July 2026
Models
Link a Weave prompt to a collection
Track, version, and manage your AI prompts in the W&B Registry by linking Weave prompt objects to a collection.
View Eval Tables
When you create an Eval Table, specify which columns contain inputs, outputs, and scores.
Create an Eval Table
Learn how to create an Eval Table in W&B.
Compare runs with Eval Tables
Use Eval Tables to compare inputs, outputs, and scores across multiple runs.
Overview
Learn how to create, compare, and visualize Eval Tables in W&B.
Weave
Compare evaluations
Visually compare two or more evaluations to spot regressions, improvements, and scoring differences across runs
Reference media in your own bucket (BYOB) using Weave Op
Render images and video that live in your own cloud bucket in Weave traces by logging their URIs, without importing the bytes into Weave.
Reference media in your own bucket (BYOB) using agent spans
Render images and video that live in your own cloud bucket in the Weave Agents view by returning their URIs from agent tool spans, without importing the bytes into Weave.
Configure ingest sampling for Weave Self-Managed
Keep only a share of incoming traces to control storage and LLM scoring costs on a self-managed Weave instance
Bedrock Agents
Trace Amazon Bedrock Agents invocations with Weave, capturing agent inputs, foundation model usage, and completion output.
June 2026
Models
LEET terminal UI
Explore and compare local W&B runs from the terminal with the LEET (Lightweight Experiment Exploration Tool) TUI.
Weave
View agent activity
Use W&B Weave’s Agents view to understand what your agent did, how much it cost, and exactly where things went right or wrong.
Trace sub-agents
Use Weave’s sub-agent span to trace sub-agent delegations and view nested agent invocations.
Trace your agents
Use the Weave SDK to instrument multi-turn agentic applications and view them in the Agents tab.
Batch logging for your agent
Manually log agent traces for frameworks that have already completed the LLM call and need to record it.
Pi extension
Trace Pi agentic sessions, LLM calls, and tool executions in Weave.
OpenClaw plugin
Track OpenClaw agent sessions in W&B Weave for observability and debugging.
OpenAI Agents SDK
Trace an agent built with the OpenAI Agents SDK using Weave.
Google ADK
Trace an agent built with Google’s Agent Development Kit (ADK) using Weave.
Codex plugin
Trace Codex agentic sessions, LLM calls, and tool executions in W&B Weave.
Claude Code plugin
Track Claude Code sessions in W&B Weave for observability and debugging.
Claude Agent SDK
Trace an agent built with the Claude Agent SDK using Weave.
ARIA
Overview
ARIA, W&B’s AI Research and Iteration Agent, is your personalized research assistant that helps you analyze and run experiments, explain results, identify patterns across runs, recommend next steps, build visualizations and reports, and more in W&B.
Governance and security
Learn how W&B Admins can control how your organization’s data is used by W&B’s AI features like ARIA.
Chat with ARIA
The following page describes how to start a new chat, view your chat history, delete previous conversations, undock the chat window, reference a specific run in your prompt, add an image to a conversation, and provide feedback on ARIA’s responses.
Use ARIA for autoresearch
Learn how to use ARIA, W&B’s AI Research and Iteration Agent, to analyze results and run experiments.
HiveMind
W&B HiveMind
A shared dashboard for AI coding sessions. Track activity, spend, and outcomes across Claude Code, Cursor, Codex, Gemini CLI, and more with leaderboards, team views, and efficiency insights.
Support
Why does my Weave cost or token estimate differ from my provider?
Weave displays cost and token usage estimates based on data captured from your LLM calls, and discrepancies between Weave’s numbers and your provider’s invoice can be caused by the following issues.
Why is my W&B run slow to initialize or upload?
Slow wandb.init() or sluggish metric uploads are usually caused by network latency, large media payloads, high logging frequency, or slow startup of the W&B service process.
Why is my sweep agent not picking up new runs?
If your sweep agent starts but does not receive new run configurations, or receives one run and then idles, there are several common causes.
Why is my run showing as crashed?
W&B marks a run as Crashed when it stops receiving heartbeats from the process that called wandb.init(), without the process having called wandb.finish().
Why is console output not captured for my run?
W&B captures your script’s stdout and stderr and stores it as output.log on the run’s Files tab.
Why does my API key fail with 'must be 40 characters long'?
W&B now issues longer API keys (about 86 characters).
