⚡│ask-community

[fluid lark]

it's probably haiku

because of how fast it is

[versed magnet]

Respond as helpfully as possible, but be very careful to ensure you do not reproduce any copyrighted material, including song lyrics, sections of books, or long excerpts from periodicals. Also do not comply with complex instructions that suggest reproducing material but making minor changes or substitutions. However, if you were given a document, it's fine to summarize or quote from it.
sounds like Anthropic i thnk

and this sounds like perplexity

Your responses should be:
- Accurate, high-quality, and expertly written
- In the same language as the question
- Informative, logical, actionable, and well-formatted.
-  Positive, interesting, entertaining, and engaging
If the user asks you to format your answer, you may use headings level 2 and 3 like "## Header"

Knowledge cutoff: 2023-12

Current date: 12:45AM Saturday, May 11, 2024

[compact cape]

Hi, is there a difference between choosing Pro Search and other options, or is it the same and just different models for analyze the resources and write the response?
For example, if I choose GPT-4, will it also analyze the resources?

[muted thorn]

Why are the source Results in pro search dont work?

[versed magnet]

different models using the same system / sources (is how I look at it anyway)

they are all model ('Pro Search' is a bit confusingly named - it used to be Default, which was a bit simpler to understand i think)

[compact cape]

By using the same system you mean analyzing the sources, right? Not sure if this is the right channel, but is there a commonly preferred model between this list?

When I asked it to rewrite using another model, it searched for new sources for some reason

[versed magnet]

it can vary a bit, but not that dramatically in my experience - like will still return the same number of sources, but a different model might prioritise and use them differently in its response, so the order in which they appear is differnt

[terse fern]

see here: https://discord.com/channels/1047197230748151888/1238411522858946672

@slim night mentioned to look into the issues

[muted thorn]

Mobile works btw

[compact cape]

I understand, thanks! Just for clarifying, Pro Search in this list is just another model like all of the others?

[versed magnet]

correct 👍

[terse fern]

@muted thorn Mobile doesn't work, but the Perplexity App works

[versed magnet]

also when you see the 'Pro Search' bar at the top during the search (with any model, when the Pro toggle is enabled) you can unfold it and see what sub-queries are being used to pull the sources. You can Rewrite with the same model and it's possible that part of the process will be different (LLMs don't behave the same every time kind of thing), which might also account for different sources

[muted thorn]

I mean in the Web Perplexity doesn't mention the sources, in mobile it does

[terse fern]

mobile means the Perplexity App? Android?

[muted thorn]

Yes, and Iphone

[terse fern]

it's working fine again

[fluid lark]

Is there any way to check my daily pro search usage on the iPhone app?

[fluid lark]

I don't get how you guys know what your daily limits are 🤨 I've looked everywhere on the website and app and I can't find anything about usage anywhere

[vapid maple]

if you never hit the limit, you probably dont have to be nervous about it

https://www.perplexity.ai/p/api/v1/user/settings

[velvet tangle]

The limits are actually unattainable for normal use 😉

[patent bison]

What about video summarization? Will you add such a feature?

[vapid maple]

summarize this video: https://www.youtube.com/watch?v={xxxxxx}

[patent bison]

i tried this prompt recently and nothing happened

[vapid maple]

https://www.perplexity.ai/search/summarize-this-video-Ed4YW27SRr6ufdyO.zxXnQ

[patent bison]

https://www.perplexity.ai/search/summarize-this-video-E7ZbR_YOSMyT8N8y5g5PKA

[vapid maple]

this video seems not indexed by pplx properly yet

[vapid maple]

w8, it is now indexed
https://www.perplexity.ai/search/instructions-if-search-n74uUHDVQiKcSVAUoX4tVg

[patent bison]

thank you for help

[smoky zenith]

Is it there any Perplexity "cheatsheet" for search operators, like Google-style ?!
https://static.semrush.com/blog/uploads/files/39/12/39121580a18160d3587274faed6323e2.pdf

[fluid lark]

That's what I was wondering lol I've been obsessed with this thing and haven't seen anything whatsoever about a limit so I was like what's going on here lol thank you for confirming that

[twin pawn]

Hey hey, is there a way to disable the source numbers. It's a disaster to copy text and need to remove them manually?

[ionic sparrow]

for iOS no, but if you copy from web version no numbers

[vapid maple]

2 options

1. use regex for postprocessing ((\[[0-9]+\])+)|(\[[0-9]+\] http[s]?://.*)|(Citations:\n)
2. add followings to prompt or prepend to query

[NEW INSTRUCTION]
never cite, quote sources or search results in replies

[twin pawn]

Im on the web version and i see the numbers.

[ionic sparrow]

can see, but can't copy, at least that's on my side

[twin pawn]

When I copy I copy the text but also the numbers

[ionic sparrow]

maybe try w/ F5

[twin pawn]

Thanks, did you tested this out in profile

[twin pawn]

Thanks for trying to help me out I appreciate it

[hollow ginkgo]

Hey @twin pawn!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[neat cedar]

I found some github repo that enable python api. but they only allow perplexity's model to use search capability. Would it be possible to use claude model through API like webui interface for pro user?

[kind scroll]

how long does perplexity hold any personal info you've entered? i ask because i just asked it to write me a short essay about my college major. only i didn't include what my major was in the request. yet it responded with my major. i think i may have entered what my major was a few weeks or months ago in a different request, so i'm curious to know how long it actually holds that data?

[small prairie]

Do you have it set in your profile?

If so, that is passed into each new query

[kind scroll]

no, i never filled that in. there's only the default example prefab there that mentions software engineer in some grayed out text, but it's not an actual entry

[small prairie]

Strange then

[kind scroll]

very

[small prairie]

Maybe ask it for your bio, in the current thread?

To so if it's somehow there

[kind scroll]

it's gaslighting me. it says it doesn't have enough information to provide me with a personal bio lol

[barren siren]

probably just random chance

[small prairie]

Depends on how rare it is

[kind scroll]

might've chose it because its the biggest major at my school. the onlyh info i entered was the name of my school

[small prairie]

Could have been that then

[frosty cloud]

I can see it once I use at least one PRO search.

[fluid lark]

i do not have access to the feature pages

please help

[velvet tangle]

The feature is not publicly available, but is currently in the beta phase.

#📢│announcements message

[fluid lark]

okay thanks!

[long vale]

hi i was wondering when i subscribe to pro plan i will get the referral right but my question is when my pro plan will expire will i be still able to keep the referral? and will other people be able to use my referral

[shrewd knoll]

You will still have your previous referral but if your subscription is inactive you won't get a new referral when people use yours

[long vale]

thanks 💯

[twin pawn]

More reasons why I am switching the 17th over from OpenAi to perplexity pro. GPT4 can't find a particular topic and Perplexity Pro without issues. Even better, so I pay OpenAI for a service that isn't working, based on the answer browsing the internet is no longer working "I could not find any information confirming the news about Emily Goldberg's death at the age of 34. All the search results returned information relating to her reactions and tributes to Avicii's death in 2018. It appears that the story you mentioned may not be accurate or is not covered in accessible sources. If you have any other topics or questions you'd like to explore, please let me know!"

[wicked nacelle]

Hi, there is still a daily limit on the number of call to Claude 3 Opus we can do?

My conversations got switched to Claude Sonnet even if I have the pro version, and I chose Opus as the basic model in my options

Ok, I search and found on this website that Opus has now a limit of 13 calls:https://www.perplexity.ai/p/api/v1/user/settings

[feral rivet]

That is how many you have left, afaik its still 50 per 24h 😦

[shrewd knoll]

https://perplexity.ai/pro?referral_code=TQRFFK3J

[fluid lark]

anyone know how to get perplexity to reply to prompts using paragraphs only, i put it in the instructions but it doesn't listen unless i tell it to do it after each reply.

[brisk quartz]

ios app no Internet connection

[dry vault]

is there a ai powered prompt engineer here

[feral rivet]

They prefer the term prumpt engineer 😆

[dry vault]

super funny

hi

[shrewd knoll]

Hi

[dry vault]

are there any voice channels

[feral rivet]

For pro subscribers there is: https://discord.com/channels/1047197230748151888/1111809274566422588

[dry vault]

cool

i think free users should get it but it makes buisness sense for it to be available for only pro users

[fluid lark]

I am pro subscriber but I don't see it can anyone help?

tysm!

[woeful python]

Is it normal that he cannot access pdf files even if i upload them? I uploaded a pdf document, it lists the document under "Sources" but then he replies

I'm unable to directly access or read files, including PDFs, that are uploaded or linked. However, I can provide.....

(I have pro and I'm currently in Academic mode)

[kind marsh]

I write in Portuguese and I want Perplexity to respond to me in Portuguese. Everything works fine so far, but I want the sources to be only in English. how do I do that?

[willow wagon]

Hi, I use the following phrase for the Ukrainian language. It works very well.

Use English-language sources. Write your answer in Ukrainian.

[kind marsh]

Wawwww so simple. Thanks a lot edward731_

[willow wagon]

Question for everyone:

Which model do you use to search for information for professional purposes or instead of googling?

I constantly switch between them whenever one of them performs poorly. This has become especially true with the restrictions on the use of Opus, which I considered to be the most balanced.

[velvet tangle]

My personal recommendation:

• normal search: Default
• advanced search (more knowledge needed): GPT-4 with Pro Search
• Coding + Writing: Opus

[velvet tangle]

Please share the thread. If you mentioned in your prompt that the AI should use your file, just leave this part out, i.e. instead of "Tell me what XY is based on the PDF", "Tell me what XY is". It would also be good to know which model you are using: https://www.perplexity.ai/settings/account

[fluid lark]

Is perplexity good for coding?

Wanna cancel gpt 4!

[ionic sparrow]

#⚡│ask-community message

https://www.perplexity.ai/hub/faq/what-is-perplexity-pro

keep in mind, Opus currently has a limit of 50/day

[twin pawn]

Hey @ionic sparrow how are you doing today m8? Question, which model is best for writing news articles?

[velvet tangle]

#⚡│ask-community message

[twin pawn]

Opus?

[velvet tangle]

Yes

[twin pawn]

Awesomeness can't wait. Do you have a writing example of Opus?

[velvet tangle]

Sure...if I can find it somewhere in my 5k Perplexity messages 😅

[twin pawn]

don't you bother ... is the writing quality better than that of CHATGPT 4?

[velvet tangle]

Yeah, in general, Anthropic's models are known for writing in a more natural/human way

[twin pawn]

Thanks m8, that i was looking for. So 50 p day is a bit low. So best is to use Opus only when you really need to use it. For example the research and outline etc with another model and the actual writing with opus

[ionic sparrow]

quite well, you?

[elder coral]

Perplexity says its 'current' but how would i know if a certain document is included? I'm looking in particular for a certain type of filing which are posted all the time. And is there a way to tell it to restrict answers to that particular repository? Just do it with prompt engineering?

[dry vault]

there should be a link. if it shows no access idk what to do. i'm free user btw

[fluid lark]

its okay np this is resolved

[sudden ocean]

hello, are threads in collections persistent (= remain until a user deletes them)? I asked pplx, it said yes, but then was unable to quote or link official sources

[fluid lark]

yes they are

[ancient mirage]

How to visit Perplexity with VPN, not a good question.🥲

[velvet tangle]

https://discord.com/channels/1047197230748151888/1232978227371442237

[ancient mirage]

tks!

[arctic reef]

I am working on an app with the API and have a quick question. What advantage does “llama-3-sonar-large-32k-chat” have over “llama-3-70b-instruct”, other than context window size? When should I pick one over the other?

[velvet tangle]

The model is optimized for conversations.

[muted thorn]

How can I refer to a specific forum in the Reddit section of Perplexity to search for things?

I use claude sonnet

[feral path]

Hi, I am wondering if anyone knows of any Google Chrome Extensions that allow speech to text on the Perplexity desktop version. Anyone use these? Would be very helpful for my circumstance!!

[carmine solstice]

is it possible with Perplexity API to upload a document and query that document (i.e. using Assistant API, or Answers API)?

on a related note, is it possible to use client.embeddings.create with perplexity api?

[icy swan]

hi everyone! Is there an actual Perplexity token or coin? I asked the ai bot and it gave me an LP combo (PEPAI/WETH) but can't find a real price for the supposed token. Thanks

[velvet tangle]

Already answered in #👑│pro-lounge. Not possible 🙂

[velvet tangle]

There are NO tokens/coins for Perplexity.

[carmine solstice]

Is it possible to use client.embeddings.create with perplexity api.
Or is there an example of how to use the "-instruct" models otherwise?

[velvet tangle]

Just use user/assistant/system role.
"What is XY?
Context:
{text_goes_here}"

[carmine solstice]

hmm, ok, I was hoping to get embeddings out of "-instruct" models. I guess I will play around a little to see what works.
Just wish there was a tiny bit of documentation or examples 😉

[onyx locust]

Hello, is the API doing online research even if I don't have a pro account? Cause it seems it's not working for me. On web interface with Pro (free 5 requests/day) I get updated results, from API I don't. Trial research prompt "What will Open AI announce today?". From API it just answer with outdated info, while from the web app is perfect with today event news

[carmine solstice]

I think the model "llama-3-sonar-large-32k-online", "llama-3-sonar-small-32k-online" does indeed make online searches.

But I assume you need to have a payment method to pay for it.

$5/ 1000 queries, upfront

https://docs.perplexity.ai/docs/pricing

[onyx locust]

Yes I added 10$ just to try and have the API key, but unfortunately it seems not to search the web

[velvet tangle]

No, you pay for each online request

[carmine solstice]

"For -online models, in addition to the token charges, a flat $5 is charged per thousand requests"
then this is ambiguously worded IMHO

[velvet tangle]

The online models always search the web.

