2025-12-04 Thu
- Rust/Control/Flow
- BAML
- Via the AI Engineer Code Summit, I found out about BAML (Basically A Made-Up Language) which is an open source project made by BoundaryML. The idea is to bring the spirit of strongly typed programming languages to prompts that require structured input and output. It comes with a VS Code extension for quick testing against multiple use cases.It strikes me as as basically a much more up-to-date, re-imagined version of what the original langchain package set out to be. Today the LangChain ecosystem does a lot more, but this focuses on the prompt layer. Like LangChain, BAML does come with integrations for most of the LLMs out there (including the ones we use). What do you think of this idea? I want to try it out.I love how retries and fallbacks are just included; no extra boilerplate needed. I also love how easy it is to try over a number of different LLMs.
Since BAML is a configuration file format, the BAML toolchain allows you to transpile BAML into native Python, Typescript, Ruby, etc code. So you get autocomplete and static analysis without thinking about it. Plus, our compiler is fast - it takes less than 30 ms to generate code!
- https://www.youtube.com/watch?v=gxckvkNg6KM
- Comparing Langchain | Boundary Documentation comparison with LangChain
- Via the AI Engineer Code Summit, I found out about BAML (Basically A Made-Up Language) which is an open source project made by BoundaryML. The idea is to bring the spirit of strongly typed programming languages to prompts that require structured input and output. It comes with a VS Code extension for quick testing against multiple use cases.It strikes me as as basically a much more up-to-date, re-imagined version of what the original langchain package set out to be. Today the LangChain ecosystem does a lot more, but this focuses on the prompt layer. Like LangChain, BAML does come with integrations for most of the LLMs out there (including the ones we use). What do you think of this idea? I want to try it out.I love how retries and fallbacks are just included; no extra boilerplate needed. I also love how easy it is to try over a number of different LLMs.