docs: move Evals and Version history into a new Operations section

[jordanc-relevanceai]

What

Adds an Operations section to the Build with Relevance AI tab, sitting between Invent and Agents.

Evals and Version history both work across more than one builder object, but each was filed under Agents alone. They now live in one place.

Build with Relevance AI
├── Overview
├── Invent
├── Operations                      ← new
│   ├── Evaluate Performance (Evals)    dropdown, 5 pages
│   └── Version history                 single page
├── Agents
├── Tools
├── Workforce
└── Knowledge

Why these two:

• Evals — evals/introduction says it plainly: "It works on both Agents and Workforces."
• Version history — "a complete record of every saved state for your Agents, Tools, and Workforces."

The Evals dropdown is preserved. Version history is one document, so it stays one document.

Moves

Old	New
build/agents/evals/* (5 pages)	build/operations/evals/*
build/agents/build-your-agent/version-history	build/operations/version-history

Also in here

• 6 new redirects entries, one per moved URL
• The three existing /evals, /build/agents/evals, and /build/agents/build-your-agent/evals shortcuts repointed at the new introduction
• Inbound links updated across 34 files (guides, changelog, pricing, quick start, RBAC, Invent, and the moved pages themselves)

mintlify broken-links passes. No page files deleted, no content rewritten.

Scope

Agents loses exactly two lines — the version-history page and the Evaluating Performance group. Tools, Workforce, and Knowledge are byte-identical to main.

Two things for a follow-up, deliberately not fixed here

1. A factual contradiction that predates this PR. get-started/pricing.mdx lists Agent Evaluations as Enterprise-only, while evals/introduction.mdx says "Evals is available on every plan." One of them is wrong. Needs a product answer, not an edit.
2. Sentence-case headings in evals/test-sets.mdx — ## How to Create a Test, ## Managing Tests, ## Example Tests. Pre-existing, and out of scope for a nav restructure.

Supersedes #804, which used Platform as the section name and also moved Tasks and Sharing.

🤖 Generated with Claude Code

[mintlify]

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

Project	Status	Preview	Updated
relevanceai	🟢 Ready	View Preview	Sep 8, 2026, 4:02 AM

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

[github-actions]

🎯 Vibe check

Reviewed: 40 files (6 with issues, 34 clean)

Scores

	Dimension	Score	What's holding it back
🟢	Consistency	8/10	One heading cap error (test-sets.mdx), one banned-word use (kb-generation.mdx), and a batch of pre-existing changelog entries with multiple violations. All new guide and evals content is clean.
🟡	Technical clarity	7/10	Hard contradiction between evals/introduction.mdx ("Evals is available on every plan") and pricing.mdx (Agent Evaluations marked Enterprise-only). An #alerts anchor in invent.mdx targets an Accordion title, not a heading — may be broken.
🟢	Non-technical clarity	9/10	The guide pages are excellent — the path-picker tabs, four-levels cards, and "Where to take it" trios give non-technical readers a clear ramp from cloning a template to full automation. No issues.
🟢	Structure	8/10	The FAQ section in invent.mdx repeats three enterprise-security answers that already appear verbatim in the in-page accordion group above. Otherwise the Evals docs sequence and the entire guides section are well-structured.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: The new guides section and the Evals documentation are both high quality — consistent structure, correct component choices, and clean prose throughout. One issue needs to ship clean: the plan-availability contradiction between the Evals introduction and the pricing page will actively mislead users deciding which plan to buy.

🔧 Issues (4)

• build/operations/evals/test-sets.mdx:13 — ## How to Create a Test → sentence case: ## How to create a Test ("Create" is not a proper noun here)

• guides/customer-support/kb-generation.mdx:7 — "the KB is the most leveraged thing in the operation" — "leveraged" is the adjective form of the banned verb "leverage". Rephrase: "the highest-impact thing in the operation" or "the thing with the most reach".

• build/invent/invent.mdx:194,195 — #alerts anchor links (/build/invent/invent#alerts) target <Accordion title="Alerts">, which is not a proper heading. Mintlify typically doesn't generate HTML anchors for Accordion titles, so these links may resolve to the top of the page. Either add a real ## Alerts heading before the accordion, or remove the anchor and just link to /build/invent/invent.