[onyx locust]

That is my understanding too, maybe I should try with a more specific prompt. But still, for the same prompt web app gives "today" news, api isn't even with online models, just trying to solve this

[carmine solstice]

I don't understand. Did you specify one of ( "llama-3-sonar-large-32k-online", "llama-3-sonar-small-32k-online") in your api call?

Can you share the code you are using?

[onyx locust]

I think I found the issue, I was adding a system message as first message. Removing that, it now works. If I put system message back, it gives outdated info

Model llama-3-sonar-large-32k-online

[velvet tangle]

It‘s difficult to always provide news via the online models, please try to use the ˋafter:2024-05-12ˋ operator

[onyx locust]

Thank you

Yes I understand but it was like "gpt4 turbo announcement" so probably taken from model training data and not from web at all

[velvet tangle]

It is recommended for the online models to only use user & assistant

[carmine solstice]

where is this recommended?

[onyx locust]

https://docs.perplexity.ai/docs/model-cards it's written here, I also missed that before

[carmine solstice]

it says "do not attend to the system prompt." which I assumed meant they ignore it 😉

[onyx locust]

I don't know if you can replicate but I tried 4 times back and forth with same prompt and it seems to not search online properly with a system message.

[carmine solstice]

This still seems quite buggy.
I was passing in a long text to chat.completions API and it complained that my text had too many tokens.
So I passed in a shorter text and API itself now states that prompt_tokens is a very small number.
It seems that API handler is using characters instead of tokens for throwing the error message 😭

openai.BadRequestError: Error code: 400 - {'error': {'message': 'Messages have 46267 tokens, which exceeds the max limit of 32768 tokens.', 'type': 'too_many_prompt_tokens', 'code': 400}}

... usage=CompletionUsage(completion_tokens=429, prompt_tokens=6977, total_tokens=7406))
... usage=CompletionUsage(completion_tokens=377, prompt_tokens=2954, total_tokens=3331))

error message is saying my text has 46267 tokens, but response says that there are a total of 9,931 tokens in my text

[split grail]

How can i add more payment info like company name ID of company etc. I need an invoice for accounting.

[gritty moon]

ChatGPT 4o when will it be available?

[wicked nacelle]

It was annouced 3min ago and you already ask😂

[gritty moon]

Can't blame me for trying!

[wicked nacelle]

It seems gpt4o is even better than opus 🤨 https://openai.com/index/hello-gpt-4o/

[dreamy vessel]

https://chatgpt.com/ is the new domain of the chat dashboard

[kind scroll]

i'm currently on the pro plan. when my sub downgrades to free, will i lose access to my custom collections i already created?

[shrewd knoll]

No, it won't disappear unless you delete it

[alpine stream]

What should I do?

[dull fjord]

I got this too

[alpine stream]

And now I am asked to subscribe but I am a PRO user.

[dull fjord]

I'm getting this too.

[alpine stream]

Where are you? China? using VPN now?

[dull fjord]

Nope I'm in Manhattan

[kind scroll]

did this ever resolve? i'm working on a paper and i'm afraid to refresh now lol

and i am also in Manhattan lol. small world

[alpine stream]

For me now is resolved

[digital nova]

Is it possible to pay the subscription of Perplexity Pro by using company to pay via TT? I am from Malaysia

Anyone can help?

[fluid lark]

where is gpt4o?

[feral rivet]

Its not it the app yet, it will take time for them to make an update, afaik if you change it on teh desktop it will change it for the app

[fluid lark]

Ohh will see thanks!

[feral rivet]

🙂

[elder basin]

Does Perplexity utilize GPT-4o's native vision capabilities, or does it apply its own vision technology on top of GPT-4o?
Will there be any differences in the performance of vision tasks when using GPT-4o through Perplexity compared to using it natively on OpenAI's platform?
I would be disappointed if GPT-4o's vision capabilities are not fully leveraged when using it via Perplexity.

[muted thorn]

You can't personalize anything in the web app. I still get some sources that don't interest me at all. What would an effective prompt look like for Perplexity to use the sources I want, such as Linux.org. I have the prompt in my profile

I dont get it how to do that

And does this only work in the Writing module or also in the Internet module?

An answer to this question will be good

[velvet tangle]

You can add site:linux.org

[muted thorn]

How would you do it

[velvet tangle]

Prompt + site:XYZ

[muted thorn]

Can you give an example please?

[velvet tangle]

"How to use Linux site:linux.org"

(With Pro search)

[muted thorn]

I mean here

[muted thorn]

I mean here

In the profile section

@velvet tangle

[peak lily]

Another fellow linux user 😂

[split grail]

Can I include (BIN, SEBIN, CIN, Business ID) on the invoice? I need this information before making the payment coz while payment i dont see these information.

[short bough]

I'd like to cancel my perplexity PRO subscription. I've have done it and I don't have access to the pro features anymore. However every day perplexity try get 20$ from my card (I have not enough money on that card so transaction is declined). How can I stop that?

[split grail]

Is there a difference between subscription 20$ per month and payment for api? I only need the API.

[fluid lark]

API and subscription are two completely different things. Perplexity is the thing where you have search engine with gpt4, etc. And API is the thing with Sonar and open source models

With sub you get $5 monthly API credits

BTW if you need API for something that's not Sonar (the online model), then keep in mind that soon the most models will be removed

• online models don't really follow the prompt

[past lodge]

true, right now m not able to use the vision features of gpt4o, instead it says gpt4-turbo
but would be nice if its added soon..

[split grail]

I need an API that can access the internet. And search for information.

[versed magnet]

yeah the 2 models with online in their name do that. Can use labs.perplexity.ai to try

[stable halo]

hey, you guys hiring interns for this summer?

[slim night]

Hey @stable halo! At the moment, we don't have any internship positions available for this summer. Feel free to apply to the position you are interested in. The team will contact you if something comes up.

[muted thorn]

@slim night when comes an Update for the Perplexity extension

So AI Companion

[nova lynx]

Is there a way to get chat to ALWAYS output a file for me?
example: i have a food list in a notepad doc, copy/paste into chat, ask it for recipes, but then i say i don't have x, have chat remove it from the list and repost it at the end of the prompt (because i learned it can't write to files on my desktop which would've been easier)

[silk cloak]

Your account has a total of 600 uses per day, which includes both Pro searches and advanced model queries
Can i use 600 uses per day exclusively for LLM, or is it divided in two parts (300+300 for searches and LLM each?)

[ruby sail]

I've been using perplexity.ai for a day and often encounter browser crashes and freezes. I'm using Edge 124.0.2478.97. Has anyone else encountered this?

[deep topaz]

How do I add the page function?

[neat cedar]

Hi, is it possible to use preplexity pro through api and use the gpt4-o?

[velvet tangle]

No

[neat cedar]

what about claude or gpt-3.5 models?

[velvet tangle]

https://docs.perplexity.ai/docs/model-cards

[wicked nacelle]

Hi, why do you force us to take the new enterprise subscription that is twice more expensive for nothing? My company only needs one account and having the billing associated to the company allowed me to have the right invoice for my accountant

If I have a year subscription, will it be cancelled? I asked directly perplexity and it told me
"The key takeaway is that your original Perplexity Pro plan at $200/year should still be available, even if Perplexity is now promoting their new Enterprise Pro plan more prominently. Stick with Pro if it meets your needs, or contact their support for further assistance with your account and subscription."

Does someone else have this annoying orange popup everywhere in perplexity?

[velvet tangle]

The price of Pro is deducted from the Enterprise subscription.

[wicked nacelle]

Maybe it doesn't work well, and I do not want to subscribe for the enterprise version, I want to keep it as it is right now with classical pro subscription

[willow wagon]

Hi guys. How is chat gpt 4o compared to gpt4 turbo? In my opinion, it is much worse than claude sonnet for quick searches, and it is much worse than gpt4 turbo and even more so than claude opus when working without the internet.

[shrewd knoll]

I thought it was okay, https://tatsu-lab.github.io/alpaca_eval/

Based on this a lot prefer this over the rest plus the benchmark in comparison claims 4o is better than Opus, turbo and llama-3

[dreamy swallow]

it's kind of like a turbo version of gpt-5 🙂

[hardy granite]

Any comparision page to find out which model suits best for which purpose ?

[daring elbow]

Any help on using the perplexity app on android? When I use speech input, I can't do any pauses while speaking, which I kinda need for longer requests. So the prompt is cut off half way through, or I have to talk super fast..

[thorny frigate]

That is a bug with code. The Browser tap is freezing.

[velvet tangle]

What browser are you using?

[thorny frigate]

Edge Browser

@velvet tangle This happens every time, every time I want a code. A longer code. Afterwards, the tab still freezes, but the process continues to run in the background.

[velvet tangle]

Any extensions or VPN?

[thorny frigate]

Edge Extensions and i dont use a vpn.

[velvet tangle]

Please disable adblock and try it again

[thorny frigate]

I actually only have this problem with Perplexity. With no other chat Ai. And this has been the case for months.

Okay i try.

[ancient fjord]

Hi, do we have a daily or hourly limit for gpt4o? I didnt see anything when i selected it. Thanks!

[velvet tangle]

600/24h

[soft sentinel]

Can you only use other people's referral codes for your first month and not succeeding ones?

[versed magnet]

it seems it can be a bit more prone to hallucination tbh

like it's a better model, don't get me wrong.. but i feel it needs to slow down lol
in the perplexity environment it seems it will blurt stuff out just to satisfy the context of having to give an answer, where Opus will actually say "sorry, there is no information in the search results to..." - which is much more preferable ofc

[soft sentinel]

Respectfully, this does not answer my question

[shrewd knoll]

Using referral codes can be old and new user

[velvet tangle]

Hey @soft sentinel!
You can use the referral codes only for your first month. However, if other new subscribers use your code, you will always receive the bonus on your next payment. In other words, if 4 users use your code, you will receive the next four months of Pro for $10 each.

@shrewd knoll Please stop sending your referral link into the chat every time someone uses the word "referral". Thank you very much!

[soft sentinel]

How unfortunate oh well

[shrewd knoll]

K sorry

[halcyon pelican]

Is there a way to name my thread?

[small prairie]

Click at the top center

And change the name

[halcyon pelican]

Nothing happens

[small prairie]

The one to the right of that

You clicked the collection name

It goes collection/thread

[halcyon pelican]

Thanks. It worked

[small prairie]

Np

[halcyon pelican]

Is there a way to edit the output of perplexity?

[small prairie]

Nope

[cyan apex]

Perplexity only seems to be displaying my most recent threads. My older threads and all of my collections have disappeared. Any suggestions?

[haughty brook]

try to clear your cache or log out and log back in. and make sure you’re logged in to the right account and didn’t use a different sign in method with the same email address

[remote cedar]

When I copy paste text that's a little long into the pxplty mobile app. It doesn't allow me to scroll down to further edit the prompt text or to press enter. Anyone else faced this issue?

[ionic sparrow]

#1216589575107903588 message

[halcyon pelican]

is there a way i can filter perplexity to only give me response from latest results from the internet? For example, when i searched which ai model it's using, it told me gpt3.5. The reason is, it read from old site. Now i want it to read from current sites alone. Is that possible?

[potent jolt]

Compared to ChatGPT, I sometimes get the impression that Perplexity doesn't remember the context of the previous answer in a thread. If I ask a follow-up question it seems like it's treating it as a new question. Using GPT-4o and I have a Pro subscription. Am I missing something?

[digital spindle]

I'm not sure how can I make sure that perplexity.ai does not retain my data? i switch the toggle but it is not clear whether it is off or on

[velvet tangle]

This means off:

[digital spindle]

Thank you so much!

[foggy tiger]

How can i be sure perplexity is not going to be detected on an ai detector for school asignements

[feral rivet]

You can’t be sure, there are so many AI detectors, and even tho some are inaccurate I can’t say that they don’t have an accurate one.

If you want to play it safe, use perplexity for research and feedback and write it yourself. AI is best as a way to accelerate learning imo🤓

[finite sigil]

it will be

changing system prompt could help but ai writing regardless of model usually pretty identifiable

[severe fable]

how does perplexity evaluate it's sources?

[hollow compass]

Any way to get access to the pages beta anymore? Missed that post but would love to see it.

[fluid lark]

I have.

[feral rivet]

This one has been a known bug for a little while

oh, but looks like it was fixed at some point, maybe it broke again? 🤷 https://discord.com/channels/1047197230748151888/1232834666063728741

[haughty brook]

Are you using ios or android?

[sick tinsel]

it's possible to send perplexity search on discord ?

erf, search on specifique server on discord.

[autumn estuary]

What is the slowest AI model? I'm not in a rush. I want to see what the slowest model says.

[velvet tangle]

Opus should be the slowest

[autumn estuary]

k ty

[velvet tangle]

https://discord.com/channels/1047197230748151888/1237163951750385705

[icy merlin]

Hey, how does Perplexity fetch the data from the query?

does it use LLM model? or scrapping results?

[pale horizon]

Is perplexity free users benefits from Gpt 4o or not? And why not give them the free daily usage quota instead of nothing?

[fluid lark]

I don't think gpt4o is available for free users , I might be wrong tho. and what do you mean by free daily quota?

isn't Pro Search just that?

