How to Build a Custom Knowledge ChatGPT Clone in 5 Minutes

Рет қаралды 208,106

Күн бұрын

📚 My Free Resource Hub & Skool Community: bit.ly/3uRIRB3 (Check “KZfaq Resources” tab for any mentioned resources!)
🤝 Need AI Solutions Built? Work with me: bit.ly/3K3L4gN
📈 Find out how we help industry experts and entrepreneurs build and scale their AI Agency: bit.ly/skoolmain
In this video I'll be showing you how to create ChatGPT style custom knowledge chatbots using LlamaIndex/GPTIndex. You can create useful chatbot agents for your business by loading in your own data, eg for customer service chatbots or a range of other uses that LlamaIndex/GPTIndex make possible through their range of dataloaders on LlamaHub.
Important Links 🔗
👉🏻 My code: github.com/wombyz/custom-know...
👉🏻 LlamaIndex: gpt-index.readthedocs.io/en/l...
Timestamps:
0:00 - Intro
0:27 - Why custom knowledge?
1:17 - Example (ChatPDF)
2:08 - LlamaIndex
5:30 - LlamaHub loaders
10:15 - Chatbot Example

Пікірлер: 302

@LiamOttley Жыл бұрын

Leave your questions below! 😎 📚 My Free Skool Community: bit.ly/3uRIRB3 🤝 Work With Me: www.morningside.ai/ 📈 My AI Agency Accelerator: bit.ly/3wxLubP

@oluchukwuobi-njoku2204 Жыл бұрын

Could you please show us how to deploy this to a web app, so it can be used in a web app?

@tombarimohamedamine3331 10 ай бұрын

Is whole thing free?

@samwilliams3929 Жыл бұрын

Cracking video ! Short, sharp and focused ! Information is spot on and really helpful. Thanks! Looking forward to seeing more good content.

@adboost_AI Жыл бұрын

The AI beast dropping knowledge bombs again! Awesome video Liam, punchy, engaging and dripping with actionable content 👏🏼 Cutting edge stuff.

@rushikshah2824 Жыл бұрын

God bless the algorithm for showing this channel to me!

@LiamOttley Жыл бұрын

🙏🏼 🙏🏼 🙏🏼

@abc77777 Жыл бұрын

Same!

@whemmakatatt5311 Жыл бұрын

Amen😂

@eljangoolak Жыл бұрын

Maybe the algorithm is God 😮

@sohamagarwal00 Жыл бұрын

really cool stuff! Musch more efficient than traditional methods of custom training the model or making custom responses. Thanks a lot!

@StudioTatsu Жыл бұрын

Thank you, I've been looking for something like this for weeks

@ishaanme91 Жыл бұрын

Extremely cool! Looking forward to more awesome content.

@ahmaddada-dq7zn Жыл бұрын

Nice one Liam, always enjoying your contents.

@LiamOttley Жыл бұрын

Much appreciated mate, glad I could help!

@greg_thomson Жыл бұрын

Amazing tutorial! subscribed

@jornreuvers1598 Жыл бұрын

Awesome video! Going to mess around with all this as my first steps into AI programming... well sorta first steps anyway!

@paulpaturle6957 Жыл бұрын

Super interesting ! Thank you, Liam ! I was wondering how can we see the model used, and if we can control the temperature ?

@webdancer Жыл бұрын

Liam, thanks for sharing this information. This is quality stuff.

@LiamOttley Жыл бұрын

Glad you enjoyed it 🙏🏼

@EnlistedBootCamp Жыл бұрын

@@LiamOttley wish i knew 5% of how to get any idea started, ugh, thanks you are golden

@tomtomatron8625 Жыл бұрын

Great pacing and demo, thank you for the tutorial.

@muradbaghirli Жыл бұрын

Hi, do I have to pay for open_ai key?

@romancandlefight1144 Жыл бұрын

Great video Respect for sharing your files 🙏

@codershorts Жыл бұрын

Liam never disappoints :)

@LiamOttley Жыл бұрын

Thanks mate 💪🏼

@roberthuff3122 Жыл бұрын

Fantastic! Thank you.

