Trying to make new features robustly and safely for the TS side. However, I don't think just unit testing is the best way. (Mocks just give a false sense of security in this case) How about integration tests using an OpenAI LLM + local data in the library + new feature (ie. RouterQueryEngine, etc)?

Does anyone know how to build a search engine like Perplexity? open to ideas

wrote a linkedin post about how I got started contributing to LITS if anyone wants to give a like or anything 🙂
https://www.linkedin.com/posts/elliotkang7_zero-to-hero-in-open-source-in-30-days-i-activity-7122393769374580736-HDar?utm_source=share&utm_medium=member_ios

How well do the Eval modules hold up on the Python side? Gonna have to port some version over to TS

any NodeJS folks in the house?

What's the success rate on getting a LLM to return JSON formatted answers when prompted to?