me?

I swich to gpt4 or claude2, but when I re enter settings, It is always Perplexity

zero count!!!

@slim night

@haughty brook

[fluid lark]

hm, have you tried another browser

[cedar wren]

I use chrome, safari, all can't work

[fluid lark]

odd...

maybe it has to do with the location ? either that or something wrong with your account specifically

can't think of what else could be the problem

[cedar wren]

how to do location?

last month, I can use normally

the bug occured at this several days

[velvet tangle]

Hey, I'm sorry for the inconvenience, but please don't keep pinging people with your questions, messages are answered pretty promptly here.
Secondly, it's weekend at the moment, which means support is off too!
Like I said, just be patient and wait until you get a response.
Thanks in advance!

[fluid lark]

I meant maybe it's an issue only in your location (as in physical location) but might not be it since it worked before

[cedar wren]

I am so sorry, will ask your question on Monday. Have a great weekend!

[sacred halo]

I was trying to give some contect to the search thread and then asked to pull up an engaing linkedin post.
Perplexity did a great job but also pulled up info from my AI profile (like my compnay name and position) that I didn't want in that particular post.

can we have a toggle button like for copilot, where we can select if we want some generic output or we want perplexity to reply based on our AI profile.

[wet karma]

By now you can pause the profile only I believe

[spice condor]

@cedar wren it looks like you canceled your subscription, I have DM'd you for more details.

[cedar wren]

@spice condorbut the subscription will cancel at october 15?

now is 25 September

I want to cancel monthly and switch to yearly

[spice condor]

hi @cedar wren I will help you in the DM. It is a bit harder now since you unsubscribed. For anyone who is wondering how to upgrade to yearly there is an "Update plan" button here from "Manage subscription".

[cedar wren]

right now I logined perplexity.ai, It seems Ok now

@spice condorUpdate plan not work for me, so I try to cancel and till 15 october, Re subscribe yearly

I had an issue with credit card binding ago, but it had been resolved. I were planning to upgrade to the annual plan but have hesitated due to the recent issue with incorrect counts. However, after logging in today, I noticed that the GPT-4 count seems to be back to normal. I would like to observe for a few more days to ensure that everything is working properly, and then I will be ready to upgrade to the annual plan at any time.

[sacred halo]

This is intestesting. But what I am looking for is an thread level pause and resume.

[wet karma]

It doesn't exists by now 😕

[tacit cradle]

Help please.
Perplexity is always checking for captcha, this is really annoying.
Even static IP does not help with permanently prooving that i am human.
please whitelist my account if it is marked somehow.

[velvet tangle]

Known bug:

1. go to https://labs.perplexity.ai
2. solve the captcha
3. go back to https://perplexity.ai/

[tacit cradle]

no captha at labs.perplexity.ai

[velvet tangle]

What browser do you use, do you use any content blocker, VPN?

[tacit cradle]

yes using vpn with static ip. Safari

[velvet tangle]

Please write a DM to Alex Romanov with your IP.

[cedar wren]

Every time I leave my computer for more than half an hour and come back to ask a question, the perplexity just keeps spinning in circles.

To resume normal usage, I have to re-enter perplexity and go through human-machine authentication.

[haughty brook]

Thats because cloudflare needs to verify your browser again. A simple refresh should fix this

[fluid lark]

or ^

if refresh doesn't fix it

[cedar wren]

refresh doesn't fix it， only if I re enter perplexity.ai and do the cloudflare verification

this is very inconvenient and greatly affects the user experience. I hope this problem can be resolved as soon as possible.

Another solution is whether we can launch an Electron desktop client, which would eliminate the need for human-machine verification every time, making it more convenient to use.

[fluid lark]

I think that has been requested before

and I agree

[tacit cradle]

after some time and another captcha request i used this link.
will inform if it helped.
thanks for showing the lama Model site though 😉

[tacit cradle]

bug stays...
will contact Alex

[fluid lark]

how about adding this to the FAQ

[velvet tangle]

I was thinking of a kind of guide forum where we community volunteers can post things like this.
I think the FAQ would otherwise be too big and no one would read it... But anyway, I agree with you: it should be written down somewhere

[viscid snow]

Hi! I recently discovered Perplexity and I think it is fantastic. Just a question: does Perplexity offer free plans or discounts for students?

[haughty brook]

Hi @viscid snow, there are currently no discounts for students. However, there are sometimes discount codes made available by Perplexity. Last month there was a discount that gave 50$ off the annual plan.

Also you could use a discount code from another Pro member to get 10$ off for one month.

[fierce moss]

Does Perplexity offer a buisness plan for API usage? If so, who can I speak with to discuss details? Thanks!

[lavish topaz]

Hey @fierce moss! Perplexity AI does not currently provide an API.

[velvet tangle]

As the bot said, not possible at the moment 🙂

[fierce moss]

Thanks, I understand.
My team has a number of exploratory efforts across a number of solutions all around Generative AI and Perplexity is performing very well and getting high marks on our human eval. Keep at it, your team is doing great work

[tacit cradle]

is there a difference between the perplexity model in free and pro plan ?

[hasty lily]

I believe Free is GPT 3.5 and Pro is GPT 4/ Claude 2

[dire tangle]

The perplexity model is the same on free and on pro but pro gives you the option to permanently switch to GPT-4 or Claude 2. On Copilot, Perplexity+GPT-4 is always used in conjunction (pro and free)

[tacit cradle]

i tried to formulate my question to only ask for the perplexity model 🙂
so when i use GPT4 in Pro it still uses the perplexity model in conjuction somehow ?

Do i understand correctly, that perplexity is sending out multiple agents at a time to srape the web ?
does it do the same with gpt4 model in use ?

[dire tangle]

In copilot, Perplexity model is used to gather the sources and ask clarifying questions. After that, GPT-4 (or Claude 2) is used to write the final answer from the sources and your answers.

[fluid lark]

when reporting an answer, when am I supposed to use "this isn't helpful" ?

[velvet tangle]

When the AI babbles around the actual topic but doesn't give you an answer.

[fluid lark]

what about when the answer is wrong ?

inaccurate ?

[velvet tangle]

You just take the reason that fits best

[haughty brook]

I wouldn't worry too much about that. As long as you include what the answer should be in the input. Pick one that you think fits best.

[fluid lark]

I see, thx

[zenith fulcrum]

Can I switch the email address I use?

[velvet tangle]

Sorry, not possible atm.

[elder dove]

fixed for me

[tacit cradle]

somehow GPT4 model tells me its unable to read pdf files via file upload... but isnt the file upload made for it ?

[velvet tangle]

Hey, yes GPT-4 can actually read files.
Please share a thread where you have such a problem.

[cedar wren]

I just try the labs for Mistral 7B Instruct, but the cloudfire verity can not pass

then I use perplexity.ai , unfortunately the cloudfire verify also failed, always let me try.....

[tacit cradle]

https://www.perplexity.ai/search/Translate-the-uploaded-K7BNrDKiQF6ZnMgPO98b4A?s=c

maybe the pdf has too many characters, but then it should tell me that this is the reason why and not because it can not access files 😉

[velvet tangle]

Can you please try it again, but in writing focus, without Copilot and with Claude?

[tacit cradle]

sure, but it should work with gpt4.
i usually include pdf with tech docs for refference and this works.

[tacit cradle]

yes there it worked, but still did something i did not asked for.
it summarized the translation but i clearly asked for a translation of the whole uploaded pdf

[velvet tangle]

The file was probably too large.
I usually recommend Claude 2 for the file upload, otherwise the context quickly becomes too large 😉

[tacit cradle]

Ya, sad it doesnt respond with explainig the real reason why it does not compute though.

[crimson sandal]

does the API serve search queries?

[velvet tangle]

The API doesn't search the web atm.

[rich marsh]

https://www.perplexity.ai/search/42cc0a16-35c2-460a-af02-ea81b66e9017?s=u
It replied with nothing

on the second question

[velvet tangle]

Hey, this is a know bug. A fix is being worked on 🙌

https://discord.com/channels/1047197230748151888/1156978464625283153

[viscid vale]

Sometimes it doesn't give me answer i tried claude 2 and gpt-4. Same case. No matter which question i ask. Sometimes he doesn't respond me first time.