@patrick.cheung Жыл бұрын

Great Video. Thanks for sharing. 🎉

@bilalmsd07 Жыл бұрын

Great video as always. Keep up the good work. Wish you very best of luck for your channel. I hope it will rock in the near future.

@Limesh Жыл бұрын

Superb Video, I loved it. ❤❤❤

@mamdouhalmheid9685 Жыл бұрын

I really like your content, thank you!

@LiamOttley Жыл бұрын

Thanks for watching 🤝

@TheHeartShow Жыл бұрын

Great vid as always!

@LiamOttley Жыл бұрын

Appreciate it g

@Warclimb64 Жыл бұрын

Thanks for making this!

@miguelangelcabrera2227 Жыл бұрын

thanks for this amazing video!!!

@smann43231816 Жыл бұрын

Thankyou, great video

@AdamPaulTalks Жыл бұрын

Yes an update to this with GPT 3.5 turbo (current model) would be incredible.

@michielsmissaert Жыл бұрын

Wow impressive video, thsnk you so much!

@bimwerx Жыл бұрын

Great content! How would you get around the response character limit using this example?

@AlbanBytyqi Жыл бұрын

Thank you. It is a vit above my head

@SimonStJohn Жыл бұрын

Hey Liam! Awesome thanks...can you do a follow-up on an addon to index a website like a blog? And have the output to include a link to the article used for the answer so users can click through to read more?

@a999haa Жыл бұрын

Hey Liam! Great video mate 🙌 Can I ask if this can generate responses in a particular json format if needed after indexing any document? Thanks again for the video!

@Mich6961 4 ай бұрын

omg thank you! You've helped me not have to categorise my grocery shopping list into fruit, meat etc. manually.

@nadinejammet7683 Жыл бұрын

Thank you, i can already see how to use it in education.

@olamilekanajao6377 Жыл бұрын

Well done Liam

@Noboy504gaming 4 ай бұрын

Great info!

@ahmedyehia4846 Жыл бұрын

amazing one

@Umuragewanjye Жыл бұрын

Thanks. for sharing the skills

@yakkalabour Жыл бұрын

Really cool video

@proudindian3697 Жыл бұрын

Thankyou so much..!!

@lucasalvarezlacasa2098 Жыл бұрын

Great video!. I have some questions: 1) When we create an index, I understand that what's going on is that somehow based on the question we know which part of the files inside the index should be used to reply to it, and this is context information given to GPT as part of the prompt. Is that the case? 2) Is there a limit in how bit this index can be?

@adil.acoustic Жыл бұрын

Amazing Liam bro..

@LiamOttley Жыл бұрын

🤝

@mohdjibly6184 Жыл бұрын

Awesome video...thanks bro

@LiamOttley Жыл бұрын

My pleasure mate

@ramp2011 Жыл бұрын

thank you for the video. What is the difference between using LlamaIndex vs Langchain? thank you

@Faisal1504 8 ай бұрын

Very Interesting

@birkopheim-3263 Жыл бұрын

Anyone having trouble with the code, llama index updated the name "of "GPTSimpleVectorIndex" to "GPTVectorStoreIndex". Just replace them and it should work if that is the error you are getting

@antfiv007 Жыл бұрын

Dear Liam, great stuff. Already subscribed and looking forward to learn from you. A question: how to increase the length of the output ? I am using it to document some code and it stops before completing the task entirely. many thanks

@taylormun Жыл бұрын

often the token limit is too low

@varunverma5049 Жыл бұрын

Thanks for the session! how do you get the info logs for token usage?

@april11729_ Жыл бұрын

wow!!!! than you so much !!!

@jgilmourtechsmog Жыл бұрын

great stuff, looking to tie this out as a slack bot to answer questions from employees for various business facing items contained in our KB

@umairx25 11 ай бұрын

Hello, if I use the Google docs loader, will the file be updated every time I update the Google doc?

@bgtubber Жыл бұрын

Very nice! Does this thing run locally? Especially if there is no internet connection available.

