Skip to main content

Introduction

Weave is a lightweight toolkit for tracking and evaluating LLM applications, built by Weights & Biases.

Our goal is to bring rigor, best-practices, and composability to the inherently experimental process of developing AI applications, without introducing cognitive overhead.

Get started by decorating Python functions with @weave.op().

Weave Hero

Seriously, try the 🍪 quickstart 🍪 or Open In Colab

You can use Weave to:

  • Log and debug language model inputs, outputs, and traces
  • Build rigorous, apples-to-apples evaluations for language model use cases
  • Organize all the information generated across the LLM workflow, from experimentation to evaluations to production

Key concepts​

Weave's core types layer contains everything you need for organizing Generative AI projects, with built-in lineage, tracking, and reproducibility.

  • Datasets: Version, store, and share rich tabular data.
  • Models: Version, store, and share parameterized functions.
  • Evaluations: Test suites for AI models.
  • [soon] Agents: ...

Weave's tracking layer brings immutable tracing and versioning to your programs and experiments.

  • Objects: Weave's extensible serialization lets you easily version, track, and share Python objects.
  • Ops: Versioned, reproducible functions, with automatic tracing.
  • Tracing: Automatic organization of function calls and data lineage.
  • Feedback: Simple utilities to capture user feedback and attach them to the underlying tracked call.

Weave offers integrations with many language model APIs and LLM frameworks to streamline tracking and evaluation:

Weave's tools layer contains utilities for making use of Weave objects.

  • Serve: FastAPI server for Weave Ops and Models
  • Deploy: Deploy Weave Ops and Models to various targets

What's next?​

Try the Quickstart to see Weave in action.