Best LLM for simple code completion with large context window Here's what I'm trying to do - this is _not_ for use in an IDE. Imagine you have a SaaS and you want your users to be able to completely control the SaaS with natural language. So you reduce all of the SaaS functionality into a set of typescript types (including function types), will documented, with every parameter accounted for. And then, when a user types a request in natural language, you ask an LLM to produce a set of function calls to be executed, and after confirming with the user, you run them. To do this you need to pass the entire typescript definitions for the entire SaaS as well as all of the user's data relevant to their query into the context window. And the LLM has to be particularly good at writing (simple) Typescript code. What's the best for this? Using GPT-4 Turbo for now, but really not up to date on the latest models and versions. |