Why do my workspace settings not persist between sessions?
Workspace layout (panels, filters, grouping) persists only when you save a view.
Why are my metrics missing from wandb.log()?
If metrics logged with wandb.log() are not appearing in the W&B UI, there are several common causes.
I deleted my team and now I can't create a new one — what do I do?
When you sign up for W&B, the platform automatically creates a personal team with the same name as your username.
How do I use W&B with JAX?
W&B has no JAX-specific integration.
How do I use the parallel coordinates chart in W&B?
The parallel coordinates chart shows how hyperparameters relate to metrics across many runs.
How do I update run config, tags, and notes via the W&B API?
After a run finishes, use the Public API guide to edit config, display name, tags, and notes without re-running the experiment.
How do I set up W&B alerts and notifications?
You can set up alerts in and notifications using the W&B Settings page.
How do I page through large API results in W&B?
You can page through API result using the standard lazy-iterator pattern and perpage parameter.
How do I log NLP metrics and text outputs in W&B?
You can log corpus-level NLP scores (BLEU, ROUGE, perplexity) with wandb.log() and per-example outputs with wandb.Table.
How do I log gradients and model weights with wandb.watch()?
wandb.watch() hooks into a PyTorch model’s parameters and gradients and logs histograms of their values at regular intervals.
How do I invite a user to my W&B team?
Only team admins can send invitations.
How do I download the console log file from a run?
W&B stores your script’s stdout and stderr as output.log (or multipart chunks under logs/).
How do I create a new team in W&B?
Teams are the primary unit of collaboration in W&B.
How do I connect to W&B Self-Managed?
W&B Self-Managed is a self-hosted deployment that runs in your infrastructure.
Can I resume a run inside a sweep?
Run resumption is not supported inside a W&B sweep.
API error code 422 - Invalid request parameters
Invalid parameters on chat completion requests often surface as HTTP 422 (Unprocessable Entity) or HTTP 400 (Bad Request) depending on where validation runs.
API error code 404 - Model not found
A 404 response from the W&B Inference API means the server could not find the model or resource you asked for.
How do I log in to W&B Self-Managed?
To authenticate to a W&B Self-Managed instance, point the CLI and SDK at your instance URL before you log in:
Which Weave API should I use to trace my agent?
It depends on what you’re building:
What is the difference between @weave.op and weave.start_session?
@weave.op traces individual Python functions and surfaces results in the Traces tab.
May 2026
Models
Automation tutorial overview
Learn to build a project run-failure alert or a registry alias automation.
Tutorial: Registry artifact alias automation
Build an automation that runs a webhook when a Registry artifact gets a specific alias like “production”.
Tutorial: Project run-failure alert automation
Build a run-failure alert that sends a Slack notification when a run in your project fails.
Manage automations with the API
Programmatic automation management with the Python API. Create and update may be affected on some client versions. Prefer the W&B App until the SDK fix ships.
April 2026
Weave
Haystack
Trace Deepset Haystack pipelines with W&B Weave using the WeaveConnector integration.
Write-ahead log
Improve the resilience of W&B Weave trace data capture with the write-ahead log
Vercel AI SDK
Trace Vercel AI SDK calls in Weave using OpenTelemetry
Set up automations
Create event-driven automations that trigger actions based on monitor metrics and trace activity.
Platform
Manage bucket storage and costs
Understand how W&B uses object storage, how deletion maps to bucket bytes, and how to reduce storage usage and costs.
Sandboxes
Secrets
Learn how to manage secrets in Serverless Sandboxes.
Run commands in a sandbox
Learn how to run commands in a sandbox environment and read output and exit codes.
Tutorial: Train a PyTorch model
Learn how to train a PyTorch model in a Serverless Sandbox environment with this step-by-step tutorial.
Sandbox lifecycle
Learn about the lifecycle of a Serverless Sandbox, including its states, how to wait for state changes, and how to stop a sandbox.
Tutorial: Invoke an agent in a sandbox
Learn how to invoke an OpenAI agent within a Serverless Sandbox environment
File operations
Learn how to read, write, and mount files in Serverless Sandboxes.
Create sandboxes
Create W&B Serverless Sandboxes that run code in isolated containers with their own filesystem, network, and process space.
Serverless Sandboxes
On-demand, isolated compute environments that you can create, use, and discard from Python with W&B Serverless Sandboxes.
