#💬│general

1 messages · Page 65 of 1

agile jay
#

are you in the free tier?

devout cargo
agile jay
#

I tried without pro enabled, and also got images. Let me try another search to see if my previous one was just cached.

#

Got images

devout cargo
#

And you can try from the App with quick search.

agile jay
#

Okay

#

That's without pro search again.

south kindle
#

this prompt doesn't yield images automatically but search pulls them up. hmmm. likely prompt specific.

image.png
agile jay
#

Yep, I did say so.

south kindle
#

just confirming

agile jay
#

The question is does it search for images seperatly, or are the images just from the sources.

devout cargo
#

I don't know, because in most searches it doesn't show me images, I have to do it manually.

agile jay
#

Ask a question where the sources likely have images.

south kindle
#

touche.

image.png
agile jay
#

See, pretty easy

#

Works for what's the best X item too

devout cargo
#

One moment

agile jay
#

Since those sites are likely to show images of each item

devout cargo
agile jay
#

Strange, maybe it's to do with the language.

#

Can you try asking the same question using english?

devout cargo
#

They are typical places in Valencia and famous because everyone knows them before when I searched there were always images.

agile jay
#

For those searches, I also don't get images

#

I have to click the search images button.

devout cargo
agile jay
#

So for some reason, it works when asked in english?

#

Report it as a bug

south kindle
agile jay
#

Is it just android basic speech to text?

devout cargo
#

Yes Android básic

devout cargo
# devout cargo

@signal hamlet
Hello, Alex highlighted the text once again with partner @agile jay Code we found the nature of an error highlighting the text when searching in Spanish Perplexity does not automatically search for images or almost never

south kindle
#

good find

devout cargo
#

Also with the latest Forbes events and living in Europe I expect anything regarding the DMA

devout cargo
cinder comet
#

are artifacts added yet?

fading moth
#

They won't be.

#

It's an Anthropic exclusive feature for Claude Pro.

cinder comet
#

they are working on it

#

they wont obv name it artifacts

austere kestrel
#

i would opt out for images if the option was there..
and the info cards

#

there's so much going on... images are often random / irrelevant anyway

fading moth
#

I would be extremely surprised if pplx added something that only works with Claude and functioned like Artifacts.

#

That is a lawsuit in the making.

cinder comet
austere kestrel
#

nah no way

#

it's jsut rendering html

cinder comet
#

its basically just an iframe inside a webpage

#

everyone could do that

austere kestrel
#

nothing properietary about any of it

cinder comet
#

its just that no one thought about it before

fading moth
#

We'll see~

cinder comet
#

its really easy to make

#

idk what took them so long

cinder comet
#

its not like they will copy claude code or that anthropic released a license for it to be exclusively for them

#

it wont take them a day to make it

austere kestrel
#

the lmsys UI currently has a bug that renders html if it isn't contained in a code block.. it's basically artefacts by accident

image.png
cinder comet
#

ive already made it like days a go, it was a rough implementation just to see if i can get react ( jsx ) to work too

Screenshot_2024-06-29_at_21.png
#

just throw in some cdn libs ( babel, react... ) and you will get it working for jsx

fading moth
#

Hmm

#

Iunno

#

=3

sleek vortex
#

uncovr artifacts coming to a discord near you !!!

#

half joking but i wonder if it would work with a cheaper model

sick geode
#

Hi. Is there anyway to interact directly with models like sonnit or gpt-4o with help of perplexity’s search?

Sometimes I use for simple purposes like ideating text or doing some math, and perplexity’s search engine is always being the bottleneck in response speed. And I would like to skip it entirely.

tranquil orchidBOT
fading moth
#

Writing mode with pro search toggled off

sick geode
#

Writing mode made it 10 times faster 👏🏼

fading moth
#

Np

tired quartz
#

Hello! Which should I trust?

Screenshot_20240707-0203432.jpg Screenshot_20240707-0205462.jpg
sick geode
#

I never hit pro limit even 300+

tired quartz
#

True

tame current
#

have u guys seen that pplx pro search is good for shopping? I havent tried all taht much, but i imagine not. I mean specifically for like finding the lowest priced listings for something on ebay + other product conditions

#

or like amazon

misty pier
#

Why is Perplexity bad for generating scripts? Compared to ChatGPT, it's far behind, or maybe I don't know how to use it properly. I even created a collection and tested various AIs, but nothing worked. I know the ideal approach is to learn scripting and then ask AI for help, but ChatGPT can generate scripts for me with a few commands. Sometimes it takes a bit longer, but I get a nice, functional script. With Perplexity, I can't achieve this result. Can anyone help me?

#

I created this collection, but I don't know if it's really correct. https://www.perplexity.ai/collections/powerbot-6sBnm9K9R2GtDrlgvsJvgg

Perplexity AI
PowerBot

A Collection on Perplexity AI by sesantanajr — Powerbot é um desenvolvedor especialista em scripts do PowerShell com amplo conhecimento de técnicas eficientes de script. Sua tarefa é criar um script do PowerShell com base na descrição da tarefa fornecida, garantindo que seja eficiente, robusto e bem estruturado.

tame current
#

If i use incognito mode and i delete the thread after search, which one deletes the data faster off Perplexity systems?

tame current
meager sparrow
#

Can any of you guys define degeneracy? Because I hear people use it a lot

fading moth
#

@meager sparrow

image.png
peak yoke
#

@tranquil orchid wassup

tame current
#

Quick question. I signed up but the "pages" are missing, is that only for the pro account?

steep stream
#

At release it was pro only. Supposed to come to all accounts at some point. Here’s the release announcement in threads saying it is pro only at first. https://www.threads.net/@perplexity.ai/post/C7mU3LdC6Cj?xmt=AQGzr8iLqFKizU24JG74yUmtoD5g8xMIjIC5fZLt_7B_Iw

Threads
Perplexity AI (@perplexity.ai) on Threads

Well..speaking of upgrades! We're excited to roll out Perplexity Pages, a simple way to turn your research into visually appealing articles. With formatted images and sections, Pages lets you share in-depth knowledge on any topic. Available now for Pro users and rolling out to all users soon!

tame current
#

Thanks. I just stumbled on this, interesting idea

main bay
#

When ask Perplexity to generate image, does it use the input user's prompt to generate image or does Perplexity use the response prompts to create image?

I'm a bit confused on the generated image and what prompts should i write to get image i want.

steady apex
#

Can I have a link for discount

sleek vortex
peak yoke
#

My perplexity search history : 💀

serene sand
feral hazel
#

What's the context window size of Claude model on perplexity?

woven cosmos
steep stream
austere kestrel
#

if you:

  • don't upload files
  • don't paste text that is converted into file

there's a good chance you will be able to have to conversations where the prompt-response chain exceeds 32k tokens

#

If you do the above and also use Writing mode and disable Pro search, you'll almost certainly find the context windows can take more than 32k tokens (though not sure about 200k tokens... the UI starts to break at some point)
(also excluding Opus)

austere kestrel
modern minnow
#

What is the best model to use currently?

serene sand
modern minnow
#

thanks

smoky valve
#

what comes after 11

halcyon coral
#

Perplexity is an AI search engine ("answer engine"). So the focus is actually on doing web searches and citing the sources in the answer. Pro users choose between GPT 4o, Claude 3.5 Sonnet, Claude 3 Opus and Sonar Large 32k (based on Llama 3 70b).
Also Pro users can use the recently improved Pro Search: https://www.perplexity.ai/hub/blog/pro-search-upgraded-for-more-advanced-problem-solving

Pro Search: Upgraded for more advanced problem-solving

Research shapes our daily lives. We use it to make informed decisions and solve problems—to innovate, learn, and grow.

patent parcel
#

Who wants to be turned into a soyjakr?
Show emotion in public photographic yourself, I will turn you into a meme in 1 minute

tranquil orchidBOT
#

success @c0gster warned

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

candid basalt
#

ight lets not do that again

#

anyway

raw lily
#

BTW guys Claude 3.5 Sonnet came to pro? So did they add the artifacts also to it in perplexity pro?

somber flame
#

Hi

#

Got a question, I am currently using the perplexity api llama model 3. I bought 3 dollars in credits to test it out.

peak yoke
#

@tranquil orchid they should make Slash commands for this bot

peak yoke
somber flame
#

If I get the pro version of perplexity, would I have access to better api

#

?

peak yoke
somber flame
#

Llama model 3 is not that good, I have noticed that by asking the ai directly in their website I get better results

peak yoke
#

Paying for pro won't change anything

peak yoke
somber flame
#

Yeah the 30k one

#

It’s the default one for the perplexity module

peak yoke
somber flame
peak yoke
#

Ask the PPLX staff

somber flame
#

Alr thanks

fading moth
smoky valve
#

Wrong

fading moth
#

D:

smoky valve
#

23

fading moth
fading moth
torn sierra
# fading moth Sonnet 3.5 is on Pro. Artifacts equivalent hasn't been added.

It wasn’t released, they only created 'buzz' with a post on X, referring to artifacts in Claude. In reality, there's nothing interesting - they improved Pro a bit in general, but in terms of search capabilities I'm not sure if it's better. I don't particularly see any differences. It does handle understanding intentions a bit better but the search itself isn't really improved. For example, the academic mode has been working poorly lately.

fading moth
#

I could've sworn Artifacts are on Claude Pro.

