Query Your CSV using LIDA: Automatic Generation of Visualizations with LLMs

[@priyanshudubey2245]

Sir please make a playlist of Natural Language Processing. Your explanation is too good and easy to digest.

[@AIAnytime]

I will do that Priyansh. Thank you.

[@CesarVegaL]

Thank you for your work. Very up-to-date, so it is appreciated.

[@AIAnytime]

Much appreciated!

[@sidindian1982]

Excellent Work Sir 🙂

[@AIAnytime]

Many thanks

[@valm7397]

You are very impressive. Thank you very much.!! Do you think its possble to use lida with Mistral ?

[@AIAnytime]

Yeah you can use local LLMs. You can use llmx or Vllm to Inference it within Lida. Look at Lida GitHub issues for their roadmaps.

[@avijit_barua]

Thank you so much creating this video

[@AIAnytime]

It's my pleasure

[@user4-j1w]

Amazing idea... Thank you for this video.

[@AIAnytime]

Most welcome 😊

[@yuvrajharsh8763]

Thanks a lot for making this video Sir.

[@AIAnytime]

You are most welcome

[@RedCloudServices]

wait..can this tool simply allow chat with csv better than previous options? does every answer have to be a graph?

[@amitsingh-rv4hz]

Nice topic and presentation

[@frankismartinez]

Just tried it based on your repo; summarization was very fast, charts were highly relevant. Gracias! Curious, how will it process a pivot table; i.e. will it understand all the dimensions? Or, will the CSV [assuming that data source type] have to be prepared for Lida's tokenization process [assuming]?

[@AIAnytime]

Thanks Frank for your comment. Yeah Pivot is something we need to figure out. Later is something correct I feel. I will reply back to this comment thread once I try out myself.

[@Abhi3661]

@frankismartinez Were you able to find anything relevant to the question above?

[@myfc8661]

Hi, thanks for the video, on the part of the open source (llama2) or other model like Groq, Gemini Pro, it is not so clear to set it up. Any chance you could drop the code here to set it up. That would be a great help. TQ

[@latentbhindi837]

Does this give good results for non-numeric data or rather a dataset which is a mixture of numeric and non-numeric data?

[@NileshKumarPandey-vr7pw]

is there any graph representation module I can use with LAMA3 with Ollama?

[@hawks900]

Thank you bro really great video, how to fix that list index out of range issue? I am getting that too

[@siddharthachakraborty5046]

The explanation is so good. Thanks for sharing this. Just had one question, even after doing pip install lida ; getting the following error of ModuleNotFoundError: No Module named ‘lida’; now is this because I am using python 3.8? Also is there a work around. Any answer would be of immense help. Thanks.

[@atulkishore9159]

Use python 3.11 and set the pycharm interpreter to the venv. The problem will be solved.

[@PurushothamReddy-ff6vp]

Hello, is there anyway to return the raster image directly into the streamlit rather than converting it into images? i lose a a lot of functionality when im converting into an image.

[@sohailshaik3323]

Hi Bro This code is not working with Azure OpenAI credentials.Can you help me out if you are aware of the solution

[@ayush_singh_23]

can you help me out in something - I want to establish a relation between multiple tables who are dependent upon each other and i want to query a search to retrieve some data that is dependent on other table . how can i establish relation between them so that my Gen ai chat interface can fetch data and process the user query and return the correct response ?

[@VaishaliAgrawal-c2h]

I am getting error: streamlit:command not found. Even after pip installing it.

[@vaibhavsinghania8368]

hello . I am getting TypeError: 'OpenAI' object is not callable. After having tried everything. Can someone pls help? sir?

[@victorreyesalvarado8329]

thank you again crack! a question here: what version of python are you using?

[@AIAnytime]

I am using python 3.10... most stable version for generative ai.

[@victorreyesalvarado8329]

@@AIAnytime thx! i will try in my pc :D

[@rameshh3821]

Does this work with SQL Agent/ SQL chain?

