LangChain Crash Course: Build a AutoGPT app in 25 minutes!
Вставка
- Опубліковано 22 кві 2023
- Get the free Python course
go.coursesfromnick.com/python
Get the code: github.com/nicknochnack/Langc...
Sign up for the Full Stack course here and use UA-cam50 to get 50% off:
www.coursesfromnick.com/bundl...
Hopefully you enjoyed this video.
💼 Find AWESOME ML Jobs: www.jobsfromnick.com
Get the Code: github.com/nicknochnack/LipNet
Links:
CTC Blog Post: distill.pub/2017/ctc
Oh, and don't forget to connect with me!
LinkedIn: bit.ly/324Epgo
Facebook: bit.ly/3mB1sZD
GitHub: bit.ly/3mDJllD
Patreon: bit.ly/2OCn3UW
Join the Discussion on Discord: bit.ly/3dQiZsV
Happy coding!
Nick - Наука та технологія
Nic is one of the best youtube tutors out there. i really enjoyed the video just as I like his other videos. It's always remarkable that he teaches such topics with such great ease and understandability, which speak volumes about his skills and knowledge on the subject matter. Thnx Nic
The pace of this video is perfect. You gave a pace that was fast enough to cover a lot of ground in a short period of time for the "watch through" understanding, but not so crazy that it is difficult to pause when you're trying to follow along with code. This was the first of your videos I watched, and it had me checking out some of your older stuff. Your older stuff definitely isn't to this same level, but it all seems like things where you've improved over time. I've subscribed, and I'm looking forward to future videos.
Thanks so much D. That's exactly what I was hoping to hear! I really started taking content continuous improvement seriously these last couple of months. Just feeling out what's working, this video has been one of my favourites I ever produced!
I second this. I too could pause the video, write the code and un-pause it easily. Really like the pace.
I have been fighting for weeks just for a simple script like this as my entry point to this tech. Now I can move past this obstacle and onto greater days! Thanks. I learned some coding about 30 years ago so I am learning everything as new again. Happy days!
This is super cool. Recently I read entire documentation of LangChain and I can say you literally covered important things in good details.
Love your teaching style! I think a quick recap at the end, going through what everything does would be beneficial.
I think this is my favorite tutorial on all UA-cam. Superb work! Thank you!!
Seriously such an awesome video. I am glad you didn't stop when the timer did!
❤Hey, Nick. Definitely, one of the best tutorials of LangChain! Better than some paid courses! 🎉 Thank you 🙏 As I’m not so versed in Python, I slowed it down to 0.75 speed😂
Amazing video and love that you provided the code. This is the furthest I've gotten beyond standard ChatGPT capabilities and since it's mostly one script done in Python instead of Google Collab or Jupter, even better for my local use case. Instant Subscribe and can't wait to see what else you do with LLMs and Langchain.
Glad you enjoyed it @NightSpyder! Plenty more to come, I'm writing up scripts for langchain every morning atm!
@@NicholasRenottehow to install faiss in python ??
Juypyter is python script too noob! Jupyter is an IDE not a language. Python is a language not an IDE. You need a UI or IDE or command line to do python. Even if it is only notepad. Notice the dude isnt running the scripts to show they work in his IDE? why? Because it dont work properly
Some constructive feedback: would love to see an example of the app in action before you dive into building the code for it. Video jumps straight to code before we really know what for. That being said, the setting a 15 minute timer and then speaking quickly to respect time is a phenomenal touch (something I’ve never seen done before). Well done man, thanks for the video
This is very good advice
This is true for almost all tutorials like this. I dont know why they dont spend a couple of minutes showing what the end product will be.
Solid feedback, seriously appreciate it. I’ve brought it into some of my more recent vids!
@nicholas it’s been over 12 years since I’ve coded. I never coded in Python but this video inspired me to get back in and I was able to build your tutorial from scratch! Thank you! 👏🏾
I'm a absolute noob in coding but you really bring light in the darkness of this world, I just join your course, thank you so much bro!!
That was really interesting to see how to put the pieces together. Your vid gave me a deeper appreciation for how they come together and what they're capable of. Thank you!
Thanks so much @Curious Future! I really wanted to show how to stack all the stuff together and actually show how it use it/make the most of it. Hope that came across!!
I'd love to see a video on indexes
Thank you for the step by step tutorial, as someone who is learning to code this is a highly valuable resource. Please create more content like this relating to AI.
