Weights & Biases, an integrated MLOps platform, introduced “W&B Weave,” a toolkit designed to assist software developers in deploying GenAI applications.

W&B Weave is devised to support the experimental workflow for building AI applications. Its main features include "Traces," which allows for full observability of application behavior by logging every interaction, and "Evaluations," which provides a customizable scoring method for the applications, helping to spot trends and make informed decisions about future iterations.

The company claims that the toolkit's benefits lie in its ability to provide software developers with a record of the experimental development process of large language model (LLM) applications. It states that the toolkit enables experimentation, continuous improvement, and progress evaluation in deployments, which is similar to how unit tests function in traditional software development.