[@aadarshjain671]

Hey there, Can we use LIDA without using any API Key , Can we fully download all the things in our system locally as the Base Model / LLM and then apply prompt Engineering or Fine Tuning on that model/LLM with our own data ( That may be confidential ). I want to perform Data Analytics Task on Specific data and therefore i am looking for the open source LLM's to download them locally as the Base Model and then Train it on Specific Data to answer my queries on the same (Not only Summarization but also Visualization ).

[@PurushothamReddy-ff6vp]

yes, but you'll need a lot of processing power :)

[@muhammadtalmeez3276]

Can it answer the questions from data like: which country has highest happy rate?

[@azer0013]

Thank you sir you are genius ❤

[@AIAnytime]

So nice of you... You are the one my friend.

[@ironman5034]

Which version of llmx to use on windows?

[@VikasRajpurohit-t2s]

Can you mention versions of all libraries

[@jillanisofttech2977]

sir can you build proper application using this model and like chat with lida and also have memoery store function please. Proper End to End project b ban jyee ga

[@AIAnytime]

Yes. That's on the roadmap.

[@jillanisofttech2977]

Thanks btw when you build this type of project i really want.@@AIAnytime

[@ChopLabalagun]

local execution requires 100GB VRAM which is crazy. 7B models takes about 21GB VRAM and its pretty slow and small CSV file like the one you provided. Pretty bad results, i would probably got the graph myself using pandas and seaborn in the same time it look running locally :D

[@AIAnytime]

I agree. Not worth using open source LLMs...

[@ChopLabalagun]

@@AIAnytime pretty sad it would great to generate data like this.

[@rishavmishra5786]

@@ChopLabalagun Hey! so you successfully ran it on a local model? If so that's great could you possibly help me out? I'll share more info if you are up for it.

[@ChopLabalagun]

@@rishavmishra5786 it depends on what model you want to run, its pretty heavy and the results are not that good. I can do better plots myself

[@pulkit8723]

@@ChopLabalagun Hey! I had a few doubts related to this. Could you help resolve them if you don't mind ? I'll share more details if you're up for it.

[@dhanunjaythota-f4u]

ModuleNotFoundError: No module named 'lida' if i run i got this error why?

[@TanmayUmathe]

I think you haven't downloaded the module. Go to command prompt and run pip install lida.

[@arkodeepchatterjee]

please make a video on how to run in cloud

[@roshansawant9891]

ImportError: cannot import name 'AzureOpenAI' from 'openai' any solution?

[@PurushothamReddy-ff6vp]

An error in your import openai, ensure that it is present in the scripts you are running your python from ( if you multiple python environments)

[@chinnibngrm272]

why getting list index out of range while generating viz can anyone know the reason please help

[@rachitsingh4913]

that means model is not generating any graph

[@kushagrasingh7731]

LIDA Is good but the only issue is it does no perform well with more than 10 columns

[@atulkishore9159]

But no real world datasets have less than 10 columns.

[@kushagrasingh7731]

@@atulkishore9159 yup pandas ai is the best bet

[@fozantalat4509]

Sir please update the github repo with steps on how to run it locally on our laptops.

[@AIAnytime]

I have shown in the video itself. We have built it from scratch. Pls watch the video completely to understand that.

[@yuzual9506]

LOL, sometimes i think your in my brain...I my work in La Sorbonne i try to give some AI tools for academic researchers, and they want something, and youy do another vids...THX

[@arkodeepchatterjee]

please make a tutorial on how to run webui in colab or cloud

[@AIAnytime]

Just deploy the Streamlit app on cloud. Watch my AI deployment series. I have shown how to do that ..

[@arkodeepchatterjee]

@@AIAnytime no I mean the web ui that is given on their GitHub page

[@dhanunjaythota-f4u]

hi

[@pubgkiller2903]

Use some complex dataset, avoid to use other's example :)