Abstract
Large language model (LLM) based coding agents may exhibit inconsistency when using software libraries, for example, by referencing outdated APIs from training data or generating calls to non-existent APIs. Some existing mechanisms for providing supplemental context may involve manual discovery and configuration by the user. A described system and associated standard allows library authors to package guidance, which may be specific to language models, directly within their distributable software package, for instance, in a dedicated directory. Tooling can be configured to discover this guidance across a project's dependencies, aggregate it, and provide it to a coding agent in a structured, contextual manner. This approach can furnish the agent with version-aware and tailored instructions for library installation, usage, and upgrading, which may improve the accuracy of code generation and facilitate automated delivery of context.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Bleigh, Michael, "A System for Providing Language Models with Contextual Guidance Discovered from Software Library Packages", Technical Disclosure Commons, (September 22, 2025)
https://www.tdcommons.org/dpubs_series/8611