@krisdavant5813 Жыл бұрын

This is very very good

@owen_silk Жыл бұрын

keep making great videos

@LiamOttley Жыл бұрын

🤝

@ExploitInsight Жыл бұрын

you deserve tones of subscribers

@nbgvfcmnbh6226 Жыл бұрын

This is awesome info. I like how you combined the art work with the story. Very cool.

@ISK_VAGR Жыл бұрын

Crazy Good Amigo

@chikosan99 Жыл бұрын

Thanks a lot!, Great Content! There is a limit of data that you can index ? there is an option with openai api to work with chatGPT4 with custom indexes?

@smudgepost Жыл бұрын

Very good. Need a nice front end and link to a vector db like Pinecone

@spzen98 Жыл бұрын

Hi, I have a problem with the OpenAI API rate limit when using large set of data. This is when loading the GPTSimpleVectorIndex. For small data sets it's okay. Can u advise?

@TheRealPlayer00 Жыл бұрын

Good guy Liam!

@LiamOttley Жыл бұрын

Thanks g x

@sirdrakey Жыл бұрын

Very cool 😎 following!

@michaelabdoofficial Жыл бұрын

Fucking brilliant man. Keep up the mad hustle.

@mohamednihal8215 Жыл бұрын

Wow! Can we use any open source llm model instead of using openai api key?

@joebanks4997 Жыл бұрын

Very good. How do we train the bot to be context oriented? If I only want the bot to have knowledge of radio controlled cars for example. At the moment I can ask this bot about what's in its index, but I can also ask it about washing machines and it will answer.

@MomenRashad Жыл бұрын

Thanks

@darkknightgaming9016 Жыл бұрын

Great video! I was just wondering if it is possible to make it less expensive, because when I use big data bases it uses a lot of tokens.

@LiamOttley Жыл бұрын

GPT 3.5 Turbo is extremely cheap, hopefully they add support for it soon instead of davinci-003

@troymcneil4015 Жыл бұрын

Do you know if there is a way to use multiple sources for the documents? For instance can you use text docs AND discord?

@adumont Жыл бұрын

Any way I can save the index for later reloading it? Can one add several type of things to the same index? Like Wikipedia pages and also pdf documents and maybe calender? All would go to the same index? Or how would it work? For the calendar, I assume it would need to be refresh on a regular basis, how does one refresh part of the index (the calendar part)?

@gorgik Жыл бұрын

What about GPT4 and davinci-003? And what is the quantity of data that can be indexed?

@henrymetzger9951 Жыл бұрын

Can i train llama to write my books? Build it to help me write it prompt based after making my choices from multiple selection and move forward. Like based ona large overview? Maybe have it write the structure in the beginning?

@aprilrobertson7450 Жыл бұрын

What do you recommend to build a construction engineering contractor estimating cooperation

@tomrauch1 Жыл бұрын

Hi, great video thanks! But I am running into an error loading the notebook - ok to post it here for resolution?

@dsalvat1 Жыл бұрын

Interesting, can it be done with a local trained model, like Llama or Vicuna 7b to keep it offline?

@paul-thys Жыл бұрын

It seems anyone will be able to do this soon. The value will be in the data. Can you use AI to gather the data to train in on?

@jimlynch9390 Жыл бұрын

I've wanted to "index" (if that's the right term) my mail files. I'll have to give this a try. I suspect it will fail 'cause there's quite a few emails going back to 2007. I wonder what the limits are? How much data can you feed it before it says "too much!"?

@DanielBowne Жыл бұрын

Great info. How did you capture page numbers from pdf and auto add to the output?

@LiamOttley Жыл бұрын

Which part do you mean? Drop the timestamp

@curtismrasmussen Жыл бұрын

Is this the best approach to export thousands of ChatGPT conversations and then be able to search that data (I'm a prolific writer)? Does it import and structure (or show you how) automatically? OR is there a better tool for doing that? Thanks in advance for any helpful answers. I'm NOT a programmer so that is an important consideration.

@rickyroffey Жыл бұрын

Is there a file size limit for a txt file? And if you upload a large file, does it take a long time to respond to the query?

