I haven't always done this, and the knowledge base used to visibly degrade over time. Reviewing a PR does not take a long time, maybe a few minutes, and this compounds over time.
LLM evaluations are very sensitive to the details of the prompt's structure. This post shows how using structured generation reduces the results' variance and the ranking shifts.
That whole structured generation line of work looks promising. I hope someone else takes this and runs evaluations on other benchmarks. Curious to see if the results translate!
This article presents a way to make structured generation with LLMs much faster than standard generation, but what I find most interesting is how it highlights the issues that tokenization entails towards the end.
It is currently limited by the time it takes to build the index. There are obvious optimizations we can apply to this, however in a production setting it does not matter much since you only need to build the index once for each (schema, vocabulary) pair.