Build Your Own Auto-GPT Apps with LangChain (Python Tutorial)
HTML-код
- Опубликовано: 1 июн 2024
- 👉🏻 Kick-start your freelance career in data: www.datalumina.io/data-freela...
This video is an introduction to the Python LangChain library. LangChain is a framework for developing applications powered by language models like OpenAI's GPT-3.5 and GPT-4. I will go over all the modules, provide examples, explain the quick start guide, and then show a demo of an intelligent assistant that can answer questions about any specific RUclips video. You can use LangChain to build your own intelligent systems, similar to Auto-GPT and BabyAGI. I believe there is an amazing opportunity right now for data scientists and AI engineers to become a front-runner using these tools.
🔗 Links
github.com/daveebbelaar/langc...
python.langchain.com/
⚙️ Copy my VS Code Setup
• How to Set up VS Code ...
⏱️ Timestamps
00:00 Introduction
00:26 What is LangChain
01:19 Job opportunities
02:55 LangChain's Modules
03:47 Models
05:37 Prompts
06:43 Memory
08:21 Indexes
08:58 Chains
11:06 Agents
14:32 Use Case: RUclips Video Assistant
27:45 Why I am all in on this
28:15 What is Data Freelancer?
👋🏻 About Me
Hey there, my name is @daveebbelaar and I work as a freelance data scientist and coach. You've stumbled upon my RUclips channel, where I give away all my secrets when it comes to working with data. I'm not here to sell you any data course - everything you need is right here on RUclips. Making videos is my passion, and I've been doing it for 18 years. I just genuinely enjoy helping people.
While I don't sell any data courses, I do offer a coaching program for data professionals looking to start their own freelance business. If that sounds like you, head over to www.datalumina.io/ to learn more about working with me and kick-starting your freelance career. 
Наука
What are you going to build with this? 🤑
can i add this to my website
@@infinityshort3988 Yes, you'll need to set up a server-side application to handle the function execution and use JavaScript to communicate between the front-end (HTML and JavaScript) and the back-end (Python). You can use Flask or Django for this.
the perfect servant? I feel you excitement. What a cool future and a little scary.
Not telling you !! Hahaha
Can you please do the same thing with an excel file? splitting dataFrames into 1000 token pieces didn't seem to make sense to me, there's a continueation of data, I might break it if I split it.
This is an excellent video. I’ve been deep diving into LangChain for about a month and this summarised pretty much the extent of what I’ve learnt in 30 mins. Brilliant! Looking forward to watching many more LangChain vids from you Dave. Cheers!
I can feel the doors of understanding unlocking with your great explanation mate! I’m gonna add this video to my assistant’s db so I can always use this awesome resource. Keep it up! I want to see a lot more from you!
This is THE best LangChain tutorial on RUclips
Man I love this video! Followed along with it this afternoon and everything worked well. Sometimes stuff just doesn't work from some video tutorials but this one really made me feel like the time I spent was worth it! Having a working 'applet' at the end! Thanks so much
Great tutorial, thanks for this.
For any python noobs who get stuck like I did with the dependencies not being found within the downloadable langchain-experiments example. Be sure to open the entire folder and not just the python file within VSCode otherwise the environment you setup won't be used.
Dude, you are killing it with these videos. Such clear instruction, so helpful Thank you. Got so many script ideas running around my head :)
Thanks man, really appreciate the kind words! My head is also exploding with ideas haha!
@@daveebbelaar I agree these are incredible, I'm very grateful that you are sharing these
This has been extremely helpful. As a total noob coder you've made it so easily to understand.
Thank you so much!
Great video! Just started working as a developer and got assigned a task that uses langchain. This is very helpful
Killer demo, straight to the point and left me with a really clear idea of how to start implementing in projects. This stuff is turning low code hackers like me into Harry Prompters 😅
Dave, amazing video! You are good at this man. Thank you!
Dude you're a rockstar. awesome quality video! Thank you !
This was an absolutely illuminating demo. Thanks so much!
So far the best video i've seen of Langchain. Thx a lot for investing your time on promulgate valuable knowledge for everyone 🙌
Legendary video my guy. I feel like this came to me at the perfect time. Thank you!
I'm a python dev. And this is really interesting! Would definitely be using this!
Very good summary and explanation, thank you!
This is the first youtube vid that I've seen that talks about embeddings with LangChain. This for me is an lightbulb moment as the small context window is the major limitation for ChatGPT atm. Giving it the ability to take new source data and chunk it to referenceable bites feels like a work-around for the context window limitation, but it's an awesome work-around nonetheless. Good stuff!
Thanks Patrick! That really opened my eyes as well. Until LLM APIs allow more tokens, this is the way to go.
2 weeks later: Claude enables 100k tokens
This is awesome, thanks Dave!
The best explained Lang Chain video!
Thank you for this, exactly what I was looking for to optimise my helpdesk so that customers can find answers faster.
Finally a video that mentiones the technical details and capabilities, not AI-influence hype!
Wonderful content... Good to get started. Best thing was that you decided to put this out in public. Thanks a lot🙏😃.