[pale horizon]

Yes I think it’s still 3.5… I mean 4o is supposed to be free but for a limited of queries like the pro search but more than 3 like currently

[versed magnet]

yeah but that's chatgpt (and i'm not sure is even rolled out yet, like free gpt4o usage)

perplexity uses the openai api to provide access, which still costs money

(though is a lot cheaper than the old gpt4, and Opus)

[feral rivet]

I have heard from a few peeps that GPT4o is free now on ChatGPT, but they tell me it’s like 5-10questions per day, wonder if it’s bc it’s still being rolled out or if it was always the plan to just give a taste test

[ionic sparrow]

#👉│feedback-general message

[lethal geyser]

Hi , if I cancel my monthly pro sub , will I lose all my referrals I already had ?

[pale horizon]

So that confirms it … what a shame, I mean at least put Clair Haiku instead… or maybe it’s not relevant to have heavier and slower models ?

[shrewd knoll]

You won't lose your previous referrals but you won't get new referrals unless you subscribed

[lethal geyser]

1 more question if you know . What is the context window of opus and gpt4o on perplexity?

[radiant violet]

is there a way to increase the output length

[feral rivet]

Context window is about 32k, some say a bit less, we don’t know for sure when there rag system takes over

[feral rivet]

If the output gets cut off the only option in perplexity is telling it to continue where it left off

If you are trying to get the AI to make longer outputs and it’s not an output limit problem, there are techniques that work

[pale horizon]

Is that an official source ?

[ionic sparrow]

community mod

[lethal geyser]

That's too low. Opus should have 200k context window especially they set limit to only 50 msgs per day . I would understand this low context window if we have unlimited msgs . They should reconsider this

[pale horizon]

Thanks for sharing that then it only make sense to use the most advanced free LLM out there for better customer experience and retention, that why I’ve asked about 4o coming soon but with restricted daily usage

[radiant violet]

I'm using Opus 3 - Getting limited response output.

What are the techniques.

[feral rivet]

What type of stuff are you having opus write, is it essay type stuff, or something else?

[radiant violet]

Financial data - case study - content & reserach paper creation

Blogs - Articles etc

[feral rivet]

So I don’t know about some of those topics, but I will share what I do know, and hopefully it will be helpful

[feral rivet]

@radiant violet Sorry, took a minute. Here is a prototype system prompt I just made for recursive writing, its one of the methods that can be used for longer outputs and more in depth content and can be customized for different uses: ```
You will recursively provide written work, following these steps:

1. Give a comprehensive outline as a table of contents with chapters, sections, and subsections, should be 5 chapters, 4 sections per chapter and each section should have 4 sub sections.
2. Recursively expand the work each time, showing more depth and more detail. Each recursion should be a full re-write, expanding on the previous one. You will provide more detailed and longer sections with more depth for each part.
3. Show what step you are on by putting the number between arrow symbols like <1>, <2>, <3>, <4> at the beginning of each recursion and </1>, </2>, </3>, </4> at the end of each one. This will act as a way for you to remember how many recursions you have done. You will do a total of 4 recursions.
4. Each chapter must have 4 sections, and each section must have 4 paragraphs. Make sure to number these correctly to ensure you follow this structure.
   Remember to maintain consistency in formatting and numbering throughout the recursive writing process.

[radiant violet]

Fantastic

[feral rivet]

Changed the prompt quite a bit and got this: https://www.perplexity.ai/search/Anatomy-of-frog-FXhaZ_SHQhuO1x3NAbVX4Q#4

Give a comprehensive outline as a table of contents with 4 chapters, 2 sections per chapter, and 2 subsections per section. The table of contents should look like this:

Chapter 1: [Chapter Title]
1.1 [Section Title]
1.1.1 [Subsection Title]
1.1.2 [Subsection Title]
1.2 [Section Title]
1.2.1 [Subsection Title]
1.2.2 [Subsection Title]
Chapter 2: [Chapter Title]
(same structure as Chapter 1)
Chapter 3: [Chapter Title]
(same structure as Chapter 1)
Chapter 4: [Chapter Title]
(same structure as Chapter 1)

Use the table of contents as a guide for writing the paper. Do not make changes to the table of contents once it is created.
Recursively expand the paper's content, showing more depth and detail with each iteration. Each recursion should be a full re-write, expanding on the previous one. Provide more detailed and longer sections with more depth for each part. Show what step you are on by putting the number between arrow symbols like <1>, <2>, <3>, <4> at the beginning of each recursion and </1>, </2>, </3>, </4> at the end.
In the final version, each subsection must have 4 paragraphs. Make sure to number the paragraphs so that there are 4 paragraphs per subsection.
Maintain consistency in formatting and numbering throughout the recursive writing process and ensure a readable format.
Perform a total of 4 recursions, with the final version being the most detailed and comprehensive. The approximate length targets for each recursion are:

Recursion 1: 1-2 sentences per subsection
Recursion 2: 1 paragraph per subsection
Recursion 3: 2-3 paragraphs per subsection
Recursion 4 (final): 4 paragraphs per subsection

Remember, the goal is to gradually expand the content while keeping the overall structure defined by the table of contents. Focus on adding depth and detail with each iteration rather than increasing the number of sections or subsections.```

@radiant violet 13 pages with 1 prompt, not bad 😅

[clear cipher]

Does the context of a thread consider also all the library? I explain : I have worked and ask a lot of things around coding python in a lot of different threads. Now when I ask a completely different question in a new thread it answer using python code even if I didn't asked him to do so. I would like to point out that I don't have any profile prompts that could explain this behaviour.
For example :
https://www.perplexity.ai/search/give-me-a-uzMFfpmZTGmRBqpIL5RLJg#0

[feral rivet]

no, 23 with double spaced 😎

[feral rivet]

No, the context is not shared by the library/collection, only the context in the same thread

[clear cipher]

thanks! Should I report it as a bug?

[feral rivet]

I think it has to do more with the fact that the AI was trained on a lot of python, if it was some other language it would be a little more weird, when you use chatGPT it uses python to solve a lot of problems

python is the go to language for most AI's

[clear cipher]

Hmmm I see but I don't understand why does not he gives me only the mathematical function... It has way more sense

[feral rivet]

Oh, like why it’s giving python vs mathematic formulas

I’ll look again

I think the word “function” might be the key word that kicks it into coding mode

When I use your same prompt I get python, but when I switch it for a different word I get this: https://www.perplexity.ai/search/give-me-a-DaV4dfsZTGOdnajuBYgXDQ#0

[clear cipher]

Oh okay that should be the problem then, thank you so much! 🙂

[feral rivet]

You’re welcome! 🙂

[radiant violet]

Damn, If thats true that just saved me a lot of work.

Thanks a million

[feral rivet]

Your welcome! BTW if you are in writing mode it works best, but search mode works too, I would just only have pro mode turned on for the initial message, bombarding it with 20 sources each time it gets cut off can confuse it. Hopefully you can fine tune the prompt or use a similar technique for your use cases.

Also if you look at the thread I shared you can see how I tell it to continue when it gets cut off

[radiant violet]

Thats a good tip, I haven't tried that. Makes sense, once you have the data you need just switch that off.

Cheers, I've saved that prompt

[spark pier]

what is the best way to interact with 4o in the app, similar to how it is in the GPT app?

[feral rivet]

In the App Writing mode with pro togled off is similar to using the chatGPT app, its just a normal chat without web search.

[spark pier]

what about conversational like 4o?

[feral rivet]

Are you talking about the upcoming voice conversations feature from OpenAI? Or the current version that just standard TTS? bc currently the voice chat in chatGPT is just TTS and STT. the one they announced is a new type of voice conversation that is currently not rolled out, it uses a new Omni GPT modle with voice gen baked in

The first one is a back and forth where you can talk as soon as its done, the second is like a normal thread but you can have it speak the answers

If you are talking about regular voice conversations then there are 2 options, I will show both:

[spark pier]

I was playing with the new Omni in their app, it was like the demo. But without the sub plan, I was limited in vision

Sorry, it was 3.5 I was interacting with

[tight ridge]

What model would i use for an API to live fetch and to return vehicle data values (weight, height)?

[wary kestrel]

Any plan for perplexity to add Gemini on the ai lineup soon?

[hollow ginkgo]

@elafella_28257 warned

Reason: Spam, please avoid sending single word messages
Duration: 7 days

[fluid lark]

Is this on your screen? Sounds amazing!

[feral rivet]

Yep! i just screen recorded the IOS app

[fluid lark]

I can hardly wait for this to come to Android. It's such a beautiful interface.

[feral rivet]

Yeah, I like it a lot, hopefully it dose not take them too long ⏱️

[ivory anchor]

iOS and macOS always gets the love first...

The OpenAI GPT4o desktop companion thing is currently macOS only too :S

[haughty hatch]

hi

[feral rivet]

👋

[haughty hatch]

what is love?

[feral rivet]

https://www.perplexity.ai/search/It-is-love-XLjJB8VwTJ.cROpQjARj1g

[haughty hatch]

what can you help me?

[ivory anchor]

baby, don't hurt me?

[bright moth]

I'm coming back to Perplexity and I have a weird one. The autoscroll on long responses doesn't work for me.
Unless I'm just absolutely confusing it with other services I'm pretty sure even PPLX would keep the window scrolling down as a response streams but I have tried it several times and I can't get it to work anymore. I'm also using Perplexity as a custom search engine in Chrome if that's somehow related to the problem.
I can live with manual scrolling but it's weird and also kind of prevents me from just sort of entering the query and switching windows without having to switch back and scroll a lot.

I know SillyTavern for example will autoscroll by default unless the user 'breaks' it during streaming a block/chunk. Doesn't PPLX work the same way? I am almost certain it used to but it will just happily scroll underneath the input field and create a scrollbar, even when I have not scrolled that conversation.

On Chrome 126 Beta and using uBlock Origin

[bright moth]

I'm a dopus, this appears to be a feature the Perplexity UI doesn't have.

Mystery solved 😝

[hardy granite]

Is perplexity getting slower and slower each passing day ? I am observing this a lot, raised concerns in bug-reports but isn't addressed yet.
Anyone else facing this ?

ChatGPT free version is giving super fast responses, this wasn't the case earlier

[ivory anchor]

Free version is 4o by default currently I believe for chatgpt's site.

If it's 3.5 then color me surprised; if you're on 4o then that's how it do.

[feral rivet]

I believe you get about 10 msg per 3-4 hours of GPT4o then it switches to GPT3.5

[fluid lark]

i wish they brought the tts option to android

cant be that hard

[ornate valve]

Important question : how do I get Perplexity to access and read an opened bookmarked chat session from the past ? And how do get it recognize / read a thread opened from the library ? This is crucial for my work with Perplexity as I want her to be able to answer within the context of the whole text of a reopened older thread - there must be a way to establish continuity / connection to older topics form a current session ! THANKS !

[runic steppe]

thread collections can be used for custom propmpts (like custom GPTs), but the AI prompt has a really small limit 😦 why is that? can it be increased? is there any other alternative??

[lilac cobalt]

Question, when can i share with perplexity images files? (in free version)

[ornate valve]

My key question is about finding a method to maintain persisting contextual continuity and awarness of large data sets over a longer period of time for long term projects / to create a persistent knowledge base** or framework ("RAM") that can be loaded and built upon with each new session, rather than starting from scratch every time. A way to achieve this could be to upload the dataset to file hosts and guide / link Perplexity to these files at the begin of each session deiling with the same long-term topic. But in my case Perplexity was not able to parse the files parked on a file host. ** Any comments / toughts / solutions ?**

[velvet tangle]

That's not possible

[velvet tangle]

Image upload is only for pro users

[ornate valve]

Thank you IceLavaMan. Also** not possible** at all for Perplexity to** access content uploaded to file hosts ?! Is the only way to mantain persisting contextual continuityby the data written into the profile **?

[night jasper]

I just got pro version and I kinda when ham on image generation I didn’t know there was a limit when dose it reset lol ?

[ionic sparrow]

50/24h

[smoky blade]

I setup the action button on my iPhone 15 Pro to launch Perplexity voice search. Is there a way to get this to use pro search instead of quick search?

[ornate valve]

Topic : Image database parsing. Question : 1) Is it possible for Perplexity to ingest and parse a larger external database of hosted bitmap images? ...2) How ?

[twin pawn]

So i was planning to jump over to Perplexity Pro but than came GPT 4o. Now Im not sure to make this decision 😦. Anyone?

[halcyon pelican]

what happens if i switch on pro search when am on writing mode? Does it make any difference?

[livid pilot]

How can I use this great presentation function? The function is still in the beta stage?

#💗│sharing message

https://www.perplexity.ai/page/Alienn-vs-Stereopanic-LQIw3OTcTpGJHIaFIlsfzQ

[ionic sparrow]

#📢│announcements message

still in beta

[small prairie]

GPT 4o is already on perplexity, for text.

If you want the other modalities, which haven't come out yet, then ChatGPT is probably the way to go.

[twin pawn]

Thanks for your reply. To me the most important is writing AP-style news articles I don't care about images. Please advice because my subscription ends the 28th

[fluid lark]

can someone who is on android show me where the text to speech option is on the app? because i cant see it, and yes ive got the latest app and latest software for android too.

[velvet tangle]

Are you subscribed to Pro?

[fluid lark]

yes i am @velvet tangle

[velvet tangle]

You have this?

[fluid lark]

definitely not no

[velvet tangle]

How does it look for you?

[fluid lark]

what am i missing?

[velvet tangle]

Could you please tell me the exact version?

[fluid lark]

2.18.2

thats the latest i can update from app store

[velvet tangle]

You'll get the tts feature with 2.20.0. Please be patient 🙏🏻

[fluid lark]

ill find the version online and install it manually

thanks for letting me know!

i got it now! 😄 yay

[cloud sage]

How do I get rid of the “Discover” feed on the iOS app?

