I built a fullstack PaLM AI app in just 2 minutes
HTML-код
- Опубликовано: 30 сен 2024
- Learn how to use Google's new PaLM 2 AI model with Firebase and SvelteKit. In this tutorial, we build and deploy a fullstack LLM-powered application in just 2 minutes.
Firebase Updates firebase.blog/...
Google I/O breakdown • 10 crazy announcements...
Terraform in 100 Seconds • Terraform in 100 Seconds
#webdevelopment #ai #tutorial
Firebase is great, I just wish there were open source self hostable options that don't require a subscription in order to use it for more than one project...
Are you being sarcastic or do you not know about supabase
There are many. Appwrite, Supabase, NHost, just to name a few... If your main focus on on ability to self-host, Appwrite has made that the simples of all the backends I have seen so far.
@@matejbaco2475 it’s to self host what web apps like front end only
Or can I deploy my flask app as well
AppWrite bro...
There's also Kuzzle and Parse, and PocketBase I guess?
Pocketbase is good!
This is false. You didn't include the npm install time, which would have more than doubled the entire duration of this project. Unsubscribed and banned.
Ya bro npm i is tacks 5min to finish in my crap laptop😂
Let's appreciate that Jeff spent 1 extra second to fix misspelled "mim" into "mom".
timeloss on the speedrun, but he recovered and brought home the W either way
How...how do you keep up on everything? You have kids! HOW! I am convinced you either have clones or an advanced AI working for you.
He has what we lack. Imbalance much
With so many subscribers now. He may have a team which does research for the videos and he just plays the final audit part.
I mean this is literally his full-time job. You could keep up on everything too if you spent hours every single day researching and prototyping instead of working a normal 9-5
he might just be coding for such a long time that he became a master at learning, applying and creating, so documentation that would take us an hour to understand and process, he is able to skim through for the important parts in just a few minutes
Main thing people need to notice is what firebase is charging for those so called plugins.
The plugins themselves are just prepackaged solutions. In this case, you pay for the 1 cloud function + PaLM usage, which is $0.0005 per 1000 characters currently.
@@beyondfireship thing is its not about how much, its about between adding your credit card or not
@@krishgarg2806 What are you implying that Firebase is going to make unauthorized payments with your card? That's ridiculous.
@@beyondfireship
Wait, 0.0005 per 1k?
That's interesting, way better than ChatGPT with his 0.002 per 1k
@su77ungrhow does that compare to gpt4?
Google definitely wants to put us, developers, on its Firebase needle. This is scaringly addictive,
its*
@@ieatthighs nobody asked
@@mattwyd2161 do i look like i care
@@ieatthighs someone is mad...
@@dykoyaa3937 very mad
So palm be making its own tutorials on how to integrate itself into more applications. I respect it.
At increased overall costs*
Funny nobody mentions that.
I will build this app in two minutes the same way as I learned programming with that 1h youtube video!
It's official, google would rather add literally anything other than basic search to firebase.
This. Flutter has it and yet for some reason one of their main services doesn't.
there is a plugin for that :-)
Closed source proprietary surveillance, oh joy
Firebase reminding everyone they're still the frontend developers best friend
This looks more like trading freedom for convenience.
@@Kevin-jc1fx I'd rather them over vercel, they have a more complete + more affordable solution
Amazing how little development it takes to do a lot of development.
i take two minute to just login…
I think it will take 2 hours for me since I am new to firebase but it’s still fast😊
Firebase is cool but it's such a pain to visualize the data and run queries on it
Bro I just spent like 40 minutes trying to figure out why the PaLM summarization was ERROR'ing out the whole time, only to see this sentence on the extensions hub page:
"Please ensure that you have already signed up for the waitlist and have been approved before installing the extension."
Awesome video but wish I knew that ahead of time :(
Bro did you get the invite?
@@ankan-dev Nope... very sad
Is there any other summarizer ai api?
Same here :D
The ability for small business to create apps that help increase efficiency is outstanding
riiight... being a pc tech of 35 years I've seen this done many times...Linux is so muc more powerfull then windows or mac... so anyway, ai will never exist simplyy because we programmed it, humans are flawed, expecting to make something that thinks for itself..... its called a baby.. and they screw up all the time... so woul ya really trust a newborn child to do.. say..... math?
How can you expect people to hire you to do something that literally takes 2 minutes? No wonder AI is taking all your jobs
Are you going to cover the new Firebase Cloud Functions Python integration more in-depth?
Pog this is incredible, LLMs are evolving so fast, it's hard to keep up
Are there any open source models like chatgpt or something similar?
Are you sure that python cloud functions runtime is new? Im sure I used it a few years ago. Google cloud functions written in python. What's new about that?
It's not new to GCP, but is new to Firebase. The main benefit is that you can easily use triggers for Firestore, Storage, pub/sub, etc.
Imagine understanding anything that is happening. Can’t relate
Firebase is the goat. Makes Vercel look like a child's toy.
Damn, I was fast to click this time. Blazingly fast one might say
every time i try to do something with firebase i google fireship (the thing i want to do). every single time.
References to Charles Bukowski, Terence McKenna, these tutorials are not only about programming.
This took two minutes the same way as going to the moon took couple of minutes. ;)
I agree, you do need experience and not just a CS degree. Tell that to a chronic homeless man who can’t get a job because the media ruined his life.
I’ll wait for your enlightened answer……………………
About the deployment, I guess Sveltekit have community firebase adapter.
Nice clickbait with the "in just 2 minutes". This took way longer just to even think about the problem, much less the solution. Better to rename it a "speedrun" because that is what this is: the fastest that you can go from start to finish after you have already spent a lot of time figuring it out.
it would take me 2 minutes to just download the dependencies.
