Build Your Own OpenAI + LangChain Web App in 23 Minutes
Vložit
- čas přidán 23. 07. 2024
- Twitter: / gregkamradt
Newsletter: mail.gregkamradt.com/signup
Link to the app: globalize.streamlit.app/ Note: I now ask for your API key now that the free $25 is up. Thanks everyone!
Code: github.com/gkamradt/globalize...
Go try it out! First $25 of API calls are free. If you see that I ask you for your OpenAI API key then my wallet ran out :)
Careful: Only use extremely temporary API keys with this app. Although I don't store any data, I can't vouch for Streamlit. Revoke any key after you're done trying it out with this app.
App made using streamlit.io/
0:00 - The Setup & Hook
1:25 - Streamlit Overview
7:31 - Building Globalize Text App
19:00 - Deploy To Streamline
21:12 - Live App
22:42 - Outro
See the code for that app here: github.com/gkamradt/globalize...
This is based on the example in the official LangChain documentation: github.com/hwchase17/langchai...
Lazy Aftermoon by | e s c p | escp-music.bandcamp.com
Music promoted by www.free-stock-music.com
Creative Commons / Attribution 4.0 International (CC BY 4.0)
creativecommons.org/licenses/...
Bless people with intellect and talent who are willing to help those who cannot repay them.
I appreciate the hard work you did and the way you explain things.
Thanks for making such content.
Thanks for the kind words
Your channel is the most relatable as far as these ai api videos go. Articulate explanations. Great content. I’d love to explore your github repositories or notebooks. Any links?
Nice! Thank you. Code links are in the description
Awesome content ! I could build a small app helping a dog shelter staff picking name for dogs by entering a short description and letting chat GPT come up with names. Only took me a few hours, including learning the tools.
Nice! That's super cool. Congrats on the tool.
it's great idea for the app. tech is not the key, the idea is
Great content, thank you. Really looking forward to seeing what you decide to cover next. Subscribed! ✅
Thanks! What're you building and want to see?
@@DataIndependent I'm looking at use cases for applying LLMs in the field of cyber security. For example, I'm currently working on an app to help teams improve the quality and consistency of their threat models.
In terms of new content, I'd like to see videos on more advanced use of agents, as well as potential alternatives for LangChain.
Love your work, thx for sharing!
Nice! Thank you
@@DataIndependent what would be super useful is to learn how I can use chatgpt to understand user's intend to to call specific services - ex: I am yelp and have data in terms of specific venues. Can you use chatgpt to understand the intend (I want mexican food and am in SF) to call the YELP API on mexican food venuesin SF . Apologies for the trivial example but this would hlpe enterprises use chatgpt to use with their existing services.
Great, Thanks very much for the content
Thanks for sharing - great project.
very nice tutorial, thx and god bless you and the world... regards from hong kong ^_^
Watching all your videos
The best content in the industry. No financial gateway to the content no pay me for thr info bs. Hats off. Real thought leaders
Nice! Glad to hear it. I've thought about monetizing it but I usually do that when people want custom solution and advice, not for the general work!
I just found your channel and really like your content...after watching this I feel like my scam emails are gonna be so well written in the future. Not your fault though haha, keep up the good work brother!
Thanks - Use it for good!
THIS IS AWESOME!!
Nice! Glad its working out
Amazing work
Thank you! Cheers!
I have been enjoying your videos as I am learning to build similar tools for a travel agency, just curious in this scenario is there any difference between using langchain prompt template and the chatgpt api with the same prompt?
There are different models underneath so there will be differences in the output. For example, if you wanted to use gpt4 then you need to use the chat method.
It's a bit of an awkward time for models right now. I sorta which the chat method didn't get introduced and we used chat history within normal models but oh well!
Great! Thank you so much!
Glad it was helpful! What else do you want to see?
@@DataIndependent I'd really want a tutorial on making a telegram bot (with memory) using chatgpt-turbo!
@@DataIndependent This is one of the best videos on streamlit. May CZcamsrs assume certain things (for instance the app runs top to bottom) as something that everyone knows. Such minor details matter for newbies like me. I must say, you have a great knack for teaching!
@@AlgorithmicEchoes Great! Thank you for the nice words.
Incredible video! I'm doing an application for healthcare and doing this using a chain (Langchain based). However, I ran into a "ValueError: Missing some input keys: {'input_documents'}" error that I can't overcome! I looked over the internet for a solution but had no luck. What would you suggest?