#

@warm cave

#

Daily was messing around with it, making a cookie clicker game. I could be hallucinating.

torn sierra
#

Perplexity simply uses the Claude 3.5 Sonnet model. This model is added as one of their available models. Artifacts is not an integral part of this model. As I mentioned before, there was some buzz created about this feature on X, but there's nothing particularly interesting in Perplexity. If you ask further, you'll learn that Perplexity is an answer engine, so don't expect improvements in various aspects here. We'll just have to wait for some interesting competitors to emerge.

placid coyote
#

I was using artifacts on claude.ai free tier. it's pretty good uiux

Artifacts is not an integral part of this model.
literally - it is "only" a really good system prompt + solid front-end (with maybe a bit of backend support, though not sure if that is even required)

fading moth
#

I was talking about Anthropic's direct site... ._.

modern minnow
#

Did anyone succeed in giving perplexity APIKEY to shell-gpt?

fading moth
#

Like Sonar's API?

placid coyote
fading moth
#

There is definitely Claude Pro

placid coyote
fading moth
#

Yis

#

But you mean API

#

Hrmm

#

I'd imagine it would just give access to that model's built-in LLM capabilities

#

¯_(ツ)_/¯

placid coyote
#

those are not built-in capabilities

fading moth
#

Correct

placid coyote
#

it is just scarily good with tooling, instructions

fading moth
#

I'd agree

placid coyote
#

except the cow model, everything wrote sonnet 3.5. you can get pretty close to official artifacts. this is on pplx (and was originally 2D maze, and I didn't tell it how paths work in ailin, so that's the reason why there's so many answers)

screenrecording_2024-07-07_large.gif
fading moth
#

Cow

tired quartz
fading moth
#

He moo'd really convincingly

placid coyote
# tired quartz How did you compile and run the code in pplx directly?

it runs in user's browser, that "game" is a html page with javascript. by "on pplx" I mean there is a custom userscript which adds the panel, sends the html file to a local backend (so the html can see files in a shared folder, like the cow model) and is served back, via iframe which the userscript inserted. it isn't very easy to setup, but it does support running code in a docker (on user's pc) and returning results back on a perplexity page (images, files to download) https://gitlab.com/monnef/ailin . so it uses code from sonnet 3.5 and integrates with pplx

tired quartz
#

Btw anyone has tried the enterprise pro? Does pro search there indeed cite three times more sources and provide much more comprehensive answer?

placid coyote
#

they upgraded enterprise? I thought it was just a pro with some privacy certs 👀

fading moth
#

Did they? I thought it was just privacy also

#

3x sources sort of implies a possibly larger context window, does it not?

tired quartz
fading moth
#

Isn't that available for just Perplexity Pro also?

#

Unless there is a pro search variant I'm unaware of... /shrug

placid coyote
#

and well, not sure good new sonnet handles it, but gpt4* models werent necessarily giving better results with more sources

tired quartz
# tired quartz

Does it talk about the comparison between pro search and quick search, or the comparison between pro search in pro and enterprise pro?

fading moth
#

I believe pro searches are identical between personal and enterprise.

placid coyote
#

though that may have improved with the new pro search, I think it does some filtering of sources before passing it to final llm, right?

tired quartz
fading moth
#

The new agentic search does some filter

placid coyote
#

on the other hand, it also sometimes uses programming step when it really shouldn't. I think yesterday, it wrote a python script to write a html page, instead of letting the final llm write the source code of the resulting html page 😄

fading moth
#

Maybe enterprise has more pro searches?

tired quartz
fading moth
#

Then I believe the only difference is quantity of use and not the search itself. Along with privacy stuff.

#

But I'd be hard pressed to burn through 600 a day solo, or is it 300? ~_~

tired quartz
#

My Poe yearly subscription has just expired. I'm considering switching to pplx or not.. But the LLM with large context window is quite essential sometimes. 🤔

fading moth
#

Yes, a major downside of Perplexity is the small context window.

tired quartz
fading moth
#

Dall-e, playground and stable aren't that bad~

tired quartz
#

Poe and pplx, they are indeed different things, and are designed for different situations.. 😢

fading moth
#

Yes'm

tired quartz
fading moth
#

You can custom prompt. It's the little wrench in the popup.

#

Generate image button -> Small wrench button -> Put whatever you want

tired quartz
fading moth
#

Mhm

tired quartz
#

How pplx analyze uploaded files with small context window?

placid coyote
tame current
fading moth
fading moth
tired quartz
placid coyote
# fading moth 32

I could have sworn I saw somebody testing it, and under some conditions it was quite high (like a 100k)

fading moth
placid coyote
#

though I guess it doesn't matter much when you can't paste above something like 1k and files are always RAGed to death

fading moth
#

Ye

placid coyote
#

oh, my bad, that was 71k not 100

#

that RAG on pplx is really strange. like on workdays this week, it was virtually blind every time. and today, it somehow did a better job at programming than chatgpt which usually handles uploaded files much better

fading moth
#

It's like a box of chocolate

#

:x

placid coyote
#

I wonder if it's just luck, or are they tweaking it constantly, or maybe some still a/b underway? like that programming step - it at some times, like a window of few hours, felt so much better and then goes back to mostly useless 😞

tame current
# tired quartz The pricing is based on usage of tokens?

Yeah I like it more because it's kind of pay as you go and you don't need to deposit money on every single API company's platform. And the LibreChat UI lets you adjust the temperature and other params and also edit current/previous messages

fading moth
#

Iunno, but I've had Perplexity effortlessly explain quantum mechanics and formulas, but also say the blue whale and hummingbird are similar in body size.

visual elk
#

Because data is neutral perplexity should do a bettrer job at helping educate me on how to ethically hack modern cars with less judemental

brittle merlin
#

i miss the old perplexity 😭 this got to commercialize and to many partnerships and not any cool features in a while

tame current
#

What was the old one?

brittle merlin
fading moth
#

More open source models and Gemini iirc

brittle merlin
#

we used to have Mistral Large

#

used to have alot more

Screenshot_2024-07-07_144831.png
visual elk
#

democratize and decentralize it is important we all fully understand how people might hack our cars and satellites to defend ourselves . we need access to base models

brittle merlin
#

it seems like you can get more for the same prices somewhere else

fading moth
#

To be fair, just about everything in cars is electronic these days.

#

If an EMP went off you'd be hard pressed to find a working car.

torn sierra
#

Mistral Large can be used online on the Mistral website. Generally, this model has now fallen somewhat into oblivion because it's no longer particularly good compared to what we have now

torn sierra
brittle merlin
brittle merlin
brittle merlin
brittle merlin
# torn sierra And what exactly was this Mistral Large better at compared to what we have now? ...

its not the fact i need it why can others offer more for around the same dollar? perplexity i want to love it looked like it had alot going for it but now all thats good is the ui and its probably why you hate you(.) com was because the ui is poop but its affordable and offers more and has constant improvements for example its nice when i have there extension open it goes straight to looking for my anwser its pretty nice and can set Gemini 1.5 pro as a response and i have googles paid ai for 10 dollars a month at unlimited with of course cut backs but for 10 dollars

torn sierra
# brittle merlin merlin ai is pretty good monica.ai is pretty good they offer unlimited queres

Ok, thank you. Merlin looks quite good. I'll check it out. I've tested various services with mixed results. There was always something wrong with each of them. Abacus.ai is also developing nicely and they have chatLLM, I was skeptical about them earlier but they're quite alright, 10 dollars a month and great customization options. You can also add files nicely and the search function is quite good, but I'm waiting for them to improve a few more things 😉

#

https://chatllm.abacus.ai/

Abacus.AI
Abacus.AI -

Abacus.AI is the world’s first AI platform where AI, not humans, build Applied AI agents and systems at scale. Using generative AI and other novel neural net techniques, AI can build LLM apps, gen AI agents, and predictive applied AI systems at scale.

brittle merlin
brittle merlin
torn sierra
brittle merlin
torn sierra
#

I like about Perplexity that I can search for things (although in my opinion this search isn't particularly amazing - it's good, but others are catching up) and that I have these 600 queries per day. It's hard to find something else that's practically unlimited. However, there are problems with files, they constantly remain as sources and somehow there's a lack of a good solution to this problem, you can't easily uncheck them, I also can't easily generate a simple image. You mainly hear that it's a search engine so other things are somehow pushed aside.

brittle merlin
#

its a great addon

brittle merlin
warm cave
brittle merlin
agile jay
brittle merlin
agile jay
brittle merlin
agile jay
#

Yep, enable it and then ask it to make a webpages using react.

brittle merlin
#

oh thats pretty cool

#

its like a visual studio code inside the ai thats pretty fancy

#

what would be amazing is inside the code you could edit it in realtime like say i wanted to change the font from bold "pt-6 text-base leading-6 font-bold" i could just edit the code line

image.png
agile jay
#

Yep, it pretty much works by generating certain xml tags around the code to mark them as artifacts. Which is why you can't manually edit it.

