Skip to content

Phoenix 21 changes authentication defaults and adds annotation kinds

Updated Phoenix

Summary

The server and related clients need a coordinated upgrade, especially for Docker and notebooks.

Category
Agent infra
Date
Pricing
Not stated in the sources we cited.

Phoenix 21.0.0 changes deployment defaults and adds DECISION and AGENT annotation kinds. Related Python and TypeScript client, CLI and MCP package releases are grouped here, along with server 20.27.0, rather than counted as separate product launches.

Check bind address and credentials first

The versioned migration guide makes an important distinction: the default bind address is loopback. Unless explicitly overridden, authentication is enabled for non-loopback addresses and disabled for loopback. Docker binds to a non-loopback address and therefore requires authentication by default; Helm already enabled it explicitly.

Trace exporters and API clients need an API key when authentication is enabled. Notebooks now honor authentication settings. The phoenix serve --host option is removed in favor of PHOENIX_HOST.

Secrets and sessions are separate concerns

Authentication no longer requires PHOENIX_SECRET, but the migration guide warns that without it someone holding the database can decrypt saved credentials. Setting or changing the secret makes existing saved credentials unreadable, requiring re-entry. Read the migration guide before changing a live deployment; do not interpret an optional setting as equivalent credential protection.

Browser sign-out no longer terminates OAuth2 sessions used by CLI or MCP clients. Client annotation updates require a server supporting the new kinds; older servers reject them. Version 20.27 also adds prompt-version timestamps and decision-model evaluator support.

See the phoenix hub.

More from Phoenix

View all

Questions

What is Phoenix?

The GitHub repository Arize-ai/phoenix is described as "AI Observability & Evaluation". Its docs list features such as LLM-based evaluations, prompt management, a prompt playground, span replay, datasets and experiments.

Is Phoenix open source?

Not stated in the cited sources. The excerpts say the GitHub API does not report a recognized license for the repository, so no license is claimed here.

How is Phoenix used?

According to its docs, you can create datasets by collecting traces or uploading from code/CSV, run experiments to compare different versions of your application, and attach evaluators to datasets so they run automatically during experiments.

Filed on Agent Radar. Claims in this note are limited to the sources above.