Only thing i miss is wildcard domain support in firebase auth
love this but tried to follow and got blocked as palm api isn't available outside the US. :(
1:14 who is mim🤔
I always chuckle when I hear "infinitely scalable"
This is the first video I have to slow down the playback speed. This demo was awesome. Thanks!
I get an error saying: Error calling PaLM API for document 'api_messages/3': 7 PERMISSION_DENIED: Generative Language API has not been used in project xxxx before or it is disabled. And then I try to find the Generative Language API on console cloud but it isnt there?
You need beta access for the palm 2 api
Darn! I thought you were going to build an app in PalmOS! 😓
trying to install the extension give me the spin of death. seems like it's not really working right now
I love your political views
Thanks for the good content
Does anyone know if the PaLM API is GDPR compliant? The main reason I can't use OpenAI for work is that their servers are all in the US and they retain all API requests for 30 days. I can't find any information from Google on data retention, storage or location though :/
If it's Google it's safe to bet the data may be stored on or at least "travels" through the US territory.
Nothing AI-related is GDPR-compliant lol. I guess it's really the time for US to declare war on EU and England in order to bring the promised closed-source AI-powered (but also with no concept of privacy or even auth scopes) paradise closer to the complete realization.
What's the restriction on text input size 🤔
my same question. Can I use text from a 100 page PDF?
Is this guy a legion? How is so productive?
I cant stop laughing how this PaLM model summarized "hi mom" into 3-lines
"Hi Mom! I'm doing well. I'm excited to start my
new job next week. I'm also looking forward to
seeing you and Dad this weekend"
I'm sure this video is without any edits.
Palm 2's answers are too toxic...
Can you do Nextjs version?
Gotta nerf Fireship now.
Man your are speeding up the video 🤣
Dude calm down this was way too fast
Hey, great video!! I love your content!! I just was wondering if it is also possible and sensible to use firebase in a server file or even as a api route.
Yes, the Firebase SDK works in both web and Node environments. Most common use case is fetching public data for SSR. Keep in mind, the server won't know the user auth state unless you explicitly setup session cookies.
There's almost no need to build apps now because people can just accomplish the exact same thing from the prompt.
Can you do PaLM api chatbot?
Is your voice AI Generated
Great Video, IDK If youcan build a Chrome extension of whatever but that is the aim.
Pleaes Jeff can you do another video reviewing more on this Firebase Server Side queries. Being waiting for this feature for years.
Can you post the sourcecode? Getting lots of errors and can't figure out where I'm going wrong :(
Python is already supported in Cloud Functions for a long time.. it’s just 2nd gen functions that now have new support for it
This looks awesome but I'm still waiting to hear back about my waitlist application 😭
I see you can add a custom context prompt, but how do you also add example responses to the extension (so it doesn't hallucinate)? Otherwise it's nearly impossible to actually use this extension in production.
i was struggling here, Thank you so much 😢 I'm very happy
What do you have on your learning platform that do not exist here on RUclips?
First 🎉
What the f**k!!!! i am feeling dizzy now 😵💫
I've had a python runtime in cloud functions for months
Bruh you should've made this in 100 seconds
There are a lot of open source text summarizers in python. I would have prefer writing my own backend on that.
This actually scared me af to be honest
Like, how, what the hell
I had to read it to understand that you were saying "vonage" and not "bondage" lol
Would be nice to add subtitles in your videos
adds more effort into editing when it's really not needed. Fireship videos are meant to be fast, you can optionally slow down the video manually and/or use the video transcription feature yourself
voice sounds fake.
How does Palm2 compare to Gpt 3.5?
Terence McKenna quote ❤
Jeff === Based
Loved the Bukowski, my dear Jeff
undefined, lmao
Is this an ad? I think you gotta be clear about that. I’ve heard real pros hate firebase, like it’s not really used beyond hobby sized projects.
Well yeah, all these cloud solutions infinitely scale your credit card along with infinite scaling of your infra. At certain point it's way more effective to do things slower but cheaper, which is obviously not what the *aaS providers want you to do.
This is AI read...Not the real deal!!
yeh,
this video made me subscribe.
Holy shista
I am a secret cat who refuses to recycle news, byebye
Based Fireship class consciousness????
Make a video on Warp Terminal!!!
This ticket is clearly an eight!
Firebase console is still a pain in the arse
Where we can try this for ?
I like your style keep rockN
dude is speedruning code
Using svelte is like cheating at productivity
Svelte is very OP
🤯
It feels like my hair is on fire
can it make cat videos ?
Firebase is back baby!
The reason why I don't use this: money
Maybe if you customer has a small user base or is willing to pay, then the time it saves you is worth it?
Anyone know pricing for this? Ive been looking but i cant find even a hint of what its supposed to cost.
cloud.google.com/vertex-ai/pricing#generative_ai_models
@@beyondfireshipty! Couldnt find that for the life of me.
One thing im seeing is that they measured cost in “characters”, specifically $0.0005 per 1000 characters. Anyone know if that stacks up to be cheaper than openAI? I cant tell if a character is the same as a token.
@@shep9194 As far as i understand, a character is not the same thing as a token, necessarily. The prompt will be split up into tokens which can be characters, words, subwords, or other segments of the prompt.
Wait so i shouldn't use firebase firestore in load function
Depends if you want to fetch on the server or client. In this example, I did it client side to take advantage of realtime features.
@Beyond Fireship yeah but the server sdk is different and i found it not to be quite documented as the client sdk,
Nice
And it’s BLAZINGLY FAST