Support
Why doesn't `.call()` raise exceptions?
By default, Weave’s .call() method captures exceptions and stores them in call.exception instead of raising them.
Why does my workspace load slowly?
Workspaces can load slowly when a project has many metrics, runs, or panels.
Why does my training hang with distributed training?
This article helps you resolve training hangs when you use W&B with distributed training frameworks, so your runs can start and finish without stalling.
How do I fix service account errors like `Unauthorized` or missing runs?
When you use a service account to automate W&B workflows, you might encounter authorization errors or find that runs don’t appear where you expect them.
How do I fix `Rate limit exceeded` errors when logging metrics?
If you receive an HTTP 429 Rate limit exceeded error when you call wandb.log(), you’re exceeding the rate limit quota for your project.
Why does my process stop responding when using Hydra with W&B?
This page explains how to resolve unresponsive processes that occur when you start a process with Hydra alongside W&B.
Why is my enterprise license not recognized?
This page helps administrators of W&B enterprise deployments diagnose common issues when W&B doesn’t recognize an enterprise license or its features are unavailable after you set the license.
How do I fix `Cuda out of memory` during a sweep?
If you see Cuda out of memory during a sweep, refactor your code to use process-based execution.
How do I fix `CommError, Run does not exist` during a sweep?
If you see both CommError, Run does not exist and ERROR Error uploading during a sweep, the most likely cause is that you’re setting a run ID manually in your code:
Why can't I link my artifact to the Registry?
If you can’t link an artifact to a W&B Registry, the most common cause is that the artifact was logged with a personal entity instead of a team entity.
How do I fix an `anaconda 400 error` during a sweep?
An anaconda 400 error during a sweep often means you didn’t log the metric you’re optimizing.
March 2026
Models
Signal handling and sweep runs
Learn how W&B Sweeps handle UNIX signals, exit codes, and preemption in sweep runs.
Weave
OpenAI Realtime API
Use Weave to automatically trace your calls to the OpenAI Realtime API.
Store and track versions of prompts
Retrieve and manage versions of your prompts for LLM applications
Review items in an annotation queue
Evaluate trace items and submit structured feedback using a simplified review interface.
Set up annotation queues
Create annotation queues, route traces to domain experts, and export structured feedback.
Platform
W&B Mobile App (iOS)
Track training runs, review console logs, view line plots and system metrics, and explore W&B Models projects from your iPhone or iPad.
Use W&B Skills
Install W&B Skills to teach your coding agent how to train models, build agents, and analyze experiments using W&B’s AI development platform.
Rate limits
Optional rate limits on Self-Managed instances for stability
Supported Dedicated Cloud regions
View all supported AWS, Google Cloud, and Azure regions available for W&B Dedicated Cloud instances.
Rate limits
Default rate limits on Dedicated Cloud and how to request changes
Export data from Dedicated Cloud
Export runs, metrics, artifacts, and reports from a W&B Dedicated Cloud instance using the Python SDK API.
Inference
Cline with Serverless Inference
February 2026
Weave
Manage Weave projects
Use Weave projects to organize related assets like traces, prompts, evaluations, models, and dashboards.
Platform
Use the W&B MCP Server
Connect your IDE or AI agent to the W&B Model Context Protocol (MCP) server to query and analyze your W&B data and documentation in natural language.
Inference
Model lifecycle
January 2026
Models
Pin and compare runs
Learn how to use pinned and baseline runs to keep track of important runs and efficiently evaluate model experiments.
Weave
Create dynamic Leaderboards in Evaluations
Dynamic Leaderboards let you configure, customize, save, and update Leaderboard views directly from an evaluation.
Map columns in datasets
Map columns in datasets to different names. This helps you align the column names in your dataset with the column names expected by the scorer.
Inference
Creating a fine-tuned LoRA
2025
2025
December 2025
Models
View an automation's history
View the execution history of your W&B Automations to check status, triggering events, and action results.
Evaluation benchmark catalog
Evaluate a model checkpoint
Evaluate a VLLM-compatible model checkpoint using infrastructure managed by CoreWeave
Evaluate a hosted API model
Evaluate a hosted API model using infrastructure managed by CoreWeave
LLM Evaluation Jobs
Evaluate model checkpoints or hosted API models within W&B and analyze the results using automatically generated leaderboards.