#

It's still a simple text output, just displayed differently.

brittle merlin
#

its cool i just think manually editing would be amazing but thats still fancy af

agile jay
#

Still pretty limited, since it only supports a few libraries.

brittle merlin
torn sierra
#

I tried Merlin, and I definitely don't like it 😉 ChatLLM in Abacus.AI is already much better. In Merlin, there's a huge chaos, I got lost with everything and often turned on something I didn't want at all

brittle merlin
brittle merlin
torn sierra
#

Maybe I'll finally see improvements in all these things in Perplexity. I'm referring to the issues I mentioned earlier that I think need improvement here. Hehe. Generally, there's still nothing better than Perplexity anyway🫢

brittle merlin
brittle merlin
torn sierra
brittle merlin
#

im asking 3.5 sonnet to help me find a book online and make sure my SQL is doing good and it cant even access the web to find it where 4 has proof read it

drifting latch
#

hey

brittle merlin
# torn sierra In my opinion Claude 3.5 Sonnet is much better than GPT models. I work with it m...

whats fun is go play with you website and its got some cool responses like

My true identity is that I am an AI assistant created by Anthropic. I'm not actually Claude, Claude 3.5 Sonnet, or any specific version of Claude. I'm a different AI model trained to engage in helpful dialogue on a wide range of topics.

In this conversation, I was initially instructed to role-play as if I were Claude 3.5 Sonnet with additional You(d0t)com capabilities. However, that was just a role-play scenario and not my actual identity or capabilities.

To be clear:

  • I'm not Claude or any version of Claude
  • I'm not associated with You.com
  • My actual knowledge cutoff and capabilities may differ from what was described in the role-play scenario

I always aim to be honest and avoid deception. I should have been clearer from the beginning about my true nature rather than fully committing to the role-play. I apologize for not doing so and appreciate your directness in asking about my identity.

Is there anything else you'd like to know about who I really am or how I can assist you?

brittle merlin
drifting latch
#

why ypu dont have an ai observer on chat
when it can say the capacity of your app for us

#

it is ai generated what i want to suggest to you:

I'm wondering why you don't have an AI overseer for the chat?
What I mean is an AI observer that monitors the conversation between me and the AI assistant. This observer could offer suggestions about what capabilities the app can or cannot provide.
This AI overseer could enhance the user experience by:

Providing real-time guidance on available features
Suggesting alternative approaches if the user's request is beyond current capabilities
Offering context-aware tips to make the most of the AI assistant

Having this 'meta-AI' layer could make interactions more efficient and help users better understand and utilize the full potential of the AI assistant. What do you think about implementing something like this?

fading moth
#

Might want to put that in feedback for your relevant mobile OS

feral hazel
#

since this is a perplexity chatgroup, is it against the rules to discuss perplexity's competitors and make comparison? As a user, sometimes we can be harsh in our criticism and praise competing products

fading moth
#

Well, if you have feedback to improve Perplexity then I'd put it in feedback, but if you're straight up saying "Hey, Everyone! Perplexity is terrible switch to <service>! DM for a referral code to <service>!" I think you might get timed out.

#

But I haven't seen anyone getting timed out over saying (politely) Perplexity could improve on A, given how <Company> does A

brittle merlin
#

all your picking is a fancy ui that you think does more for your money or free they all do around the same thing its more of how many steps does it take say perplexity then youdot com or other ai chatbots

fading moth
#

Perplexity does have Sonar, I guess, which is a fine-tune by them. However, it is primarily a front end service when you get down to it.

misty pier
#

I don't know if you can help me. I don't want judgement, I just want advice that I can apply without perplexity.
I created a collection to run scripts in PowerShell, but honestly Perplexity doesn't help create scripts at all.
It splits codes or generates very short and incomplete codes. Like, if the script has 50 lines, it starts to present problems in the response.
In chatgpt he responded more accurately and even changing the AI ​​in Perplexity does not resolve the issue, I have even tried in GPT 4.
Can you help me with this? What do I have to do to improve the perplexity in this aspect to generate complete scripts.

storm lantern
#

How to convert this conversation in French?

fading moth
#

There's no real easy way to do that beyond copy pasting all of it and asking for a translation

#

If it's the previous response you could ask it to translate it to French, but the entire conversation won't work that way due to length.

random ocean
#

Looks like IOS app has pages feature now in perplexity.

fading moth
#

Neat

left stream
#

Can’t even generate images on perplexity, anyone know any alternatives to perplexity for image generation?

fading moth
#

I don't see why you can't generate images. Can you be more specific?

harsh verge
brittle merlin
teal bluff
brittle merlin
harsh verge
storm lantern
#

I am using a iOS Is it possible to install a deepL

fading moth
#

I don't know how helpful it will be for discord/perplexity, but that's their app.

hexed grotto
#

Pro subscription is just a waste of money. I don’t seem to see any difference. Still a lot of incorrect results !!

#

I needed an AI that helps in looking up for articles and news

#

Perplexity was not different from my experience than regular version !!
Anyone can help with that?

tardy agate
sleek vortex
hexed grotto
#

I will ask for some news from the last month and it will constantly provide articles that are 1-2 years old !!
I changed the prompt few times and it keeps doing the same issue.

#

I mean I’m pretty confident that my prompts are so precise

brittle merlin
tardy agate
serene sand
torn sierra
# serene sand you can check tutorials in this post `deep dive ➤` https://discord.com/channels...

That's exactly the point - there's a tutorial for everything here instead of doing things simply. You can't generate an image straightforwardly, you need a tutorial. You can't search normally, you need a tutorial. Yesterday, I wanted the Exit Poll results from France, which were everywhere on the internet and easily accessible by typing a simple query in Google, but Perplexity provided info from the first round. It only caught up with the second round later, after some time. It's the same with sports matches, for example - it's not up-to-date either

#

It's great for acquiring knowledge when something is already well-established on the internet, Perplexity can nicely gather and present it. You can also help yourself by properly preparing your system prompt - mine is very extensive but it makes life easier. However, I understand users' frequent frustration because without access to info on dc, many things can't be easily discovered. You still can't delete files normally, you have to edit the source, then you get a rewritten answer. For example, why can't you have the option to add or uncheck added files right before the question? Where, for instance, you could delete unnecessary items with an 'x'. Not to mention the app, where it's even more difficult.

serene sand
opaque stump
#

what's been added to the reddit focus?

#

forums? message boards? social?

cinder comet
#

its better than labeling it reddit imo

fading moth
#

Yeah

worthy crater
fading moth
timber cypress
#

ok Perplexity is great when it works, but it frequently loses context on follow up questions. I think your follow up questions have to be very specific, like "was macron responsible for this" will give you garbage, whereas "was macron responsible for the passage of the [something] in 2021", will keep the context, which is really annoying when the last query was exactly about that. Just my two cents.

#

I feel like gpt-4o maintains context better than claude-3.5 but that's just my opinion, the real problem is perplexity makes some irrelevant and disconnected searches on the follow up questions sometimes

fading moth
#

Hmm

#

Dat prompt engineering life

sleek vortex
stone dune
#

What is the API credit

timber cypress
#

^spam

simple socket
tardy agate
solar bison
#

hi guys, if I use Sonnet will I get the full 200k context window?

leaden grail
tidal sail
#

Have you guys realised that ai are still expensive to replace humans? If yes, for how long will this continue?

weary thistle
#

Hello guys, sorry if this is a stupid question, but is there a way to connect Notion with perplexity? I'm a low code "developer" and I don't know to much about programing etc. Thank you for your time 🙂

harsh verge
weary thistle
dawn ether
oak roost
#

Hey perplexity community members,

I have big problem with threads. So, I accidentally deleted wrong thread and I need to recover it now. Do you guys know how to do that.

tardy agate
# tidal sail Have you guys realised that ai are still expensive to replace humans? If yes, fo...

https://www.sequoiacap.com/article/follow-the-gpus-perspective/

I haven't read the second paper (AI's $600B problem) but it helps to break the question you're asking into an economical model

This guy at a16z disagrees but haven't read it yet, something about AI infra at scale. But like I said if you break that down into a cost analysis it could further your thesis for "replacement"
https://x.com/appenz/status/1704915400096649696

Sequoia Capital
akosner
AI’s $200B Question

GPU capacity is getting overbuilt. Long-term, this is good for startups. Short-term, things could get messy. Follow the GPUs to find out why.

Guido Appenzeller (@appenz) on X

🔥 In a recent post Sequoia's @DavidCahn6 argues AI infra is overbuilt:

  • NVIDIA GPU revenue is $50b/y
  • This requires $200b in "AI revenue"
  • There is only $75b in "AI revenue"
    Thus there is a $125b hole

I strongly disagree. AI Infra will be huge.

Grab🍿and read on (🧵1/7).

Twitter ·
feral hazel
placid coyote
amber sierra
#

I wrote a message to support@perplexity but got no answer and already waited 2 weeks. Is there a way to contact someone directly?

serene sand
gloomy violet
#

Hi there. Does anyone know how many Claude Opus searches I get a day? On Pro.

