We are excited to release SAFA's generated documentation for Tensorflow. You will be able to view: -Project Overview (Overview, Features, Entities, Dataflow) -Code Files and their summaries (975 Code Files) -Functional Requirements (371 Functional Requirements) -Features (182 Features) -Mapped Relationships with explanations Learn more at https://safa.ai |
That said, I have to commend the clarity of your design. You focus on reconstructing high level artifacts like stories, requirements, and features. LLMs do an incredible job at generating these items. But you also go beyond, and visualize the relationships.
Traditionally only large enterprises could afford to generate these kinds of documents, and even then they were slow and out of sync with real requirements.
Now, not only can I use these processes for small or single worker enterprise work, I can use these methods to have excellent planning and documentation for open source work.
When we hear someone complain that AI doesn't make their dev work better, that's usually because they're not using AI to extract requirements, plan tests, stub out work, etc. They're not putting the AI in the position to be a good PM, architect, tester, etc. They're just throwing the AI at meager, poorly documented contexts, as if the AI is a junior doing a coding interview. Then they catch it in a mistake and kvetch, as if the failure was not poor leadership on their part.
I think systems like this - especially when open sourced - will give teams the kind of structure and process they need to make use of AI assisted engineering, and make the AI more like a first class member of the team, in the way skilled AI users already do, but with less effort. It's really exciting!