Abstract
In the field of software engineering, maintaining up-to-date and accurate documentation can be challenging as code evolves, and manual update processes may result in documentation being incomplete or inconsistent. To address this, systems and methods are described for an automated documentation maintenance pipeline that may be driven by one or more large language models. The pipeline can be initiated by a code change event in a version control system. It may analyze the code modification and retrieve relevant context from a persistent memory store, which can contain, for example, a vectorized documentation corpus and a historical log of approved changes. Using this context, the system can generate a proposed documentation update, which can then be presented to a human for review. The outcome of this review can be recorded, creating a feedback mechanism that may be used to adjust future system outputs, potentially reducing manual effort and contributing to documentation consistency.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Kuligin, Leonid, "LLM-Driven Self-Improving Agentic Pipeline for Software Documentation", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12016