feat(mcp): improve metric discovery, add quiet-saturation eval scenario

[karl-power]

Why

Eval transcripts showed agents solving investigations without ever touching metrics — not because metrics weren't useful, but because they were the only signal behind a discovery wall. After list_sources, logs and traces are queryable immediately (key columns are in the catalog), while a metric source showed only opaque table names: querying it cost 1–3 extra calls (describe_source/list_metrics) just to learn what exists. Under a turn budget, agents rationally skipped it — even when explicitly nudged (we tested this: prompt- and output-level hints naming the exact metrics were ignored in 7/7 deliveries when metrics weren't on the efficient path).

What changed

• clickstack_list_sources: metric sources now include metricNamesPreview — up to 10 recently-reported metric names per kind, sampled from the team's own tables — plus a usage note that metrics are queried directly via clickstack_table/clickstack_timeseries with metricType + metricName, no describe hop needed. Best-effort under a 3s wall-clock budget with a concurrency pool and per-table dedup; omitted silently on timeout.
• metricNames.ts (new): the metric-name sampler extracted from describeSource.ts into a shared module, with a widening lookback (24h → 30d, first non-empty window wins) so sparse or batch-emitted metrics still produce a sample. describe_source uses the same sampler, so its per-kind sample gets the lookback too.

Measured wins

Benchmarked with the hdx-eval framework on quiet-saturation (connection-pool leak where metric history is the efficient diagnostic path), branch vs main, identical seed/anchor/prompts, claude-fable-5 + claude-opus-4-6, 3 runs/cell, two independent batches.

First tool call whose args name a target metric (the load-bearing pool gauges), every run, both batches:

arm	batch 1	batch 2	range
branch / fable-5	6, 5, 4	5, 5, 6	4–6
branch / opus-4-6	6, 7, 6	6, 7, 7	6–7
main / fable-5	8, 10, 9	8, 8, 8	8–10
main / opus-4-6	16, 14, 13	14, 13, 21	13–21

Zero overlap across all 24 runs (branch ≤7, main ≥8; p ≈ 0.001 by permutation): the preview removes the discovery hops entirely (list_sources → query), where main inserts list_metrics/describe_source detours or trace-grinds first.

Headline metrics from the second batch (which ran exactly this minimal build):

	branch/fable	main/fable	branch/opus	main/opus
Combined score	96%	91%	82%	84%
LLM-judge (weighted)	96%	92%	92%	84%
Programmatic checks	96%	96%	97%	93%
Pre-symptom onset attributed (distinguishes_true_onset)	100%	100%	67%	33%
Tool calls (mean)	13.0	13.0	19.7	20.7
Tool errors (mean)	0.0	0.3	2.3	0.7
Metric adoption	100%	100%	100%	100%

• fable-5: +5pp combined here, +9pp in batch 1 (12.3 vs 15.7 calls) — consistently equal-or-fewer calls, zero tool errors, faster wall clock.
• opus-4-6: the freed budget converts to depth rather than speed — judge +8pp and the pre-symptom onset attributed twice as often. Its combined score dips below main only because of a 12pp tool-error penalty from unrelated agent-side query-syntax mistakes (one run retried an identical malformed select 4×).
• Metric adoption is 100% on both arms — the change lowers the cost of the metric path rather than nudging inclination; agents reach the load-bearing signal roughly twice as fast.

[changeset-bot]

🦋 Changeset detected

Latest commit: fdeeac5

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 4 packages

Name	Type
@hyperdx/hdx-eval	Patch
@hyperdx/api	Patch
@hyperdx/app	Patch
@hyperdx/otel-collector	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
hyperdx-oss	Ready	Preview	Aug 27, 2026 1:50am
hyperdx-storybook	Ready	Preview	Aug 27, 2026 1:50am

[github-actions]

🟡 Tier 3 — Standard

Introduces new logic, modifies core functionality, or touches areas with non-trivial risk.

Why this tier:

• Diff size: 687 production lines changed (Tier 2 max: < 250)

Additional context: 8 file(s) in private internal-tooling packages, excluded from the line count

Review process: Full human review — logic, architecture, edge cases.
SLA: First-pass feedback within 1 business day.

