Boundary
Boundary, formerly known as Gloo and Gloo Chat, is a Seattle-based B2B company specializing in AI development tools, including the BAML programming language and its VSCode extension. It is recognized for its innovative features like converting production errors into test cases and providing type-safe monitoring through Boundary Studio.
Company Overview
Boundary, previously known as Gloo and Gloo Chat, operates from Seattle, WA, USA. The company has a small team of 2 members and functions within the B2B industry, particularly focusing on AI development tools. Founded by former engineers with a decade of AI system-building experience, Boundary participates in the Y Combinator W23 batch. Their operations span across the United States and parts of Canada, including remote and partly remote work arrangements.
Products
Boundary offers innovative tools for AI development, including the BAML programming language, a VSCode extension, and Boundary Studio. BAML is designed for defining AI functions with zero boilerplate and supports type-safe outputs in Python and Typescript. The VSCode extension for BAML assists developers in creating structured outputs from large language models (LLMs) within a familiar coding environment. Boundary Studio provides an analytics dashboard for tracing, labeling, and measuring AI performance. Additionally, Boundary plans to introduce React components to collect type-safe user feedback.
Key Features
Boundary provides several unique features aimed at AI developers. Their VSCode extension for BAML includes a playground for LLMs, enabling instant testing and enhancing the development experience. One-click conversion of production errors into test cases allows for quick correction and redeployment. Furthermore, the system supports type-safe logging of BAML functions and their inputs and outputs to Boundary Studio for comprehensive monitoring.
Supported Technologies
Boundary’s BAML compiler is primarily written in Rust and supports various models, including those from Openai, Anthropic, Gemini, and Mistral. It also allows for custom model integration. The company’s toolchain is test-driven, designed to guarantee type-safe LLM outputs without boilerplate code. BAML’s features are built to support both Python and Typescript, ensuring wide compatibility and usability within the AI development community.
Documentation and Support
Boundary provides detailed documentation and a frequently asked questions (FAQ) section on their website to assist users. These resources cover various topics, such as BAML's utilization of LLMs, the programming language used for the BAML compiler, supported models, deployment process, security, and comparison with other frameworks. Additionally, installation instructions for MacOS, Windows, and Linux are available to help developers get started with BAML and its VSCode extension.