Next Steps

[curran]

• Modify VizHub to use the package in https://github.com/vizhub-core/vizhub/blob/main/api/src/endpoints/editWithAIEndpoint.ts
• Refactor the library to decouple the actual LLM invocation step from the core of the library - would make testing more straightforward
• Change the tests so they don't do the crazy mocking of packages
• Add tests that actually call out to an LLM
• set up a file-based LLM response cache to avoid hitting OpenRouter every time
• Build out code editing benchmarks!