Waiting for video on indexes , superb video buddy keep bringing awesome content like this and if possible just take your time and don't rush on topics so that more people can understand well, rest upto you :-)
im a complete beginner with langchain and this was the perfect intro! please keep it coming! 🔥
I'd love to see a video on indexes! I think thats a very powerful tool for integrating LLMs into an existing data-based application
On the way soon!!
Not sure about that - indexes need to index the data, any data based application will outdo that faster than you can realize. THAT SAID: ChatGPT-4 is able to use tools and such a tool can be an API to access a database. I had a nice discussion with ChatGPT about OData - can fully use it. So, use the indices for i.e. the employer handbook and the documentation where to find stuff, then use tools to access this stuff. No need to load all the data in to the ai, particularly data that changes frequently (i.e. orders in an online shop), stick to the regulations the AI needs to know that are static (i.e. goods return policy).
bhai khuud bna yyar mujhai ek hindi mai ayesa channel chahiye
@@ThomasTomiczek awesome insight here. Thanks! 🙏🏾🏆🔥
Yes index extension of this project would be great and using functions. Many thanks
Great video - as usual!
It is mind blowing how python libraries are so powerful - that we can do so amazing POC applications in about 30mins where it would take week of teams work couple years ago... not even talking about NLP part...
Thanks for sharing and motivating me to try it out for some weekend projects!
Kinda crazy right?! Insane what's possible now!
Thankkkkssssss !! I was looking that tool inside a lot of LLM github projects but not understanding exactly doing what!! Everything makes sense now!!
Great video! I just discovered your channel! Please continue more work like this, because you're approach is very easy to follow and extremely concise!
You got it John, plenty more to come!
Excellent video! Very helpful and love how it's structured. Thanks so much for making it!
Hey Nicholas, super and clear I really like your videos, and easy to understand the logic behind them. Thank you so much!
I spent a few hours with you this morning writing my first script. Your script, though I didn't realize you'd be chopping and changing stuff everywhere. I got it in the end and learned lots. Cheers mate
Just found your videos, finding them so beneficial, I’ve subbed my friend, looking forward to diving into your content
Another great video 🙏 ! In a future video can you expand on showing how to incorporate other AI API services? You touched on this subject briefly in previous videos on how each AI system API can leverage on one each other to produce the best results. Seems like LangChain NeuralSeek would be perfect for this. Maybe a LC and NS combo vide?
Great work Nicolas, very clear explanation and easy to follow.
You prompted the script chain with the prompt again rather than the output from the title chain. Very interesting stuff, thanks for the information.
Ive watched many of your videos now and i gotta say you are a great teacher and im so glad you do these videos :) big thank you and greetings from germany! I Learned so much valuable Information i cant thank you enough :))))
Great video I followed step by step and it was really helpful in understanding how to develop with Langchain. Did you do a video about indexes?
Beautiful, thanks! Can you do a video on CGPT connection to outside applications, like vscode, word, excel, etc? Similar to what AutoGPT is doing
This is great, still trying to understand embeddings and indexes, so yes, please do a video on them!
Amazing video, everything was well explained and the result was awesome, thank you so much!
Wonderful Tutorial! Clear and concise. Gg as always Nick 🤞. Waiting for more LLM content.
🙏🏽 thanks so much, plenty more inbound!
Fantastic video. If you decide to do a follow on video, I'd like to see the next step being using vector databases/stores and creating a ChatGPT web app that can answer questions from a collection of PDF or Microsoft Word or HTML documents. i.e. Summarize documents , rewrite documents, or answer questions based on the contents of documents.
This is really incredible. Amazing work. Super simple to follow along with! I’m going to do this tutorial after work
Awesome, let me know how you go with it Christopher! Thanks for checking it out 🙏
Simply awesome. Thanks for making this.
Awesome crash course! Thanks so much!🙏🏾✌🏿🔥🏆
This is one of the best videos to understand Langchain, It was perfect for me, thanks.
Just found your channel. Epic content brother. Thanks for taking the time to share! Can you show how us how we would actually upload something like this to a website (as opposed to being on your localhost)?
Hey, would it also be possibel to adjust the prompt to something like "create a youtube title and a fitting sctript for {input}" instead of using a sequential chain? If yes why would you still use a sequential chain, is it cheaper or more efficient or whats the point?