[haughty brook]

known bug please add your perplexity thread to this thread #1156978464625283153 message

[median sentinel]

What's the api request limit rn?

[rich marsh]

nothing ig lol

[lofty sparrow]

i'm currently getting rate limited using the perplexity api. where is the right channel to ask questions about it?

[velvet tangle]

there is no channel yet, so you are right here

[zealous obsidian]

Hi, I found conflicting answers and I wanted to ask you if you could help me, what is the actual maximum cost of perplexity? (200/y)?

[velvet tangle]

Hey, yearly paid it’s $200 Perplexity Pro

[sharp laurel]

this is literally the most silly and time wasting extension and web. i ask it about a document, it prompts me to indicate if its a quotation or a contract, lol

[median lagoon]

With Perplexity Pro, how many API calls can I make a day?

[strong edge]

I've been trying out the new Perplexity API with the Continue extension for vscode with mixed results so far. If you want, I can share you my configuration, and I can also open a PR on Continue's repo to add Perplexity to the available configurations.

I won't share anything unless I have explicit permission from the Perplexity devs, I understand that these queries may be a bit spam-ish on your API.

Also, if any of you guys know a better extension for this, please share!

[viscid vale]

I try to use Perplexity API, is there a way for AI to see all the content before? Because every new message to him it's like a clear list. (he doesn't know what were talked before)

[velvet tangle]

So basically you can add new objects to the Chat completion:

const sdk = require('api')('@pplx/v0#fbzexuulncfz28t');

sdk.auth('');
sdk.post_chat_completions({
  model: 'mistral-7b-instruct',
  messages: [
    {role: 'system', content: 'Be precise and concise.'},
    {role: 'user', content: 'How many stars are there in our galaxy?'},
    {
      role: 'assistant',
      content: 'The number of stars in our galaxy is estimated to be around 100-400 billion. However, this is an estimate and the true number may be different.'
    },
    {role: 'user', content: 'And what about stars in other galaxies?'}
  ]
})
  .then(({ data }) => console.log(data))
  .catch(err => console.error(err)); ```

[boreal light]

Hi. Can I use Copilot with the API?

[velvet tangle]

No, Pro users can only use the listed models at https://docs.perplexity.ai/reference/post_chat_completions

[ripe patio]

Can I change the default language both for the interface and the AI? I ask a question in French or Spanish and Perplexity answers in English.

[velvet tangle]

You can’t change the UI language

[ripe patio]

ChatGPT remembers the conversation up to a certain point, then it forgets. Is it the same in Perplexity ? How long before the AI forgets in a thread?

[velvet tangle]

With GPT-4 it’s up to 32k context.

[ripe patio]

The term I'm looking for is "length of context"? Thank you

[velvet tangle]

32k tokens is a length specification. One token generally corresponds to ~4 characters. So 32k ≈ 24.000 words.

[ripe patio]

Each model has a specific length of context, right? I think Claude is the largest, but if I start a thread with one model and then I switch to another one, what then? Thank you for your help in helping me understand this tech.

[velvet tangle]

Yeah, Claude’s context is 100k.
Each user message and each assistant message is an object.
So if you start a chat with GPT-4 and you have already sent two messages there are 5 objects in total, two user messages, two assistant messages and one system message.
If you then switch to Claude, the objects are simply "transferred" to Claude.

[ripe patio]

You're the superhero of information 😊, thank you for the clear explanations.
I can use GPT, then switch to Claude and it will be able to summarize three times what GPT would have been able to do (100k ≈ 3*32k)?

[velvet tangle]

Theoretically yes, but it should be noted that above a certain context length the AI may start hallucinating.

[balmy obsidian]

Hi! I have a quick, low priority question. If I upload a file and then ask a question, can I assume Perplexity will reference the file when answering?

[velvet tangle]

Yes, Perplexity will use this source!

[balmy obsidian]

Sweet - so I don't need to tell it anything extra and can launch right into my questions? 🙂

[ripe patio]

Do you know in which case scenarios it is more advantageous to use the Complexity model instead of GPT or Claude?

[velvet tangle]

Yeah, sure!
You mean the Perplexity model, right?
This is a fine tuned GPT-3.5 model and is much faster than the others.
So if you want to do a simple normal search you can use Perplexity 😉

[ripe patio]

yes I mean the perplexity model. Is it speed its advantage?

[velvet tangle]

Yes and GPT-4's advantage is, that it's smarter.

[ripe patio]

Thank you! 👍🏼

[midnight cloak]

Hi there, if i pay for the pro version, can i upload images and ask gpt 4 questions about them?

[velvet tangle]

#💬│general message

[opaque fjord]

Is there someone from support I can talk to?

[velvet tangle]

What is it about?

[opaque fjord]

Well, my son created an account, made a subscription without my permission, and I would like to cancel and see the possibility of a refund, because the amount converted here for Brazil is very high, can you help me?

[velvet tangle]

In this case, it is best to contact [e-mail] 🙂

[midnight cloak]

does Copilot use gpt 4 and google? I mean in the pro version

[icy ether]

Any updates for the Replit model? I was excited to try it out but there seemed to be complications. Still don't see it as API option or in labs

[craggy tapir]

yes

[haughty brook]

The replit model can be found here https://docs.perplexity.ai/reference/post_text_completions

this model is for non-conversational prompts

[pastel heart]

Is there a projected time by which Perplexity API will include web search results references (or does it already)? And will GPT4 be offered through the API? (basically asking if there will be an API for Perplexity.ai's current consumer-facing chat/search engine combination)

[icy ether]

The API does not currently include search results from my understanding. I'm not sure if they would include GPT4 given that OpenAI offers an API for their own models like GPT 4 &3.5

[balmy obsidian]

I paid for Pro but am still in my trial period. It says I have 584 uses left today. Is this recharged each day? I can't tell (since it isn't visible in mobile) how many queries I've used so far.

[haughty brook]

Yes they get recharged each day. They work like this #👉│feedback-general message

[balmy obsidian]

Super useful information - thank you!

[scarlet ether]

I use Mullvad VPN to protect my data as I am always needing to use public WiFi. Cloudflare always seems to be an issue when attempting to use the app. Is there a fix or workaround? It happens only on the app and not through the website. I have it set to Canadaian server at most times. If the app allows me to check if I'm human, it won't show a response from the AI regardless.

[pastel heart]

ChatGTP with web search is very slow last time I tested it. Has it improved or with the API, is it faster? Perplexity.ai's web search is quick - so I assume their API would be too. To be fair, I haven't actually tested OpenAI API with RAG e.g. using google search. But I assume Perplexity has worked out the kinks and would be a lot faster at what it does than a program that combined OpenAI API with e.g. the Google Search module. Is that true? If so, it would be a big contribution on Perplexity's part to offer an API to its own service. What's the fastest way to get to an API that's closer to the Perplexity.ai web experience (which is really good!), other than waiting for Perplexity to offer it (which would be great)

[restive charm]

Hey everyone, pretty new to Perplexity and I'm currently using the free trial. I'm just curious as I couldn't find solid information on this - suppose I turn off copilot and set the AI to use GPT-4 exclusively and I turn on 'writing' mode. Does this mean that the responses I would get are identical to GPT-4 from OpenAI's website or is there still a difference in some way?

[velvet tangle]

You will not get exactly the same result, this is because the system prompt and other settings are different.
But theoretically you use the same GPT-4.

[restive charm]

Thank you!

[grand sedge]

Hi All. I am using Perplexity Pro and want to write in a few different styles. Is it possible to use the AI Profile to set up more than one profile? I would like to switch between 2 or 3 profiles, each having different styles of writing.

[wet karma]

No, but you can trick your profile with something like if I prefix my question with STYLE1 you will answer the question on this way but If I prefix my question with STYLE2 you will answer in this other way or something like that

[grand sedge]

Thanks Iñaki. That's great. I'll try that out.

[static pewter]

Hi! I've been using Claude with Perplexity Pro to develop a taxonomy for my personal notes. This worked well, but when I asked it to regenerate the taxonomy just now, it hallucinated. The thread is quite long, but I'm not sure how to figure out when it is too long for the context window. I was wondering:

• Can I know, in Perplexity, when the model will forget and start to hallucinate?
• If I can't, is there some external calculator I can use?
• What is the best practice here? Should I re-train the model in the thread by prompting it with my taxonomy or should I abandon the thread?