Thank you so much Dave, I just subscribed, this info is amazing.
glad to see ur video on langchain!! awesome work as usual!
Thanks! 🙏🏻
Thanks for very clear and concise video!
You tutorial was rally awsome I was looking long time like this tutorial . Thank you sir !
excellent exposition of langchain, super helpful thanks for sharing and looking forward to more 👍
Awesome sauce. Thanks Dave!
Thanks for the video, such a good lesson for beginners! :)
Excellent and detailed Video Dave, thank you for sharing, much appreciated👍🙏
You're welcome Aziz!
Eye opening content. TY Dave.
Amazing, thank you for sharing your knowledge
Lekker bezig man!!!, thanks for sharing 👏
Great video! Thanks for sharing!
Thank you for sharing. Extremely interesting.
your video is superb and well beyond the usual LangChain chitchat video. Keep up the good work and start expanding your company...
Great video! Even a lawyer can understand it. Thanks!
Hey Dave, I really love this video. I had heard about langchain, but with so many new things popping up I didn’t give it a lot of attention until now.
In terms of things to build with this, I’d like to use all of my past long form journal entries to create a virtual representation of myself.
I’m on Upwork, but my domain of expertise is embedded systems. I have a little experience with data science, but not enough to easily market myself in that way.
Anyway, thanks for sharing this info about langchain! Looking forward to more videos
I used a text file of a few months of journal entries, and I find it has a problem.
The response seems to just be based on a narrow view. As if it only takes in one of the journal entries and then can only talk about that one. It makes general questions like "what are some highlights of your travels" very bad.
Is there a way to make the context more broad? @daveebbelaar
So Exciting 🔥 🔥
An excellent channel and your videos are truly great. Thanks! I've been exploring langchain for a few days. I see that it can be an alternative to give my application architecture's the concept of pipes that I need, but I just realized that doesn't integrate natively with SpaCy, so I'm looking to see if I can make the integration using Agents to connect SpaCy with the transformer models in a clean and maintainable way.
This information is worth tons of gold... Thank you a lot!
Thanks! 🙏🏻
Amazing video . one concern i have for startup idea would be the API costs to let people access it publicly
Wow! Thank you :)
Nice and helpful vid! The ecosystem of LangChain really feels like the missing toolset for developers to connect anything you want with those LLM's all popping up. It also really helps to build businesses not only relaying on a single LLM which might be quite risky imo.
Exactly!
Excellent! Subscribed.
thanks a lot! awesome video
Wow, this video really Informative, Looking forward to more content about this!
More coming soon!
Your videos are awesome. I was able to build a flutter app to work with a python backend running in replit, using fastapi to serve API endpoints. In my app, I can upload a PDF file and chat with it using an agent with memory. It works fine. However, I need to allow multiple users, each one to have its own agent with its own memory. I have no idea how to acomplish this. What path would you recommend?
Hey Dave it's request to you that pls make more videos like this... I have some Questions
1) How can we make it interact with basically anything on our computer screen(Cause there are a lot of work that should be automated most of them do not even have API and can we use chrome extensions cause they are really helpfull)
2) How can i use my finetuned model as a model.
HI Dave! Thank you for your very detailed explanations. I watched also your video about auto-GPT. I am just wondering how could I use all Plugins with your langchain-experiment repo?
I never comment under videos but this was just great. Goed gewerkt pik.
Thanks haha!
awesome video!
Sounds very promising! What companies are usually interested in this type of stuff? I mean the use case is interesting but not sure how to use it in a commercial way...
I am completely new to coding but, even I feel like I can replicate this thanks to your amazing explanation.
That's awesome, thanks!! Good luck with your coding journey 💪🏻
Mind Blown!
This is a great video, it was very inspiring. I just get worried about cost controls when I see that it can connect with all of those external APIs. I'm assuming there's free tier usage, but that's one thing I'd be concerned about when it comes to small businesses and their budgets if API calls get out of control. Great video either way. I'll just have to figure out how to put in API call controls.
Thanks! You can set spending limit or call limits on most APIs
Very clear tutorial. How do you do it with text documents instead of transcripts from RUclips. 1. Read the text from a text file 2. Carries out the document splitting. 3. Get a query from user . 4) writes the response to another text file. Do let me know if you with to collab on this.
Use the Directory Loader documentation. It's the same
This is great video, thanks for your effort. Do you plan to make a video about DemoGPT?
AMAZING!!
Excellent
Is this the best way though? What if I want to, for example, build a C# or nodejs application to do something similar with my company data. Would I use langchain or build the logic into my app? Serious question.
Great video
The multiplication in 14:02 (year of python creation * 3) is wrong. Is llangchain able to call multiple tools in a single prompt? After the wikipedia info it should have called llm-math.
Awesome vid btw!
Thanks for this
Super!!
Thanks for the video. It is inspiring. When using a Loader like Confluence can this be done without sending my private content to openai?
You should definitely create a udemy course on this subject. I would definitely buy.
Thank you for sharing this, this is amazing! I hope I'm not too late to the LLM party 😂😂😂
We are just getting started
When trying to ceate a database why does it show you are out of tokens; after running the video url when I am running the db command the error pops up
Great vid! A funny thing though: in the python year *3 example, the model didn't use the llm-math, and hallucinated the calculation (5783 instead of 1991*3=5973). 👻👻
THAT'S a VERY important issue though!
