How to add local Pandas datframe in Agentic workflow for Analysis?

[mraguth]

Is it possible to get an local pandas DataFrame in agentic workflow and ask an agent to analyze the structured data using Python (as suggested in this link)? I love this concept and am trying to expand it to real-life examples by adding more agents. I'd appreciate any advice and sample code.

I am using this link for building agent: https://langchain-ai.github.io/langgraph/tutorials/multi_agent/multi-agent-collaboration/#define-tools

I am using this link for Python analysis: https://python.langchain.com/v0.1/docs/use_cases/sql/csv/#chain

[simoneb]

We are also trying to do something similar. An agent returns some structured data, e.g. a DataFrame (or anything else really, but not a string) and this data should stay in memory and doesn't need to be passed to the model, because it's not textual data, and it would be too large to be passed around in interactions with the model. How do we store this data in memory without serializing it, and just provide it as inputs to an agent?