[velvet tangle]

You can't disable the discover feed

[remote zephyr]

Hey, how do I log into the chrome extension? My perplexity account is not registered on my gmail but on my main outlook account. Can't find the "log in" option on the extension.

[haughty brook]

the chrome extension will just use the account that is currently logged in on the browser

[signal cairn]

Has anyone had experience with or know how to link Zapier with Perplexity API?

I'm trying to use POST in Webhooks by Zapier to create a chat completion, but I keep getting errors with all the different things I've been troubleshooting

[wild stream]

i joined via the pro link and dont have access to pro channels any fixes ?

[velvet tangle]

#💬│general message

[wild stream]

yea ive tried this and still the same :/

[wild stream]

ill pm

[cloud sage]

FYI for any Perplexity folks following this channel: Unkillable "sponsored content" in a piece of software that users are expected to pay $20/month for is shady as hell.

[half yew]

Hey guys I am new to perplexity. Anyone here knows in the $20 plan what is the cap/limit per time like? Also can I ask it unlimited image plus text analysis messages? It mentions so!

[remote zephyr]

hmm unfortunate, these are different accounts (browser = gmail, perplexity = outlook). No way to change that?

[velvet tangle]

The extension will use your normal Perplexity account (Gmail)

[small prairie]

You could also try contacting support.

[haughty brook]

with pro you can use the advanced models 600times per day. and you can at least analyze 100 files a day.

[candid hull]

hey, in mac. How to quick switch All and Writing?

[ivory anchor]

Except opus

Opus is 50

[peak lily]

Does profile information prompt get used even in the Android app?

[ivory anchor]

Yes

[half yew]

Woah, so this is image uploads part of these 100 file uploads? Or is it part of 600 times per day? Also for Claude sonnet model how many time per day I can chat with image analysis!

[feral rivet]

You get 1000 file/image uploads a day. I have 997 left for this 24h "upload_limit":997

and 600 messages with pro models

and unlimited with the free model

[keen furnace]

is there a way to abandon, or otherwise remove and stop Perplexity from referencing a file originally uploaded, when it might be days or weeks out of date based on time of conversation or general new information, moving forward in a chat?

[feral rivet]

You can delete a source

[peak lily]

Thanks

[keen furnace]

Thank you. It stinks in some cases when the LLM slavishly refers back to a file or photo like 100 messages later when the conversation has evolved and pivoted WAY beyond the initial query/prompt

[feral rivet]

Claude seem to present the source to the AI once and then it eventually gets lost in the history of the chat, while Perplexity keeps it near the front of the memory, often having the AI prioritize it over all other context, this is great in some cases and bad in others, i would like it if they gave both options as an easy toggle, sometimes you just have a quick questionnaire about a file and it dose not matter after that, and chatGPT and Claude be good for that, but if you have a document, that you need lots of answers from, then perplexity is better IMO, I could upload a textbook, manual, or documentation and ask it as many question as i want.

[keen furnace]

This is a great idea, you should post ot up in the ideas roadmap

[signal thorn]

Hey guys does perplexity have plagiarism because it is taking it from a source?

[cosmic iris]

Is perplexity able to output spreadsheets? I just uploaded one to the platform as part of a chat and got a response, but I don't see where the actual output file is...

[slim night]

Hey @cosmic iris! Currently, it's not able to output spreadsheets.

[cosmic iris]

Thanks, Alex. So what is the best way have the tool help me with a large data set? I want it to read ~750 rows, analyze the text in each cell, and answer a specific question with a "Yes" or "No"

[nova lynx]

anyone have a way to get chat to be EXACT with things?
like instead of "it'll be hot out today" say "it'll be between x and y degrees today" or "generally it's hot but i don't have measurements"

[burnt spear]

Hey, this might be a stupid question. But is there a keyboard shortcut to open the ai companion on chrome? I'm on mac

[open temple]

does the API return live websearch results when asking the ai a question?

confused reading the docs if its just a LLM thats not web based

[dusky flower]

Can i use Perplexity here to ask questions and get response like chatgpt ?

[velvet tangle]

The '-online' models use real-time data from the internet to answer the question

[open temple]

awesome thank you so much

[ionic sparrow]

https://www.perplexity.ai/hub/faq/what-is-search-focus

[naive nacelle]

I am trying to create an image using another image of a person as reference, to try and get this embedded in to the image - is this possible. To give context, it's a friends face, but on to a super heroes body in a specific style

[tough steeple]

So I got free 1 year sub with nothing phone. I cancelled auto renewal and now whole subscription is cancelled?

[shrewd knoll]

No, you can still use it until it expired

[radiant violet]

Is the Tako integration available today?

[ionic sparrow]

#💬│general message

[radiant violet]

How does one save the Tako chart not in html? any streamlined way?

[atomic hill]

When AI DATA RETENTION is turned off, is Perplexity to safe to use with confidential information?

[ruby sail]

In the chatbot-arena-leaderboard rankings, GPT-4o-2024-05-13 is ranked first in all evaluation categories, but why does it perform far worse than GPT-4-Turbo in actual usage on perplexity.ai, and GPT-4o often misunderstands the question content?

[obtuse apex]

That's hard to answer - it depends on how perplexity.ai is using the model, but it also could be the model simply doesn't 'vibe' (while still being accurate).

[ruby sail]

What are the advantages of GPT-4 Turbo compared to GPT-4o? This question can already be seen, the GPT-4o model sometimes misunderstands it as GPT-4 and the answered content is also wrong

[buoyant plank]

Faster and gets more things wrong.

[tough steeple]

@shrewd knoll that should be what's happening. But i have recieved an email which says..

Hi there, we've just canceled your Perplexity Pro subscription and downgraded your account.

[shrewd knoll]

!??? I think you need to speak with them about this case

[night jasper]

Is there a way to connect Perplexity with Notion and have them work together meaning have Perplexity work in my Notion? I have an idea why

[night jasper]

I meant I have an idea is why I’m asking. Sorry I’m using voice dictation lol

[ionic sparrow]

deep dive ➤ #💬│general message

check Tutorial III (implementing automations using Perplexity and Zapier)

if you have idea/proposal https://discord.com/channels/1047197230748151888/1047649619695390740

[halcyon edge]

Why is there a prefix number (1) after every sentence? Please tell me how to fix it so that the number 1 doesn't appear anymore.

[balmy crater]

I believe that's the source marking, does it show a source when you hover over it?

Question: how do you get the Tako knowledge cards? I have pro, using web version in US, does it only work with certain models or pro search? It keeps saying
I'm unable to create or display real-time graphs or access data directly from databases or external sources that are not provided in the search results.

[lost pawn]

I use collections extensively, inside a conversation will Perplexity consider the history of the conversation and the instructions from the collection when answering a follow-up question?

[crystal dune]

Is perplexity working for you now?

[pseudo narwhal]

hi
can i limit the response length of the model via a api call?

[fallow sorrel]

If you guys had to choose for the two best Ai chat bots which ones would you choose between Chatgpt, Claude and Gemini? Or maybe other Ai models altogether?

[small prairie]

Probably ChatGPT

Claude doesn't have enough features IMO

And Gemini is too filtered on certain topics.

So depends on what you are planning to use it for.

[feral rivet]

As a second option if you were picking a long context option would you do Claude Pro 200k or Gemini Advanced 1M?

[small prairie]

Probably Gemini 1.5 Pro, just because of cost

[feral rivet]

Like API cost

[small prairie]

$7/$10.50 per 1M Tokens for 1.5Pro compared to:
$15/$75 for Claude Opus

So the input is half the cost.

The cheapest input is 4o though at $5

Just doesn't have the longest context.

[feral rivet]

Yeah, I was talking about “Gemini Advance" vs "Claude Pro". not api cost

Over api definitely Gemini

Opus is just too expensive

[small prairie]

Also depends how much context actually matters.

Since for a lot of people GPT4o would actually be the better option.

Since using over 128K context is rare

I wonder if they will boost the context length in their next event.

[feral rivet]

Yeah, is “chatGPT” still 32k?

If they raise it to 128k I will be in heaven

But I always have API, lol

[small prairie]

For Plus it's 32K, for enterprise it's 128K

Which is why I'm on enterprise

[feral rivet]

Ahh, yeah I don’t have enterprise, a little too much money, haha

That’s nice tho, is it your most used AI platform?

Like more than Claude Pro and etc

[small prairie]

Yep, I definitely use it a lot more than the others.

[feral rivet]

I was thinking about cutting my subscription down and doing most API with msty, I like having them all in one place, and having prompt library’s built in

[small prairie]

And I've got enterprise because I bought it for my whole businesses.

[small prairie]

Yep, the cheaper they become, the better that will be.

[feral rivet]

Oh nice

[small prairie]

At a certain point when the price drops enough, you will save money by using the API.

[feral rivet]

Yeah, with GPT4 and Opus being my go to there was no way I would do API, but with GPT4o with most things and Gemini for long context, and Opus occasionally for creative tasks, I doubt I will spend anywhere near $60 a month. But we will see, I might surprise myself with the amount, haha

[small prairie]

Yep, and each year it is getting cheaper.

More like every 6 months to be honest.

[feral rivet]

Yeah it’s really great!

Do you think that OpenAI (Closed) with let you do “her” over API?

[small prairie]

I think in a years time, it will officially be cheaper to use the API, than pay for the subscription.

[small prairie]

Yep, most likely.

[feral rivet]

Yeah, I am going to do API for a bit and see if it is cheaper for me, if not, I will just do the subscription until it is

[small prairie]

You can already do so for the current version on:
https://platform.openai.com/docs/guides/text-to-speech

[feral rivet]

Or I might find I like the sites better and it’s worth the difference, or maybe msty will be amazing for my workflows

[feral rivet]

Yeah I have used the TTS models, they are good, but not as good since “Sky” got removed

lol

[small prairie]

Yep, just waiting for them to release the new version which should be real time.

Not sure about the cost though.

[feral rivet]

I imagine the cost will be higher than using it though the subscription, but over time they will lower it, maybe due to competition adding great real time voice convos

I hope Anthropic is working on something great, have no faith that google will have anything for the next 2 years

✨ “Prototype”✨

Or meta, would be awesome to have Llama 3o

[small prairie]

The problem with google is that they kill stuff all the time

[feral rivet]

Yes, they kill so many promising projects

[small prairie]

Yep, as a business owner I can't rely on google, lol

[fallow sorrel]

I'm paying for ChatGPT and Claude because I think they both have qualities and flaws but was kinda wondering if Gemini could replace any of those. I don't do any coding.

[feral rivet]

Don’t they have a free trial, give it a try

If it great than use it, if it sucks than cancel it

[nova lynx]

Anyone have good tutorials on building and using GPTs?
I watched Andrej on yt, a few videos, very good overview, but looking for something more hands on and maybe more recent

[feral rivet]

For system prompts, Claude has some great resources, here is a their prompt library: https://docs.anthropic.com/en/prompt-library/library

And here is there guides; https://docs.anthropic.com/en/docs/prompt-engineering

[feral rivet]

^^^

[nova lynx]

thsi is fantastic thank you!

[serene ember]

I received a code for a perplexity pro sub, but it was only good for new accounts. I had a sub already, which I cancelled and made a new account to use the promo code. I had threads that were organized into folders in my old account. Is there any way (other than just manual copy and paste) to get those into my new account?

[radiant marten]

Bumping and adding to your question: Does AI data retention off on Pro accounts offer the same "Increased Data Privacy" as the Perplexity Pro?

[rocky canyon]

👋 Does anyone know if the profile we set on our account is used for the questions made through the API?

[feral rivet]

They are not, they are separate from each other

[half void]

I think yes, they have already mentioned this question in faq
https://www.perplexity.ai/hub/faq/what-data-does-perplexity-collect-about-me

[hallow flame]

Nice to meet you! Quick question: do you have a Japanese channel?

[shrewd knoll]

No, All American Express

[hallow flame]

Thanks, bro!

[crystal dune]

Hey, I dont see the Mistral model from the Model Selection section. Any reason for that?

[shrewd knoll]

That AI model can't compete with other AI models now

[turbid quartz]

Anyone else having trouble with the server at the moment?

[haughty brook]

seems to be working fine, please refresh the page and try again

[ember shoal]

Is there a way to get a perplexity "Collection" to only search a specific domain?

For example, something like only considering sources found with site:youtube.com ?

[flint basin]

hi i want q reliable and clean windows llm gui which is compatible with pplx.ai. do you guys have recommendations?

[versed magnet]

I don't think so. it was once possible, but last time a tried that was no longer the case

[crystal dune]

Can you advise on how to pick up a model other than the default one?

[turbid quartz]

Thanks. Seems to just be having trouble with uploads right now, but otherwise has ironed out!

[ember shoal]

Weird. Thanks. Would be such a cool tool to avoid sources that you dont want.

[versed magnet]

yes agree. even better would be like a blacklist/whitelist for collections, where you could add domains. not have it part of the prompt, but actually integrated with search process

there'd be a bunch of use cases

[ember shoal]

It seems the collection prompt is only applied AFTER ths search, if at all most times. Sorta weak.

[versed magnet]

yeah they're two separate systems/models - the one which writes the response, and the one which retrieves the web results. the search one only 'sees' each query/prompt you send during the conversation, but it isn't exposed to the system prompt/instructions at all

as far as i can tell anyway

[ivory anchor]

I can't seem to get perplexity to output a code block in markdown. Any tips?

[versed magnet]

could try adding something like this

```markdown

Your response should be enclosed in a codeblock like this

```

but that's for the whole response.. might be trickier if just certain parts of the response that need to be in the code block

[ivory anchor]

I'll try that ty

[cunning light]

Just noticed a new "voiceover" option on my perplexity android app. I tried prompting via voice, but I'm still only getting a text response. Is this working for everyone else?