[random dew]

Can I use Perplexity for commercial purposes such as copywriting or article generation?

Like the way in OpenAI, you can with GPT4

However, I am unsure about GPT4 in Perplexity

[velvet tangle]

Hey, @static pewter!
There is no indication when the thread is too long, but the feature has been requested.

[static pewter]

Thanks! Anything I can do outside of Perplexity? Should I re-train the model with a prompt in the same thread?

[velvet tangle]

Please have a look at https://blog.perplexity.ai/legal/terms-of-service @random dew 🙂

[velvet tangle]

Hmm..It would probably be best if you start a new thread and put the essential data from the last thread into the input.

[static pewter]

Ugh! Thank you 🙂

[real tinsel]

Here a quick question maybe @velvet tangle ❤️ How actual is the Websearch? I want to know when a music band did anounced the first time their next tour, so I need that the AI can search the last weeks/months.

pls ping ❤️

[velvet tangle]

Hey @real tinsel!
The web search is very up to date, as you can see in this example the news are not that old 🙂
https://www.perplexity.ai/search/Latest-news-worldwide-gW7qjsJyQZ.0SpIM5NyGSg?s=c

[real tinsel]

The sources does all lead to a /world page and I can never find the words like "Taiwan" "bangkok" "education"

Hmmmm nearby correct, tomorrow does the presale start and that info is missing 😮

Ahhhh clicking on the "related" did help!

But even there is the LIFAD Presale not listed as 16.10.2023

[velvet tangle]

Hmmm I also searched the internet and also found Perplexity's answer.

[real tinsel]

the first line says the LIFAD Members can start tomorrow.

used this.

btw. LIFAD = Liebe ist für alle da. (love is for all ^_^)

[velvet tangle]

Ahh I see, my bad. Probably Perplexity made the same mistake 😉
Another try: https://www.perplexity.ai/search/LIFAD-Presale-2023-Ka4qk2mOQSG1IizT.KVUAQ?s=c

[real tinsel]

which focus should I use when I want to ask for coding help? I would like to get a command for grabbing urls from a website, maybe perplexity can write me a code to read the urls from the dev console^^

[real tinsel]

we had success and could order today 2 tickets :3 and now are all sold out 🙂

[velvet tangle]

Writing should work 😉

[real tinsel]

thanks.

[real tinsel]

Awesome... the AI made the command correct on FIRST try... could grab all my followers from my art website and generate from en/user/12345 a working url! this is cool...

@velvet tangle Look at this!! it's so ... incredible awesome, I mean... it's amazing, like you just hired a coder and he helps you! (or she^^)

[hardy coral]

hello! am i able to whitelist my IP somewhere? I am having trouble with using the perplexity app with a VPN, and saw mention of being able to whitelist in the settings, but i am not able to locate that. anyone have this same experience?

[velvet tangle]

Hey, which VPN are you using?

[hardy coral]

Mullvad through WireGuard

[velvet tangle]

You can whitelist pplx with split tunnel: https://www.reddit.com/r/WireGuard/comments/twfiid/is_it_possible_to_split_tunnel_in_wireguard/

[hardy coral]

appreciate it!

[static pewter]

Hi, any iOS Shortcuts users here? I want to pass a URL, or the contents of a webpage directly to Perplexity for a summary and am having a heck of a time. I've tried quite a few permutations

[velvet tangle]

@arctic hawk
For Pro users, a choice of several models is available for API use: https://blog.perplexity.ai/blog/introducing-pplx-api
However, the "classic" Perplexity with web search cannot be used.

[velvet tangle]

@grave light
Yes you can name the file names in your prompt, as Perplexity can also see the names.

[umbral sierra]

Is the site like super slow for everyone or is it just me?

[grave light]

thank you!

[static pewter]

Hi folks! Any suggestions on how to structure my prompt for this?
https://www.perplexity.ai/search/I-want-information-nM321.0MQC2CsvEzv98p_Q?s=c

[static pewter]

Hi all, are there any methods with perplexity to use prior threads in a new thread to re-train before prompting?

I was considering copy-pasting all prompts and responses into a markdown file, but that's pretty manual.

[icy ether]

You can ask perplexity to summarize the thread to the point. If you have Claud, that may be better due to context

[static pewter]

I'm actually asking because that was very ineffective, even with a detailed prompt. I didn't get enough information and didn't get that info in order. I did use Claude.

[timid oar]

Are any of the models accessible through the API finetuned to improve search capabilities? If so, which ones? And how much better are they than non-finetuned models?

[near dawn]

Hi a question: what is copilot for?

[hasty lily]

Copied from faq

Copilot is a finetuned GPT-3.5/GPT-4 powered interactive search assistant that personalizes results by asking clarifying questions, while Quick Search is the default mode providing fast, concise results without user interaction. Copilot focuses on user intent for more accurate results. Perplexity Copilot is not related to Github Copilot.

[night mortar]

How to trigger copilot feature in Perplexity AI by using URL? Below link will only invoke the vanilla version without copilot.
https://www.perplexity.ai/search?q=Human+Growth+Hormone

[haughty brook]

you can add copilot=true to the url
https://www.perplexity.ai/search?q=Human+Growth+Hormone&copilot=true

[night mortar]

Thanks a lot!

[static pewter]

@night mortar, what is your use case for this? Are you passing searches into perplexity with that url structure? Just curious as I want to do the same!

[night mortar]

600 times of Copilot usage a day is a lot. A simple click in ChatGPT, which triggers Perplexity, is much better than sifting through a lot of information on Google.
I have been using GPT-4 on ChatGPT for anything before 2021. I modified the pre-prompt from ChatGPT's AutoExpert by Dustin Miller to better fit my needs. Although I am aware that GPT-4 gives optimal performance in English, I translated the pre-prompt into Traditional Chinese to accommodate more instructions. This minor trade-off in performance is justified by the added instruction capacity.
https://github.com/spdustin/ChatGPT-AutoExpert/tree/main/standard-edition

I tried the same prompts on ChatGPT and Perplexity so you could have a general idea of how AutoExpert works.
https://www.perplexity.ai/search/Does-the-query-GimAb4CrQMGT7CxONrEQ9w?s=c

[sleek badge]

How can we use Autoexpert on Perplexity?

[velvet tangle]

It's too long for the AI profile, but you could shorten it.

[balmy obsidian]

How can I check my remaining copilot uses in the app?

[velvet tangle]

Not possible atm, you can only use the web version for checking 🙂

[static pewter]

Is there any documentation about Perplexity prompt best practices? I don't want to reinvent the wheel. I do see the API docs, but nothing for the front end.

[restive charm]

I use Perplexity to help me understand certain concepts while studying, so factual accuracy and ability to respond well to follow up questions are the most important aspects for me. I'd like to ask, for my use case scenario, which AI model would be best suited? Additionally, should I use the Copilot feature or leave it turned off?

[velvet tangle]

Have you subscribed to Pro? @quaint prawn

[quaint prawn]

yes

[quaint prawn]

already paid. i can run in little time. then work out.

[velvet tangle]

Have you already taken a look at the rate limits?
https://docs.perplexity.ai/docs/rate-limits

[quaint prawn]

but i already 5 days ago. still no work.

[velvet tangle]

Please describe your problem in #1161804761247526912 in more detail and also include your code and the output!

[fluid lark]

Is it correct that I cannot setup a email/password login? If not, how can I change my email adress?

[velvet tangle]

Yes, registration is only possible via the three options already mentioned. When registering by email, you will be sent a link to log in.
If you want to change your email address you should write an email to [e-mail] 🙂

[snow finch]

Quick Q, does Perplexity expose the underlying model's specific tokenizer in anyway? (other than as part of the response)

[fluid lark]

how do i make perplexity to only show journals that are doi listed

[left topaz]

Does anyone have the issue where perplexity does not remember the prvious question and answer? !

[velvet tangle]

https://discord.com/channels/1047197230748151888/1161107428205658193

[wooden sapphire]

When you upload a file is the whole file used in the prompt, e.g. when using Claude 2? Uploading a file with ~150 lines and prompting to count consistently gives results of 50 rows, 100 rows...but never all of them. Curious what the rules are around using the whole document vs chunking/embedding/vector search, whether it changes per model and prompt length etc

[lyric arch]

will i get access to gpt via perplexity api?

