Repository: github.com/wjbmattingly/youtube-txtai Help video for Anaconda and Environments: ua-cam.com/video/mIB7IZFCE_k/v-deo.html (a little old but still very useful)
Thanks for the great video on txtAI, just loved the way you explained it. Thanks. I would like to see more txtAi + Streamlit app kind of videos that you mentioned in the end of this video.
I failed installing txtai[all] with errors in fasttext block (cannot build wheels for fasttext). With pip install txtai, there was no error but still txtai is not "detectable" in vscode when importing. This is on windows 11 with visual studio (c++ build tools) installed. The version of python is 3.11 and the virtual environment is placed in drive D: (physically separated from drive C because I use SSD for C: and HD for D:). What could be the problem? Thank you.
This video is so helpful for people like me who are taking baby steps towards NLP , I would really love to see how to update a txtai model, the github code for untitled.ipynb might be corrupt I couln't see the code , however wasn't a big issue I could write the code for drawing inferences by myself after seeing your video
Hi another awesome little video that not only shows a great use case but how to get up running with the code. Thank you. In answer to your questions yes to all! Integrated with streamlit absolutely as this would be how I would apply it. And understanding how to update the model would be great. Also say I had two written texts broken down into smaller documents could I return a tag with the results to see which text the document was returned from?
Thanks! Awesome! I will put that video together that shows how to do that. As luck would have it, I have just done this for another project. The nested index will also be included.
Big next thing is an appropriate UI for users including showing them the source of each finding. Also using a vector database that doesn’t rely on memory which won’t work at scale. This is very cool though!
Great video as always, thank you for the tutorial! :) Just one quick question: I don't know anything about semantic search nor TextAI, but is it language indipendent? Or has it been trained on English?
@@python-programming oh well, that would be awesome! You don't have to though, it's fine :) I was just wondering whether there was a transformer model for Latin!
If want to build system to research papers data base to determine new research originality. What the best type of files csv or json to apply machine learning algorithms on it
@@python-programming yes I want to sit it on the web ... and it will contains abstracts and titles for the research papers.. and it supposed to help students to determine researches with most similarity ot their proposed research.. I am applying your lessons about South Africa data set. You used json in the beginning . So I'm still learning and didn't complete all your lessons.. Your lessons helped me so much. I'm grateful for it 🙏
I don't think you mentioned whether txtAI sends my/your data off to the cloud somewhere for analysis or refers to any external APIs or providers which would be important if dealing with sensitive information? Or who/what is behind it and whether it is legitimate? Paid/free? Or if there are licensing restrictions?
Confirming that txtai is all local and doesn't send your data off to the cloud. You can download a model, disconnect your internet and everything will still work.
Thanks for the valuable videos. I have involved in semantic search mapping text as query and image/ other pdf docs as output. I have followed the approches for unstructured images/pdf/other extensions---->Tried like OCR based text extraction from images , pdf text extraction for pdf files only, bert embedding and doing clustering the images. Any other inputs or approaches using libraries from your end will be helpful for semantic search on unstructured data/images/pdf. Whether txtai is open sourced helpful for QA between images and text. Kindly reply.
Im looking to create a web scraper thing in python that basically makes me find what im actually searching for. I want it to include something ive named table search. If anyone cares to ask, ill explain what that is/ how im thinking of making it work.
There are a lot of great vector database options available. txtai strives to make it easy to get up and running fast. It has built-in vectorization, vector storage, hybrid search and an LLM workflow framework for retrieval augmented generation (RAG). Everything runs local, no external APIs are required.
Repository: github.com/wjbmattingly/youtube-txtai
Help video for Anaconda and Environments: ua-cam.com/video/mIB7IZFCE_k/v-deo.html (a little old but still very useful)
Also in this video, I reference the previous video. That video will go live next week instead.
Please we need more videos on TextAI.
Your channel became my favorite!!! Kudos!!
Great video! I might have an opportunity to use this at work instead of some NER I was doing.
Awesome!! Glad to hear it was useful!
Such a wonderful tutorial on txtai please do more video of txtai and paperai.
Thanks!! Will do!
Thanks for the great video on txtAI, just loved the way you explained it. Thanks. I would like to see more txtAi + Streamlit app kind of videos that you mentioned in the end of this video.
Thanks! So glad you enjoyed it! I will be doing more in the near future!
