docs(TSP-1444): document alarm notification channel failure visibility

[claude]

Summary

• Adds a Notification channel health section to the Alerts page documenting new failure-visibility features from PR #17781
• Documents the yellow Channel issue badge that appears on alert rule cards when a notification channel has a permanent delivery failure
• Documents the warning banner shown inside the alert modal's channel configuration step when editing a broken-channel alert
• Documents automatic fallback-to-email delivery when all configured channels fail
• Adds a <Note> to the Notification channels section explaining the OAuth account removal warning dialog (ships to all users)

Linear issue

https://linear.app/relevance/issue/TSP-1444/

Test plan

• Verify the new "Notification channel health" section renders correctly on the alerts page
• Confirm all headings are sentence case
• Confirm <Note> callout is a single paragraph with no bullets
• Verify internal links to Slack and Teams integration pages still work

🤖 Generated with Claude Code

[mintlify]

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

Project	Status	Preview	Updated
relevanceai	🟢 Ready	View Preview	Sep 15, 2026, 3:03 AM

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

[linear]

TSP-1444

[github-actions]

🎯 Vibe check

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

Scores

	Dimension	Score	What's holding it back
🟢	Consistency	9/10	alerts.mdx:91 — "agent and eval views" uses lowercase "agent" when referring to the Agent product UI.
🟡	Technical clarity	7/10	alerts.mdx:113 — "/" command reference is unexplained. alerts.mdx:63–67 — Teams setup omits the required app-installation step that the Teams integration docs say is mandatory.
🟡	Non-technical clarity	8/10	alerts.mdx:113 — "using the / command" reads as jargon with no explanation for someone unfamiliar with the prompt editor.
🟡	Structure	7/10	alerts.mdx:75–83 — sequential setup procedure uses a plain numbered list instead of <Steps>. alerts.mdx:69–71 — <Note> wraps a destructive-action warning; should be <Warning>.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: Solid, well-organized page — good component use, consistent sentence case throughout, no banned words or British spellings. The main gaps are a vague "/" command reference that will confuse anyone new to the prompt editor, and a <Note> that should be a <Warning> since it describes how a user action (removing an OAuth account) can silently break alert delivery.

🔧 Issues (4)

• alerts.mdx:91 — "The badge appears in both the agent and eval views" — "agent" should be "Agent" (referring to the Agent product UI feature). Also consider whether "eval" warrants capitalization if Evals is a named product feature.
• alerts.mdx:99 — "emailing the alert rule's original creator" — unmentioned edge case: what happens if the creator's account is deactivated or they've left the organization? Worth a brief gotcha note.
• alerts.mdx:113 — "using the / command to give the agent guidance on when to escalate" — "/" is unexplained. Readers who haven't used the prompt editor won't know what this means. Describe it: "type / in the prompt editor to insert a reference to the tool."
• alerts.mdx:63–67 — The Microsoft Teams card says "Connect your Microsoft Teams account via OAuth, then select a team and channel." The Microsoft Teams integration page states that setup requires two steps — OAuth connection AND installing the Relevance AI app in Teams — and that "completing only one is not enough." If the app installation is also required for alert delivery to work, this description is incomplete and potentially misleading.

🧩 Component suggestions (3)

• alerts.mdx:75–83 — The "Setting up an alert rule" section is a sequential 7-step procedure. Plain numbered lists don't signal procedural flow the way <Steps> does. Wrap in <Steps> for visual progress indicators.
• alerts.mdx:69–71 — <Note> wraps a warning about a destructive action: removing an OAuth account can silently break alert delivery. This belongs in a <Warning>, not a <Note>. The callout copy is already a single short paragraph, so no rewording is needed — just change the tag.
• alerts.mdx:117–130 — The "When to use escalation" <CardGroup> has four non-navigable cards (no href) with single-line descriptions. These are thin enough that a bullet list under a ### When to escalate heading would be cleaner and easier to scan. Cards without links work best when each item has enough visual weight to justify the container — four 8-word lines don't.

🏗️ Page structure (1)

• alerts.mdx:7 — The intro paragraph ("Set up rules to get notified when your agent encounters an issue, reaches a threshold, or gets updated.") is a word-for-word copy of the frontmatter description. The description serves as the meta/subtitle; the opening paragraph should expand on it — e.g., briefly explain the two distinct features on this page (alert rules and agent escalations) so readers know what they're about to read.

⚠️ Contradictions (1)

• alerts.mdx:63 says Teams setup is "Connect your Microsoft Teams account via OAuth, then select a team and channel." microsoft-teams.mdx:32–41 says setup requires two separate steps — OAuth connection AND installing the Relevance AI app in Teams — and explicitly states "completing only one is not enough." If both steps are required for alert notifications to reach Teams channels, the alerts page undersells the setup requirements and will leave users confused when alerts don't arrive.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	3
Total lines processed	~373

Files read: build/agents/build-your-agent/alerts.mdx (131 lines), build/agents/build-your-agent/triggers.mdx (122 lines), integrations/popular-integrations/slack.mdx (60 lines), integrations/popular-integrations/microsoft-teams.mdx (60 lines)