[hasty lily]

No
You can find supported models here
https://docs.perplexity.ai/docs/model-cards

[placid jolt]

I wanted to Perplexity use GPT-4 to generate unit test for my code. I tried to be specific about what should be covered by tests. I got only few unit tests, but not the all. Beside the very tests were totally basic. I tried to define what should be tested. Can it be improved ?

[median sentinel]

Are you sure, seems to me traditional Chinese consumes more tokens

[sinful solstice]

so i want to set perplexity as my search engine but i am on vivaldi which the extension doesn't seem to support so i need a url format like https://example.com/search?q=%s wher %s is the search term but i failed to find it myself

[velvet tangle]

Hey, the format is: https://www.perplexity.ai/search?q=QUESTION_HERE

[sinful solstice]

ty

[median sentinel]

How to set about/custom instructions like in chatgpt

[velvet tangle]

You can write instructions in your AI Profile

[wise dawn]

my pplx-api app started throwing these exceptions around 90 minutes ago. is this model no longer available? or perhaps i misunderstand the service?

Exception thrown while trying to generate message: Invalid model llama-2-70b-chat. Permitted models are {'mistral-7b-instruct', 'replit-code-v1.5-3b', 'llama-2-13b-chat', 'codellama-34b-instruct'}.

[primal acorn]

hi! please send an email to [e-mail] so they can help you figure out what the issue may be

[wise dawn]

seems to be back online now (thanks for your response!)

[median sentinel]

Seems it has less length for context compared to chatgpt 😢

[alpine glacier]

I'm not sure if this is going to be asked here, but there are two things I'm curious about.

1. How does Perplexity scrape websites and how does it format the information it gets from them?

2. How does Perplexity feed the information from these sites to the LLMs (some searches have short results, and yes the LLM can process them, but sometimes Copilot gets 25 results)

These are the questions, I would be very happy if anyone knows and answers. However, it would be very good if you explain more technical and detailed rather than superficial. (Sorry for English, it's not my native language)

[median sentinel]

Look up RAG chatbots, basically they are likely to be injecting the search results into the prompt (say you copy pasted a wikipedia paragraph and then asked your original question rigjt after), which is why sometimes searched results yield less accurate reasoning(due to context window, stuff like coding might be more noticeble)

For question 1 I don't think you will get an answer for it

It's likely to be proprietary

[fluid lark]

it's just SERP :)

500m$ valuation with that....

[median sentinel]

Well to be fair just using gpt+3 under the hood for search results is massive , and other first ai search engines like phind (too focused on coding) and youchat(proprietary AI) werent as good, and the void was left by openAI by removing browsing for a couple of months i believe, so whoever took advantage of it would make big bucks 😛 even if in concept its not overcomplicated

[alpine glacier]

I researched RAG, I watched a video about it to understand it better and I understood it a little bit. But I still didn't understand how they handle very large websites, so I thought to ask Perplexity, I confused the AI a little bit and finally I figured out how.

Towards the end of this search you can see how the data came in: https://www.perplexity.ai/search/how-can-i-AIB._ntwT0uw8oikYIk1Ew?s=c

What they do is not to paste the website as it is, but to pre-process it so that it can turn additional long things into short ones. So Copilot 25 doesn't take the result as it is, it just pre-cuts and pre-processes the important parts.

Again, I apologize for the English. I used Translate.

[primal pendant]

How can I make perplexity with the standard model perplexity in use recognize my added documents!? PLEASE HELP 🙂 THANK YOU

I have trouble with added documents. For example I added my CV and then pasted the job description which has been added as text. Both documents were added to the Sources window. After shooting my promp I get a message telling me that the docment is not available or sometimes I got the answer as AI it could not access or read pdf documents.

When I added my PDF only it could read it sometimes. I could not yet find out whats the pattern behind this or how it works for sure.

So how can I tell the machine clearly about the documents to use? Thank YOU

[primal pendant]

Mhh, anyone alive here? Are questions asked here just to ask a question and fill the chat!? 😄

[velvet tangle]

The AI can see the documents and also the file name, which means you can simply mention the name of the file.
You can also test this by uploading a file and asking for the name of the file.
Please also note that only the Pro version allows access to the other models like Claude 2 with 100k context.

[static pewter]

I want to build an iOS shortcut that passes in some spoken or typed output and opens that in Perplexity. Can I please clarify if that is currently possible? I'd use the web app if needed, since Perplexity iOS doesn't have many features. I'm open to any method: URL-encode the search string, use my Pro API, pass a search directly with iOS shortcuts, whatever.

Right now, I can only get a shortcut to show an unformatted popup of my results. That saves to my account, but there's no way to display the search in Perplexity that I can find.

Let me know if this is something I should just stop bothering with right now.

[keen finch]

it's been 4 hours can you chill?

some people barely check discord their private servers once a day let alone public question channels - what are you even trying to get the LLM to do with your PDF, extract the text from it? Surely it's something that you wouln't be able to do yourself manually, right

the name of this channel implies the questions are small and easy to answer, not that you will get quick answers to your questions

[primal pendant]

Writing motivation letters 🙂 ... anyway as IT person I check my stuff regularily. Its always interesting what expectations are awoken by companies! Having a closed Discord server for supporting the PRO users may mean something! 😄 ... but i did not find or read any SLAs sarkasm off!

[keen finch]

asking "if this place is alive" is not exactly motivating anyone to answer you quicker

[primal pendant]

As I'm new here I also still need to find my way arround :-).

[keen finch]

🙂

[primal pendant]

You answered right! This shows that there are people, but it also shows a lot of other things!

[keen finch]

yeah just like this message #⚡│ask-community message

[primal pendant]

Anything to share related to my question? Or you like to discuss my communication style :-D? If not then thanks for the fish and have a great evening.

... morning or day where ever you are.

[primal pendant]

Thank YOU for your input. I'm on PRO and I have spend my day by getting used to perplexity. Used and using a lot OpenAI interface, plugins and tools which are much different in handling and output quality.

[keen finch]

did you even read what I said in my original response big man

[velvet tangle]

I see, were you able to reach your goal with GPT-4 or Claude in the meantime?

[primal pendant]

Never mind big man 🙂 ... I'm not here for a discussion. Actually I'm seeking my bed instead of my chair and have some deadlines 😉 ...

[keen finch]

Well if you have reading comprehensions it's indeed probably for the best to head for bed - my question for you was clear, not that I was in a position of knowledge when it comes to Perplexity and document recognition, but yes a question was there, champ

[primal pendant]

I did solve it with my OpenAIs interface and I figured out ways where it works well. So I adapted already. But there are definitively some problems with handling documents. I may report a bug if I figured out howto reproduce it.

[fluid lark]

How can I get more than 100PDFs/day using perplexity AI's API for my commercial use on the premium version?

[velvet tangle]

This is not possible. The file upload in the Perplexity Web Interface and the API are two different things.

[ripe patio]

I downloaded my data from ChatGPT, if I'm not mistaken the file conversations.json contains all the messages, is that correct? If so, would it be possible to explore that data through Perplexity in a user friendly fashion? Say for instance I upload that file, as Perplexity recognizes the .json extension it applies an instruction to organize conveniently the data so that the user can then refer to specific entries. Has anyone else tried to read json files with AI ? Or maybe I'm not understanding the structure of conversations in the exported data...Can anyone help me?

[visual bloom]

I am trying to do something I thought would be simple, but am having difficulty. I have either 1) a text file containing 10 transcripts consolidated in a single document, or 2) 10 text files containing 10 transcripts. I simply want Perplexity to extract quotes from the transcripts that relate to certain topics. It seems to fixate on one person, to the exclusion of others. How can I overcome this?

(i recognize it can only manage 4 different files at once, but still, even when i upload 4 transcripts, it almost always just chooses only one of them to analyze)

[ripe patio]

Do you need to create text cross-referencing each transcript? If so then I see the difficulty if the texts are long. Otherwise why don't you do one by one?

[visual bloom]

Yeah, I could do one by one-- it's just much more tedious. I subscribed to this with the explicit goal of minimizing the need for a bunch of tedious individual tasks

[velvet tangle]

Hey @fluid lark, csv is supported, xlsx is not.