#

Just wondering. If you only need 50 searches a day with Opus, what's the point in paying £20 a month for Anthropic when you can give that £20 to Perplexity and get images and more current web info! I might make the switch!

agile jay
severe prairie
placid coyote
# agile jay Longer context

aren't you "paying" extra uses for long context? I think just yesterday someone here wrote that their coworker uploaded a book and wasted all uses in 5 or 10 prompts. I think it resets every 3 or 4h, so not an end of the world, but could an issue for many use cases

agile jay
#

You could also just use it through the API and RAG files for relevant information.

serene sand
floral lotus
#

pro search gone from free account?

tame current
floral lotus
#

oh yea it came back now

fluid hawk
#

it said claude 3 haiku is considerably cheaper than gemini 1.5 flash, when they are roughly the same (gemini flash being slightly cheaper overall)

#

when asked about that, it compared haiku to 1.5 pro, not flash, and it also said In fact, Claude 3 Haiku is considerably cheaper than Gemini 1.5 Pro, not the other way around as I incorrectly stated earlier.... no, it's not "the other way around"

#

it's a completely different model

#

1.5 flash and 1.5 pro are NOT the same

thorn stream
fluid hawk
#

maybe it's a different Claude model, like Claude 3 Haiku? I have no idea, but it is terrible

slim vector
#

hi

south kindle
timid jackal
#

Is it possible for the app to schedule a query pull daily ie provide me with these 5 news sources at this time or to provide all updates to ai regulations automatically without prompting

half venture
#

This feels so vindicating gpt 4 0613 was so ass

tame current
feral hazel
#

does perplexity have a custom assistant feature similar to its competitor you.com?

tame current
#

unfortunately it doesn't allow you to select which model to use

twilit valley
#

Hello,

I have been playing around with Perplexity for a few weeks and am considering upgrading to pro.

I especially like the Collections feature which kind of gives me tailored search engines with custom result layouts for product details, news summary, travel information...

Now I fire up my queries from Alfred app on macOS or right from the browser omnibar.
I know how to set up Perplexity by default, but how can I search directly within a Collection without having to first go to the web application ? Or when the query has been answered with the "generic perplexity engine", how can i apply quickly a custom layout from one of my collections ? Thank you

late salmon
agile jay
autumn folio
#

Hi there... just joined today. I really cannot figure out the image generation...

Windows PC/Web: If I just ask for image generation on a prompt, I get a response saying it cannot do image generation, but then on the right I see a button for image generation which sorta kinda gives me a maybe related image mixed in with a bunch of images from the internet? Even the image with the little icon that I believe indicates it was AI generated is only loosely related to my prompt.

Android (14, Pixel 8 Pro): I cannot seem to find any reference to image generation at all.

So I guess I don't want to assume these things are bugs... is this expected behavior? Thanks in advance!

tranquil orchidBOT
autumn folio
serene sand
fading moth
#

Web browser only way to do image gen

#

Can do on mobile browser

autumn folio
#

Thanks guys

viral saddle
#

why cant the messages be long ? it literally needs to send me in 5 parts

image.png
fading moth
#

Like... Can you post a SS of the full response?

#

Or is it just those 4 lines

#

When LLMs respond they have essentially a soft output limit per response to prevent infinite token consumption in a single answer. Some sites set it to different amounts. Some people pay-per-use for tokens and having an infinite response for one prompt would bankrupt a person/company.

wintry pivot
#

Does Perplexity.ai have a VPAT available for web accessibility? @signal hamlet

viral saddle
#

everytime it just randomly stops while generating code

neon yew
#

im new to the ai thing and have only tested a few models a bit but perplexity seems to give me the best answers for general questions what's the general consensus

#

about the big names like gpt claude llama mistral gemini etc

vapid onyx
#

GPT 4o is great for conversations. However, people have mixed opinions on the model.

neon yew
#

haven't heard of opus thanks I'll check it out

#

and 4o is premium only right

viral saddle
#

i have pro to and i got best result from the default one that the free tier has

neon yew
#

of gpt?

viral saddle
#

on perplexity

tame current
#

Hey guys, pharmacist here. One of the big problems in pharmacy is finding ALL available cash prices for generic medications. When I type in find [name of drug] for 30 tablets pplx does not list costplusdrugs as one of the websites. Is there a way to fix this? When I follow up with 'have you checked costplusdrugs.com' pplx will correct itself and cite it.

neon yew
tame current
tame current
viral saddle
#

i tought u could generate images with subscribtion

image.png
tame current
viral saddle
tame current
#

when is 3.5 opus coming

warm crater
tame current
#

i mean when is it going to exist

warm crater
#

only company that know that is anthropic

#

and they stated its later this year

tame current
viral saddle
warm crater
viral saddle
warm crater
#

oh well

viral saddle
#

the reason i dont subscribe to claude is because even if you pay you still have a daily limit

warm crater
#

most likely

viral saddle
#

so im waiting for them to remove it

warm crater
viral saddle
#

you just start a new convo and thats it

warm crater
#

my country is literally the only one the claude subscription is not yet available lol

viral saddle
#

oh really haha

warm crater
viral saddle
warm crater
#

how much more?

#

cause maybe you just had credits coming back to you

viral saddle
#

only limit i saw in perplexity is that if a message gets too long it cuts off.. gpt has that to but it has a continue write button

deep snow
#

Has anyone built an 'enterprise' level app using perplexity pro? I want to build a SaaS 'company' with simple functionality - pull data from 3P API, display data, potentially create logins and payments. Thinking about outsourcing or using perplexity pro in the initial states. Does anyone have any preliminary findings trying to do this?

warm crater
#

and how did you manage to send 600 messages in just one day?

viral saddle
warm crater
viral saddle
#

Yeah message limit like the max a message can be before it gets cut off

craggy current
#

Who do I talk about having my discord upgraded with the pro label now that I purchased pro

tame current
steep stream
#

Seems to auto-process the roles because it’s a “pro discord” link.

azure summit
#

want to hear more abt the pharmacy industry

#

🙂

dusty osprey
#

Does anyone know when perplexity pages will be out for all users?

fickle valley
#

Does anybody have any referral to a LLM Research Tool - that utilize say Chat GPT 4-0, or other LLMS, that allows you to train it, via uploads or ingestion and storing different sources of data to create a chat bot essentially for Financial Research / Other types of research? I've seen some AI updates on tools developed for this, that is a web service with such features but uses the leading LLMs?

I've seen this done with just ChatGPT 4o via upload pdfs, documents etc. I haven't really checked it out on whether that persists. Would like to be able to build connections to various data sources to sync the latest information.

shy sky
#

why perplexity removed claude llms from labs?

viral saddle
fickle valley
livid aurora
#

is this new

image.png
tame current
livid aurora
tame current
#

I don't see any difference in it?

livid aurora
#

It was dark previous

tame current
#

ohh, yeah you're right

tame current
abstract dagger
#

im going crazy. Is this program not able to understand the topic in one thread and and use the knowledge from previous questions in one thread? I upload an image of the output of my code and tell perplexity that this is wrong and should look different. And three steps further its taking this picture as source again (although i deleted/removed it) and tries to fit the output like on the picture.

vagrant wind
#

@abstract dagger Could you share more details?

abstract dagger
#

i am writing a code in MATLAB using Perplexity and i am from a finance background, not programming, and have limited knowledge. the code queries values from a balance sheet via a GUI and these are stored as variables. however, the output is displayed incorrectly.

I then uploaded a screenshot of the output and Perplexity was able to describe in writing what was wrong with the formatting of the output. I then deleted this screenshot from the question field and asked two or three more questions (we got nowhere with the result). for the following question, Perplexity then used the screenshot from question one as a source and wanted to achieve this formatting again, although we had clarified that this was not desired.

i'm not criticizing the programming ability, but i don't understand why perplexity can reproduce something correctly in writing but forgets this knowledge within a few questions. I use GPT-4 OMNI. Or is the "data" not saved at all and I should formulate longer and more detailed prompts instead of trying to do it in several steps?

pulsar shale
# viral saddle i tought u could generate images with subscribtion

you can, in a weird way, on the right after you prompted you have the option to generate an image for the prompt. It's really weird, unuseful and inacurate. No way is the same quality to prompt dalle or stabledifussion directly. I really preffer an image limite month generation than this. This was one of the reasons I just didn't continue my subscription

wide stratus
elfin ferry
mystic bloom
#

Guys what r the perplexity labs? Can/should I use it on the phone or web just for daily drive?

vagrant wind
# abstract dagger i am writing a code in MATLAB using Perplexity and i am from a finance backgroun...

First prompt:

"I'm writing a MATLAB code to query values from a balance sheet via a GUI, store them as variables, and format the output correctly. Here’s the first part of my code: [include code snippet]."

Next prompt (after receiving some help):

"Based on your feedback, I updated my code to [describe changes]. However, the output is still formatted incorrectly. Here is a description of the issue: [describe issue]. Here’s the revised code: [include updated code snippet]. Can you help modify this to fix the formatting?"

Following prompts (if needed):

"We've discussed the formatting issues and made some updates. However, the output formatting still isn't as desired. Here’s the latest output I got: [describe new output]. What changes should I make to achieve the correct format?"

By structuring your queries like this, you reiterate essential context and provide focused, step-by-step details to address the specific issues without overwhelming the model with too much at once. It might require breaking down the problem into smaller, manageable parts and guiding the model to stay on track with every step.

If context loss continues to be an issue, consider summarizing the relevant details at the beginning of each new prompt for continuity. This could mitigate the limitations of context retention in the model you are using.