Just imagine you want to create an A.I. that assists lawyers, feeding it legal documentation. Now, in THAT case, you want to make ABSOLUTELY sure that the model ONLY comes up with replies it finds in your docs and will ALWAYS 100% return "I don't know" when it can't connect the dots with the material it's been provided.
Also, you might not only want to get an overview, you're most probably interested in retrieving ALL necessary information on a given topic.
Sadly, this is never really discussed - I'd be REALLY grateful if anyone could point me to a video where those issues are actually addressed.
Great video, really clearly laid out! As somebody new to web development. Could you explain how I could take this python code and then build it into a web app? Any tips on what to focus on learning or really good web dev tools to pair with lang chain? My goal is to be able to build something like this and then quickly make it accessible online. I have one idea in particular in somewhat of a niche
I have future video's planned for that, so stayed tuned!
@@daveebbelaar awesome can’t wait, love the content!
@@daveebbelaarwaiting on this! I am currently in the stage of introducing AI to my web app and would love your guidance
Would it be possible to take an example like that and do a full tuto from the python script to a web app plz ? :) Can't find this on internet if someone have resources for this I'm interested.
Very interesting video btw
Thanks bro for this video. I'm new to all llm , I don't know where to start. I just can write python . Any advice?
Is it possible for a model to have existing knowledge of like gpt-3.5-turbo and also add those videos or files knowledge? Because when I do this, my davicnhi model loses every knowledge and than it knows only knowledge about file or video provided.. haha
One thing is for sure. If you are making an AI/langchain course I am on it!!!
Hi Dave, this is good video, how secure is the data that we input to GPT indexes ? Is there any option to restrict the data usage and access ?
keep in mind that everything you sent to openai will be processed by there servers (usa)… you can use azure open ai to be gdpr compliant in europe… if you are using a chroma db you can store ur data on your server.. keep in mind that you should better not use any internet database. especially in the us while you are processing critical data
@@digitald74 Great answer, thanks!
Is there a way of getting the time code with the transcript? That way a search could be done for the video segments when a particular subject was discussed.
That is possible with Whisper 👌🏻
Amazing video, i have a newbe question, i´m learning python and i wanted to ask you how you configure visual studio code to show you the right window, do you have a video teaching that? Thank you very much.
Thanks, this video should help: ruclips.net/video/zulGMYg0v6U/видео.html
After the similarity search, I got kinda lost between the definition of docs_page_content and the injection of {docs} in your "template parameter". I was under the impression that you would have passed docs_page_content instead.
Did you figure this out? I'm going to try and debug it tomorrow as this baffles me. When I swap docs out for docs_page_content it no longer works too.
Cool!
What type of Mac are you running?? I have a MacBook pro Intel. Been having trouble running code with LLM. Any help or suggestions on this would be really appreciated!!
I have an intel mac too. What's the specific issue with running the code?
can i use csv loader and create pivot tables using langchain?
What do you use to display the output on the right?
@masidawoud your link disappeared. Can you message me?
This is great. How would I make a web ui for this?
how does he get each line to execute on command? i have to run the entire script. i press shift+enter (on PC) to run a selected line, but it always gives me 'NameError: name 'xyz' is not defined'
Check this out: ruclips.net/video/zulGMYg0v6U/видео.html
what i am missing is someone implementing a custom agent.. like for some kind of api that is not available at the moment
At the moment i am stuck to install FAISS (with "pip install faiss" in my conda env. Downgraded python to 3.8.16 as in the vido but still get this error and also could not found another solution.
ERROR: Could not find a version that satisfies the requirement faiss (from versions: none)
ERROR: No matching distribution found for faiss.
Has someone an idea?
For future videos it would also be nice to see how you created the env. and in which errors you run.
Thanks!
🙏🏻
Great. Do I need to have an Openai API key? Looking for a free solution please. Thanks
How to handle the query function execution and deploy on AWS?
is there any chance of langchaining for free by not using the official chatgpt channel but via local lm models provided for the oobabooga webui?
Great video bro. Where can I find more reading material on this
Thanks! Soon on my channel, but for now, the LangChain docs: python.langchain.com/
What is the difference between Conversation Chain and Open AI Chat End Point? For me, it is effectively the same. I love Langchain but I don't think it works for every use case that I have had. Works fine for simple conversational agents.
I cannot tell you exactly right now. Have to dive deeper into this for that. In general, LangChain is a framework that supports all kinds of models, not just OpenAI's. So, you can also implement, for example, open-source LLMs that don't have a chat end-point.
Hey Dave 👋
Great tutorial Dave 🚀If anyone’s looking for more Python tutorials, we’ve released Loguru logging, task scheduling, and more to help the community too 💪
Just a question. If you are using OpenAi, what will happen to your data? I mean it's going to be saved on openai server's right ?
openai.com/api-data-privacy
Can you show how to post online with the python code
how can i let it use my data in database like mongodb?
Can you make tutorial on converting this to a website