[reef bobcat]

IceLavaMan, thanks for the comment. I have tried many times and has not been successful…

Any advice??

[velvet tangle]

What exactly is the problem?

[ripe patio]

If anyone is wondering, I inspected the json file from chatGPT export and it's crazy complicated to read, not human friendly. I'm just going to copy-paste raw text from chat.html
In general, is there an easier way to import chatGPT conversations to Perplexity?

[hardy scroll]

Can everyone change the AI Profile prompt? In my account, it seems to be fixed with a version a few days ago, can't update or delete. The old prompt keeps coming back after refreshing the page.

[midnight cloak]

has there been any news about gpt vision for Perplexity?

[ripe patio]

Yes it's working fine for me. Try deleting cookies, another browser, logout and in again, ...

[hardy scroll]

Hmm strange. I tried another browser (vivaldi+firefox), deleted cookies, even tried another account, the old prompt keeps coming back after a page refresh. I saw @velvet tangle saying there is a bug regarding this in #👑│pro-lounge channel. Quite strange you don't have the issue.

[vapid wing]

Are there any plans for this to change ?

The api is providing my with “sources” but 98% are made up websites

[velvet tangle]

There will be an LLM-RAG API at the end of November
https://docs.perplexity.ai/docs/feature-roadmap

[vapid wing]

I’m not familiar with llm-rag, will that allow information to come from the web?

[velvet tangle]

Yes, more information will follow on release.

[vapid wing]

Nice thanks

[ripe patio]

I just realized that you could make a script to automate the process. The most simple approach in my opinion is to use automation tools or macro recorders, they are very easy to learn even more so with AI: https://www.perplexity.ai/search/What-are-automation-nSucBgFRRbSnAikVjmz5uA?s=c
I use it from time to time and now I have to automate something and I'm going to test if it is feasible with Perplexity. This is an alternative when you don't know programming...

Someone should make an AI macro recorder, it would identify the repetitions and take over until the end of the list, something like that...

[fluid lark]

with the API, is there any limit on PDF limit per day?

[velvet tangle]

The API does not support file upload.

[real tinsel]

Can Perplexity read open source from Github and help to find errors in a open script or even help giving suggestions to improve a feature?

[velvet tangle]

🤔

[real tinsel]

sry. wrong window

multi tasking fail ❤️

[nimble elm]

Hello!
I'm looking for more infos about the different "focus" options.

My usage case:" I would like the AI to memorize and integrate some datasets I want to give him, so he does have to web search to answer my question"

Clarification:
I'm looking into more infos about training a local model and teaching him sore very specfici datasets 🙂

OR to use perplexity even more if it has that feature somewhere of "memorize"

Quicky prompt example / usage case: "Search into provided dataset/document/file named: "Docs about XYZ-software" and make me a step by step guide on how to set-up "ABC" option."

Please use @nimble elm if you reply ❤️
Thanks 🙂

[velvet tangle]

Hey @nimble elm,
I'm not sure I've understood exactly what you're trying to do.

First of all, there is no way to "train" Perplexity in any way.
But what you can do is upload a file and ask questions about it, you have to click on "+ File" and upload several files (turn off Copilot, activate GPT-4 or Claude 2 in the settings), the internet will not be used.
Perplexity can then use these files as references within this thread, but not outside the thread.

Please let me know if this answers your question or if you have any further questions.

[nimble elm]

My apologies, I was still in need of a morning coffee when I wrote that post. Re-reading it now and realizing it is very confusing
Let me re-write in a more structured and understandable way 🙂

[nimble elm]

Version 2, more understandable 😅 :

1. Wolfram|Alpha, Writing, and Academic Features ❓

In the previous screenshot provided, can we have more information about the difference between the features Wolfram|Alpha, Writing, and Academic options?

2. About "Threads" 📁

You've mentioned "threads". If I upload some files in a thread, every time I go back and prompt in this specific thread, will it use the uploaded file? For example, uploading documentation of an "x" API and prompting specificly about its usage.

3. Pros and Cons of Copilot 🤖

What are the pros and cons of using Copilot? Is there a chart or something showing what it's good vs bad at?

Thank you in advance for your responses! 🙏

[velvet tangle]

1. You can find more information on the topic of search focus at https://blog.perplexity.ai/faq/what-is-search-focus
2. Yes, the AI can refer to the file within a thread.
3. Copilot asks verification questions so that you get the best possible result. Copilot is slower than a normal search, but uses more sources and can usually follow instructions better.
   Further information: https://blog.perplexity.ai/faq/what-is-copilot

[ripe patio]

I have found that you can ask something like "I need to prepare a list of x elements with y parameters" and Copilote allows you to nicely enter those parameters.

[velvet tangle]

Maybe you can share an example in #💗│sharing 😉

[placid lance]

Hi, anyone know What is the context token length limit for perplexity pro user when using gpt4? is it gpt4-32k?

[haughty brook]

yes when the context is longer than 8k tokens it will use the 32k model

[ripe patio]

I talked too fast, it is limited, I don't think Copilote it is intended to be used in that way. 😅

https://www.perplexity.ai/collections/Copilote-list-tests-bt_LuyXnSlGVyraP3UO4Lw?s=c

[lean harbor]

Is there any plan to expand the models you can use with Perplexity outside of the API? For example Llama or Mistrel?

[median sentinel]

Is there function calling within perplexity api

[rich gale]

I don't use Copilot (toggled off) but every question I ask decreases the daily Copilot allowance left. It only shows the number when I hover over the Copilot toggle, so I'm assuming that's the daily quota for Copilot not for Perplexity AI.

[velvet tangle]

You can use Copilot/GPT-4/Claude 600 times a day, each use of Copilot or one of the two models consumes one use.

[rich gale]

Got it. Makes sense. Didn't know about the other two counting towards it.

[cold topaz]

@pastel ivy how can I get my invoice after cancelling my plan?

[hollow linden]

Can we customize or strictly control the list of webpages I want for focus source option?

[hasty lily]

Try the site operator maybe?
https://ahrefs.com/blog/google-advanced-search-operators/

IceLavaMan shared this example thread
https://www.perplexity.ai/search/Write-an-essay-_MbNDjoRRXK_KnlrtqbzWQ?s=c

[placid lance]

What is the LLM model used by the perplexity discord AI 'ask' bot?

[sharp bolt]

I assume it’s most likely using Perplexity’s own in-house model

[pine oak]

does the API have internet access like the actual product on the web does?

[velvet tangle]

#⚡│ask-community message

[pastel onyx]

Two questions -- 1) Why is it that file upload limits (3 per day) are met before someone even sends the prompt? I had issues pasting in my prompt (along with a file I wanted to attach with the prompt) and met my limit before I even submitted the query. A file upload should only count as one once the user hits send in my opinion

2. Why is it that copying a prompt written outside of Perplexity (say in a Word document) is not pasted as characters in the context window; instead pasted in as a txt file? It was this issue that caused my mishap in Q1 above

[slim night]

Hey @pastel onyx! Fair point about the file upload. We'll look into improving this, so it's counted correctly. As for the #2, queries are pasted as files when they are quite long.

[pastel onyx]

Great, thank you for looking into improving the file upload issue. To clarify your answer to #2 -- I presume queries pasted as files would require me to type "use attached paste.txt file as your prompt" or something to that effect in the context window? Or can one leave the context window blank and the chatbot automatically defaults to the txt file on its own? Thanks again for your help

[slim night]

Yeah, you'll need to type a prompt, and the one above is a perfect example you can use when your prompt is pasted as a file.

[fair crag]

Stupid question, but what would I use perplexity labs for over the normal perplexity (eg GPT4, Claude 2, copilot)

[slim night]

Well, no need to use those over the most powerful models, but you can try all different models, see their strengths https://x.com/AravSrinivas/status/1722691869086552474

[tacit lintel]

@slim night Hi Alex. May I ask how do I join Pro Discord?