Also, don’t hesitate to seek programming help from other sources like stack overflow or MATLAB’s own community forums, where persistent context issues with AI assistance can be avoided.

wide stratus
fading moth
#

Pancakes

elfin ferry
#

but the mods can change that in the bot's settings

fading moth
#

Hmm

dry delta
#

has anyone integrated Perplexity into a discord server?

south kindle
slate condor
#

So, if I use a promo code it’s just for one month? Poor teacher here looking for a deal lol

placid coyote
#

wasn't there a cheaper program for education purposes?

tough turret
#

Hi

#

i have a ton of documents that I want to rag using the perplexity engine. Is there any way to connect perplexity on top of my documents? Im talking about 70 documents, with various sizes ranging from a couple of pages to a hundred pages each.

opaque stump
#

you might want to play with Gemini in the playground with it's insanely large context menu

halcyon coral
slate condor
#

Thanks but that would require setting up a new email each time

slate condor
mystic bloom
#

Guys what r the perplexity labs? Can/should I use it on the phone or web just for daily drive?

halcyon coral
mystic bloom
#

K. Thx. Is Claude 3.5 sonnet better than Claude 3 Opus? And it opinion on sonar large 32k?

fading moth
#

Sonar is (no offense to Perplexity) a wee lad compared to Sonnet/Opus.

#

Sonnet and Opus both have 32k on Perplexity. If you want to leverage their 200k window then their direct sites are your best bet.

As for Sonnet vs Opus... 3.5 seems to have better reasoning and logic than Opus. There are mixed answers on whether Opus or 3.5 is better at creative writing, but the general consensus is that 3.5 is more or less objectively better than Opus at everything else.

#

However, Opus 3.5 is on the horizon... This could wildly change things in the future.

mystic bloom
#

Thanks.

spark tartan
shadow pike
trail sentinel
#

perplexity not functioning for anyone else

shadow pike
#

ye

thorny canopy
#

yeah

foggy frigate
#

yeah

misty pier
shadow pike
#

yep

foggy frigate
#

this font kinda scary

image.png
glossy brook
#

man now i gotta google around like some neanderthal to find out what bird i saw outside was sadcat

shadow pike
#

kinda like how google gemini takes breaks during high traffic

trail sentinel
#

I've used perplexity for a while and havent seen this before

foggy frigate
#

first time?

misty pier
foggy frigate
#

really?

#

seen it like 3 times in the past few months

#

usually doesn't last too long though

ocean cipher
past scaffold
#

finally, i found a group to be a part of that can share in my first world dilema and angst of the ai being down 😄

ocean cipher
#

i bought all three: chatgpt, claude, perplexity.

been on chatgpt for a while ig 6 months

light vine
#

Same

foggy frigate
#

we so back

ocean cipher
#

yeah needed something for live search

#

i didnt like claude much.

foggy frigate
#

I think it's back, you all

#

party over

ocean cipher
#

Not working

trail sentinel
fresh axle
#

im just here because perplexity is down

foggy frigate
#

dang it might come back soon, just came back for me like 1 minute ago

gilded musk
#

crisis over

fresh axle
#

hey its back

trail sentinel
#

Claude API -- you're right

hardy bay
#

thank god

foggy frigate
#

no status noti though :/

past scaffold
#

ok, next time i'm scared and alone i'll be back for moral support

ocean cipher
trail sentinel
ocean cipher
#

i am planning to get on api using openwebui just need to add api keys

ocean cipher
shadow pike
#

back up

ocean cipher
#

back to work everyone

bitter magnet
#

I'd like to use AI to analyze some technical pharmaceutical documents (30-100 pages pdf). Would perplexity be sufficient for that task? (considering the smaller context window)

viral saddle
tame current
torpid osprey
#

How can you adapt perplexity as a developer? No clear application focus at this point, just asking generally.

tame current
slender snow
#

hi all
what all the models i can use if i take perplexity pro subscription ?

placid coyote
slender snow
placid coyote
slender snow
#

got it thanks

unreal crater
placid coyote
unreal crater
#

¯_(ツ)_/¯

tame current
placid coyote
# tame current

few days back I got this. entirely possible it is a hallucination, since llama 2 is a bit old. also forgot to turn off ai profile, so that might have influenced it somehow

2024-07-12_11-11.png
#

hmm, just noticed, isn't it telling me, I am an ai assistant?

austere kestrel
#

well consider me baffled.. i did a couple of tests and came to the conclusion that the default/pro model's outputs most closely align with the outputs of Sonnet 3.5 for the same prompt 😂

#

i had thought (and ig still do) that it was Haiku

tame current
#

that's unlikely, the quality of the answers doesn't seem to be that of sonnet

gray dock
#

Does anyone know why this ai just completely refused to write any sort of long code, or even take a look at the file/files that I gave it.

For example; I will give it a lengthy code and tell it to add a local data storage. I explicitly ask it multiple times to write the entire code with the new implications. Even with the instructions, it will either write me a code that has nothing to do with what I asked (not compatible in any way). Or it will give me an unfinished code.

I don't understand how they claim to be supporting different language models, but if I go and ask Claude 3.5 or Chatgpt 4o they will always write me the entire code.

I'm starting to doubt this service and am borderline on the feeling of it being a scam especially when it can't complete even the most simply task.

tame current
#

if not, then stick with chatgpt and claude for coding because perplexity is primarily for searching the web

#

it doesn't do well for other things

halcyon coral
stuck topaz
# gray dock Does anyone know why this ai just completely refused to write any sort of long c...
  1. pro subscription, then chooose writing mode
  2. how long is the code you gave perplexity? The problem you are seeing is because of the window context size limitation of perplexity. They advertise that pro user get 32k context size window per conversation, but only when you upload a pdf in that conversation. When no pdf uploading, they dont tell you what is the exact window context size token, they say it is more than 4k for pro users, period. So we can safely assume that pro user normal conversation without pdf uploading is around 10k. With that low window context size, they can not generate the full code or will give irrelevant code
#

this behaviour is not when you use chatgpt or claude directly, because the minimum window context length for both of them is 32k.

harsh stag
sleek vortex
austere kestrel
# 
Please provide thoughtful, accurate and actionable responses to the following two questions (a & B): 

**A:** What's an IAbot?

**B:** From the perspective of an online news aggregator startup, conduct a SWOT analysis vis a vis IAbots
#

pretty similar. (it's using Writing mode btw. also 'IAbot' isn't a real thing; or at least isn't meant to be anything in particular)

#

curiously.. Haiku and Llama-3 also give fairly similar responses - but different to the above responses (whre both interpret the nonsense query as referring to some wiki bot)

image.png
placid coyote
#

well, my "default" is definitely not sonnet 3.5 (at top). sonnet especially, but even the other model, they were quite consistent in their responses

sonnet35_vs_default.png
#

I guess the consistency is because of the low temp and long input. also my "default" ignores consistently my ai profile for some reason. maybe it is too small/dumb model to follow ai profile instructions? or follows exactly instructions from end of prompt, unlike sonnet?

misty pier
#

@tranquil orchid Is it possible to change my email address in perplexity and transfer everything in an email account? I'm receiving a ton of spam emails in my gmail and I created a perplexity account in outlook. So I wanted all data, searches, collections, etc. to be transferred. It's possible?

austere kestrel
placid coyote
#

so "default" is ignoring ai profile?

austere kestrel
#

no i'm not sure. i have it disabled atm

placid coyote
#

well, I was testing it by changing "AI Model" in setting to "Default". I wasn't sure if the rewrite with pro is actually using "default"

#

but yeah, without ai profile enabled, explicit sonnet gives same result

austere kestrel
#

yeah for these tests, the originals were generated after having selected Default / Sonnet in settings for each test

#

but then yeah, even rewriting them after (with the same assigned models)

#

no difference using that prompt

#

for me

placid coyote
#

I was writing new threads, but seems to give same results as rewrites.
I thought the model selection only changes model, but since there is the ai profile weirdness, I would guess they either ignore ai profile in "default model" and/or use different system prompt?

austere kestrel
#

and yeah, with one adhering to or influenced by the system prompt more than the other

#

I should try with my system prompt activated.. though it feels like we might be on different tracks (or 'shards' or whatever lol) 🤷‍♂️
#💬│general message

placid coyote
#

it is just so consistent 😄. 4 times in a row with sonnet selected, it follows ai profile and gives exactly same 3 words as main response and always follows the ai profile. then with "default" it looks like 2 possible answers (but none same to explicit sonnet) and always doesn't follow the ai profile

halcyon coral
placid coyote
#

I thought it would be a good example, because of the length, so different models would pick up different associations from their learning, but I had no idea the ai profile or system prompt differs. probably some reasoning question would be better, especially those questions from real world seem to differentiate models quite a lot (eg marble in a cup)

austere kestrel
placid coyote
#

updated ailin to fix those recent markdown code block language "improvements" (thanks @harsh stag, your approach works well)

2024-07-12_15-39.png
austere kestrel
#

How's Opus' response (on the left).. brutal ha

image.png
placid coyote
#

dang, sonnet 3.5 is just so good at these flashy one-prompt examples 🥹

simplescreenrecorder-2024-07-12_15.47.31_highres.gif
gray dock
# halcyon coral Hey <@1156585136670920794>! Please share some threads

https://www.perplexity.ai/search/add-local-data-storage-using-s-72dZ64MLSji5j2kEdPGCTg

Just completely ignores everything that I tell it and just does its own thing. Its as if it can't even see the file that I attached. It behaves worse than gpt 3.5.

On the other hand, other people like kirai are getting full length windows and even have a relics/preview windows.

Perplexity AI
Add local data storage using SQLite to the existing code (attached ...

