Evaluate integration with AI models

[SoylentBob]

There are some use cases, where it might be useful for gavel to ingest code or entire components and use an AI model to analyze them.

Recent post on the Martin Fowler Blog: https://martinfowler.com/articles/legacy-modernization-gen-ai.html

An approach may be to include an optional dependency on a llamafile such as: https://github.com/Mozilla-Ocho/llamafile

Some ideas I have in mind:

• We could analyze whether a component actually implements a well defined pattern (e.g. a repository, a service, etc.) as defined by some document or public knowledge
• Find out, whether the domain layer actually implements the rules that have been defined in the requirements
• Ask the AI questions regarding the assessed architecture

[SoylentBob]

This repo describes how to set up a RAG process with llama files: https://github.com/Mozilla-Ocho/llamafile-rag-example/

[SoylentBob]

https://github.com/cameronking4/Annotated-AST-For-LLM

[SoylentBob]

https://medium.aiplanet.com/understanding-and-querying-code-a-rag-powered-approach-b85cf6f30d11?gi=11d2dde6ba9e

[SoylentBob]

https://severalnines.com/blog/vector-similarity-search-with-postgresqls-pgvector-a-deep-dive/