found this cool way to stop losing info in random design docs and runbooks. most of us have a massive mess of architectural notes and incident reports scattered everywhere, making it impossible to see the full picture. you can actually use python to pull all that unstructured data into a single
knowledge graph so everything is connected. instead of manually tracing how the checkout service relies on the payment api, the graph does the heavy lifting for you. it basically maps out your dependencies and database links automatically from your existing text files.
it saves so much time during on-call shifts when you actually know what will break . i was thinking about trying to run
pip install spacy
to get started w/ the extraction part. does anyone else use something more advanced than just basic python scripts for this? it feels like
manual documentation is a lost cause at this point. automating the connections seems like the only way to survive a growing codebase.
article:
https://dzone.com/articles/python-graphrag-knowledge-graph