graph-persistence

This is a very simple project that shows the setup of a memory mechanism for a human-in-the-loop system.

MemorySaver is an ephemeral tool and does not persist data across sessions, thus the data in this example will be saved in a SQLite database.

%%{init: {'flowchart': {'curve': 'linear'}}}%%
graph TD;
	__start__([<p>__start__</p>]):::first
	step_1(step_1)
	human_feedback(human_feedback<hr/><small><em>__interrupt = before</em></small>)
	step_3(step_3)
	__end__([<p>__end__</p>]):::last
	__start__ --> step_1;
	human_feedback --> step_3;
	step_1 --> human_feedback;
	step_3 --> __end__;
	classDef default fill:#f2f0ff,line-height:1.2
	classDef first fill-opacity:0
	classDef last fill:#bfb6fc

Loading

Utilities

poetry add langgraph langchain-community python-dotenv black isort grandalf langgraph-checkpoint-sqlite

Note

The DevContainer was added to analyze the checkpoint.sqlite content, however, the extensions used in vscode are not compatible with the format being saved in the checkpoint and metadata columns and the values will be displayed as hex, instead of JSON.