> ## Documentation Index
> Fetch the complete documentation index at: https://arizeai-433a7140.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# 10.27.2026: Decision-Model Evaluators and Project Onboarding

> Run TypeScript classification evaluators on AI SDK decision models, get a setup guide in empty projects that clears when the first trace lands, and read annotation scores by how far they sit from the target.

TypeScript classification evaluators, including every built-in, now run on AI SDK decision models such as OpenAI's `gpt-6-luna`. New projects also show a setup guide until their first trace arrives, and annotation score colors now show how far each score sits from its target.

# Classification Evaluators on AI SDK Decision Models

October 21, 2026

**Available in @arizeai/phoenix-evals 2.7.0+ and @arizeai/phoenix-otel 2.3.0+ (TypeScript)**

Pass a decision model as `model` to any classification evaluator. A decision model picks one of the labels directly and returns a probability for each, so it is much faster and cheaper than a chat model judge.

* **Probabilities in `metadata`.** `metadata.probabilities` maps each label to its probability, and `metadata.confidence` is included when the provider reports one. Decision models do not return an `explanation`.
* **Refusals throw** the AI SDK's `Experimental_DecisionRefusalError`. In `runExperiment`, a refusal is recorded as that evaluation's error.
* **Decision calls are traced.** Decide spans record the decision state, question, and answer, and `@arizeai/phoenix-otel` 2.3.0 traces AI SDK decision models. `isDecisionModel` and the `DecisionModel` type are also exported.
* **Evaluation models change too.** TypeSafe's Jev now also goes through `experimental_decide`, and message prompts are sent as plain text instead of JSON state. Use `@ai-sdk/openai` 4.0.89 or later with OpenAI decision models.

```typescript theme={"theme":{"light":"github-light-default","dark":"github-dark-default"}}
import { openai } from "@ai-sdk/openai";
import { createCorrectnessEvaluator } from "@arizeai/phoenix-evals";

const correctness = createCorrectnessEvaluator({
  model: openai.decisionModel("gpt-6-luna"),
});

const result = await correctness.evaluate({
  input: "What is the capital of Australia?",
  output: "Sydney.",
});
// result.label === "incorrect"
// result.metadata?.probabilities -> { correct: 0.02, incorrect: 0.98 }
```

> [!WARNING]
>
> Decision models can't work through a problem step by step. Use them for judgments you can read off the text, such as relevance, faithfulness, and toxicity, and keep a language model for checks that need a calculation.

> **[Classification Evaluators](https://arizeai-433a7140.mintlify.site/docs/phoenix/sdk-api-reference/typescript/packages/phoenix-evals/classification.md)**
>
> Decision models, probabilities, and refusals

# A Setup Guide for Empty Projects

October 21, 2026

**Available in arize-phoenix 20.24.0+**

A project with no traces now shows its setup guide over the Spans, Traces, and Sessions tabs. The guide disappears as soon as the first trace arrives, without a page reload, and the table underneath is already filled in.

* **The tab strip and Config tab stay usable**, so you can configure the project before sending traces.
* **Skeleton loaders** now stand in for tables, charts, and panels while they load across the app, and their wave animation stops when your system asks for reduced motion.

> **[Tracing](https://arizeai-433a7140.mintlify.site/docs/phoenix/tracing/llm-traces.md)**
>
> Send your first traces to Phoenix

# Graded Annotation Score Colors

October 21, 2026

**Available in arize-phoenix 20.24.0+**

Score colors now get stronger the further a score sits from the annotation config's threshold or midpoint, in the direction the config optimizes. Before, every score was just green or red.

* **Applies everywhere scores are colored**: annotation summaries and badges, span previews, metrics tooltips, evaluator previews, and the experiment compare views.
* **A score exactly at the threshold is neutral** instead of unfavorable.
* **Bounded freeform scores without a threshold** now get colors in the experiment example details view too, as they already did in the compare table.

> **[Annotations](https://arizeai-433a7140.mintlify.site/docs/phoenix/tracing/concepts-tracing/annotations-concepts.md)**
>
> Annotation configs and optimization direction

# Server Telemetry for Self-Hosted Deployments

October 21 to 23, 2026

**Available in arize-phoenix 20.24.0+ (frontend errors) and 20.26.0+ (service.version)**

Self-hosted operators get more signal from the Phoenix server's own telemetry.

* **Browser errors reach the server log.** Uncaught errors and unhandled promise rejections in the Phoenix UI are logged at `ERROR` level on the `phoenix.client_errors` logger. Identical errors are reported once per page session, and reports are rate limited per client IP.
* **`service.version` on server telemetry.** The server's OpenTelemetry resource now carries the Phoenix version next to `service.name`, so you can split server traces and logs by release.

> **[Configuration](https://arizeai-433a7140.mintlify.site/docs/phoenix/self-hosting/configuration.md)**
>
> Server instrumentation environment variables

# Additional Improvements

October 21 to 27, 2026

**Available in arize-phoenix 20.24.0–20.26.1, arize-phoenix-evals 3.9.4+ (Python), @arizeai/phoenix-client 7.16.3+ and @arizeai/phoenix-evals 2.7.1+ (TypeScript)**

* **MiniMax-M3.1-Flash** replaces MiniMax-M2.7 in the playground's MiniMax model list, and you can still type any other MiniMax model name.
* **The prompt Versions tab loads every version** as you scroll, instead of stopping at the first 50, and its counter shows the total.
* **The icon-only Add to Dataset action** in the span details header now has a tooltip and an accessible name.
* **Cancelled code evaluations delete their sandbox**, so a sandbox no longer keeps running on the provider account until it expires.
* **Python evals honor overrides**: `template_format` now applies when copying a `PromptTemplate`, and sync and async client kwargs now override the shared LLM kwargs.
* **TypeScript `toSDK` keeps `description`, `default`, and other keywords** next to `anyOf` in prompt tool and response-format schemas.
* **`createEvaluator` keeps the `metadata`** returned by custom TypeScript evaluator functions.