Thanks in advance.
so cool, so clear! thank you a lot! looking forward to your indexes video~
You are awesome. Thank you. Love the little cuts and analogies. Well done mate.
Stoked you enjoyed it @Everett. Tried to visually explain a little more than usual, hopefully that came through!
@@NicholasRenotte 1000%
Awesome, thanks. As a beginner in python, this was super helpful.
Hey Nick, can you make a playlist of best practices for prompt engineering?
Thank you for the really amazing video. I would love to see a video about indexes in LangChain too! It must be a very powerful tool right now.
You got it!
Amazing Video as usual man, really loved to see this and understand how you can do amazing stuff with LangChain. However, the title is a little misleading as I thought it would actually be related to the AutoGPT repo and using langchain with that.
Would love to see more though on langchain, especially indexes
Ah my bad, what AutoGPT does is the same in architecture as what is shown in the vid. That being said when I’m back at my desk I’ll prep a vid that shows the comparisons and how they can work together!
Index code is done, just gotta get back home to make the vid!
Thanks for explanation. Looking forward to it
Amazing content. Just subscribed. Best wishes from New Delhi, India
Omg this was exactly what I was looking for. More please :)
Fantastic presentation, so pleased with how you spelled it out for me. I've been scratching my head a little bit with Langchain these past couple of weeks!
I was in the same boat @Paul, tried to break it down as well as I could for people that wanted to build with it!
ditto, i thought i was being a little bit stupid but im glad other find it hard too, gives me a bit of confidence to continue.
@@NicholasRenotte Nicholas, what's your thoughts on system messages? Is it worthwhile using this or just putting your message direct in the prompt template?
More video about langchain. You have explained the best out of everyone on UA-cam, cheers mate
Plenty on the way @Steve!!
Can you do a video on GPT index and the difference with LangChain?Or how to integrate the two frameworks.
Just amazing!! Loved this video. Quick question - how would you change the code if say the prompts are in a file? Instead of having to write a prompt every time, I would like to add the prompts to a file, have the app read the prompts from the file and run in the background. How can I reuse the code to do that? Again, thanks for the great video!
Wow, this is amazing! Save tons of time and easy to understand. Well done Nicholas!
Glad you enjoyed it @Meta Intelligence!!
Dope video! Thanks for keeping it quick and spicy, I learned heaps mate
I'm curious about indexes but not sure the value add?
I'm excited to build an app to simplify nutrition tracking using cronometer. It's a pain to type everything in all the time and figure out measurements. I'd love the leverage this to do that for me
Ayyyyy Spenny! Indexes allow you to add your own data to the LLM chain. You could bring in your own chronometer measurements into the pipeline for example if you had them stored in word or CSV.
Not sure if the pace or how well you discuss the why of things but i finally get it as a non dev person
Beautiful! Thank you!🌞
Damn, another video in my todo list.
Awesome crash course guide!
YESS, let me know how you go with it @Md!
Hi Nicholas, I am restarting my development journey after a decade.. Your videos are really inspirational and easy to follow, thanks for this. I had one query, how do I import the langchain documentation in VSCODE so that it auto completes python imports and code?
As always the best crash courses ! 💪🏻
Cheers Kev!! 🙏
Amazing video! Thanks you so much Nicholas.
Amazing! Thank you for awesome tutorial
Excellent job and very useful. Thank you! Liked and subscribed.
Awesome content. Definitelly interested on the indexing.
Thanks a mil, you got it, starting planning on it now!
Hi Nicholas, this was great 😃 I am looking forward to eleminate OpenAI dependency and use llamacpp bindings for local. There are few documentation errors of langchain otherwise I am almost there.
Going to make a vid of how to do it from my hotel room when I land for ya!! It’ll prob be a little shorter but it should set you up to do it!
@@NicholasRenotte yup 😃
Keen on this video too 👍
Fantastic video as per usual!
Thank you very much for your sharing, it has given me a lot of inspiration. I am currently preparing a programming assistant for a software. I have prepared 10 sample programs and stored them in a JSON file. I hope that users can ask questions using the chatbot and get relevant responses. However, I am facing several issues at the moment:
1. I am struggling with how to upload the JSON file to Vetor Store. Currently, my approach is to convert the JSON into a CSV file, but this method is not yielding satisfactory results compared to directly uploading the JSON file using relevanceai.