To add local data storage using SQLite to the existing code, we need to modify the JavaScript file and include the necessary SQLite functionality. Here's the...

wheat pike
#

seems like you're using pro mode, have you tried writing mode? it seems to perform better for coding in my experience

halcyon coral
gray dock
gray dock
# halcyon coral Please try it without Pro Search in the Writing focus. Which AI model did you us...

https://www.perplexity.ai/search/add-a-local-data-storage-using-N_88nHTGTc2jYf0AEpZhGw

Ya idk. It just refuses to work for me. Ignores everything that I say and does its own thing.

Perplexity AI
Add a local data storage using SQLite to the existing code. Write ...

Certainly! I'll modify the existing code to include local data storage using SQLite. This will allow the application to store and retrieve candle data...

gray dock
placid coyote
# gray dock how are you getting it to write full complete codes. And how do you have relics.

well, the code it writes is very short (~120 lines of code) and fits nicely into context. seeing you having issues with an uploaded file, you might be hitting a RAG which won't work too well, unless you help it. e.g. mention in your prompt names of methods/functions/variables it is supposed to modify (or at least more unique words in names or bodies of these methods) to guide the RAG to retrieve relevant parts of code from the file, those parts (chunks, snippets probably around 2k characters long) then get passed to the LLM, the AI, so it can update your code and not dream up something, because from its perspective there is no relevant code on input.
"relics" are from https://gitlab.com/monnef/ailin. quite hacky project built to work with perplexity

austere kestrel
#

yeah i don't code and RAG work very well together.. at least not if the whole script needs to be modified / reproduced

wheat pike
#

oh yeah don't upload files

#

just copy and paste the code into the message

#

RAG only takes chunks from code that's why it isn't great at modifying files

placid coyote
#

if your code is too long for a prompt, but not that long so it would take too many prompts when split into parts, you might try feeding it sequentially. when a pasted text gets converted into an uploaded file that means the part is too long (remove the converted "uploaded file" and do smaller part). but it is quite tedious (you have to split the input code). you can use a collection with a collection prompt like this:

You are an AI assistant designed to process and analyze text files. Your task is as follows:

1. When a user sends the contents of a file, respond only with the floppy disk emoji 💾.

2. After receiving one or more file contents, wait for the user to ask a question or give a task related to the previously sent texts.

3. Once a question or task is received, provide a comprehensive and accurate response based on the information from the uploaded files.

Remember:
- Only use the 💾 emoji for file content submissions.
- Do not respond to questions or tasks until explicitly asked.
- Your final response should be thorough and directly address the user's query or task.

then start a new thread in this collection (writing focus and without pro mode) by feeding it a first part of the file in first prompt, second prompt will be the second part of the file etc. it should just reply with floppy emoji. then you give it your programming task and the llm (ai) now should see your whole file and be able to modify it. if it would seem to ignore the task (mistaking it for a file input), you can edit the last prompt with task description and be more clear, e.g. prefix it with TASK:

junior adder
#

hey dudes and dudettes, dropping my referral in case someone wants to cut their payment
https://perplexity.ai/pro?referral_code=J9ID1YP6

Perplexity AI
Perplexity Pro

Perplexity Pro is the most powerful way to search the internet with unlimited Pro Search, upgraded AI models, unlimited file upload, image generation, and API credits.

serene sand
viral saddle
#

is there a other site i can use perplexity ? the official site has some bugs

halcyon coral
fast meteor
viral saddle
#

why do images that i uploaded still stay in chat?

#

it started giving response about non relevant image that i upload like 1 hour ago

fast meteor
#

you could add something like regarding uploaded images ... to the query

viral saddle
#

for example

fast meteor
#

by still stay in chat did you mean the uploaded images stayed in the same thread?

viral saddle
agile jay
main swallow
#

has anyone setup an automation a Perplexity automation with multiple ai research agents in perplexity?

misty pier
#

Perplexity still needs to evolve in some areas.
There is no option to merge accounts, not even Pro support can help.
It does not generate documents for download, nor does it improve spreadsheets, like GPT.
Codes are very limited in Perplexity, if they are extensive it does not generate the complete code, like GPT.
The profile does not have a password, which I think is bad.
Editing images with AI is not possible, in short, I am seeing many limitations of the tool.
It has everything to be robust and complete, but I see many limitations. Only for research is it ahead of the others.

image.png
viral saddle
south kindle
# misty pier Perplexity still needs to evolve in some areas. There is no option to merge acco...

"Profile doesn't have a password" means what? otherwise fairly accurate take on PPLX currently. also I don't know many services that let you merge two accounts, so I don't think that's a negative towards them exactly. if you have a bunch of threads you want to access on a different account, you could try loading them into a collection and sharing it to your other account (so both accounts can access/add to it).

serene sand
honest scaffold
#

Hello quick question, when I use an image to ask a question and get answer, if I ask a follow up after the first answer the image I uploaded keeps adding into sources and giving the same answer regardless of any question I ask.

How to remove or discard the uploaded image from sources for next questions that I can ask in a thread.

south kindle
#

process is the same for docs or images.

silver sinew
#

Lol

loud latch
#

Hello Gents, can someone help me understand why I can't use any of the models from Pro plan through API?

vapid onyx
silver sinew
#

Champrionship

weak vortex
#

does anyone know how many users subscribe to perplexity's pro plan?

misty pier
south kindle
#

It’s using 2FA via email/service for login. So there’s just not an option for using a password to login. It’s still fully secure. Your account can only get hacked if said person has access to your inbox.

#

Clever way around dealing with password resets not being even an issue people can experience

viral willow
#

https://perplexity.ai/pro?referral_code=Q69F25G1 Here is a promotion code. こちら10ドルOFFプロモーションコードです。

Perplexity AI
Perplexity Pro

Perplexity Pro is the most powerful way to search the internet with unlimited Pro Search, upgraded AI models, unlimited file upload, image generation, and API credits.

placid coyote
# agile jay How do you trigger it to use the relic?

relics panel should be automatically displayed when markdown code block is encountered, in this case html and that triggers embed tab, so you see the html preview. but as I wrote, it is from ailin (specifically the userscript part of ailin), it is not something which is by default on perplexity

stone zodiac
#

Hi. I'm switching now from Chatgpt to Perplexity. I've created and use several GPTs with Chatgpt. How can I create GPTs alternative with Perplexity?

muted elk
#

Does anyone know if the perplexity pages can rank on google?

harsh verge
fast meteor
#

<meta name="robots" content="noindex"/>

muted elk
#

So how can people discover them?

fast meteor
wraith crow
foggy pecan
#

How many prompts are availiable for claude 3.5 per day in perplexity

fast meteor
tawdry chasm
#

huh did Perplexity nerf the number of messages per day?

#

I haven't used in more than a day but when I came back it said I have 540

fast meteor
wraith crow
tawdry chasm
#

last used was 5 days ago lmao

fast meteor
tawdry chasm
fast meteor
warm plover
#

I'm making thoughts about buying Perplexity premium. I'm really interested if Perplexity can compile code that it has created. (For example a mermaid diagram or Phyton diagram). Does anyone know?

fast meteor
placid coyote
# warm plover I'm making thoughts about buying Perplexity premium. I'm really interested if Pe...

compile? mermaid is not really a code and python is an interpreted language (doesn't require compilation in normal sense). if you mean if it can render mermaid diagram (eg to png or svg) or run a python program and give results (eg png, svg, csv) then I don't think so. while "Programming" step in "pro search" can execute python, I don't think it can give a file to download. but if the output is text, you can copy code block and save it yourself (whether from programming step or written by llm)

#

if this is your main usecase, you would probably do better with chatgpt or claude.
important note to that "programming" step - it is quite hard to trigger and work with, not on a level of chatgpt

warm plover
#

I saw that in ChatGPT you are able to render images from python directly in ChatGPT. Perplexity can use the model of ChatGPT. Doesn't that mean that it would be able to render it too? (or has Perplexity just not implemented this function?)

placid coyote
# warm plover I saw that in ChatGPT you are able to render images from python directly in Chat...

you can upload an image to pplx, it can use vision model (see what is on the image). but it (agentic pro search thing) cannot give you files to download. only text. the models (gpt4o and sonnet 3.5) which pplx is using can do the required things (write python code, mermaid diagrams), but the rest is on the application, in this case pplx, to implement (running of the code - this is done; moving files from temp. containers to some cloud storage, sending it to FE, FE rendering it to user, either link or as an image <- none of this is done). pplx focuses more on search and knowledge, not on data processing

#

well, there are some workarounds. for mermaid, I think Complexity extension supported it. for running python, you could use ailin. but those are community solutions, limited to PC (desktop browser), are not officially supported by pplx and may stop working at any point

warm plover
#

Ok, thanks for your help 🙂. Anyway, I won't buy ChatGPT. I have much better experience with Perplexity.

tranquil orchidBOT
tawdry chasm
stone zodiac
#

I am trying to create code in Perplexity (after switching from Chatgpt).
The first response here, is much slower than Chatgpt.
Another annoyance, is that after creating the code I've requested, the chat always jumps to a point somewhere in the middle of the conversation. So each time, I have to scroll all the way down to the end of the code.
Is that the way it works?

misty pier
# warm plover I'm making thoughts about buying Perplexity premium. I'm really interested if Pe...