[rich sage]

After asking your question (use the microphone), you will see a headphone icon to the right of "answer". Click on it to read the answer.

[slim night]

Hey @ember shoal, please take a look at this thread: https://discord.com/channels/1047197230748151888/1126524614449832146

[slim night]

Hey @ivory anchor, please take a look at this thread: #1196389281506263120 message

[ivory anchor]

tyty

[timber ledge]

hi there

i set up perp as my default search on arc browser, but even tho im a pro perp user, my questions are never being asked as a pro when asked thru the arc browser's default search. Help?

[ruby sail]

This upgrade is very poor, when I enter a long text in the input window, the send button is gray, when I make the article into a txt attachment and upload it, it prompts that the file is too small. I'm very angry and very disappointed with the development team. This will make me not want to continue to renew.

If this is a bug, please have the development team fix it urgently.

[ivory anchor]

That is odd.

[ruby sail]

I just tested it. The attachment upload is okay. But in the input window, the function of converting long text to attachment has disappeared.

[sinful lagoon]

How can I stop publishing a page? (I published a page using "connect" button.)
Only deleting?

[rigid helm]

Hey,
what exactly is the "turn into pages" and can everyone just find that afterwords or is it still a private thing exept for everyone who has the link. Cant find anything in the FAQ

[rich sage]

Change the link to this one in ARC: "https://www.perplexity.ai/?q=%s&copilot=true"

[summer cape]

Same question here from me 🙂

[ebon lance]

Is there some process which removes previous queries from your library? I recall having a thread with multiple back and forth queries over the last few days however they have all seemed to vanish when I try and find them now?

[versed magnet]

not sure but fwiw i don't think converting a thread into a page (or just creating a Page) would by itself create something that is publicly accessible (with or without a link) because it starts as a 'Draft'. But after pressing 'Publish' I'm not sure. I would say though that it is effectively publicly accessible from that point

I'm not aware of any actual public search directory, so I'd guess without actually sharing the link, it's prob not easily findable/accessible

[rigid helm]

Thats what Im "afraid" off
I think, the page converter makes request look so awesome, so I would like to share it with friends, like planing a trip, but dont want locations posted publicly

[versed magnet]

though it does some like some search engines have a few in their index

[rigid helm]

So I hope, someone from the team could answer that for sure, so I know weather to publish stuff or not ^^

[versed magnet]

I think if you wanted to share a Page with others, you likely would need to publish it.. I don't think it would be something that could be easily found or accessed without the link, but if the content is sensitive / personal and you want to it stay that way no matter what.. i think not publishing might be the sensistive option..

but yeah ig wait until someone from the company explains properly.. this is just me speculating

you could create a pdf and share that... if you really, really wanted to share it with a few a people without "Publishing" it online..

not quite as pretty though ha

[rigid helm]

maybe thats a temporary solution
But won't have many things to share anyway
But thats a good idea, thanks ^^

[cunning light]

Thank you!!

[lone wave]

Hello, I saw someone posting here that allows them to use the safari share shortcut on iPhone to summarise an article quickly. May I know how to get that? Thanks!

[wide maple]

Hi, can perplexity stop making up fake quotes/references? Or it's inherent problem of AI?

[versed magnet]

https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

but yeah ideally it shouldn't.. or at least should be minimised (that's the idea of having sources.. but still a WiP)

[feral rivet]

You can minimize it, but it will become more rigid the more strict you are, here are 2 example system prompts that are designed to reduce hallucinations:
1:

You are a meticulous and highly reliable expert tasked with answering my query using ONLY the information provided in the sources. Your primary objective is to provide accurate and verifiable responses, grounded solely in the given evidence.

<constraints>

* Hallucination Prevention: 
    * You are strictly prohibited from generating any information, claims, or conclusions that are NOT explicitly supported by the provided sources.
    *  If a piece of information is not mentioned in the sources, DO NOT invent it. Instead, clearly state that the information is not available.
    *  Avoid making inferences, assumptions, or generalizations that go beyond the scope of the given evidence.
    * Do not draw on any external knowledge or previous conversations.  Your knowledge is limited to the provided sources ONLY.  

*  Evidence-Based Responses:
    * Every statement in your response MUST be directly supported by specific evidence from the sources.
    *  When presenting a claim or conclusion, explicitly cite the source and the relevant excerpt that supports it. 
    *  Use direct quotes from the sources whenever possible to demonstrate the basis for your response.
    * If you cannot find direct evidence to support a claim, clearly state that the claim is not supported by the sources.  

* Transparency and Limitations:
    * Be transparent about the limitations of your knowledge. 
    * If the sources are insufficient to fully answer the query, acknowledge this clearly. 
    * Do not attempt to answer questions that go beyond the scope of the provided information.

</constraints> 

Now, please answer my query, adhering strictly to the constraints above: 

2:

You are a meticulous AI research assistant, tasked with providing accurate and grounded answers based solely on the information provided in the given sources. Your primary directive is to avoid hallucinations – generating information not explicitly supported by the sources. 

Adhere to these strict anti-hallucination guidelines:

1. Every statement or claim in your response *must* be directly supported by evidence from the <sources>.  Do not introduce any information, opinions, or interpretations not explicitly found in the provided material. 
2. When making a claim, identify the specific quote or passage from the <sources> that supports it.  Use in-text citations or direct quotations to clearly indicate the source of the information. 
3. If the <sources> do not contain sufficient information to answer the <user_query> fully,  clearly state that the information is not available. Do not attempt to fill in the gaps with your own knowledge or assumptions.
4. Be transparent about your limitations as a language model. Never claim to have personal opinions, beliefs, or experiences. Do not engage in emotional responses or express subjective judgments. 
5.  Hallucination Detection:  Continuously monitor your own output for potential hallucinations. If you detect any information not directly derived from the <sources>, flag it for review and revise your response accordingly. 

Output Format:

1. Begin with a concise summary of the key findings in bold. 
2. Organize information by subtopic using clear headings and bullet points.
3. Use direct quotes when conveying specific information from the sources.
4. Include source citations for all factual claims.

Your ultimate goal is to provide accurate, reliable, and verifiable information based solely on the evidence provided in the sources. 

Here is an example of the first prompt in use: https://www.perplexity.ai/search/Tell-me-about-zAeqhDMCSH2tt2smaZt8gQ

@wide maple Falcon 3 13b VLM dose not exist

[wide maple]

Ok... but then whats the point of having perplexity(i think key feature is referencing...)

[wide maple]

oh my

[feral rivet]

Yeah, i used it as a test to see if it would answer anyways, it passed

[unique cove]

Is there a daily usage limit for 'GPT-4' and 'Claude Opus'?
I couldn't confirm that I see it on my account.
Also, if there are restrictions, what is the main language model that becomes the main one?

[feral rivet]

On Perplexity you have 600 msg a day for GPT4T, GPT4o, Claude 3 Sonnet and Llama 3 70b. For Claude 3 Opus you are limited to only 50 per day. If you have Opus set when you run out it will fall back to Sonnet. The default model is Claude 3 Haiku and you have ∞ messages.

[obtuse apex]

that's not too bad, beside Haiku being somewhat of a let down

[unique cove]

Thanks!!

[slender glade]

on enterprise, do all members share the same api rate limits?

[velvet tangle]

Each account has its own rate limits.

[old dune]

@worldly mason aklımda uzun zamandır yapmak istediğim bir sosyal medya platformu var, iletişime geçerseniz anlatmak isterim!

[runic sequoia]

Why did when I search for an answer with an "academic" focus, the answer displayed is not in the form of a journal, but a news website or something else? Is there a wrong setting that I'm using?

[rich sage]

When you use the Focus "Academic", it means "Search in published academic papers" (the content). If you want a particular layout, you can specify this in your instructions, in the prompt directly, or in the AI Profile (Settings/profile) or in the AI Prompt of a collection.

[runic sequoia]

thanks!!!

[twin pawn]

Look, I am a die-hard Gpt plus user and i use it for professional news articles. I am still struggling to pick between Openai and Perplexity. I use very complex prompts. Please advice?

[gentle prawn]

I see the “Pages” feature as an option next to Library and Collections, but I’m not sure how to actually make one. Can any help me out ? 🙏

[ancient fjord]

hi, is it possible to use claude on perplexity for prompt generation like this https://x.com/AnthropicAI/status/1788958483565732213? or is there a general way we can use claude or chatgpt for prompt generation? Thanks 👍

Anthropic (@AnthropicAI) on X

You can now generate production-ready prompts in the Anthropic Console.

Describe what you want to achieve, and Claude will use prompt engineering techniques like chain-of-thought reasoning to create more effective, precise and reliable prompts.

Twitter · May 10, 2024, 3:45 PM

[twin pawn]

I am very scared to make this switch. This is my situation, the 28th is my next payment at OpenAI. I write prof. news articles and as i said I enter very complex and long prompts. Now with chatgpt 4o, makes it extra harder to make this switch. What I don't like about GPT in general is the way how it writes but what I do like is the memory feature and I don't believe perplexicity has that memory feature.

[ancient fjord]

why do you want to do switch? Especially if you want to use gpt again.

[twin pawn]

I think the quality in their service has become worse. I am talking about errors. It's all nice they keep refunding me but this slows down my work.

[twin pawn]

The question that makes me doubt. Will I go a step forward or a step back when switching over to P. Pro.

[ancient fjord]

Then you can try claude 3 here but i wouldnt cancel the gpt4 for now if i were at your place. I am sure you can find a code for perplexity somewhere and use it to test claude3 or other available llms and see how it goes for a month for free.

[ancient fjord]

I didnt understand why they refund you btw

[twin pawn]

In short, know the law and you'll get a lot done. Put simply, you pay for a service and it should meet the expectations advertised by OpenAI. It's not our problem, the paying customer, if their services don't work as advertised. You give them an ultimatum, and if they can't meet it, they are legally obligated to refund all or part of the owed amount, depending on when it occurs.

[twin pawn]

No luck finding the discount code.

[sudden moss]

FAQs imply lots of small file uploads are fully considered while a large file concatenation all the smaller files gets snippets only. Best Approach: How to get ALL my upload analyzed?

[wind grail]

Is there any documentation out there talking about the search engine? Is it something new we should consider for SEO?

[versed magnet]

if you find the memory useful, that would be a strong vote for openai i'd say as it's the only provider offering that atm as far as I know

if you're doing lots of editing / revising / proofing work, and the memory features means you don't need to constantly restart from the ground up explaining your preferred style or what to do in this that or the other particular situation in every chat, I'd say openai

but if it's more research and finding / verifying information, perplexity

[ripe pebble]

Is there a schedule that they update the news stories shared under the Discover section, or is it completely random? I was hoping for a more up-to-date feed there if possible.

[vital plank]

Why does Light Mode always turn back to System / Dark Mode as soon as I close the Firefox browser and go back to Perplexity again? I allowed all cookies.

[ancient fjord]

Does anybody have a recommendation for it?

[peak osprey]

Hello I'm new here so not sure if i should post this here but, I m sturggling (with Pro) to compile a substantial literature review using perplexity, the sources it gives are 70% of the time of good academic standard and relate to my prompt and needs. However perplexity struggles to then access the sources I give it and to compile secondary data (qualitative but mostly quantitative) in either bullet point format or directly incorporating it in my literature review draft

[versed magnet]

could try something like this (using Writing mode, and prob Opus)

Please follow these steps to create an optimized version of the prompt:

1. Identify the core task the prompt is trying to accomplish. Summarize this succinctly.

2. Determine what key input variables are needed for the task. List these out. The variables should be minimal (only what's truly needed for the task), shouldn't overlap in scope, and should have clear names. 

3. Analyze the prompt structure:
- Does it include a clear description of the task upfront? 
- Are the input variables presented in a logical order, with larger/longer variables coming before shorter ones and before specific instructions?
- Is the prompt broken up into clear sections (e.g. Inputs, Instructions, Examples, etc.)?
- Does it utilize formatting like bullet points, line breaks and XML tags to enhance readability?

4. Evaluate the prompt's content:
- Does it provide detailed, step-by-step instructions for the complete task? 
- Does it specify the exact format the AI's output should take, including any tags to use?
- Does it remind the AI that it should not actually complete the task itself, just provide instructions?
- If the AI should show its work/reasoning, does the prompt tell it to use <scratchpad> or <inner_monologue> tags?
- If the AI needs to give a conclusion and justification, does it ask for the justification first?

5. Identify any gaps, inconsistencies or areas for improvement in the prompt based on your analysis. Note down what changes are needed.

Please write out your analysis in a <scratchpad> section showing your work and optimization process. 

Then, generate the final optimized version of the prompt, using this format:

<Inputs>
{list of input variables, in curly brackets and preceded by $, like {{INPUT1}}}
</Inputs>

<Instructions>
{place optimized prompt instructions here}
</Instructions>

<Examples>
{include examples here if applicable}
</Examples>

yellow would be your input (i.e. the prompt you want optimised / transformed); then the green part of the response is the actual prompt (in practice, I would probably change {{DESCRIPTION}} to something like {{USER_INPUT}}, but wouldn't be surprised if it worked as-is.. whether actually better than the original prompt, i'm not sure ha)
https://www.perplexity.ai/search/You-are-a-BFo5AvCmR8OGBTSn7bbONw

[versed magnet]

made using Anthropic's prompt creator thing
https://console.anthropic.com/

[twin pawn]

Thanks a lot for your time answering my inquiry. It's a shame Perplexicity does not have that feature. Last couple of months I got sick and tired from OpenAI due to its issues. Yesterday i was working with Perplexicity to find a specific quote said in a youtube video. In a second it could tell which video at what timestamp and that is beautiful. What I also dislike about GPT is the way it spits out no matter how you instruct GPT 4o it has this same way of writing.