2. In my own setup, I am using openai's GPT3.5 along with Pinecone and openai embedding. These configurations are similar to relevanceai except for Pinecone. Should I consider switching to another vector store?
really enjoy your videos! great content =)
Hi bro!
This video is amazing!
Could you create an indexes video to keep learning?
Great Job!!!!
Hi great class thank you. Please let me know which openai model this api example uses? gpt-3.5? Is that possible to config it at the code?
Curious about maintaining confidentiality though with the autogpt stuff.
My main concern with using gpt for anything serious is leaving any potentially sensitive information while working on creative projects or research projects.
I you could adopt this to work with more localized models, but most of us don't have machines capable of that even with these 3rd party llms.
I have no clue about programming but i learned quite a few things from this video so thank you very much!
Amazing tutorial... thank you for sharing with us
do you need to buy credits with openai for this? I'm just a student trying to make a small project, and I'm getting rate limit errors?
Would it be possible to use OpenAI/Azure OpenAI to tweak prompts before giving it to my actual chatbot? For instance if I write a poorly formulated prompt into my chatbot, it is rewritten and by ChatGPT before given to the chatbot. Thank you in advance :)
Thanks a lot! It's much simpler than I thought ;)
Would love to see indexes. Possibly larger memories like a couple of philosophy books and semantic search. Associating some context to user input and answering some cool questions. SocratesGPT?
Thanks for posting this video, was great and really liked how you iterated on concepts (from easy to more complex). I added this line and it seems to help with the script getting cut-off.
llm = OpenAI(max_tokens=1024)
Thanks for the love on the video guys!! Let me know if you have any feedback. Really working on improving the video quality! 🙏
Thanks man great and practical video, nice job. Does llm transform wikipedia research somehow or it just pure text from wikipedia in the wikipedia section output?
Great work thanks for the course
Hi. Enjoying your presentation style but a bit over my head at this point. Couple questions. Can you recommend a learning path to get to here from a decent understanding of basic Python? 2. I would like to develop an AutoGPT (agent?) that can generate a spreadsheet with the contact information for every dementia care facility in Canada. How would you go about doing that? I'm envisioning an autoGPT agent that scrapes Google results for each zip code in Canada. Anyway, anything that can get me moving in the right direction would be great. Thanks.
excellent tutorial. Thank you!
finally I see a video about langchain straight to the point!
YESSS, stoked you enjoyed it! Wanted to make it as practical as possible, went a little longer than I wanted but 🤷🏽♂️
@@NicholasRenotte It was really great! don't mind how long it would be as long as is well explained
excellent. great job man
I love the way you explain everything, Just the a question though is there a way to display full script, I have been going over this video for 2 days, the output is displayed on half, and some of them stopped in between a sentence. Is there any way to overcome this? Thanks for your help.
Wow!! Really Good Explaining, Very interested learn more about it
great video. I did have one question. In your video, the script seems to be cut off at end. I assume that this is because of a token limit of the model. Can you please explain.
Hi Nicholas ! Can you make a video about application of deep learning on medicine ( like how to create a cure for diseases etc )
Thanks 😊
How do you think we should approach combining unnecessary chains. For example, in this case title and script both could be asked in only 1 api call but it would be much messier to work with. Do you think we should go for cost efficency or deal with chain combining after we get the whole script working, or not do it at all.
Just wish you wouldn't let yourself be restricted by the artificial 15 minute shenanigans 😊 awesome video!
Amazing! Best 27mins
Great video, langchain appears to be the solution to a LOT of the issues I'm currently facing. I definitely want to see a video with the indexes!
You got it @KJ! Thanks for checking out the vid!
In the scenario of conversational robots, how to limit the token consumption of the entire conversation?
For example, once the consumption reaches 1,000, it will prompt that the tokens for this conversation have been used up.
This is amazing ! thank you for the video ! i was wondering if this can be also a solution to create reports out of an excel file following previous reports in a form of PDF if so how ? thank you !
Very informative
Greetings, a newbie question. Based on this video I have a hard time understanding the benefits of using LangChain. It seems I can accomplish the exact same thing with two subsequent OpenAI API calls. First generate the title based on the user input, grab the response and then feed the response as the input to the next call. What am I missing here??
Fantastic presentation!
So rad! Thank you!!
Excellent video and walk through.👍
Thanks a mil @Chris, glad you liked it!
Really nice video. Subscribed. ❤