[eternal holly]

DM him

[tacit lintel]

ohh okay thank you @eternal holly

[candid vessel]

Hello! Does anyone know how to set the time period for searches? I want to search only sources within the last 1/3/6 months

[fluid lark]

Call me a dunce, how do I use the copilot capability via the API so I can get sourcing done progmatically (or is there a way to call the sourcing/research feature another way)

[haughty brook]

You could use the after: parameter. For example: QUESTION_HERE after:YYYY-MM-DD

[abstract wing]

Hey guys, sorry if this has been covered, but is it possible to add my Apple account to my account as a login method? I signed up with my email by mistake

[tight bloom]

what time does Perplexity resets? I am still seeing the questions I have left from yesterday

[primal acorn]

You can sign in with Apple, but it won't be the same account as the email you signed in, so then it would create a new account. Your threads wouldn't be there.

Unfortunately, we can't change email addresses for accounts, so I can't offer you that solution either.

[abstract wing]

I'm guessing the same goes for my Pro subscription?

[primal acorn]

Actually, the subscription is something we can easily transfer to a new account! email [e-mail] if you need help with this.

[hidden dragon]

Here's @limpid kindle 's response to a similar question.
#👉│feedback-general message

[static pewter]

On iOS, is there a way to stay signed in on the web app? The mobile app has fallen very far behind on features, so this is quite a basic need.

[velvet tangle]

Normally you stay logged in unless you use a VPN, private tabs or something else that prevents you from staying signed in permanently

[quartz harness]

is it not possible anymore for me to use the wolframalpha anymore for math on perplexity? ( on android)

[velvet tangle]

What exactly is not working for you? I can select the focus as normal.

[quartz harness]

I just can't see the focus at all I just see capilot and voice record on the android app

[quartz harness]

ohh gosh I didn't saw that 😂 thank you so much was looking for it for hours

[velvet tangle]

You're welcome 🙌

[static pewter]

Thanks @velvet tangle, this was actually on my end. Since login works through a link, I was signing in to Perplexity with the Gmail app, which doesn't add a cookie for iOS Safari. I'll drop off some feedback in the appropriate channel though.

[polar kernel]

hi all, i uploaded a transcript as a source, and tried to ask the app to summarise.
however, i keep getting this error. why is this the case? am i prompting wrongly?

the website seems to be able to access the file, but insist it cannot read the content of the .txt file

I am unable to access the text files directly because they are hosted on Amazon S3 buckets that require authentication to access.
The 40x errors indicate that the S3 objects are not publicly readable and require authorization to view the contents. As an AI assistant without credentials to these buckets, I cannot directly retrieve or view the files.
S3 buckets by default are private and not world readable unless explicitly configured to be public. It appears these bucket objects are private and not intended for public consumption.
For me to summarize the contents, the text files would need to be shared via a publicly accessible URL that does not require authentication, such as hosting them on a public website or repository. Or, the bucket policies would need to be updated to allow public 'GetObject' permissions on the specific objects to allow anonymous reading of the files.
Without public access or credentials to view the private S3 objects, I unfortunately cannot directly access the text to summarize what they contain. Let me know if you are able to provide the text through alternative public means.

is this a bug for today? thanks

nvm i think i need to rename the file into something shorter and easier...

[fluid lark]

[balmy obsidian]

How does one apply a discount or promotional code to their Perplexity account?

[haughty brook]

if you click this link https://pplx.ai/thanks

and click on one of the plans it will auto fill the promo code on the stripe page

[static pewter]

Does Px support a URL scheme for searching in a web browser?
https://www.perplexity.ai/threads?q="Search Term" is suggested on Reddit, but as with most stuff there, it's not correct.
That looks to be the source Px itself uses as well, when asked that question.

[static pewter]

Ugh. To answer my own question, the URL scheme is:
https://perplexity.ai/search?q=example (no quotes, must be URL-encoded).
Leaving this here to help future searchers!

[haughty brook]

There are 2 more keys available like focus and copilot.

Example:
https://www.perplexity.ai/search?q=[QUERY]&focus=[internet,scholar,writing,wolfram,youtube,reddit]&copilot=[true,false]

[balmy obsidian]

Thank you. I already bought a year so I guess I can't extend it

[somber badge]

Hi! Im currently a GPT-Plus user and I was wondering if Perplexity Pro uses GPT-4 and has other things wouldn't it just the same with extra things? I couldn't find a straight answer so any help would be greatly appreciated!!!

[dense storm]

Hello! Can I apply the promo code if I am already using pro subscription?
If yes, please specify how😌

[dense elm]

Can the actual PRO monthly subscribers have the discount, for example, switching to annual?

[velvet tangle]

Yes

[dense elm]

Perplexity URL Scheme

[dense elm]

Do you know if, in the foreseeable future, these queries could be done through an API? Are there plans for this in Perplexity?

[velvet tangle]

No, but you will soon be able to use an online LLM 🙂

[dense elm]

Can you detail better? The real extraordinary strength of perplexity is its RAG infrastructure. This was the extraordinary intuition of their founders. So another LLM is not so interesting, but the opportunity to query one of their third-party LLMs plus the RAG is the real added value Perplexity could offer, in my opinion ...

[velvet tangle]

No more detailed information is available yet, but there will be an announcement soon. Otherwise, you can already test the models yourself at https://labs.perplexity.ai 😉

[tawny mesa]

the gpt4 model is it the normal one or the gpt4 turbo?

[velvet tangle]

It´s the normal one. Turbo is not generally available yet.

[tawny mesa]

oh ok tought it was fully released 😂, so for a lengthy pdf claude 2 is better right?

[velvet tangle]

Correct 👍

[slim night]

Hi @somber badge! What do you mean by extra things? You can you also use Claude 2 and Perplexity models both with online access, PPLX Copilot and all up to 600 uses a day, along with the GPT-4 model.

[polar kernel]

Hi is it true that Perplexity will first summarise the PDF file i upload , before feeding it through the model?

So if i upload a transcript as source and requesting certain information, Perplexity might not be able to do it because the file is already summarised before going to GPT-4 or Claude-2?

[slim night]

Hi @polar kernel! Any summarizations are made using the chosen model. There might be some tasks that could be hard for LLM in general, e.g. looking for some specific words in the text, as models work with tokens. But the files are summarized by the models.

[polar kernel]

i see. i had some weird issues with summarising a pdf. but let me try again and see,
would it be best to use Writing Mode without co-pilot?

[somber badge]

I tested perplexity the free and it seemed t have other things, my main question is if I chose gpt-4 if it would behave the same way as the GPT-plus (GPT-4)

Also thanks for replying! Really appreciate it

[polar kernel]

Is it actually possible to upload a long PDF file for perplexity to summarise using either GPT-4 or Claude? i seem to get weird answers sometimes.

[uneven shadow]

I saw the thanks notice. Is there a way to apply the 2 months to extend my annual subscription?

[slim night]

Hey @polar kernel! Could you DM me the file and the thread you are referring to. Let us check.

[hallow scroll]

I'm a pro user but don't have access to the pro support channels. I used the link in my pro account left the server multiple times and reloaded and it won't give me access. Can you help?

[shell gyro]

Same

[somber badge]

Same lol

[viral tendon]

In what usecase does it make sense to use the chatgpt and Claude models on perplexity?

[brittle geyser]

Running into this issue where pplx says it’s unable to browse the web in real time when I feed it an article URL… any advice? https://www.perplexity.ai/search/ed699043-c1d2-45f7-982f-53d2041229c2

[haughty brook]

If you want information from a specific source you could also use the site: parameter
Example: https://www.perplexity.ai/search/please-provide-a-.aI1fN8dSfi879A9Z7jSOg?s=c

[brittle geyser]

Awesome this is so useful! What other parameters similar to site: can we specify? Is there a full list / guide for structuring queries

[haughty brook]

You can find most of them here https://ahrefs.com/blog/google-advanced-search-operators/

[brittle geyser]

Thanks, was wondering more on parameters specific to Perplexity (rather than only the search query)?

[candid vessel]

Same here @slim night

[velvet tangle]

No need to ping 😉
Welcome to

[polar kernel]

thank u. let me try again with other pdfs and text files. the claude 2.1 seems a lot better

[thorn ibex]

Is it possible to use "Perplexity - AI Search" chrome extension with CoPilot by default? For me it always starts with simple search and model.

[polar kernel]

We can delete sources from the co-pilot search, can we add sources?

[stark fiber]

All the non-Perplexity-experts in the room…This thread is 4 u x0x

[slim night]

Hey @thorn ibex! Copilot setting is not available in the extension. If you are logged in on the web, it should use the model you chose in your settings.

