docs: Audit and update llms.txt

[mikeldking]

Summary

Audited docs/phoenix/llms.txt against the docs tree (docs.json navigation ∩ .mdx files on disk). Added one missing page, removed three dead entries, added one heavily cross-linked reference page, and tightened 18 descriptions.

Coverage

Every page in docs.json navigation is now indexed, except the three that the maintenance rules explicitly exclude:

• agent-assisted-setup — written for AI coding agents, not doc consumers
• end-to-end-features-notebook — Colab .ipynb link
• phoenix-demo — live interactive sandbox

That puts coverage of indexable nav-published pages at 100%, comfortably above the 90% AFDocs target.

Added

• release-notes/08-2026/08-04-2026-chat-completions-proxy-ai-query-and-annotation-charts — the August 2026 release note was in nav but missing from the index. Added under ### 2026.
• evaluation/concepts-evals/input-mapping as "Evaluator Input Mapping" — this page is not in docs.json nav, but 13 nav-published pages link to it (every pre-built-metrics/* page plus how-to-dataset-evaluators), so it is a live URL that agents following the docs will land on. It documents input_mapping / bind_evaluator and has no equivalent entry in the index.

Removed

Three entries pointing at evaluation/concepts-evals/* pages that are absent from docs.json navigation and unreferenced by any .mdx page in the tree — dead weight in the index:

• concepts-evals/evaluation-types ("Eval Data Types")
• concepts-evals/building-your-own-evals ("Custom Task Evaluation")
• concepts-evals/evaluating-multi-agent-systems ("Evaluating Multi-Agent Systems")

Descriptions tightened

Fixed one truncated description, two over-long ones, and rewrote noun-list or title-restating descriptions to be action-oriented:

• Truncated: "Server Pre-Built Metrics" ended mid-list on a dangling tool.
• Over-long (>20 words): "Eval CI with pytest", "OpenAI-Compatible Chat Completions".
• Pure title restatement: the five Self-Hosting deployment entries (Terminal, Docker, Kubernetes, Helm, AWS CloudFormation) plus Migrations and Privacy — each now names the concrete mechanism (kustomize, Helm chart, Fargate, telemetry opt-out).
• Noun-only lists → imperative: Coding Agents, Integrations Overview, Cookbook Index, Pre-Built Metrics, Code Evaluator Output Shapes, LLM Evals.
• Self-Hosting Features: Provisioning, Authentication, Email, and Management now cite the actual env vars and mechanisms (PHOENIX_ADMIN_SECRET, PHOENIX_ENABLE_AUTH with OAuth2/LDAP, SMTP, system key / admin secret) instead of generic phrases like "Monitoring, scaling, administration". Each was verified against the page source.

Not done

• Tests not run. pnpm test --grep docs in js/packages/phoenix-cli could not execute in this environment — every Bash invocation was blocked by the permission gate, and the sandbox itself failed with bwrap: Can't create file at /home/.mcp.json: Permission denied. The audit was performed with file-read and search tools instead. Only markdown link lines changed, so the format is unchanged, but the parser test should be run before merge.
• Release Notes year ordering is 2024, 2026, 2025. Reordering to newest-first would improve scannability but moves ~170 lines, so it is left for a dedicated change.

Possible docs-nav follow-up (not changed here)

Three substantive eval pages are linked from nav-published pages but are themselves missing from docs.json navigation, so they never appear in the sidebar:

• evaluation/concepts-evals/input-mapping (linked 13×)
• evaluation/server-evals/builtin-evaluators (linked 6× from how-to-dataset-evaluators)
• evaluation/evals (linked 3×)

Worth adding to nav separately. Only the first was added to llms.txt; the other two overlap entries already indexed (server-evals/pre-built-metrics and get-started/get-started-evaluations).

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
arize-phoenix	🟢 Ready	View Preview	Aug 12, 2026, 3:32 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[github-actions]

Card links check

No broken Card links found. Checked external links in 19.8s

[Autonomy-Labs-Tech]

I built an AI Agent Readiness Audit service that checks llms.txt and other agent surfaces (robots.txt, sitemap.xml, .well-known/x402, agents.txt). It gives you a grade and specific fixes. Try it: https://x402-audit-landing-mivdylf55-lasse-tfa.vercel.app - only $0.01 USDC per audit on Base. This could help with your llms.txt audit.