Give Agent more context

[shroominic]

Input more context like

• all the files in the current working directory (cwd) or
• all the available variables in the codebox kernel to improve muilti step actions like analyzing a dataset or doing stuff with multiple files at once.

[absin-lt]

Context awareness (it can refer to your previous messages) is something I need. Some context: A huge use case of code interpretor is multi-exchange as in multiple messages sent to the LLM appended with actions performed. For example analytics on raw data extracted in csv or xlsx or sql and so on. @shroominic It seems that the latest version does not remember conversation context, unlike Open AI's code interpretor. Is this assumption fair and is this the issue where you are thinking of adding this feature?