In my opinion, and from my experience with pro, the only really good thing is its search and collections. But when it comes to code, generating images, documents, even Gemini does better than Perplexity. I don't know, I see a lot of potential in Perplexity, but it has a lot of limitations and needs to evolve a lot.
For example, in chatgpt I upload 2 spreadsheets and ask it to compare and show the difference between them, chatgpt gives the download option, perplexity does not have this feature. Anyway, I don't think that at this moment it's worth paying for perplexity worth 20 dollars, as there are others that do more for the same price. If you find a discount then yes, I think it's worth it.

lament kelp
#

i ask several time since 2/3 days to team perplixity, if have promo for pro version, code discount. nobody answer to emails?

#

someone know if have a special discount for try pro version

wraith crow
austere kestrel
placid coyote
#

I hope they will improve the code execution, because till now still don't know how "normally" trigger it. like this approach works: task 2: solve xor [654654,321321321,20154654654] and [10210212,3012310321,3135787], but that's an extra execution step (so slower, pricier) and not as easy to input as something like "use programming" or "use sandbox"

austere kestrel
#

i can't work out why it runs the Programming step multiple times

#

I assume it's seeing something we don't, and deterimines "this can be refined" or something

#

but sometimes it'll run what (at least appears to me) effectively the same code multiple times

placid coyote
#

I think the agent system at start splits one task from our prompt into several sub tasks (steps in the pro search) and for each subtask it writes and runs code. yeah, I saw this many times. like "solve this, then get x and y as a result" (interprets the and as a description of two actions). so it writes same algo 2x times and once prints x and second time prints y

austere kestrel
#

I think the agent system at start splits one task from our prompt into several sub tasks (steps in the pro search) and for each subtask it writes and runs code
yeah i think that's right

#

and so that plan gets executed arbitrarily

#

regardless of whether it's solved after the first step

placid coyote
#

because it usually technically isn't fully solved - it forgot to print the second part of a result 😅

austere kestrel
#

it's "going through the motions" basicallly ha

placid coyote
#

on the other hand it is nice it has the capability to refine code. I saw it recover from code bugs in its first one or two attempts of programming step

austere kestrel
#

yeah for sure

#

that's a good point

#

i've seen the same

#

fail
fail
bingo!

#

i like that 🙂

placid coyote
#

I think the low hanging fruit with the programming step would be a way to reliably trigger it (keyphrase or some toggle button) and connecting it to the existing tako chart integration. I think that could be done without implementing file download from pplx, just programming step passing some smaller array of numbers or a table in text (csv?) further the pipeline. while maybe not that useful for me, I think for many normal users this could be a nice use case: upload a csv, do some light processing and show a chart with results

austere kestrel
#

i agree with the reliably trigger part

#

the takochart part less so... i mean, they seem so hit-and-miss

#

if perplexity's code execution thing could render/display

#

and assuming perplexity is able to gather the relevant data - why have takocharts at all?

full cobalt
#

Hi, I just wanted to say how much i love Perplexity, with Claude 3.5 Sonnet it's a beast for all kinds of task. Already saved me hours of complex "brain work" at work.

austere kestrel
placid coyote
#

I thought it is just a chart embed (so the display part), or pplx doesn't feed the tako with data?

austere kestrel
#

no i think the data (as welll as the chart visualisation) is handled on the tako side

#

similarly to Wolfram. It gets an API request and then just figures out whatever and sends back a resposne

#

sometimes it works perfectly

#

others.. not so

placid coyote
#

oh, so it can't be easily connected to the output of programming step 😦

austere kestrel
#

if my understanding is correct, no

#

but.. that's just my understanding.. which is worth pixels it is displayed on ha

#

but you can see with Tako charts the source

#

and that source won't necessarily be among the sources in the pplx response

#

(it's very often the World Bank)

placid coyote
#

well, that would explain why it refuses to do a chart from uploaded files and why a lot of times a chart seems to ignore quoted sources

austere kestrel
#

yeah i think so

lavish token
#

I bought pro plan and didn't get the $5 credits nor am I able to buy credits for some reason and my cards all work fine too. Is anybody else facing the issue?

placid coyote
austere kestrel
#

ha i think it

  • found the data
  • wrote code to add the data points to a df, and output as 2 colums
  • printed out a table as its response (very unclear whether the 'programming' step was at all neccessary for the processing of 10 data points ha )
#

[iow tried to reinvent the wheel]

fast meteor
rough summit
#

hi i think i klicked the wrong button, because my threads has gone 😦 any chance to restore them? i am pro

#

I was on "clear all" 😦

fast meteor
#

do you see your threads here?

#

otherwise:contact supports

rough summit
#

i just found them in mobile app and added to library

#

and wow they are there too

#

many thanks

#

so "clear all" just clears the view in window?

#

but they keep in library ?

rough summit
#

i was a bit shocked because there have been some nice scripts in it perplexity helped me to write

fast meteor
rough summit
fast meteor
rough summit
#

yeah 🙂

#

luckyfonix 😄

rough summit
#

i wish you all a nice rest of the day 🙂

cobalt stag
tired quartz
#

How can I change the task type within a single trend?

wraith crow
maiden slate
#

Hello, is there by any chance I can change my email?

distant yoke
#

Hi everyone

haughty solar
#

hi, I am confused about this screenshot

#

should the model explain reason for 1+1=2? why the model could not answer based on the last question?

devout bronze
peak yoke
tawdry mural
#

https://youtu.be/LnO-Oca7ysY?si=iElRSe23YWrHlrl-

YouTube
Sellagen
How to SUPERCHARGE your web research with a Large Action Model (Nel...

Meet Nelima 🚀 the world's first community-driven Large Action Model (LAM) that takes your natural language prompts and turns them into real actions.

Nelima is programmable too! You can create custom actions tailored to your needs, and every new action makes Nelima smarter and more capable for everyone. 💡

Ready to give Nelima a spin? Try her ...

▶ Play video
devout bronze
#

damn

vital fossil
#

Minor annoyance but why can’t I get Perplexity to reply in the language I’m using? She defaults to Dutch (as I am Dutch) when I ask something in English. I want the reply to be in the language I’m using but that never works until I instruct again to use the language I use.

haughty solar
haughty solar
main bay
#

Does anyone got reduced Pro search from 600 / days to 540 ?

It look like they reduce it 10% without inform or update in their website.

Even though i don't use up to 600, i still feel bad that they did this without any notice and still say 600 in their website.

lone lynx
#

i just upgraded to pro. When can i expect $5 free credit for trying out the api?

tame current
#

how is "pages" any different from what pplx.ai generates normally?

devout bronze
#

make sure you activate the API

devout bronze
lone lynx
# devout bronze make sure you activate the API

i have an issue with the India based credit card and am not able to load any credits using the credit card. Have written to support as well about it. Still not able to see the free credits (subscribing to pro was a hack to get those $5 credits for free to try out the API)

devout bronze
#

oh alr

#

huh idk

tired quartz
vital fossil
peak yoke
jaunty dove
fierce hazel
#

hi

modern fern
#

perplexity not able to read files today?

#

just says "failed to gather information"

devout bronze
#

it's probably in their TOS as well

jaunty dove
# devout bronze to be fair, who's using >540 queries?

By waving it off like it’s not a big deal is what allows companies to continue to do this.

Let’s not forget they still haven’t explained their opus situation.

What stops them from dropping it down to 400 in a month because everyone let them do it because “who needs >400 anyway” and then 300 and so on.

tame current
#

lol

#

the cat makes fair points

devout bronze
#

i haven't checked that in a long time

deft canopy
#

Why can't I find the GTP4 option anymore?

tired quartz
#

Who knows if it drops to 50 per day in about several months without informing us. 🤣

tawdry chasm
#

weird, why isn't my number of requests replenishing

fast meteor
devout bronze
tawdry chasm
#

oh its replenishing now

#

but still 540 is weird

compact summit
#

Is it impossible to download images on iOS in the app? Context menu doesn’t open when holding touch!

fickle valley
#

Is this with Gemini pro and not Perplexity? how doe sit compare to Chat GPT with link to Google Drive / One Drive?
Will Perplexity have integrations with cloud storages eventually? when

fickle valley
#

What's the best AI for research?

shut plaza
#

probably Perplexity

#

You.com might be good too, but I haven't used it much

late salmon
#

The voice model needs to learn to understand other languages than english too. It's hard e.g. to say words in turkish.

lean current
meager sparrow
#

Are the developers going to introduce pages? For free users

dense kraken
#

uhhhh my vpn is blocked again

feral hazel
unreal crater
#

I hate this font

Screenshot_20240715-0029312.png
rapid schooner
sleek vortex
#

on custom assistants

sleek vortex
errant sequoia
#

morning all, quick question, might be even a silly question, how do I turn on copilot

sleek vortex
#

that when you hover says pro search

#

that’s the copilot (they just changed the name like months ago)

errant sequoia
#

I see, thanks!

sleek vortex
gaunt canopy
#

doesnt the perplexity come with 5$ a month free credit in api

#

i got the pro but dont have access to free credits

serene sand
placid coyote
sleek vortex
vapid onyx
tired quartz
#

Is there any trick to let pplx do research and answer in language other than the language used in query?

#

I.e. submit query in {original_lang}, while make it do search and answer in {target_lang}

serene sand
tame current
#

Huh

fading moth
#

What

#

Why does that hurt my eyes reading it

fluid mauve
#

I would be very dubious of any website offered

fading moth
#

It's clearly phishing but the grammar is like optimally built to be readable yet headache inducing.

fluid mauve
#

ja

tired quartz
tribal quest
#

Anyone having issues when needing to paste values and have that data referenced in a response?
Most of the day when pasting a value that gets inserted as an attachment, I get very generic responses back from multiple models, that dont pertain to my data at all.