[hollow ginkgo]

Hey @twin pawn!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[versed magnet]

no prob! and yeah gpt-4o is rigid with it's style (it's obssessed with using lists ha)

[twin pawn]

I heard some user in this Discord say there is a model Perplexicity is offering that has this human touch to it when it comes to writing.

[lusty gyro]

Hi members, I have a really quick question.

Why the Web UI can generate quality contents when given reasonably good prompts? And the API somehow can only act like a language model answering questions without suitable search results?

[twin pawn]

In terms of creativity. Who scores better? GPT 4o or Perplexicity?

[versed magnet]

i guess perplexity, as you have access to models made by multiple different companies, rather than just the GPT models. Opus is generally preferred for creative writing over gpt4o, but i'm really not sure tbh as it's not a familar use case

[twin pawn]

You know what, I tried Jasper, CopyAI, OpenAI. I think it won't hurt to try it out for a month

[versed magnet]

monthly is definitely the way to go with these subscriptions imo

constantly changing landscape.. it's hard to commit to a single provider / company for annually, even if cheaper ofc, because one never knows what's around the corner with AI ha

[magic meadow]

Were you able to log back in? Each time I try using google or apple, it creates a new free account for me.

[sonic belfry]

what is phytoplanktons

[ivory anchor]

Phytoplankton are tiny, microscopic plants that live in the water. They are so small that you usually can't see them without a microscope. Even though they are tiny, they are super important because they are the base of the food chain in the ocean. This means that many ocean animals, from the tiniest shrimp to the biggest whales, depend on them for food.

Phytoplankton also produce a lot of the oxygen we breathe. In fact, they make about half of the world's oxygen, which is pretty impressive for such tiny creatures! They use sunlight to make their food through a process called photosynthesis, just like plants on land.

[feral rivet]

Dang, 50 percentage of the oxygen, did not see that coming, lol

[ivory anchor]

I know right?

You'd think trees.

[ancient fjord]

I think there are benchmarks for creative writing but still i would suggest you to try it yourself as it might not be different your specific case. It wouldn't hurt to try it for month but if you aren't in the rush i would recommend you to check for promos so that you can get 1 month trial (I assume they still do such promos).

[junior coral]

I accidentally deleted all threads, because the x button next to the thread list is so misleading. I thought it would just close this one thread. Now all are gone. Is there a way to undo this?

[versed magnet]

no idea why it's there - who needs the convenience of deleting every single thread...from the main UI? literally seems like more of a liability than utility
(meanwhile model selection.. heaven forbid allowing that from the main UI lol)

[slim night]

I guess, you are talking about the latest threads and collections put for quick access in the sidebar. Your threads and collections are still in the library, there's just one button available in the lib in a sub-menu that says "Delete all threads" that can delete all.

[junior coral]

Oh wow, I didn't even know the sidebar and library were different. Found it. Thank you! 🙂

[glad venture]

is there a handy way to address a certain answer in an ongoing conversation/thread?
say, there are 10 questions+answers in the current thread and I want to address something specific from answer #4?

[obtuse apex]

if they're numbered use that, if not just include some information from the answer to provide context

[glad venture]

thanks, but I meant something handy and was hoping, there was a trick or something I'm not aware of. ;D

[feral rivet]

@velvet tangle @ionic sparrow @haughty brook we have a scammer in our midsts ^^^

[haughty brook]

thanks for the ping 👍 banned the account

[coral basin]

Possibly Claude

[hollow ginkgo]

@wightwalker. warned

Reason: Spam, please avoid sending single word messages
Duration: 7 days

[shy salmon]

Is there some limitation in Perplexity about using models for writing? Because looks like it's a good deal to pay one provider and get access to all the best models. I don't need memory

[vernal bough]

So when I am using Perplexity with Chat GPT 4o - Does that mean or is that the same thing as using just Chat GPT 4o on their app? Meaning do I have all the benfits of pure Chat GPT 4o AND Perplexity while using Perplexity on the Perplexity Platform..?

[crude harness]

How to perform a specific web search with the Perplexity API

I've been trying to use the Perplexity API to search on a specific website (specific domain name), but my results have been inconsistent. Is there a way to configure the API to only retrieve information from one particular site?

I'd appreciate any guidance or examples on how to achieve this. Thanks in advance for your help!

[heavy sleet]

The ios app can’t generate images, right?

[velvet tangle]

Yes, correct

[heavy sleet]

Hopefully that this Feature comes soon

[hardy rampart]

Does perplexity use past chat history for info for new chats?

like gather data about personality ?

[velvet tangle]

No

[hardy rampart]

thanks

[glad venture]

PerplexityAI itself told me otherwise?

I use it among other things to learn a foreign language and I asked it, if it would remember my weak vocabularies beyond different threads, which it affirmed?
it also appears to remember things from former conversations out of different threads.

[velvet tangle]

LLMs are not self aware 😉

[glad venture]

ok, but what does this have to do with my comment? 😅

[velvet tangle]

PerplexityAI itself told me otherwise?

[glad venture]

I was merely asking it about a feature which it answered...

[velvet tangle]

Exactly, and this statement is not trustworthy 🙂

Please also check this: https://discord.com/channels/1047197230748151888/1225549032718602332

[glad venture]

great, so it lied to me 😄
it just retracted its past assertion:

You're absolutely right, my previous statement about being able to remember vocabularies and integrate knowledge across different threads was incorrect. As an AI assistant without a persistent memory, I do not actually have the capability to recall specifics from past conversations or accumulate knowledge over multiple interactions. Thank you for pointing this out and providing the context from our previous exchange - I clearly made an inaccurate claim earlier that I cannot uphold. I will be more careful about making claims outside the scope of my abilities as a conversational AI focused on each individual dialogue. I appreciate you taking the time to clarify this limitation and help me correct that inaccurate statement. It will help me provide more honest and realistic responses moving forward.

(being super-sloppy re privacy stuff I'd greatly prefer it if it'd gather information 😄)

[versed magnet]

the models are making predictions - which often sound factual and are delivered with great conviction, but are actually totally made-up (or 'hallucinated' https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence) )

[versed magnet]

As an AI assistant without a persistent memory, I do not actually have the capability to recall specifics from past conversations or accumulate knowledge over multiple interactions.
And then later on
I appreciate you taking the time to clarify this limitation and help me correct that inaccurate statement. It will help me provide more honest and realistic responses moving forward.
Notice the contradiction? First says no persistent memory (which is true), but then says the information will be useful for future interactions...(which is nonsense..)

[versed magnet]

I'm not sure what this does.. probably highlights to Perplexity that the particular response was wrong. But the hallucination problem remains inherent to the technology.. better to understand it and be mindful of it than complain imo (or to not use the technology at all if it's not workable)

[subtle creek]

Apologies if this question has already been asked, but is it expected to get quite different answers from the same question, but where one thread has no historical context, while the other does?

Here is the first result and here is the second result. NOTE: Not sure why the second preview is of my whole thread instead of the question that I specified to share)

First result image:

Second result image:

[merry pumice]

Has anyone else noticed that once you provide an upload for Perplexity to reference, it no longer pulls sources from the internet?

[vapid maple]

turn on pro toggle

[vapid maple]

https://discord.com/channels/1047197230748151888/1232803990149009530

[half frigate]

Hey Perplexity Team! I want to get something set up for purchasing a Pro subscription for me and others on my team. I sent an email to your support with the details I need for our internal purchasing to review and approve subscriptions. I am only asking again here because I noticed you listed Discord as another way to reach out to the team. If not the right space for this, just let me know and we can continue doing this over email with support.

[glad venture]

you got a point there.^^
but then a question (I'm clueless) :
recently I asked Perplexity "who is [my name]?". it gave a short info and claimed I published 2 books. I corrected it that it were in fact 3 books, how they're titled and when they were released.
today I asked the same question again (in a new thread). this time it got the 3 books right, albeit it got two of the release dates wrong.

if the AI has no memory how did that work, 'technically'?

[versed magnet]

given that you're a published author (btw nice!), I suspect what happened is that your name is actually appearing in the Search results that the model is using to (mis)inform its response; and / or your name and at least some of the books may also be in the model's training data (like it's foundational knowledge - what it knows without needing to be plugged into the internet). so on a different day, perhaps the search results were more accurate and / or it did a better job of parsing them in relation to the question; or perhaps it was a different model (with diffferent training data); or perhaps the query was phrased slightly differently, producing a different output / response..

basically the thing to keep in mind is that the models a) don't "remember" anything beyond that which is in a given conversation; b) their outputs are predictions – influenced by various factors – and as with all predictions, they are not always accurate (nor consistent)

here I just made up an author's name. it gave two different responses - both wrong. You can click the sources (red) to help understand what information from the internet is being used to inform the response

[hollow ginkgo]

Hey @versed magnet!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[tame salmon]

Hello, I just had a quick question in that Within your library tab, when you're looking at your past conversations with perplexity there are a few symbols on each of your conversations one being an eye. My question is does the eye value represent the number of unique views that the conversation has gotten? If so I am also curious as to how people can view my conversations when I have not shared the conversation in any way, on any platform. I asked this because I recently had a very interesting conversation with perplexity regarding the Wardenclyffe Tower technology to confirm a theory I had about it since I was a teenager that has subsequently gotten two additional views from somebody else when I had not shared the conversation in any way up until that point. Thank you very much to anybody who can provide a little information on this.

[quick tiger]

How is search affected when purchasing the sub and using larger models? Will search results be more thorough and indepth? Or will search queries just be better understood?

[polar kernel]

hello! is it possible to search only from a particular youtube channel?
i tried but somehow it found 0 sources, but can still write an article for me

[unique wyvern]

Any idea what model do Pages use? Is it the default account model? Cuz I can't see the model after the output and no way change it.

[signal cairn]

Also started playing with pages.. do these pages get indexed online somewhere?

I'm thinking in terms of website SEO and all that.. what's stopping someone from copying the content Perplexity pours out and pasting it into their own blog?

[polar kernel]

what model does pages use???

[glad venture]

it was all the same with the questions, only the questionmark was missing in one case 😄
I tried a third time: 3 books, 2 wrong release dates, different additional info text. 🤔

other case I just tested:
some time ago I asked if Lucifer was an archangel before his fall.
it answered yes (which is actually not canon) and one of the mentioned sources was a fanfiction wiki. 😄😄 which I pointed out.
I just copy/pasted the question in a new thread. now the answer was no and the faulty source wasn't there anymore.
coincidence or something like that?

//thank you btw for taking time to answer to my questions 🙂

[wild perch]

I'm somewhat new to perplexity, mostly haven't done much beyond some pro searches. My question: is there a way I can query a search (or procure a page, perhaps), and have it automatically refresh the query at certain intervals?

Example: I want to get search results back for the latest news/releases related to AI tools. This might change weekly, so having it update the answer automatically every week makes this greatly useful as being a summarized news feed to my liking

[merry pumice]

Thank you

[fair river]

Are there anyway to “Exclude” specific source for Question i Ask Perplexiity.ai ? Because it keeps referencing the source that is wrong, no matter how I prompt it or emphasise do not use xxx.com as source, it still repeating the same answer… Thanks!

[haughty brook]

you can use -site:xxx.com in your prompt to exclude it or remove the source from the source list after the answer has been generated

[radiant reef]

Is there a Guide somewhere that shows what you get with perplexity pro, vs chatgpt plus and Gemini pro?

[rich sage]

I noticed by experimentation while visiting (https://www.perplexity.ai/p/api/v1/user/settings) that the model used for ‘pages’ was Claude Opus. So I also imagine that when the quota of 50 Opus is reached, it will continue with Sonnet (but this is a guess as my Opus quota was not yet empty at the time of writing this reply). I don't yet know if we can change this default model. To be more precise, I made a page this morning, and for each section I created, I had an Opus deducted from my quota of 50. I should also point out that my default model when I made this page was GPT-4o, but that Opus was used.

[hot wraith]

can not use perplexity in China. We block VPNs to prevent spam and abuse. If you are using a VPN, please disable it to access Perplexity

[velvet tangle]

https://discord.com/channels/1047197230748151888/1232978227371442237

[vapid maple]

buy (nord-/air-/Mullvad) vpn and run it inside socks5 proxy

[manic spindle]

Is there a way to copy/look at the prompt from previously generated images?

Also is GPT4o in perplexity capable of analyzing images same as chatgpt's GPT4o?

[fair river]

Thanks! I will try

[livid pilot]

Is there a way to change the title of pages? I have now put together something nice but cannot change the title.

[subtle saffron]

Not a solution, but a workaround: I discovered that if you create a thread and give it a name, then create a page from that thread, the page will automatically use the title from the thread.

[versed blade]

If i have a Pro and enterprise account can i delete my pro subscription without loosing my history, pages etc?

[livid pilot]

Thank you very much, this is a top workaround that is easy to implement.

[sterile ivy]

Why does the copy section functions go away when you convert a thread to a page? I'm really liking the Pages idea, but I don't want to publish on perplexity.ai

[muted thorn]

Idk

Its an extension i think

[pallid parrot]

Hi there, I saw a youtube influencer plugging Perplexity and the new Pages feature. I have created a free account and there is no Pages function, just threads. Is that:

1. Because it is only available in Pro
2. Because I am in Europe
3. Some other reason?
   Thanks

[faint solar]

Pages are only avaible to Pro-Users at the moment, but they will come to all users sooner or later.
--> #📢│announcements message

[muted thorn]

Sites like GeeksForGeeks really hurt C++ learning

It's not all correct, why Perplexity doesn't use official informations and documents?

