docs(TSP-1291): expand Monitor dashboard task status filter documentation

[claude]

Summary

• Renamed "Conversation status filter" to Task status filter in the Monitor dashboard setup steps
• Documented all available agent task statuses with descriptions (completed, escalated, pending approval, execution limit reached, errored)
• Added guidance on when to use status filtering and why (focus evaluation credit on specific outcomes)
• Added a Note callout documenting new workforce dashboard behavior: evaluations trigger on every state transition, and status options are tailored to the resource type

Linear issue

https://linear.app/relevance/issue/TSP-1291/

[linear]

TSP-1291

[mintlify]

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

Project	Status	Preview	Updated (UTC)
relevanceai	🟢 Ready	View Preview	Jun 11, 2026, 3:23 AM

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

[jordanc-relevanceai]

Fact-check fix pushed ✅

Corrected the agent task-status filter list — it listed two statuses that don't exist for agents:

• Execution limit reached is a workforce-only status (WORKFORCE_STATUS_GROUPS), not an agent one — removed.
• Errored isn't a real status label — removed.

Replaced with the actual agent failed-state labels from AGENT_STATUS_GROUPS / STATE_LABELS: Timed out, Unrecoverable error, Exhausted retries.

Also fixed the workforce note: evals do not trigger on every state transition (active/transient states are intentionally omitted — observability evals on them would grade an in-flight trace with no output). Kept the accurate point that workforce dashboards expose different statuses (e.g. Execution limit reached).

Source: apps/builder-app/features/evals/dashboardStatusFilters.ts (relevance-api-node).

[github-actions]

🎯 Vibe check

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

Scores

	Dimension	Score	What's holding it back
🟢	Consistency	8/10	Bold label inside <Info> callout (line 8); bullet list inside <Accordion> (lines 359–362) — both violate CLAUDE.md rules. No banned words, no British spellings, capitalization is clean throughout.
🟢	Technical clarity	9/10	UI element names are precise, edge cases are well-covered (truncation behavior, credit calculation, 10-Check limit). The "Filter tasks" option in Monitor step 5 could name the actual status values rather than just saying "leave blank to evaluate all tasks."
🟢	Non-technical clarity	9/10	Concepts are explained before instructions, the progressive structure from overview → sections → deep-dive works well. No jargon drops.
🟡	Structure	7/10	<CardGroup> used for the Best practices section — CLAUDE.md says this component is not appropriate for best practices that read naturally as flowing bullets.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: This is a well-crafted, comprehensive feature page — specific UI references, good edge-case coverage, and a logical flow from concept through how-to through reference. Three CLAUDE.md compliance items to fix (two component-content mismatches, one wrong component choice) before it's fully clean.

🔧 Issues (2)

• build/agents/build-your-agent/evals.mdx:8 — **Rollout Status**: is a bold label inside a callout. CLAUDE.md explicitly prohibits bold labels inside callouts. Remove the bold label: <Info>Evals is rolling out progressively, starting with Enterprise customers. If you don't see this feature in your account yet, reach out to your account manager to discuss access.</Info>

• build/agents/build-your-agent/evals.mdx:359–362 — Bullet list inside an <Accordion>. CLAUDE.md says accordion content must use flowing sentences, not bullet lists. Rewrite as: "Credits are calculated by adding the Agent task run, the simulator (which uses an LLM to generate the user persona's messages), and every Check that runs on the conversation — each Check, especially LLM Judge, makes its own LLM call."

🧩 Component suggestions (1)

• build/agents/build-your-agent/evals.mdx:323–342 — <CardGroup cols={2}> for the Best practices section. CLAUDE.md says CardGroup is not appropriate for best practices that read naturally as flowing bullets. These six tips are practical guidance, not navigable destinations or parallel feature choices — a plain bulleted list under ## Best practices serves the reader better and matches the pattern used in sibling pages like tools.mdx.

⚠️ Contradictions (0)

No contradictions found between this file and the context pages read.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	2
Total lines processed	~527

Files read: build/agents/build-your-agent/evals.mdx (390 lines), build/agents/build-your-agent/build-overview.mdx (26 lines), build/agents/build-your-agent/tools.mdx (111 lines)