https://www.boundaryml.com logo
heyy, pls i am working on a very
# general
t
heyy, pls i am working on a very deterministic generation prompt, do i put the constraints in the prompt, as description to the fields in the output schema type or both??
v
as with all things on LLMs, sadly the most conclusive way to know in your specific scenario is to test it. personally, I've found using descriptions works best
t
thanks, i tested it with both and i am satisfied with the results
pls another issue i am facing is I'm trying to use baml inside of wasm with rust but the runtime I'm using requires wstd crate to be used for http requests instead of reqwest crate. Pls could you point me to any resource that shows the raw request that baml sends so i can construct my own http request using wstd crate? Thanks
v
hey! are how you using BAML inside rust btw? openapi?
i'm almost done with our rust sdk and hopefully will ship it asap
just touching out a few things like async streaming semantics
t
yeah, I'm using openapi but I am using it inside of golem cloud - a wasm runtime for durable execution which as at now doesn't support the reqwest crate so i need to use the wstd crate. Pls how can I achieve this easily?
v
oh 🙁 sadly we don't yet ship BAML w/ support for that 🙁
what target are you building for?
I can see what it would take to support that
likely best route would be to run the baml openapi server on your own machine somewhere
then ping it from rust however makes most sense!
but idk if rust openapi has support for wstd. that would sadly be an openapi question (which i can try and help with but likely not until this weekend)
t
I am actually working on a tight deadline for the project, so I'll try and figure out fairly easy solution for this. Anyway, thanks soo much for your speedy replies, I really appreciate.
maybe, i could copy the raw requests from the baml playground vscode extension and implement them using wstd for now. That should be my best case only it would be a bit tedious