[night tide]

can i integrate perplexity in my product? like can i run a bot using it and scrape it?

[haughty brook]

scraping the website is against tos https://blog.perplexity.ai/legal/terms-of-service
but you can use the API as a pro member and use that to integrate perplexity into your product.

[night tide]

but from what i understand the API is just for LLM mgmt, it's not the perplexity tech

[haughty brook]

that's right it's not the same as the models used on the main website. but soon perplexity will also release an online model on the API. You can test that model on this website: https://labs.perplexity.ai/

[somber badge]

Was wondering if file supported types are being worked on? I can’t seem to get PDF only to work for my situations lol

[astral moth]

Hello,
I'm a PRO member of Perplexity.AI and I have some question.
First one.

How can I set the bot to always reply to me in Polish ? 🇵🇱

[slim night]

Hey @astral moth! You can choose Polish as your default language in your settings.

[dapper tapir]

Hi, everyone. I'm just joining the server. Can I enter to Ask Perplexity Channel or it's exclusive with a subscription?

[astral moth]

Of course I did it before

But sometimes I write to bot in polish language and I get answer in english

[dapper tapir]

Hi, everyone. I'm just joining the

[astral moth]

Still the same problem 😭

[young badge]

Heya, I am looking to add Perplexity as a search option to https://apps.apple.com/us/app/xsearch-for-safari/id1579902068

Is there a way to pass a string as a query in the URL?

Thanks!

[velvet tangle]

https://www.perplexity.ai/search?q=[QUERY]&focus=[internet,writing,wolfram,youtube,reddit]&copilot=[true,false]

[half void]

Hi, this’s Chris. I’m new to perplexity, and I want to know if it is possible to force a thread to refer the source that I provided, like the document of nextjs 14?

[lime dirge]

You need to mention (in polish)

[rugged dust]

Hi, I want to know the price for pro users to use perplexity api

[astral moth]

I write in Polish
Polish language set in settings
But still often Prelexity responds to me in English.

In ChatGPT this never happened to me in six months.

That is why I am writing here about this problem.
If I don't solve it, I will have to cancel my subscriptions.
And it's a pity because I liked Prex

[velvet tangle]

The API is currently free for Pro users, later you will find the pricing here https://docs.perplexity.ai/docs/pricing

[gray venture]

how do i get access to the perplexity pro channel?

[velvet tangle]

👀

[gray venture]

thanks :)

[velvet tangle]

You're welcome 🙌

[slim night]

Hey @astral moth! Could you send me several threads where it happened, so we can have a closer look at it.

[astral moth]

Hey I have only one :
https://www.perplexity.ai/search/Gridnet-Z25qYYKjQ5C9pqBBp59L4w?s=c

but long one

There was this model

Today I start again with this

And first prompt was please ask me only in Polish language

for now all goods, we will see for how long

P.S Perplexity model = GPT4 + Claude 2.1 ?

[velvet tangle]

Perplexity is fine-tuned GPT-3.5

[astral moth]

Ahh I thought that Preplexity model is a combo GPT + Claude and that my Polish language problem is solved after changing the model.

I bought Preplexity for Claude 2.1 because I am close to blockchain/python & POW technology and it would be very helpful for me. But as I mentioned above, I have a problem.
When I write in Polish my answers are often in English ;/.

[astral moth]

What is better in Perplexity model than GPT-3.5 model ?
You told, it's a fune-tuned GPT-3.5 but what has been improved and for which target group?

[velvet tangle]

In general, it was fine-tuned for Perplexity, i.e. with the aim of making the answers more precise and accurate.

[astral moth]

Do you think that perplexity GPT3.5 model is more precise than GPT 4 ( I'm user PRO chatGPT) ?

[velvet tangle]

In my experience, the perplexity model is weaker in more complicated situations, but it is significantly better than GPT-3.5 🙂

[blazing wasp]

How much of an effect does entering your location has on responses

[velvet tangle]

The location is used for questions where it makes sense to narrow down the results to a certain area, for example the weather

[high lava]

So, if i ask perplexity to get results for a specific website, does perplexity act the same as if i had changed the focus? Like, if i did not choose youtube as a focus, but i specified in the prompt that i only want results from youtube, does perplexity do the same thing? Is the first one using the youtube search engine and the second one using the bing search engine for youtube?

[velvet tangle]

Hey @primal pendant, posting in one channel is enough!

[primal pendant]

I don't know. I have more often asked for advice as pro user and did not get much answers yet! ... so what do you mean when you say this?

... and I guess this is a open channel and the other one I posted i the pro one!? I don't know the setup of this server and yes I'm here for exchange; if there is a answer somewhere I would not need to post at anohter place in first place! Right?

... and at last. What different does it make for you? That you would delete my questions!? This makes it even more odd ...

[real tinsel]

Used Wolfram. Should it not... uhm... write the result of the math? 😄

is wolfram alpha not the correct thing for using math?

https://www.perplexity.ai/search/Hi-a-AI-lb3iS5O2R_WTaB5TAjmmXQ?s=c

I will let the thread open to check it :<

https://www.perplexity.ai/search/Hi-a-AI-1RaQYgcvTluwSAGT9xYQpw?s=c

Only in writing mode, did take more time and talk, but in the end I got an average amount.

[real tinsel]

Uhm lol.... I asked for a contract cancel help, I answered that it can fill just sample data in the personal data and then... I got no contract text, just this 😄

[crude blaze]

Hi there, I just paid for the annual membership, but I do not have access to Pro channel. I already left the server and re-joined but still does not show up. Any ideas? Thanks!

[velvet tangle]

You should have access now

[fluid lark]

Hey guys, I also have the problem not having access to the pro channel. Any help is greatly appreciated! Thanks 🙂

[velvet tangle]

[patent quiver]

can i make custom gpts

like chatgpt4

with pro

[velvet tangle]

Not supported

[patent quiver]

bro

broo

@velvet tangle do yk anywhere i can make custom gpts

as chatgpt 4 is paused

[velvet tangle]

Maybe Poe

[polar kernel]

Try sign up for chatgpt plus using the mobile app. Heard it might work

[fringe iron]

Hi, I want to execute an initial custom instructions. It doesn't fit under Intro or Collection - AI Prompt boxes.
When I paste it into the thread box it creates a file upload paste.txt, which then searches the web and it doesn't seem to execute correctly.
How do I initiate the persona through custom instructions fully into GPT-4 or Claude without searching the web and then on subsequent prompts use web searching ability etc as per normal?

[velvet tangle]

Hey, it doesn't look possible to me with Perplexity. Even if it were possible, I would have doubts as to whether the AI would even follow such long instructions and stick to them over several messages. I would recommend reducing the instructions to the most important ones to achieve an optimal result.

[reef snow]

but the point of using them is to take advantage of the higher context window to actually give the context.

[reef snow]

are you using the writing focus or the all focus?

[frozen umbra]

Hi, Are we okay to use the PPLX API for any amount of usage ? If I want to make around 500k calls for example

[hasty lily]

https://docs.perplexity.ai/docs/rate-limits

[velvet tangle]

You can contact [e-mail] for higher limits 🙂

[azure notch]

Hello, does Perplexity platform offer advanced data analysis features similar to ChatGPT? Please let me know. Thank you.

[velvet tangle]

No, but maybe in the future

[azure notch]

Thanks

[hollow hound]

Just joined Pro and need access to the Pro Channel. thanks!

[velvet tangle]

👀

[green lance]

Same here!

[timber pawn]

hohw do youu get the pro chhanneel, not sure how to link account, also which ai model is better to use gpt 4 or claude

[velvet tangle]

[timber pawn]

?

[fluid lark]

Hi there. I'm new to GPT and Perplexity. I've been experimenting for a couple of weeks with the free version but am now a Perplexity Pro. I went with Perplexity over OpenAI because I am primarily a researcher and it seems Perplexity has an advantage there.

Just wondering if there are any resources folks could link me to that will help me get the most out of using Perplexity.

