Thank you for your suggestion.

Indeed, such functionality is often beneficial in data analysis software. However, PyGWalker follows a distinct design approach. It does not manage all aspects of the GUI, particularly those related to dataframes. This design choice grants users greater flexibility to customize features directly in Python.

For example, in your case, you could conceal certain columns either by modifying the dataframe before passing it to PyGWalker or by adjusting the field_specs parameter.

apart from the GUI functionality (which wouldn't be a great effort to implement and the benefits outweigh design restrictions) - what about the export / return analysis data? the export or "function return data analysis" in JSON or at least CSV would also come in handy. Reasoning: the analysis process doesn't stop with pygwalker, it is a tool in a whole process, where it receives data and forwards analysis outputs. It should be considered as "embedded" tool in a process with "API-based" data input and output to avoid media breaks (as currently with the "standalone" architecture). Otherwise, other tools will fill in that important gap and pygwalker could loose traction.
What's your opinion on these considerations?

export csv

pygwalker had never been and never will be an standalone architechture, it is designed for embeddable analytics without breaking the current workflow since day one.

Not only it can export csv, but dataframe, sql, view state, painter values as well. We are making more tutorial videos to introduce these features, but welcome to explore them by yourself.

Remeber, pygwalker is a component, not a system, you can use it in jupyter notebook, colab, notion etc. It can interactive with its env and connect the workflows. "Standalone" architecture is the most improper tag to pygwalker.

export csv

I am aware of the export button and apologise for not being clear enough - I was suggesting an "return results" function, when f.ex. closing pygwalker (part of an analysis chain). To give you an outlook - in order to automate, make complex analysis more comprehensible, pygwalker is "only" an (important) step of a whole analysis chain, f.ex. I am trying to incorporate pygwalker in an autogen (Microsoft llm agent framework) "dynamic" process, but as I cannot return any results (only separate file via export.csv, manually) it is of limited final use ...

Firstly, you can get the result as dataframe in python API of pygwalker. Secondly, I am not sure I understand what you want to implement with pygwalker, you want some AI agent to make "drag-and-drop" operations to analysis the data in pygwalker? Can you share more about what you want to build so I can provide more helpful suggestion based on pygwalker?

Firstly, you can get the result as dataframe in python API of pygwalker. Secondly, I am not sure I understand what you want to implement with pygwalker, you want some AI agent to make "drag-and-drop" operations to analysis the data in pygwalker? Can you share more about what you want to build so I can provide more helpful suggestion based on pygwalker?

ad result as data frame in python API from pygwalker)

• please excuse my ignorance, but can you please point me to this API?
• I only saw so far API calls for sending data to pygwalker, but not receiving data from pygwalker ...

ad use case)

• I have a streamlit application, which has autogen framework under the hood with two agents and a number of functions.
• one of these functions is to invoke pygwalker and send the data to be analysed as pd data frame
• what I am now seeking for is to receive the analysis back as answer to the function call (as pd data frame or json or ...)