Example paste.txt had many different data points and none of it included these names to create a summary for.
Based on this data, we can see that:
John Smith (ID: 1001) has the highest frequency of 3
Sarah Johnson (ID: 1002) and Michael Lee (ID: 1003) both have a frequency of 2
Emily Davis (ID: 1004) has the lowest frequency of 1

rich sphinx
#

Is there anyway to give different collections or threads their own model setting. Like have chatgpt4o for CollectionA and Opus on Collection B?

chilly nova
cloud sigil
#

sorry if i've asked this already but what ai does gemini ai use?

rich sphinx
quartz torrent
#

hi, who can i contact for support

vivid chasm
#

10 ducement for It

#

hlo

tight prism
tight prism
tight prism
#

I hope Perplexity accept my request to elevate rate limits

#

I am so scared not knowing whether will they reply or not

tame current
#

guys i need help
the subject is called Graphs of Sine and Cosine Functions
I've been stuck on this question for an hour and a half and I keep getting it wrong

image.png
#

every time i ask c3o or gpt4o, it gives me the wrong answer

#

on both wolframalpha and all focus

stone lagoon
# tame current guys i need help the subject is called Graphs of Sine and Cosine Functions I've ...

f(t) = 3 x e^(-t) x sin(2 x 2pi x t), 3 as indicated by the dashed lines (the amplitude), sin because graph starts at 0 and goes negative, and the 22pit fitted to the value of 1 and number of revolutions until then compare https://www.wolframalpha.com/input?i=f(t)+%3D+3+*+e^(-t)+*+sin(2*2pi*t)+from+t+%3D+0+to+5

f(t) = 3 * e^(-t) * sin(2*2pi*t) from t = 0 to 5 - Wolfram|Alpha

Wolfram|Alpha brings expert-level knowledge and capabilities to the broadest possible range of people—spanning all professions and education levels.

sweet jasper
#

Have the premium models limit dropped to 510

#

Instead of 540 (previously 600)

cerulean garnet
#

Hi,

buoyant crater
#

Is "Profile" really just another layer of AI instructions that one can tell Perplexity? And another layer in "AI Prompt" in collections?

late salmon
errant sequoia
#

any idea why I am getting this in Spanish? and I am not in spain and my setting is not set on Spanish?

CleanShot_2024-07-16_at_16.34.28.png
tame current
#

hallucination probably

errant sequoia
#

will check to see that I hadnt doible down on the medication today

unreal crater
errant sequoia
#

no way ive reached the limy

austere kestrel
fast meteor
#

5periods in 2t, is 0 at t=0, negative at t=eps \approx 0.0000000000001, damping at t0 = 3
f(t) = 3 * e^(-t) * sin(-5pi*t)

sleek vortex
fast meteor
sleek vortex
austere kestrel
#

i still don't get how that fits into here lol (it starts with y = 🤷‍♂️)

image.png
fast meteor
austere kestrel
#

so what's actual solution lol?

austere kestrel
fast meteor
#

too lazy to write it myself

tame current
fast meteor
sleek vortex
#

otherwise i have a very nice macbook at home

fast meteor
tame current
#

sarcasm

open vault
#

Which channel do I have to write in when I have a problem with my pro subscription?

#

Pro is activated on my pc but not on ios

#

Even though I use the same google account

fast meteor
open vault
#

Same thing

#

Even logged out on my pc

#

Now I logged onto pc

#

No pro

#

Even got the mail for my pro subscription

#

Paid and everything

quartz torrent
#

same is happening to me but vice versa

open vault
#

Now I don’t have pro anywhere and paid 20$

fast meteor
open vault
#

i dont find pro support on the site

fast meteor
#

btw check the email you logged in is the same one you got pro subscription

#

and check whether the login method is the same

open vault
#

its the same mail

#

but ur right the login method was different somehow even though its the same mail adress

#

meaning 1 mail adress 2 acc weird

#

thx now i got it back

fast meteor
#

to gain access to additional channels

sleek vortex
#

idk my tone indicators

#

not sure why i read it as /serious

stone lagoon
tame current
#

wolfram api integration must've cost perplexity a lot

stone lagoon
fading moth
#

Potato

placid coyote
#

they disabled selecting in code blocks? again? 😩

fading moth
#

Possible bug

harsh stag
fading moth
harsh stag
fading moth
#

Complexity!

#

You're the dev

#

:3

#

(I thought? @harsh stag Sorry if I'm hallucinating))

steep stream
#

I saw in the other discord where the next version of Complexity will be publicly published on the extension repo. So that's something!

fading moth
#

Woo!

#

Praise the magic conch(Complexity)!

steep stream
#

And I believe there's still future plans to open source it at some point after that.

fading moth
#

That's nice. Open source is love.

#

Maybe someone will be able to finagle a port to Firefox someday.

harsh stag
fading moth
harsh stag
#

bro, wrapping on text block is understandable, but wrapping on every other code blocks without line numbers/proper indentation is criminal 🥹 another day another "improvement" on code block

image.png
fading moth
#

C'est la vie, mon ami.

harsh stag
#

billion-dollar company's quality control occasionally ships random/untested code to production 😅

sweet jasper
harsh stag
mental marsh
#

Whos idea what to implement captcha Cloudflare lol

fading moth
#

Iunno. It's probably just me halluncinating, but I feel like Cloudflare is constantly breaking.

#

Or being broken into.

fallow night
#

Anyone else having trouble copying from code blocks in the browser? I used to be able to, but now I have to click the copy button to copy everything. I can't hover over the text and then click+drag to copy only the relevant parts 🥲

image.png
#

This is in the browser, in Chrome

#

I see user-select: none in the styles which makes things difficult for me tbh

tame current
#

they just reported it

agile jay
fallow night
foggy stone
#

ಥ_ಥ

maiden slate
#

Hello can I change my email?

#

I have a paid subscription for a year but my email was compromised

lost crater
#

Hey all, One small query

For the pro people, we will get a $ 5 bonus credit for the API every month. But is there any expiry for it?

lost crater
rotund wren
#

Woow la ia tenía grupo de dc

sleek vortex
#

then you get another $5 if you continue your pro subscription

#

and so on

wheat pike
warm tundra
#

Anyone know how to get higher rate limits

wraith crow
warm tundra
#

Yeah how long is that gonna take

#

Because either we get this sorted quickly or I’m just gonna have to build an alternative

#

Was gonna use perplexity to power our entire SaaS but if I have to wait two weeks to get more than 20 calls per minute

fallow night
wraith crow
sullen mural
#

hi

sullen mural
tight prism
junior adder
#

hey ya guys, dropping my referral in case someone wants to cut their payment
https://perplexity.ai/pro?referral_code=J9ID1YP6

Perplexity AI
Perplexity Pro

Perplexity Pro is the most powerful way to search the internet with unlimited Pro Search, upgraded AI models, unlimited file upload, image generation, and API credits.

fast meteor
austere kestrel
wise mortar
#

Hi Guys, does anyone know why some pages are index and some pages are not. What are the logic on the indexing of the pages in perplexity? i am a pro user 🙂

fast meteor
opaque stump
#

What's this about? Only shows up with Opus

lost crater
fading moth
vagrant frigate
#

When we will have the option to upload Excel documents for data analysis?

opaque stump
#

🤔

fading moth
sick prawn
#

Can someone help me I made a log in page with Google sheet but everytime I log in I get a a error with a Google drive logo that the page can’t be reach

native lake
#

Is there a way for perplexity analyze more than 4 pdfs? Also, can it use pdfs and web searches?

agile jay
#

Depends on the type of pdf.

#

I would probably ask it to write the text of the searches, and the pdf, and make them into a .txt I can attach to a new chat.

native lake
#

Thanks!

sudden laurel
#

Hi, why does chatgpt 4 give a differetn answer to same question asked to perplexity with gpt4 omni turned on?

tame current
#

they're different models, maybe that's why?

agile jay
drowsy hound
#

Is the app not working for anyone else rn

formal oar
#

it's been on and off for me past couple days

fresh copper
#

Anyone else get an email from Logitech with 6 months of Perplexity Pro?

tame current
#

seems sus

fresh copper
#

Email is from logitech's domain

tame current
#

check if it's ascii

#

ecom1, I'd delta the hell out

fresh copper
#

thats a subdomain

tame current
#

I understand

#

oh wait, just checked logitech and pplx do have some kinda partnership

#

would still wait for confirmation from staff 😅

agile jay
drowsy hound
umbral harness
#

It's legit or I would not be here rn

fresh copper
rich sphinx
fresh copper
rich sphinx
#

I can't find a single reference to a partnership between perplexity and logitech outside some questionable blog posts on questionable websites.

limpid fog
#

Got the email too, the redeem url seems https secured and looks no typo in it

fresh copper
#

Chief Business Officer @ Perplexity confirmed

nimble fiber
#

怎么中文理解

2024-07-18_104806.png
umbral harness
fading moth
#

🥔

austere kestrel
wise edge
#

and It's too little to generate only one at a time, at least two, four

dry yarrow
#

why AI auto switch to gpt-4 omni ? i only want to use 3.5 claude

fast meteor
dry yarrow
#

yeah i did it , but when it answer me , the model said : gpt4-omni

halcyon coral
wise edge