@novimnovember Жыл бұрын

Hi, great video! Totally new to python here. If i want to host this in a website, how do i do that? Thanks.

@nemesis851_ 2 ай бұрын

Does this training of the knowledge base, cut into the token limit?

@MrOmsa12 11 ай бұрын

Does the ChatGPT api key required for this to work? And if so does it cost money to use llma?

@NickMart1985 Жыл бұрын

Can something like this be used with sensitive business data? Is OpenAI consuming and storing all of this data as its being indexed?

@user-cs2xi1mo8z Жыл бұрын

does this method prevent chatGPT from using/learning from the indexed data?

@fabioarobne Жыл бұрын

What's the difference between that and fine tuning?

@OnChainEpic Жыл бұрын

Hey Liam, sorry not sure what Jupiter is, is that what your running the code in? Assuming we can run this locally? Also how would you integrate this into something you built, by referencing this new model or what within Open AI? Would like some more details on the code your using and integration???

@LiamOttley Жыл бұрын

This is all Python code running in Jupyter notebook. Super easy install with the anaconda launcher. Deploying apps is a bit trickier so you’d probably want to play around with things on your own as I am then hire a developer once you’re happy with it to create a product out of it

@SeanietheSpaceman Жыл бұрын

can we teach it to iteratively improve its own code?

@bombibombi7258 Жыл бұрын

Hello, How can I increase response message length (maybe token limit I guess) . After I train data and ask question answer is not complete and token shows always not more than 2500 600 etc

@codeHacker693 Жыл бұрын

Hi Liam, thank you very much for the tutorial! You are amazing. Do you know how to reduce the amount of token we have been using?

@taylormun Жыл бұрын

use GPT 3.5 Turbo

@codeHacker693 Жыл бұрын

@@taylormun thank you!

@Uzurpat0r Жыл бұрын

Can you use multiple data sources?

@thayaganthusiyanthan9708 Жыл бұрын

Helpfull and cool video, how can I interngrate this with a disocrd bot which will message a user back?

@evaagustine7962 Жыл бұрын

Hi Liam thank you for the great knowledge, may I ask is it possible to save the chat history to a variable and extract to docs/txt? If it is possible, could you show me the way? thanks!

@LiamOttley Жыл бұрын

Yep sure can! Probably easiest if you just put my code into chatgpt and ask it to edit as you need than me type it all out here!

@lonniesims868 Жыл бұрын

very informative video! Could we get a video on langchain soon? 👀

@LiamOttley Жыл бұрын

Quite a big beast to tackle, hard to not make it too technical for most of my viewers :/

@lonniesims868 Жыл бұрын

@@LiamOttley understandable, you have been one of the best teachers when it comes to AI and how to leverage it! If your able to do a video on it in the future it would definitely help a lot. Until then I’ll be waiting for your next video!

@Sleeperknot Жыл бұрын

In this example, where does the transformer model sit? At facebook servers or locally?

@ashishrathore7783 Жыл бұрын

The code isn't working anymore, got stuck with an error on GPTVectorStoreIndex. The libraries you have used have been modified.

@Democracy_Manifest Жыл бұрын

Replace with GPTSimpleVectorIndex

@jeanpeuplu3862 Жыл бұрын

Not sure if I got it wrong, but it means we have to pay for an openAI key, it's not free, is it?

@livb4139 11 ай бұрын

isn't this limited by tokens from the openai api? like if i want to load 500 doucments it wouldn't work right? if so is there any way to run these models locally

@Coco-pk4sx Жыл бұрын

Is there a way to do it in NodeJs? Since I'd want to use it in a website

@bianka1994 7 ай бұрын

hi! great video! I tried to use the code provided in the notebook, but I got an error saying the openai API key is incorrect. maybe it's expired?

@naodivulgado Жыл бұрын

Great video! Is there a maximum amount of information (number of tokens) to provide to chat GPT using LLama?

@naodivulgado Жыл бұрын

@kirtg1 Жыл бұрын

Thanks for the video. i want to create a chatbot that can access a 26 book ms word or text file of the same books. can this method do this? i am not a programmer. Thanks.