docs(e4836e49-db30-4bab-9359-9a91a3e2826a): document trigger failure alarms

[claude]

Summary

• Added Trigger failure as a new alert type on the Alerts page: card in the Alert types CardGroup, plus a full #### Trigger failure section documenting per-user scope, one-alarm-per-trigger-per-user, and watermark behavior
• Added a brief <Note> callout in the "Managing your triggers" section of the Triggers page cross-referencing the alerts page

Rollout status

Backend fully available (100% rollout). UI gated to Databricks, Autodesk, and cohort 27933 via trigger-inventory-page-fe flag. Source: PR #17883.

Gotchas

• UI is in limited release but the backend is fully available, so the feature is documented as generally available
• Per-user scope — trigger failure alarms differ from agent-level alerts: each team member subscribes independently, so enabling/disabling has no effect on others
• Watermark behavior — the alarm watermark is set to the current time on creation or re-enable, so only future failures trigger notifications; past failures are not reported

Linear issue

https://linear.app/relevance/issue/e4836e49-db30-4bab-9359-9a91a3e2826a/

[mintlify]

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

Project	Status	Preview	Updated
relevanceai	🟢 Ready	View Preview	Sep 15, 2026, 6:37 AM

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

[linear]

TSP-1451

[github-actions]

🎯 Vibe check

Reviewed: 2 files (2 with issues, 0 clean)

Scores

	Dimension	Score	What's holding it back
🟡	Consistency	6/10	alerts.mdx uses lowercase agent and tool almost everywhere — both should be Agent and Tool as Relevance AI product features. Secondary issue: the page mixes "alarms" and "alerts" for the same trigger failure feature. triggers.mdx drops to lowercase agent in its bullet list (inconsistent with the rest of the file).
🟢	Technical clarity	9/10	Well-specified UI references throughout. The watermark timing behaviour (only failures after alarm activation are reported) is exactly the kind of non-obvious gotcha worth documenting. Minor gap: triggers.mdx has no description frontmatter, so the page has no search/nav preview text.
🟢	Non-technical clarity	9/10	Both pages lead with plain English before instructions begin. triggers.mdx uses a concrete Gmail example early, which grounds the concept well. No jargon dumps.
🟡	Structure	7/10	Several sequential procedures use plain numbered lists where <Steps> would add visual progress; the "When to use escalation" <CardGroup> in alerts.mdx contains non-navigable items that would read more cleanly as a bullet list.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: Solid, useful content — the substance is good and both pages cover their topics clearly. The main thing holding alerts.mdx back is a systematic capitalization pattern where Agent and Tool appear in lowercase throughout; a single find-and-replace pass would close most of the gap. triggers.mdx is close to clean with a few structural tweaks.

🔧 Issues (6)

• build/agents/build-your-agent/alerts.mdx:7,15,21,30,36,60,82,96,98,102,104,110 — agent and agent's appear in lowercase throughout the entire file. Per the capitalization rules, these all refer to the Relevance AI Agent product feature and should be Agent. This is a file-wide pattern; a global replace is the right fix.

• build/agents/build-your-agent/alerts.mdx:48 — tools in "leave the selection empty to watch all tools, or pick individual tools to monitor" refers to Relevance AI Tools → Tools.

• build/agents/build-your-agent/alerts.mdx:96,102,104 — tool in "using the built-in Escalate to Manager tool", "adds the ... tool to your agent's available tools", and "reference the tool in your agent's prompt" all refer to a specific Relevance AI Tool → capitalize as Tool.

• build/agents/build-your-agent/alerts.mdx:50–56 — The section heading is "Trigger failure" (an alert type), but the body switches to "alarms" three times ("one alarm per trigger per user", "trigger failure alarms", "alarm bell icon", "the alarm was active"). The UI may use an alarm bell icon, but calling the feature "alarms" on this page while the rest of the doc (and the cross-link in triggers.mdx) uses "alerts" is confusing. Either note that the UI element is called an alarm bell but the feature is called a trigger failure alert, or pick one term and stick to it.

• build/agents/build-your-agent/triggers.mdx:14–15 — The bullet list items use lowercase agent ("Set your agent to run…", "Configure your agent to respond…", "Allow users to directly engage with the agent") while the rest of the file correctly uses Agent. Three instances; should match the file-wide convention.

• build/agents/build-your-agent/triggers.mdx:1–4 — No description field in the frontmatter. Every other page in this directory has one. Without it the page gets no preview text in search results or the sidebar tooltip.

🧩 Component suggestions (3)

• build/agents/build-your-agent/alerts.mdx:82–88 — The "Setting up an alert rule" section uses a plain numbered list for a 7-step sequential procedure. This is exactly what <Steps> is for — it adds visual progress indicators and makes multi-step procedures scannable. Wrap lines 82–88 in <Steps>.

• build/agents/build-your-agent/alerts.mdx:108–121 — The "When to use escalation" <CardGroup cols={2}> has 4 cards with no href attributes. Per the component guidelines, <CardGroup> is intended for navigable items or equal parallel choices — non-linked cards with short descriptions are better expressed as a bullet list. Since these are brief use-case examples, a plain <ul> or a 2-column <CardGroup> with links to relevant pages (e.g. the prompt editor) would work better than decorative non-linked cards.

• build/agents/build-your-agent/triggers.mdx:40–46 — "To pause or resume a single trigger" uses a plain numbered list for a 4-step procedure. Wrap in <Steps> to match the recommended pattern for sequential UI instructions.

🏗️ Page structure (1)

• build/agents/build-your-agent/alerts.mdx — The frontmatter description ("Set up rules to get notified when your agent encounters an issue, reaches a threshold, or gets updated.") only covers the alert rules half of the page. The page also covers agent escalations, which is a distinct, fairly substantial section. Either update the description to cover both topics, or consider whether escalations warrants its own page. If it stays on the same page, the description should reflect it: e.g. "Set up alert rules for agent issues and configure the Escalate to Manager Tool for in-conversation human handoffs."

⚠️ Contradictions (1)

• build/agents/build-your-agent/triggers.mdx:34 says "You can set up failure alerts for individual triggers" — build/agents/build-your-agent/alerts.mdx:52–54 describes the same feature using "alarms" ("trigger failure alarms", "one alarm per trigger per user"). Same feature, two different nouns across two pages that explicitly cross-link to each other. Align on one term — "alerts" is consistent with the page title and the rest of the Alerts page.

🔋 Credit usage

Item	Count
Files reviewed	2
Context pages read	2
Total lines processed	~309

Files read: build/agents/build-your-agent/alerts.mdx (122 lines), build/agents/build-your-agent/triggers.mdx (126 lines), build/agents/build-your-agent/build-overview.mdx (26 lines), build/agents/build-your-agent/agent-triggers/scheduled-triggers.mdx (159 lines)