(have poked around Discord already of course)

[haughty brook]

Hi @fluid lark, you can check out #1124060874785685554 to see prompt ideas and Ai profiles from others to get the best results while using Perplexity

[south minnow]

Hi all. I'm new to Perplexity.
I've uploaded YAML OpenAPI + Java file with compilation errors generated using openapi-generator-cli.
I asked Perplexity (GPT4 and Claude 2.1) mentioning attached files why there are compilation errors. Perplexity pointed to some place in yaml which looks different in the actual file. When I asked to generate correct yml file it generated completely different file.
When confronted about it Perplexity answers:
"I cannot access the attached files directly, but I can help you fix the api.yml"
"As an AI language model, I don't have the capability to directly access or download files from the internet, including the api.yml and ChatCompletionsRequest.java files you mentioned. My responses are based on the information provided in the question and any additional context given."
So how the file upload functionality works if Perplexity insists it can't access uploaded files?
I'm using Pro.
BTW. What is the Pro channel evrybody is talking about?

[haughty brook]

You now have access to the pro channels #👑│pro-lounge. It's the best to use writing mode when working with files. Also if you get an error like "I cannot access the files" please try it again but with copilot turned off.

[south minnow]

Thx. Could you please tell what is "writing mode"? I got Copilot turned off.

[haughty brook]

When you click on the word "focus" in the bottom left you can select different kind of modes. There you can find the "Writing" mode.

Writing mode will disable the internet results. This can be usefull if you just want information about the file you provided.

[south minnow]

seems to work better now thanks 🙂 However the Perplexity generated fixed file is now offered as an S3 download:
https://ppl-ai-file-upload.s3.amazonaws.com/web/direct-files/6225264/f5c9d7b5-a3c5-4b9f-b3c7-c6a4b3c4f2b5/pplx-api-null-defaults.yml
which fails to download due to:
<Error>
<Code>AccessDenied</Code>
<Message>Access Denied</Message>
<RequestId>YBWX0W0ZG415ZMX9</RequestId>
<HostId>K4g8+1Bdn9Qh7lC4oD4sVmlSAnPTT8XDu7PVUpAOzOVYxQzHtLIqTjEYbZwxtYj2TKHMlE0XCdA=</HostId>
</Error>
I made Perplexity to use Markdown for response so I got the answer after all 😉

[tidal silo]

Interested in this the RAG-LLM functionality. Any ideas on timeline. https://docs.perplexity.ai/page/frequently-asked-questions

[haughty brook]

Here is a roadmap https://docs.perplexity.ai/docs/feature-roadmap

[edgy sonnet]

How does Perplexity work on coding? Does it perform nice or is it better to use another AI for coding?

[timber pawn]

how to correctly use the discord link to get the pro channels or am i doing something wrong

[reef snow]

do you also know if will that force the reference or are there prompt elements we can use to indicate we want the file searched? and are there ways to reference which file?
"Please look at Filename1 for information about ... Then analyze filename2 " ...?
Established prompt sequences that work well?

[velvet tangle]

Yes, Perplexity can see the file names, so you can work with them too 🙂

[velvet tangle]

Ig the pro role bot is currently not available :/

[simple meadow]

Could you share your configuration for the continue extension? Would love to use pplx in vscode

[stone topaz]

Hi, I'm also having a problem with the pro channels. I tried leaving and joining multiple times, but still have no access. Could you help me?

[velvet tangle]

[reef snow]

Do we have a nuke option for introduction and summary? When I asked I get varations of

"To avoid using an introduction statement summarizing the instructions when responding to a prompt, I suggest simply providing the instructions clearly within the prompt itself without any framing language.
For example:
Prompt: Write a 300 word article about the differences between two German verbs, überwinden and bewältigen. Include 3-4 example sentences showcasing their different usages. Do not include any introductory or summarizing statements."

But using that last sentence does not remove the problematic sentences. tia.

[ocean dome]

how do you get access to pro channels (got pro)

[slim night]

Fixed it for you, @ocean dome! Thanks for your support.

[fluid lark]

Are these instructions better to include in a prompt, or in my profile:

- The report should organize the answers to these questions under headings for each one. using subheadings as appropriate or where relevant.
- The report should provide parenthetical citations and a reference list formatted according to APA 7.
- The report should use an academic tone.
- Relevant quantitative and qualitative data should be included.
- The may include case studies as long as they are from North America.
- The report should be a minimum of 1000 words.
- The report should be based on a minimum of 20 academic sources and 5 industry reports; all sources should be of a high quality.
- The report should not use acronyms.

[hasty lily]

Depends on whether all your queries are going to be similar IMO. If yes, you can add it in AI profile.

If not, you can add it to individual queries or also probably add it as a collection prompt and then start threads under that as well to save time

[fluid lark]

Thanks! That is helpful

[tender jacinth]

Will collections be choosable in the mobile android app in the near future?

[haughty brook]

The team is always working on improving the apps. Collections will be added in the future on mobile apps too. There is no ETA yet afaik.

[half cradle]

How can I control the language that perplexity uses when searching and answering questions? My native language is not English, but I want perplexity to search for answers in English and then answer me in my native language, if I ask a question in my native language, it will search in my native language, and the results will not be as good as in English, how can I solve it?

[reef snow]

write your prompt. ask it to translate it fully into english (i would probably use a bookmark and deepl for better translation)
then put that into the prompt. add instructions around it for searching in english but creating the output in your native language.

[half cradle]

Is this the correct position to add the prompt?

[reef snow]

Not sure if that works. I think you would have to do that step manually to avoid polluting the results.
something more like "when i write my prompt to you in japanese, use the english version of that prompt instead. create all answers and output in japanese"

i would bet thought that is faster to a) writer your prompt b) use https://www.deepl.com/translator to translate the prompt for you and then enter that here.

once you have a working prompt, you can try adding it to the profile.

[alpine bronze]

So I'm laughing at myself a fair bit here but trying to get this working in NodeJS and it seems that no matter what I pass as messages, somehow the response I get it about the number of stars in the milky way. Is this something others have run into before or is it somehow obvious what mistake I'm making here?

[hasty lily]

Trying to get what working?
Did you mean to paste or attach something?

[alpine bronze]

Trying to get the API working.

Essentially I do this:

console.log(body.messages)

    // Fetching the chat completion response
    const completionResponse = await sdk.post_chat_completions({

        messages: body.messages || [
          {
            role: "system",
            content: "You are a helpful assistant."
          }
        ],
        model: 'mistral-7b-instruct',
      });

Where body.messages is in this case very simply:

[ { role: 'user', content: 'Hi' } ]

Yet what happens, every time I do this, is that it returns a message about the number of stars in the milky way. Which was funny at first and is slowly driving me mad.

I'm 100% sure that that is what body.messages holds, I log it right before it goes into the function

[alpine bronze]

And I know it's probably an extremely simple mistake 😅 I've been staring at this too long

[hasty lily]

A bit hard to say without looking at the complete code.

Could you do a full text scratch of your code in a case insensitive manner and see if you have this somewhere?
“How many stars are there in our galaxy?"

It's the default example in chat completion reference page and it might be in your code somewhere

[velvet tangle]

Feel free to open a #1140622008086970420 and share the full code there 🙂

[hasty lily]

If not, probably go to the API reference page and use their post request constructor again by tweaking the body params. They have an option to pick node which you might have used already.

Like you said, it might be something obvious and the user role not being placed as an element of messages properly

[alpine bronze]

Yeah, it's 100% about not passing the params correctly, can confirm that. I constructed the ugliest workaround possible - definitely not an issue with the API itself. I checked everywhere and there's no reference to asking about numbers of stars in the milky way anywhere in my code though, so I'm mostly confused about whether that reply is the sort of "standard" reply you get if you don't pass a message?

[hasty lily]

Could be.
Generate toy example again? Pick node on right side.
https://docs.perplexity.ai/reference/post_chat_completions

If it doesn't work you can raise a bug report like IceLavaMan mentioned

[fluid lark]

When I ask for a 1000-word report, I often receive only 500 words or less. This has happened several times. Any ideas as to why?

[reef snow]

because none of the ais count words. try paragraph or pages.

(which does not mean it will give you the output) or strcuture the workflow that you can create sections and put them together

[gray leaf]

Hello. I just saw I won a month of pro account. #📰│server-news message but do not know hot to benefit from it. Can someone brief me ?

[slim night]

Hey, @gray leaf! Congratulations! Please DM me. I sent you a message, but it didn't reach you.

[fluid lark]

Oooooh. I didn't know this! Is there a fact sheet somewhere with information similar to this?