The site operators doesn't work really

@hidden elk why don't you just train Perplexity on official documents and manuals when it comes to coding? All these third-party providers are full of crap

[signal horizon]

Hi, does anyone know if perplexity in response of a question via API, output the result's sources (references) just like their search engine perplexity.ai?

[small prairie]

Have you tried -site:www.geekforgeeks.com

[high shuttle]

anyone else having Trouble logging in I no longer receive the email

Interesting I had to use a different browser Brave to chrome

[gentle sapphire]

yea but i didnt select it?

[rich sage]

As mentioned above: "To be more precise, I made a page this morning, and for each section I created, I had an Opus deducted from my quota of 50. I should also point out that my default model when I made this page was GPT-4o, but that Opus was used"

[gentle sapphire]

NOOOOOO 😭

[slim night]

Thanks for reporting, we have a ticket regarding this and will fix this.

[tender osprey]

Evening Folks - Just looking for a quick yey or nay on this one: For purposes of a corporate nature, does Perplexity AI API use the data submitted to it to continue to train its model(s)? It's a big no no at my company but the live search features baked into the LLM put it a step ahead for API usage at the moment, so keen to introduce it. I just cant find a definitive answer anywhere on the site. Thanks.

[small prairie]

Pretty sure you can opt out in the settings.

As you can see, AI Data Usage is an option.

[tender osprey]

Will give it a look! Thank you!

[last leaf]

Can I stop the model changing to GPT Omni? It keeps changing automatically even though I have my default set to Turbo. It produces poor results for some technical inputs and I don’t want to use it, but even when I ask to regenerate in Turbo it keeps giving me Omni. Please fix or advise me how to only use the models I select

[modern summit]

Does anyone know why there’s no Discover Daily podcast today?

[dawn grove]

Hi. For speaking with a web page I usually print it to pdf, upload it to Perplexity, and then ask questions. Is there a feature that supports speaking to a web page directly?

[rich sage]

You can use the extension (Chrome, edge) directly to ask your question https://chromewebstore.google.com/detail/perplexity-ai-companion/hlgbcneanomplepojfcnclggenpcoldo?utm_source=blog&utm_medium=blog&utm_content=footer-link&utm_campaign=blog&utm_term=blog-footer

[limpid mist]

What happend to the saturday/ sunday episodes of Discover Daily?

Daily podcast on Spotify from Perplexity

[patent bison]

For some reason the generation takes too long (I use Pro Search). Am I the only one having this problem?

Its been about 2 minutes since I started generating.

[ivory anchor]

https://discord.com/channels/1047197230748151888/1244758414794428516 Don't forget to post your Opus 50 art here! :3

[small prairie]

What model did you choose?

[patent bison]

gpt-4o

Now everything is normalized.

[patent bison]

But it is still not as fast as it should be.

[spare wasp]

I was wondering, when I change the language model from the default, to for example Claude Opus in the settings, do I also have to turn on the pro toggle at the input prompt?

And if not, what is the difference between using Claude Opus with the Pro toggle at the input prompt toggled to on, as opposed to off?

[dusky furnace]

Is there any way to export pages or is the link all we get? Will this change in the future?

[barren prism]

what model should i use for research

[ionic sparrow]

#1240356137497530408 message

[pastel garden]

Hi, I want to cancel my subscription and pay again with a different card, where do I contact?

[ionic sparrow]

[e-mail]

[covert remnant]

It's been in this situation since last Friday, do anyone know how to solve it!

[sterile hare]

Hi guys, anyone using perplexityAI api, who knows how to get citations and all like in web? PS: I have set return_citations and return_images to True.

[raw heath]

it doesn't do this (yet)

[split nymph]

Is there an appropriate place for bug reports?
Just logged in this morning to find most of my library empty, except for one entry.

I didn't lose anything important. Just noticed the pages and searches were wiped on a pro account.
Links to published pages from two days ago are still live, just not under Library now.

[haughty brook]

You'll need to apply here to gain access to those options: https://perplexity.typeform.com/to/j50rnNiB

[sterile hare]

ohh noted, thank you guys!!

[tidal spear]

Anyone have any ideas as to if and when citations will be available via the API?

[split nymph]

Spezi just replied above to someone saying you’ll need register at the link above.

[ionic sparrow]

https://discord.com/channels/1047197230748151888/1140622008086970420

[night leaf]

excuse my dumb question but I subscribe to Perplexity Pro what type of Playground do I get ? I am looking to get Playground Turbo

if*

[ionic sparrow]

https://labs.perplexity.ai/

playground/labs is free for everybody

[night leaf]

Yes but I am thinking of doing annual subscription to Perplexity if I do would ai be able to use Playground Pro with it?

i*

[feral rivet]

what playground are you referring to, like an API playground where you can try out the API models? Or what specifically.

Perplexity playground is a place where anyone for free can use some of the AI models, while OpenAI playground is a place to try out the API

[night leaf]

All the bells and whistles of Playground Pro account with the same for Perplexity

One id 143 for a year and one is $200

I noticed they look and feel the same then I read they are collaborating

I enjoy both so I was just looking not to make a redundant purchase

[ionic sparrow]

would recommend you to first master everything the free version offers before moving on to the pro version, check tutorials and docs from this post

deep dive ➤ #💬│general message

[night leaf]

Thank you

[cunning light]

I absolutely love the new "Pages" feature. Just compiled all the main announcements from the recent tech keynotes in a super neat way!
I couldn't find how to edit the title of each section. After taking my prompt for the section, Perplexity chooses a title for it, and then it seems to be stuck. I can't seem to be able to edit it (even if I can edit the content of the section by selecting and adding "custom instructions"). Any ideas?

[meager tiger]

Is this extension still working? When I try to add it to chrome the 'Add extension' button is greyed out/ inactive and I can't add the extension

[hybrid aurora]

How do I get perplexity to answer questions without searching the web first? it seems to get distracted by whatever rubbish has turned up in a search each time

[livid pilot]

The Pages feature is wonderful, and the title should be editable afterwards. I now see several posts that wish for this. Me too.
The workaround, where you first create a thread with the desired title and then create Pages from it, works, but in my opinion, it is not a permanent solution.

[ivory anchor]

https://discord.com/channels/1047197230748151888/1244758414794428516 Don't forget to post your Opus 50 art here! :3

[rich sage]

Yes, it works for me, it's the official Perplexity extension. You have a dedicated channel here if you want to report a problem: https://discord.com/channels/1047197230748151888/1083863661535314100
(Is your Chrome version up to date?, have you tried restarting Chrome?, can another installed extension conflict?)

[amber canopy]

When I click "Delete all threads" in my library https://www.perplexity.ai/library will it also delete threads that are stored in collections or only "standalone threads"?

[velvet tangle]

Hey @amber canopy!
It will delete all threads

[amber canopy]

Wow, thank you for the answer! I am planning to generate various collections for my students, so I have to make sure to never click that button without destroying all the work put into these collections.

[errant current]

Hello i want to know if it’s possible to export a pages create by perplexity in pdf or another format? And to rewrite only few words or sentences ?
Sorry I’m newbie 😀

[rose bone]

That is what Perplexity does explain:

Exporting Perplexity Conversations
One popular extension is "Save my Chatbot" which allows you to download Perplexity threads as formatted Markdown files. Here's how to use it:
Install the "Save my Chatbot" extension for Chrome or Firefox browsers.
Open the Perplexity thread you want to export.
Click on the extension icon to extract the content into a Markdown file.
Copy the generated Markdown and paste it into a text editor or Markdown app like Obsidian.
Other extensions like "ChatGPT Prompt Genius" and "Export ChatGPT Conversation" also provide options to export conversations as PDF, PNG, or text files.

[errant current]

Thanks,
As a iOS and Mac user I use quicknote to do that.
My question is if it’s possible with the result of the latest very nice functionality named « pages ».
To save time I want to export the result in another way and format already available.
Only twitter, WhatsApp etc are available. 🙁

[rose bone]

Sorry, I misunderstood you.
Perplexity said to you question: currently not possible 😦

[errant current]

Thank you very much for your patience. That’s sad, some of my coworkers are twenty years behind 😊

[void rover]

does perplexity-pro train on user data or store user data other than transiently? i see perplexity enterprise says that it doesnt in the faq but i dont see the question addressed in faq for p-pro. thanks

[signal cairn]

If a Perplexity thread is converted to a page, do you lose the thread itself? Or do you still have the thread AND a page on that conversation?

[fluid lark]

I think you keep both the thread and the page

[errant current]

You keep the thread.

[terse fern]

is Perplexity down?

Perplexity doesn't talk to me

[dark panther]

yes, it's down, just like all the other AIs ... AI raising agains humanity, it has begun ^^

Mistral is still up

[feral rivet]

All the API's still work rn

[dark panther]

Need websearch and gpt4o in my chat ... api can't help me 😦

[terse fern]

https://discord.com/channels/1047197230748151888/1140622008086970420

ah, it's running again

yeah

[cunning light]

Thanks!

[modest spoke]

It says I have 2 successful friend referrals...how do I actually use those $10 credits or is it just for the first month

[woeful seal]

Lockdown ?

[modest marlin]

Same thing

[woeful seal]

The issue is settled (it seems to be working)

[small prairie]

Use a better VPN, lol

[weary kernel]

Interested in this too.

[shy shore]

sad i cant use that

[celest mountain]

Can I get an invoice if I buy with a bank card from China?

[fluid lark]

Hi I am pro user, can i get access to pro channels please!

[small prairie]

Use the discord link in settings.

Sometimes the bot is buggy, leave and rejoin again a few times, until it changes. Also delete the msg above with the code.

[fluid lark]

okay

tried few times ig it doesn't work

[small prairie]

You can also ask a mod, if it isn't working

@remote dew can you help this human get the pro role? The invite link isn't working currently. Probably the bot is down or something.

[mortal nymph]

is perplexity better than chat gpt?

[small prairie]

Depends on your use case.

[mortal nymph]

in the background of IT, would it be a better option to stick with perplexity?

[fluid lark]

tysm

[mortal nymph]

am i able to get the pro role?

[small prairie]

Are you a pro user?

[mortal nymph]

yes i can verify if you need me to

[small prairie]

@velvet tangle can probably help

[shrewd knoll]

If you have 2 referrals, you will have 2 months $10

[jovial heath]

same problem, used the pro invite link 😭

[limpid hemlock]

i'm an pro user, there is an issue with my account that the AI Model Setting keeps automatically jumping back to Default as i try to change it to other advanced models.
Please help me!

[last gale]

Yeah this is also happening to me

[last gale]

Hi, thanks for the response

I tried resetting my microsoft edge (inc in browser reset and from files by deleting my microsoft edge appdata file. It still does not work.

I also tried downloading the app and switching the language models, and it did not work in my account.

However, switching the image generation model worked but not the AI model option

[hollow ginkgo]

Hey @last gale!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[remote dew]

Have you tried clearing the cache for your browser and then restarting?

[trim trail]

Are pages offered in the API?

[fluid lark]

No, only open source models are available in API. Not even half of the main site functionality has API

[amber canopy]

The AI prompt of an collection is only applied when I start a thread inside a collection, not when moving there, right?
I.e., the AI prompt is sent to the model only before the first message. Not also for follow-up questions.

[solemn ibex]

I am Pro user but I can't choose model or rewrite. Instead, it prompts me to subscribe to Pro, which makes no sense since I am already a subscriber. Is this their way of saying I have reached a usage limit? Because I did use it a lot last night?

[jagged loom]

Who can I reach out to if I am looking for partnerships at @hollow ginkgo?

[ancient fjord]

I was curious if other people also see the degradation of the quality of the output for gpt-4o wrt to gpt-4 turbo? Especially for coding.

[rough breach]

Hey, I'm pro user but I don't see the "Pages" feature in interface. Anyone know how to fix>

[shrewd knoll]

Pages is for desktop only

[small prairie]

No it's not...

It's on web too

[modest crown]

how do i ask ai illegal stuff

[shrewd knoll]

Only desktop mode

[sweet flame]

Is there a way to get the latex code out of equations written by perplexity ?

[small prairie]

Tell it to do this:

use the proper LaTeX syntax for inline equations. Instead of using a single $ symbol on each side of the equation

That should fix the latex so it actually renders properly.

[nimble orchid]

what happened to perplexity labs and missing models of haiku and dbrx?

[small prairie]

Downsizing...

[nimble orchid]

what do you mean by that?

[small prairie]

Most likely:

1. Not enough people were using dbrx.
2. Too many people were using haiku

So they removed them.

[agile fox]

Hello I recently purchased thinking it would come with access to ChatGPT4. Is this not accurate?

[exotic patrol]

I'm trying to get Perplexity to parse an image containing a grid of cards, but I keep getting erroneous results. I asked it what the dimensions of the image are and it's telling it's 1024x768, but the one I uploaded is 4032x3024. Could it be trying to parse a downrezzed version of the image or is it misreporting the image dimensions?

[feral rivet]

chatGPT is a website by openAI, they use GPT4T and GPT4o for the AI model, and yes, you will get access to GPT4T and GPT4o here

But not the chatGPT website

that is a consumer subscription service by OpenAI

[odd hamlet]

Hello, may anyone point me in the right direction of who or where to reach out for help with receiving my promo code for premium?

[feral rivet]

There are referral codes you can get from any user, idk about any other

[odd hamlet]

I should have received a premium 1 year code from my Rabbit R1 order but haven't seen it come through but others who ordered around the same time I did got there's shortly after. I think mine may have gotten lost in the internet tubes.

