Building an AI Data Assistant with Streamlit, LangChain and OpenAI | Part 1

Imagine accelerating your machine learning projects with an AI assistant that will save you hours and hours of work.

In this video, the first in our series, we are building an AI-powered assistant that will transform the way you explore and analyse data. Say goodbye to complex data analysis processes and hello to a more intuitive and interactive experience!

This video is part of the series Building an AI Assistant to make your data science life easier in which we will develop an AI assistant using Streamlit, LangChain and OpenAI’s GPT models, designed to help users with their data science projects. This AI assistant will streamline the entire process of a data science project, including exploratory data analysis (EDA), model selection and prediction, saving valuable time and resources.

I'll walk you through the entire process, from installing the required libraries to solving a machine learning problem using AI. By the end of this series, you will have a powerful tool at your disposal, ready to assist you in every step of your data science journey.

If you want to take a deeper dive in data science, check out our library of courses on DigiLab Academy

————————————————————————
🎵 Music
————————————————————————
Creative Commons / Attribution 4.0 International (CC BY 4.0)

————————————————————————
📌 Timestamps
————————————————————————
Intro - 0:00
What’s covered in this video - 00:39
Setting OpenAI key - 03:10
Running Streamlit - 04:01
Importing required packages - 4:24
Titles headings and subheadings - 5:35
Writing text - 7:00
Sidebar - 7:57
Further text formatting - 11:03
Adding a divider - 11:45
Integrating HTML - 12:36
Adding expanders to the sidebar - 13:26
Buttons - 14:25
Integrating a CSV file uploader - 15:35
Session state - 17:08
Converting CSV file to dataframe - 18:46
Loading our LLM - 20:05
Generation information using our LLM - 21:10
Creating our Pandas agent - 23:46
Using Pandas agent to answer specific questions about the data - 24:43
Using Pandas agent to answer questions about a specific variable chosen by the user - 28:04
Caching - 29:20
Creating visualisations - 37:00
Answering user questions - 41:54
Answering more user questions - 45:33
What’s next - 45:55

————————————————————————
📌 Resources
————————————————————————

————————————————————————
👱🏻‍♀️ Connect with me
————————————————————————

————————————————————————
🏷️ Tags
————————————————————————
Building an AI assistant to make your data science life easier
Simplifying your data science journey with an AI assistant
Crafting an AI assistant using Streamlit, Langchain and OpenAI models
Enhancing data science efficiency through an AI-driven assistant
Creating an AI assistant to ease your path in data science
Developing a data science ally using Streamlit, Langchain and OpenAI models
Developing an AI assistant for smoother workflows
Designing an AI assistant to simplify your data science projects
Making data science easy with the aid of an intelligent assistant
Making data science effortless with the implementation of an AI assistant

————————————————————————
✨ Hashtags
————————————————————————

[drxdre]

Love the video, the pace, the explanation style and the focus on aesthetics. I hate all the videos about how easy it is to make a streamlit app in 100lines of code and then they show it and it looks like a 5 year old made it. I've used plotly dash for quite some time and never made the switch. Would you recommend using streamlit instead of dash for 'applications' where there are more buttons/user actions vs dash which usually seems more for live dashboard visualisation and adding 'ui' elements often feels quite clunky

[willtaylor]

Now THIS is a real tutorial video. I will be watching this several times over the coming weeks.

[prnmdid]

Ultimate video that i was looking for. Google never suggested me since 2 months. Today perplexity recomendded your video. I must say this is the best one. Thanks 🙏. Can u make a video on
Building a RAG pipeline on schema of a SQL database like Postgres and chunk it & embedding it on to pgvector extension to load only relavant schema in order to optimise tokens and prompt size & then passing natural language to sql with defog SQL Coder and futher give insights like how this video works. Thanks in advance

[mccloudmedia]

This was fantastic! I havnt done any programming since I was a kid back when I was coding in VB3. I’ve been thinking of getting back into programming and I’m just blown away with how easy it’s become, especially with building a nice user interface. Thanks for posting this! I feel like a kid again on Christmas Day! 😂

[Neo-xqwt]

Oh man! That's awesome. Explanation, blog, timestamps, etc. Where in the world could you get such a quality learning? Here.

[arpanoverload]

I couldn't tell it's your first time until I read @Me-in-a-Nut-Shell 's comment... in fact, I was pacing to open up vscode and follow along beacuse that's how relevant and intriguing your content is... subscribed, looking forward to the next one. Remember us when you get your first YT button :)

[rajrajabhathor]

Awesome Ana .... but the create dataframe is undergoing some changes to "experimental" version - and hence doesn't seem to work...

[chukwumanwakpa]

Thank you so much for this great content. However, i think on bringing on the part 2 there should be a session that speaks about the files that we ought to create for the projects code to functional optimally just as yours.

Sadly, my lines of codes as accurate as yours happens to give no diaplay at all when i run it via the localhost. I saw that you used a perhaps linux or Mac OS but mine here is windows. Please look into this also. Thank you for your great work.

[nkwachiabel]

This tutorial is extremely helpful. Question though, have you thought about saving all analysis, questions and answers to a "custom word document" which can be used as documentation? I believe this tutorial is important for auditors and if they can chat with documents and files, and save these questions and responses as documentation.

[dbiswas]

Simply awesome. Your explanations are so clean and detailed. Can’t wait for the next episode. ❤ Do you have any video on Azure OpenAI ?

[smiley]

thank you so much!!! you deserved more subscribers!!

[adiy]

Great explanation @digiLabAcademy. Thank you. ..subscribed!.. It will be nice if you can also do a video on using an open source model like llama-2.

[pyalgoGPT]

keep posting ore such amazing tutorials

[JohnStevenTapican]

How is this different from Data Analysis feature inside ChatGPT? And what use cases I could possibly use this?

[SurajKumar-ufit]

Can you make a video do the same thing using Google Gemini Pro API Key? Please

[itailironne]

nice! how did mange to write the code and then see what u write in the streamlit? is it only available in mac?

[dandulokeshvarma]

you did an awesome job😍
can we do the same using database instead of csv file.

[leslietientcheu]

Hi, Please can someone help me?
I have the following problem when compiling my code: ValueError: An output parsing error occurred. In order to pass this error back to the agent and have it try again, pass `handle_parsing_errors=True` to the AgentExecutor. This is the error: Could not parse LLM output: `df.head() | Date, Open, High, Low, Close, Adj Close, Volume`

[adityasakalekar]

How is the reasoning capability of the model ?

[jormun]

would it be possible to replace Open-AI API key with a local model from huggingFace?