Add Alarms and Auto-diagnosis by Inventor documentation to evals page

[claude]

Summary

• Adds a new Alarms section to the evals page (build/agents/build-your-agent/evals.mdx), positioned between the Monitor and Publish sections
• Documents the alarm creation flow (threshold, notification channels, cooldown logic) with a settings table
• Documents the Auto-diagnosis by Inventor opt-in feature: what it does, how to enable it, firing behavior, and its Enterprise + feature-flag gating

Details

The Alarms section includes two H3 subsections:

1. Creating an alarm — numbered steps, settings table, firing/cooldown behavior
2. Auto-diagnosis by Inventor — opt-in toggle, prerequisites (alert channels, Inventor Enterprise plan), feature-flag notice via <Info> callout

Gating noted: Inventor is Enterprise-only; auto-diagnosis is behind alarm-auto-diagnose-ui and alarm-auto-diagnosis-trigger-be feature flags (both default-off).

References

• Linear: https://linear.app/relevance/issue/TSP-1359/
• Engineering PR: RelevanceAI/relevance-ai#16459

[linear]

TSP-1359

[mintlify]

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

Project	Status	Preview	Updated (UTC)
relevanceai	🟢 Ready	View Preview	Jul 21, 2026, 7:09 AM

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

[github-actions]

🎯 Vibe check

Reviewed: 1 file (1 with issues, 0 clean)

Scores

	Dimension	Score	What's holding it back
🟡	Consistency	8/10	Lowercase "agent's" / "agent" in the Cost and billing section (lines 249, 380, 382, 389) while "Agent" is capitalized everywhere else in the document.
🟢	Technical clarity	9/10	Notification channels table in the Alarms section lists "email, Slack, or both" but omits Microsoft Teams, which the linked Alerts page documents as a supported channel.
🟢	Non-technical clarity	9/10	No issues found.
🟡	Structure	8/10	Best practices section uses <CardGroup> for items that read naturally as flowing bullets (CLAUDE.md explicitly flags this pattern). No closing CTA for a comprehensive concept+how-to page.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: This is well-written, comprehensive feature documentation with strong use of tables, accordions, and embedded demos. Coverage of edge cases (truncation behavior, cooldown, per-node Tool simulations for Workforces) is genuinely useful. The two main issues to address before merging are the notification channel omission in the Alarms table and the capitalization inconsistency in the Cost section.

🔧 Issues (2)

• build/agents/build-your-agent/evals.mdx:249 — "the agent's tool runs" → "the Agent's tool runs". The document uses "Agent" (capitalized) throughout when referring to the Relevance AI product feature, but the Scenario Results table description drops to lowercase.
• build/agents/build-your-agent/evals.mdx:380,382,389 — Same capitalization inconsistency in the Cost and billing section: "the agent's own run", "the agent's own execution", "The agent's own run during the test", "the actions from each tool the agent runs", "when the agent ran" — all should be "Agent's" / "Agent" / "Tool".

🧩 Component suggestions (2)

• build/agents/build-your-agent/evals.mdx:396–415 — <CardGroup cols={2}> for best practices. CLAUDE.md explicitly lists "Best practices that read naturally as flowing bullets" as a case where <CardGroup> is not appropriate, and these six items (Start simple, Be specific with Checks, etc.) do read naturally as a bulleted list. Convert to a <ul> or <Steps>-free bullet list.
• build/agents/build-your-agent/evals.mdx:341–343 — The <Info> callout packs three unrelated concerns into one: (1) the alert-channel prerequisite, (2) Inventor's Enterprise-only status, and (3) a feature-flag caveat. CLAUDE.md calls for "a single short paragraph." Consider moving the prerequisite sentence to a <Note> immediately above the "To enable auto-diagnosis" steps, and converting the Enterprise + feature-flag sentence into a separate short <Info> or inline parenthetical so each callout has a single focus.

🏗️ Page structure (1)

• build/agents/build-your-agent/evals.mdx — No closing CTA. This page functions as a concept + how-to hybrid for a major feature, and CLAUDE.md says concept pages should end with a CTA pointing readers somewhere useful. Natural next-step pages are /build/agents/build-your-agent/alerts (setting up the notification channels that Alarms rely on) and /build/invent/invent (Inventor context for auto-diagnosis). A short ## What's next? section with two <Card> links would round off the page.

⚠️ Contradictions (1)

• build/agents/build-your-agent/evals.mdx:323 says Alarm notification channels are "email, Slack, or both." But build/agents/build-your-agent/alerts.mdx documents three channels: Email, Slack, and Microsoft Teams. The Alarms table in evals.mdx needs to include Teams as a third option to match what the linked Alerts page describes.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	2
Total lines processed	~628

Files read: build/agents/build-your-agent/evals.mdx (457 lines), build/agents/build-your-agent/alerts.mdx (111 lines), build/invent/invent.mdx (60 lines, partial)

[mintlify]

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

Project	Status	Preview	Updated (UTC)
relevanceai	🟡 Building	–	Jul 21, 2026, 7:08 AM

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