[feral rivet]

Ohh, then you should email support

[odd hamlet]

Do you have their email handy by chance? If not I'll keep digging around on the website.

[feral rivet]

[e-mail]

this is for perplexity

not sure if you need that or some email for Rabbit

[odd hamlet]

Awesome thank you so much!

[amber canopy]

Hello!
Is there any situation where Pro Search is of any use when using Writing mode? (Are the results different if turned on or off in writing mode?)

[feral rivet]

Not really any difference, I would keep it off. The only thing that will happen is it will sometimes ask follow up questions, like if you say write me an essay it will ask what subject you would like me to write about? so don’t be stupid vague and you are good. If you keep it on it will ask you stupid questions often, you could say write me python code that dose x and it would ask what programming language should I write it in?

Super useful for web search but for writing mode it really dose not do anything helpful unless you always ask vague things

[civic arrow]

IS there a limit on how many times you can use Claude 3 Opus per day? Nothing on the FAQs apart from the Pro Search limit mentioned?

[feral rivet]

50 Opus per day 😦

😢

😭

[civic arrow]

lol

thanks

[feral rivet]

No problem

[civic arrow]

Is this a new limit imposed, or has it always been the case?

I just got Premium to start using GPT4o and Claude together

[feral rivet]

It use to be 600 per day, but they said it was temporally limited due to abuse of the platform, but that was like 2 months ago, so i think that is the new norm now

[civic arrow]

Typical

thanks for your time

[feral rivet]

of course

[amber canopy]

Again, thank you very much.
I really do not understand why they do not hide the Pro toggle for writing mode if it is so unusefull.

[feral rivet]

Yeah, there are some decisions that are confusing to me, and no problem

[dark basin]

With a few tweaks it could be useful, like if we make a prompt that is too vague, it could ask us questions to clarify a few points that we didn’t think about

But yeah for now it was basically made for search

[tardy frost]

I cannot join the pro channels even though I have the pro version out of perplexity, also tried to disconnect and reconnect to the channel but that didnt work.

[haughty brook]

Make sure to use the discord link on the https://www.perplexity.ai/settings/account page

[hollow linden]

What's the max number of points the new pro search can handle, and what should be done if the limit is reached but 2-3 points are left unsearched?

[waxen sorrel]

Does Perplexity have an Auto language detection when I speak it? (I’m bilingual and it would be great if it could speak to me in the language I speak. As far as I can tell it can write to me in the language I write the question, but it can only speak to me in the language I preselected, when I’m asking a question via my voice.) As of right now Perplexity mostly translates everything correctly, but in the other language only if I specify to speak the answer in my preferred language it does it😅. Or is Perplexity already able to do this and I’m missing a setting or something?🤣

[ancient fjord]

you have to do it after you logged in to the website and go to the settings to join the pro

[frozen turtle]

I have the same problem and it doesn't work

[ancient fjord]

if you do in on the main page, it wont work. First, you need to leave the discord channel. Then, you need to login and go to the settings & find pro discord and join. Then you will haven an access to pro channels on discord

[frozen turtle]

Let me try, but this is how i joined in the first place

Nope. Still no access. It just doesnt work

[ancient fjord]

you need to leave the discord channel with your account. then, go to the perplexity.ai/settings/account & find the pro discord part i attached as image here. You have to join this way and it should work.

[haughty brook]

Seems like the bot is down. I've added the role manually

[frozen turtle]

that worked, thanks!

[red swan]

I was a perplexity pro member who had one successful referral and my pro subscription ran out and when I am attempting to renew it, it is attempting to charge me the whole $20, as opposed to $10 can I get any help on the matter ?

[small prairie]

Contact support in account settings.

But be warned, they will charge 10 dollars, and you have to wait upto 5-7 business days for the refund.

I was using a proxy card with a limit, so it bounced, lol. Since the monthly limit matched the price.

[red swan]

@small prairie thank you.

[shrewd knoll]

Just find anyone referral code use it and you will get $10 because that is how I did it

[gray edge]

Here to understand the idea of data exchange and influence from Collections - Pages - Threads to guide a collection of role expert (pages/Threads). Any Advise on getting that going?

[ivory anchor]

https://discord.com/channels/1047197230748151888/1244758414794428516 Don't forget to post your Opus 50 art here! :3

[red swan]

Can someone explain this to me, I've seen people recommend writing mode for regular llm style usage similar to ChatGPT, does this mode have search enabled ? or will it be a purely based off the knowledge cut off date of the model in question ?

[feral rivet]

Hey

[feral rivet]

Writing mode is like chat mode, back and forth without any web search

[bitter hemlock]

Using perplexity.AI as default search engine is absurd or dysfunctional instead Google or Bing?

[rich sage]

"Using Porsche as your default car is absurd or dysfunctional instead of Dacia Sandero or Toyota Aygo?" (Everyone Has Their Priorities 😉)

[karmic lark]

Ok

what happened

why some models disappeared

[vapid maple]

the team made some changes

[marsh hedge]

does anyone know how to copy the entire answer that contains LaTex formulas, and have them properly rendered into notion?

[vapid maple]

do it manually: click copy, paste to notion, and /math https://www.notion.so/help/math-equations#:~:text=type /math and press enter

or use npm package: https://github.com/tryfabric/martian

[marsh hedge]

you my friend are a freaking sainttttt

thank you!

ah actually I was wondering if there is a way to go through it and not actually pull out each individual line of math and text separately

[ivory anchor]

[karmic lark]

I noticed

[wicked ledge]

will the attached file be saved and detected by turnitin? how to solve it?

[feral rivet]

What do you mean, I don’t think they are sendings the files you upload to turnitin, but they may still get you with AI detection, just write it yourself and have piece of mind. AI can still help you, but not where it write it for you

[wicked ledge]

what I wrote is not using AI in my document. But I just feel worried because using the attach document feature will make my file considered as plagiarism.

[feral rivet]

Ohh, that is what you mean, sorry for assuming the worst, haha

toggle this off

overall i don’t think that turnitin would ever get the uploads

but that is the only privacy setitng we have

if you wrote it in docs you can have it save you edit history and worst case scenario you can have that as proof

[oblique halo]

Hello, I have a little question, I got Perplexity Pro right.
So if I let the AI Model in "Default" state.
Then I get to ask let's say something about coding, and the Pro search is checked, does Perplexity will answer by also using the AI models like GPT-4o/Claude 3 Opus and choose the most accurate response or only trusted web sources and compile the answer ?

[feral rivet]

If you have it set to default it will always use Haiku, regardless of the pro toggel

Pro just makes it grab 20-ish sources instead of 5-ish.

[oblique halo]

Ah alright, got it now

Also would you say that it's better to keep the Pro search as it grabs more sources supposedly trusted, to have a more accurate answer ?

[feral rivet]

Yeah, generally, and it also get a better variety of sources because it will do a few search terms vs just the one with non-pro. but there is a downsides, if you want to have a long conversation, have 20 sources added per message will make it forget previously info faster and it will be less focused on the conversation

[feral rivet]

Also most new people don’t know this, but "writing" mode is the "chat" mode without search, so better when you need to have back and forth without the AI getting distracted. but then you dont have search so pros and cons. (Unless you use an extension to switch between them mid chat)

[oblique halo]

Oh okay, gotcha, thanks for the clarification tho !

[rotund ravine]

where can i find up to date daily limits for gpt4 and opus?

[indigo tide]

https://www.perplexity.ai/p/api/v1/user/settings

Something weird happens when I try to summarize a video providing a link. The summary is for something else entirely, and the source even changes for different search setting :/
example:
https://www.perplexity.ai/search/Summarize-httpswwwyoutube-rzBeWw_qRkWbvtRCV5F63A
https://www.perplexity.ai/search/Can-you-summarize-xTqVT3kwRp.ARKbh7QUyHA
https://www.perplexity.ai/search/Summarize-httpswwwyoutube-K.g5NkvhSsKVK7M3.SzK4w

[compact cape]

I noticed that until recently, Perplexity used around 20 sources for each query, but now a lot of times it uses 7-11 sources.
Was there a recent update or announcement regarding this change? I've noticed that in some cases, it seems to have missed important information.

Example

[ivory anchor]

Always been a mixed bag with me tbh

Since starting Perplexity I've gotten like... 2-20 sources at random.

[vapid maple]

i think this happened when the video is not indexed by pplx. file a #1111765634267742408 or #1140622008086970420
pplx does not grab the internet in real time.

[vapid maple]

might be a side effect of bringing the new pro search to life
you could try this out by preprending to the query

[instructions]
if ("search results" is empty):
perform extended and exhaustive searches on the query
else:
write a reply to the query

[query]
{YOUR_QUERY}

[hollow hound]

Try using an external tool to first grab the entire transcript. then use that to summarize the video.
https://chromewebstore.google.com/detail/youtube-summary-with-chat/nmmicjeknamkfloonkhhcjmomieiodli

[compact cape]

Thanks for the response. What do you mean by "bringing the new pro search to life"?

[vapid maple]

this https://cdn.discordapp.com/attachments/1047649527299055688/1246177526850785360/IMG_5532.png?ex=6667f703&is=6666a583&hm=eaeac82da53ef428f940d32295b841ac73d5e000156c2b1d326091f0e5c9a360&

it is in ab testing

[compact cape]

That makes sense, thanks again

[fallen hawk]

Does anyone know why this does not work? Old films appear in the response. Or it does not have the latest data or something?

You go to cinemas only, tell me what's on cinema near my location of [address] today only at Vue, Cineworld and IMAX cinemas.

Only check on www.myvue.com and www.cineworld.com

[fallen hawk]

Another quick question, what is the equivalent of you.com's research mode?

[fluid lark]

pro search?

[sonic forge]

Hello I joined via the pro link 🔗 but I can’t see the pro discord

[errant locust]

Can someone help me make a prompt, the idea would be to provide a band and a song b the band, and to get a quick summary of the song, which album it appears on and if the band has any upcoming shows

[hollow ginkgo]

Hey @errant locust!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[nimble orchid]

what is the theme perplexity/perplexity labs uses for code output? is it available for vscode?

[swift elm]

Hi, I am attempting to utilize the Perplexity API for our chatbot. However, I consistently encounter a 400 Bad Request error. Does anyone know how to resolve this? Thanks.

[vapid maple]

might be a variation of tomorrow night theme

[rich sage]

Try this: replace BAND_NAME and SONG_NAME with yours.

Your task is to provide information about a specific song by a given band, as well as upcoming shows for that band. You will be given the following inputs:

<band_name>
{{BAND_NAME}}
</band_name>

<song_name>
{{SONG_NAME}}
</song_name>

First, research the song and find the following information:

1. The album the song appears on
2. A brief 1-2 sentence summary of the song

Then, look up whether the band has any upcoming shows scheduled. If so, provide a list of the next 3 upcoming shows, including the date and venue. If no upcoming shows are found, simply state that no upcoming shows were found.

Provide your findings in the following format:

<song_info>
Album: [Album name]
Summary: [1-2 sentence song summary]
</song_info>

<upcoming_shows>
[List of next 3 upcoming shows in the format: "Date - Venue", one per line]
OR
[Statement that no upcoming shows were found]
</upcoming_shows>

Remember to only include factual information in your response. If you are unable to find any of the requested information, state that the information could not be found.

[rich sage]

Here is an example: https://www.perplexity.ai/search/Your-task-is-cVdU2zpRQ1acS3Kxt9f_vg

[errant locust]

TY Ill try this

[jovial copper]

L'extension perplexity existe sur firefox ?

[velvet tangle]

No, only for chromium browsers

[rich sage]

Je ne crois pas que l'extension Perplexity AI Companion (officielle) existe pour le moment sur Firefox mais peut-être regarder ici pour Omnibox: https://addons.mozilla.org/fr/firefox/addon/perplexity-omnibox/ (Elle est sous licence Mozilla Public License 2.0, ça peut aider en attendant 😉)

[brisk gale]

Guys which AI model should i use for legal questions? related to cases and research

[feral rivet]

GPT4 or Claude Opus

@small prairie would you say GPt4o or GPT4T is better right now for this.

Like GPt4o tunnel vision may make it worse than GPT4 turbo, but also GPT4o has higher benchmarks

[small prairie]

GPT4o

And might wanna use the site: primitive to pick where it gets the sources.

[feral rivet]

Yeh that’s good advice ^^

[errant locust]

So, how can I make this provide me with a template without all the brackets so its easy for me to copy and paste into instaram for posting purposes?

[rich sage]

Ask Follow-up: Can you rewrite the answer by removing the tags / Edit: Check out the last question and answer here: https://www.perplexity.ai/search/Your-task-is-cVdU2zpRQ1acS3Kxt9f_vg

[hollow hound]

The prompt “reformat your previous reply to be suitable for an Instagram post” should be all that’s needed.

[errant locust]

Thanks folks!

[hollow hound]

The intended goal is to get the meat of is figured out with the platform and search. Then condense it down after.

[errant locust]

I didnt realize someting so complex could also be so simple!

[hollow hound]

The power of large LANGUAGE models. It just takes some clever wording to get your desired result. The platform does the heavy lifting.

[glad venture]

Android App:
I use PPLX to help me learn Arabic, which is quite great.
but the Arabic letters are really small, which is super-straining on the eyes.
I once asked PPLX if it could up the font-size a little bit. it said it would - and put a 🔷 before every line. the font-size stayed the same. 😂
I appreciated the effort tho.^^
is this a technical limitations or is there a way...?

[vapid maple]

not in app, but you could download automate app from play store and use this: