pretty cool, we are working/learning on something similar and its good to see were on the right track but LangGraph is a tough nut to crack so far. Any pointers in learning more about meta-data creation/filtering ? Best of luck with your business brother.
@@levbszabo Multiple: as part of our start-up we are working on several projects that could benefit from a LangGraph back-end. Our first project is an innovative RAG based solution for enterprises while our second project focuses on e-commerce. The e-commerce market is massive and something we researched extensively over the years. We already build several MVP's using low-code platforms but decided to dive deep and do it from scratch which is fun but difficult. Were recording the whole process and hope to be inspiring others to follow their ambitions.
Seeing your follow up question now - this is what I used in this example python.langchain.com/v0.1/docs/modules/data_connection/retrievers/self_query/ if you are missing the meta-data, then there are unsupervised learning methods you can use to pre process. I've set up a free community where I am happy to share resources + answer questions if you are interested course.journeymanai.io/ best of luck!
@@levbszabo appreciate the reply. I signed up for your community and really like your course outline (sadly not able to afford it right now but it looks well structured). Out of curiosity can you talk a bit more about unsupervised learning for meta-data creation ? that's new to me.
@@awakenwithoutcoffee I appreciate it! I’ll be posting some sub modules for free on here as they become available. Some unsupervised tools you could use **LDA** identifies topics from text data, outputting themes that can be used as metadata. **K-Means** clusters feature vectors (we could map the product descriptions to a vector using openai vector models (first) , grouping similar items to assign metadata like categories.
Hey , thank you for the explanation, wouldn't be better if we use other tools such as PythonREPL to exctract information from this csv file / dataframe ?
Hey, my hesitation with PythonREPL is that it is much less deterministic and has a tendency to error out. But a tool that can 'slice/dice' the dataframe is definitely useful