What is the max length of the output that can be achieved? Is there a token limit for Streamlit?
Cool, thanks.
cool!! tnx!!
Nice!
Thanks for the tutorial sir. Can u please help how to set environmental variables.
I Appreciate your hard work in explaining the things in a simpler manner
But i had a doubt how can i connect this to a database ,like a git repo and ask it to give me if there are any duplicate code present in the repo ,here duplicate code means :
1.Functions which have same name and perform same action
2. Functions which have different name but perform the same action
3.Functions which perform the same action but have a different approach to do it
Will i able to achieve this using langchain , your response is highly appreciated.
Soooo interesting! めっちゃ面白い！(japanese)
It’s been a while since your last video , would you make a video about deployment, how to deploy the code into a web app , that we can use
Amazing! Than you so much.. how would you upload a .csv insted of copying from clipboard?
Good question! Ask chatgpt and they'll give you the code to load up a file
Again, thanks a lot for your amazing content. What is the difference between defining a function to get the variable (like get_text() in your example) and defining the variable directly? ChatGPT told me that the difference lies in flexibility and reusability of the code. Is it the case here? Thanks in advance :)
Yep that is the case! and if we ever want to attach other code when we get the variable it's easier if we package it in a function
Great video! Can you also make a video on finetuning the model with custom datasets?
Nice! I can look into it.
What’s the tactical example you’d want to walk through?
Taking data from tables using them to create prompts
@@petersmith8153 would also love to see this video made
@@DataIndependent i would also love to see a video like this!
thanks, for ebook scanning, is it required to use pinecone? can it work without using pinecone and just connecting openai llm with the text/pdf ? if both can work, the only difference would be in speed?
Depends how long the ebook is. You can do it with out a vector store but that’ll cost a lot of tokens
Thank you so much, love your content. I searched everywhere , but i was not able to find , what is the difference between gpt index and langchain? My basic shallow understanding is that both are libraries to help querying through open ai by using embeddings. Gpt index uses langchain. But i don't have clarity, If you can tell me difference clearly, i will be very grateful
Awesome - and yes I get this all the time. I'm adding it to the list of videos to do. Right now it's #2 on the list.
@@DataIndependent thank you 😊
How to get the User to provide the API key instead of using your own ?
Hey, if we wanted to learn web front end more deeply, what language would you recommend?
JavaScript and a framework (like react) are popular
following line for line and having issues with module not found on the "from langchain import x" lines showing as module not found. Running in anaconda and checked that the requirements are satisfied for langchain running pip. Any guidance appreciated
Hm, sounds like there may be an issue with your environment? When you run pip show langchain does it come back with a result?
@@DataIndependent it does, but I’m using anaconda and I’m using a python venv. Don’t know if that makes a difference. Thanks for coming back to me
Also tried a conda forge and also forcing the settings in the interpreter in pycharm to use specific versions of openai
Appreciate this series, I would like to make a request about calling a REST API through LangChain, This API request need an authorization to access the data. Is it possible to make a call to a system using REST API? Please make a video on this or provide some pointer how we can do this with the help of LangChain. Thanks in Advance.
How long have you been doing data science related matters?
I've been doing data for ~7-8 years. Ranging from analysis, ETL, analytics stacks, managing teams and stakeholder management. Only more recently AI and Langchain because it's a beautiful mix of introductory w/ creating value.
Can we build a hotel booking bot using langchain framework with LLM?
Totally - Do you want to be the one accepting bookings? Or be the one doing the bookings?
@@DataIndependent Accepting the bookings.
You don't like word wrap do you?
It's a bad habit I need to break
13:51 Regarding your prompt: “There’s probably going to be some purist prompt engineers out there that say this isn’t correct”… Why? What’s wrong with it? Missing some optimization or what?
Looking back at it I actually like the prompt. I think it was lack of self-confidence which I'm getting better at.
I wish the documentation was better and gave actual real examples. It turns a chatbot that ends up being less than 100 lines of code into a week long project
I really don’t understand why are you bothering for such silly things? Just past into Bing Chat Creative mode this prompt and you’re done. please convert this text to formal letter improving grammar and puncuation: [your-letter]
That may work for one-offs (I love quick MVPs), but you won't build scale by copy and pasting