• changelog.mdx — Pre-existing entries from May 1 and earlier contain multiple content-standard violations that have accumulated. Likely out of scope for this PR to fix all of them, but they should be cleaned up soon:

  • Banned words: "powerful" (lines 111, 237, 410, 454, 482, 514, 572, 604), "seamlessly" (lines 290, 408), "effortlessly" (line 822)
  • Emoji arrows (➡️) used as bullet points throughout entries from May 1 back through January 4
  • Exclamation marks closing paragraphs: "Start syncing your Confluence documentation today!" "watch the magic happen" (line 251), "Your AI workforce just got a whole lot more collaborative!", etc.

  The newer entries (September 3, July 22, June 29, June 12, May 29, May 21, May 13) are all clean — this degradation is confined to the older content.

🧩 Component suggestions (1)

• build/invent/invent.mdx:327–358 — Three FAQ entries ("How does project isolation work for Enterprise customers?", "Which integrations can Invent access in an Enterprise project?", "Where can Enterprise Invent data be sent?") are verbatim duplicates of the three Enterprise security accordions at lines 90–102. The in-page accordions are the better presentation since they're grouped under the ### Enterprise security controls heading with context. Remove the three duplicate FAQ entries and keep just the structured accordion group. CLAUDE.md says not to duplicate content — link to it instead.

⚠️ Contradictions (1)

• build/operations/evals/introduction.mdx:98 vs get-started/pricing.mdx:164 — The Evals introduction FAQ says "Evals is available on every plan." The pricing comparison table marks Agent Evaluations as ✗ for Free, Pro, and Team — Enterprise-only (✓). The original changelog entry (February 9, 2026, line 588) also describes Evals as "Available exclusively for select Enterprise customers." These three sources need to agree. If Evals has since expanded to all plans, update the pricing table; if it's still Enterprise-only, correct the introduction. Either way, the pricing page and the Evals introduction are currently giving users opposite answers to the question "do I need Enterprise to use this?"

✅ Clean files (34)

build/agents/create-an-agent.mdx, build/tools/create-a-tool.mdx, build/workforces/create-a-workforce.mdx, build/operations/evals/checks.mdx, build/operations/evals/running-evaluations.mdx, build/operations/evals/monitor.mdx, build/operations/version-history.mdx, get-started/pricing.mdx (aside from the Evals contradiction noted above), guides/customer-success/account-health.mdx, guides/customer-success/getting-started.mdx, guides/customer-success/qbr-prep.mdx, guides/customer-success/renewal-expansion.mdx, guides/customer-support/getting-started.mdx, guides/customer-support/response-drafting.mdx, guides/customer-support/ticket-triage.mdx, guides/marketing/campaign-analytics.mdx, guides/marketing/content-repurposing.mdx, guides/marketing/getting-started.mdx, guides/marketing/lifecycle-campaigns.mdx, guides/revops/data-dedup.mdx, guides/revops/getting-started.mdx, guides/revops/lead-routing.mdx, guides/revops/pipeline-hygiene.mdx, guides/sales/competitive-intelligence.mdx, guides/sales/crm-data-enrichment.mdx, guides/sales/getting-started.mdx, guides/sales/lead-scoring.mdx, guides/sales/meeting-briefs.mdx, guides/sales/personalized-outbound.mdx, guides/sales/prospect-research-use-case.mdx, guides/solution-engineering/demo-prep.mdx, guides/solution-engineering/discovery-summaries.mdx, guides/solution-engineering/getting-started.mdx, guides/solution-engineering/rfp-responses.mdx

🔋 Credit usage

Item	Count
Files reviewed	40
Context pages read	0 (cross-references were all within the changed file set)
Total lines processed	~5,500

Files read: build/agents/create-an-agent.mdx (92 lines), build/invent/invent.mdx (360 lines), build/operations/evals/checks.mdx (123 lines), build/operations/evals/introduction.mdx (105 lines), build/operations/evals/monitor.mdx (55 lines), build/operations/evals/running-evaluations.mdx (110 lines), build/operations/evals/test-sets.mdx (132 lines), build/operations/version-history.mdx (89 lines), build/tools/create-a-tool.mdx (103 lines), build/workforces/create-a-workforce.mdx (46 lines), changelog.mdx (855 lines), get-started/pricing.mdx (271 lines), all 28 guide files (~96 lines each, ~2,688 lines total), docs.json (navigation scan, ~100 lines relevant)

[NiamhRelevance]

One small thing on the structure: is "Operations" a term we're using elsewhere? I haven't heard or seen it before, so just checking whether it's an agreed label before it lands here.

Also wondering whether evals and version history sit naturally together. Evals are more about testing and monitoring quality, and version history is about rolling back a change, so someone looking for either might not think to check Operations first.