Weights & Biases Weave
Weave is a toolkit by Weights & Biases for developing Generative AI applications. It helps log and debug LLM inputs/outputs/traces, build evaluations, and organize information across the LLM workflow. It is for developers building with language models who need observability and rigorous testing.
✨ Key features
- Log and debug LLM inputs, outputs, and traces
- Build rigorous, apples-to-apples evaluations for LLM use cases
- Organize information across experimentation, evaluations, and production
- Trace any function with @weave.op decorator
- Integrates with OpenAI, Anthropic, Hugging Face, etc.
- Free tier available with Weights & Biases account
🎯 Use cases
- Track and debug API calls to OpenAI or other LLM providers
- Monitor generation calls from Hugging Face models
- Validate and trace data transformations in LLM pipelines
- Build and run evaluations for LLM-based features
- Organize traces and evaluations for team collaboration
📦 Installation
🧰 Requirements: Python 3.10 or higher and a Weights & Biases account (free tier available).
pip install weave
🚀 Usage
import weave
weave.init("my-project-name")
@weave.op
def my_function():
# Your tracked code!
pass
⚠️ Good to know
The codebase is being cleaned up; Weave engine and boards are paused as focus is on Tracing and Evaluations.
❓ FAQ
How do I trace a function?
Decorate the function with @weave.op after initializing Weave with weave.init().
What Python version is required?
Python 3.10 or higher.
Do I need a Weights & Biases account?
Yes, you need a Weights & Biases account, but a free tier is available.
Can I trace calls to OpenAI or other LLM providers?
Yes, you can trace any function, including API calls to OpenAI, Anthropic, Google AI Studio, and Hugging Face.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.