I failed installing txtai[all] with errors in fasttext block (cannot build wheels for fasttext). With pip install txtai, there was no error but still txtai is not "detectable" in vscode when importing. This is on windows 11 with visual studio (c++ build tools) installed. The version of python is 3.11 and the virtual environment is placed in drive D: (physically separated from drive C because I use SSD for C: and HD for D:). What could be the problem? Thank you.
This video is so helpful for people like me who are taking baby steps towards NLP , I would really love to see how to update a txtai model, the github code for untitled.ipynb might be corrupt I couln't see the code , however wasn't a big issue I could write the code for drawing inferences by myself after seeing your video
Hi another awesome little video that not only shows a great use case but how to get up running with the code. Thank you. In answer to your questions yes to all! Integrated with streamlit absolutely as this would be how I would apply it. And understanding how to update the model would be great. Also say I had two written texts broken down into smaller documents could I return a tag with the results to see which text the document was returned from?
Thanks! Awesome! I will put that video together that shows how to do that. As luck would have it, I have just done this for another project. The nested index will also be included.
At 2:40 it says txtai needs Python 3.7. What is meant is Python 3.7 or higher.
So there is no need to install Python 3.7.
Big next thing is an appropriate UI for users including showing them the source of each finding. Also using a vector database that doesn’t rely on memory which won’t work at scale. This is very cool though!
Thanks for introducing the txtai.
BTW the model works with python 3.9.9 on Mac.
No problem! Thanks for that update on mac! I only have linux and Windows machines. Purchasing a mac is on my to do list.
@@python-programming How did you get the data set?
How did you generate the data file? would I have to build my own data sets for this?
Great video as always, thank you for the tutorial! :)
Just one quick question: I don't know anything about semantic search nor TextAI, but is it language indipendent? Or has it been trained on English?
Thanks! Oh that is a great question. I did not speak about that. You will want to select a sentence transformer model for your language
But it is language agnostic as a workflow
@@python-programming cool, thank you very much! Do you know how many languages these models cover? Is there one for classical languages also?
@@AndrewPeverells a good deal! What language(s) are you looking for? I can do some research and maybe demo that in a video for you.
@@python-programming oh well, that would be awesome! You don't have to though, it's fine :) I was just wondering whether there was a transformer model for Latin!
If want to build system to research papers data base to determine new research originality. What the best type of files csv or json to apply machine learning algorithms on it
Great question. The file really is not important. It comes down to preference and use case. Will it sit on the web? JSON may be better.
@@python-programming thank you.. 😊
@@python-programming yes I want to sit it on the web ... and it will contains abstracts and titles for the research papers.. and it supposed to help students to determine researches with most similarity ot their proposed research.. I am applying your lessons about South Africa data set. You used json in the beginning . So I'm still learning and didn't complete all your lessons..
Your lessons helped me so much. I'm grateful for it 🙏
Is this library safe ? In terms of data breaches ?
I don't think you mentioned whether txtAI sends my/your data off to the cloud somewhere for analysis or refers to any external APIs or providers which would be important if dealing with sensitive information? Or who/what is behind it and whether it is legitimate? Paid/free? Or if there are licensing restrictions?
Last I checked, it is all local. The creator is very pro open source.
Confirming that txtai is all local and doesn't send your data off to the cloud. You can download a model, disconnect your internet and everything will still work.
@@neuml thanks for responding!
Thanks for the valuable videos. I have involved in semantic search mapping text as query and image/ other pdf docs as output. I have followed the approches for unstructured images/pdf/other extensions---->Tried like OCR based text extraction from images , pdf text extraction for pdf files only, bert embedding and doing clustering the images.
Any other inputs or approaches using libraries from your end will be helpful for semantic search on unstructured data/images/pdf. Whether txtai is open sourced helpful for QA between images and text.
Kindly reply.
Im looking to create a web scraper thing in python that basically makes me find what im actually searching for. I want it to include something ive named table search. If anyone cares to ask, ill explain what that is/ how im thinking of making it work.
You're in South Africa? If you're in Cape Town I can show you around.
Thanks! I am actually and would have totally taken you up on that but we are leaving soon. Coming back next winter though!
Would be nice to search by phrases not just a word
what makes txtai different from the alternatives?
There are a lot of great vector database options available. txtai strives to make it easy to get up and running fast. It has built-in vectorization, vector storage, hybrid search and an LLM workflow framework for retrieval augmented generation (RAG). Everything runs local, no external APIs are required.
Make more interesting video on txt ai
Will do! Thanks for the tip
The GH link is 404
Oh no! I will make it public when Inget to a computer. Thanks for letting me know!
Fixed! Thanks again
👋
Please more videos. You’re my favorite channel now!!!