Currently, I am trying to create traceability links between the docs and the code, which will serve as the foundation for drift detection.
Using a subset of the LangChain documentation as a test case, I have achieved ~50% precision, ~90% recall, and ~1 min recovery time.
Is FluentDocs dead?
To keep it short: no. To provide fast-moving teams with more reliable drift detection, I have pivoted to a deterministic-first approach using static analysis for accurate drift detection and LLM only where needed.
fluentdocs.dev/