GitHub - BoundaryML/baml: A programming language to build strongly-typed LLM functions. BAML includes a VSCode LLM Playground and tracing / analytics out-of-thebox.
Mar 26, 2024 - github.com
BAML is a programming language designed to extract structured data from Language Learning Models (LLMs). It provides a type-first approach, allowing developers to define types for inputs and outputs in BAML files, and use a generated Python/Typescript client to call LLMs with those types and templates. BAML also offers tools for type safety, prompt engineering, and observability. It supports multiple LLM providers and allows for fast iteration loops, robustness for production, and the inclusion of comments in prompts.
The BAML compiler is written in Rust and does not use LLMs to generate code. It is free and open-source, with paid capabilities available through Boundary Studio for monitoring, collecting feedback, and improving AI pipelines. BAML is developed by Boundary, a company based in Seattle, WA, which is currently hiring software engineers.
Key takeaways:
BAML is a programming language designed to extract structured data from Language Learning Models (LLMs).
The language takes a type-first approach, allowing developers to define types for inputs and outputs in BAML files.
BAML offers features like type safety, fast iteration loops, robustness for production, and support for multiple LLM providers.
The BAML compiler is written in Rust and does not use LLMs to generate code, ensuring a secure and efficient development process.