Stats

• Production files changed: 3
• Production lines changed: 687 (+ 638 in test files, excluded from tier calculation)
• Branch: karl/improve-metric-mcp-adoption
• Author: karl-power

To override this classification, remove the review/tier-3 label and apply a different review/tier-* label. Manual overrides are preserved on subsequent pushes.

[greptile-apps]

Greptile Summary

The PR makes metric names directly discoverable from MCP source listings and adds widening lookback sampling shared with source description. It also adds the quiet-saturation evaluation scenario and informational adoption-check reporting.

Confidence Score: 5/5

The PR appears safe to merge because no blocking failure remains.

No blocking failure remains.

Important Files Changed

Filename	Overview
packages/api/src/mcp/tools/sources/listSources.ts	Adds bounded, concurrent metric-name preview collection and direct-query guidance to source discovery.
packages/api/src/mcp/tools/sources/metricNames.ts	Extracts shared metric-name sampling, widening lookbacks, and optional unit and description enrichment.
packages/api/src/mcp/tools/sources/describeSource.ts	Reuses the shared widening-lookback sampler for metric source descriptions.
packages/hdx-eval/src/scenarios/quiet-saturation/generate.ts	Adds deterministic telemetry for diagnosing gradual connection-pool saturation.
packages/hdx-eval/src/grading/programmatic.ts	Evaluates informational adoption checks while excluding them from the weighted adoption score.
packages/hdx-eval/src/reports/aggregate.ts	Carries informational adoption metadata into aggregate reports.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
A[clickstack_list_sources] --> B[Metric source tables]
B --> C[Sample metric names]
C --> D{Recent names found?}
D -->|No| E[Widen lookback]
E --> C
D -->|Yes| F[metricNamesPreview]
F --> G[Direct table or timeseries query]

Loading

_{Reviews (9): Last reviewed commit: "Merge branch 'main' into karl/improve-me..." | Re-trigger Greptile}

[github-actions]

Deep Review

Scope: MCP metric discovery (clickstack_list_sources metric-name preview, new shared metricNames.ts sampler, describeSource.ts reuse) plus the hdx-evalquiet-saturation scenario and informational adoption checks. Base cf8e7e72.

✅ No critical issues found. The concurrency/abort/timeout design in attachMetricNamePreviews is sound (per-task try/catch swallows individual sampling failures, Promise.race(drained, abortedPromise) bounds wall-clock, clearTimeout in finally, cached promise is always awaited under a guard so no unhandled rejection), the enrichment SQL is parameterised via the chSqlIdentifier/String DSL (no injection surface introduced), and the all-informational division is guarded (programmatic.ts:111). Every one of the 20 prior review comments — cache-key collision, summary-kind leakage, shared-cache truncation, anyLast→argMax, PQueue, break-mode timeout, 1h initial window — is resolved in the current code.

🟡 P2 — recommended

• packages/api/src/mcp/tools/sources/listSources.ts:114 — the new integration tests cover recent-data, wider-lookback fallback, summary exclusion, and empty-tables, but not the wall-clock-budget path: nothing forces the 3s AbortController to fire and asserts the catalog still returns with metricNamesPreview omitted, nor exercises the per-(connection,db,table,timestamp) dedup cache.
  • Fix: add an integration test that makes preview sampling exceed the budget and asserts list_sources still returns successfully with the preview absent, plus one asserting two sources on the same physical table trigger a single sample.

🔵 P3 nitpicks (2)

• packages/hdx-eval/src/grading/programmatic.ts:111 — when every adoption check is informational, totalWeight is 0 and the score falls to 0 rather than a neutral/full value; harmless today because no scenario is all-informational, but the result is surprising if one ever is.
  • Fix: treat an all-informational check set as a full/neutral score and document the intent, or assert at least one scoring check exists during rubric validation.
• packages/api/src/mcp/tools/sources/describeSource.ts:357 — the per-kind metric-name sampling runs DISCOVERABLE_METRIC_KINDS (5) in Promise.all, each doing the widening lookback (up to 3 sequential windows) with no server-side break cap, so a cold/empty metric source can issue many queries under the 10s budget.
  • Fix: consider bounding the describe-path sampling the same way the preview path does, or capping the lookback breadth for metric sources.

---

Reviewers (2): previous-comments, orchestrator diff analysis (correctness, reliability, performance, testing, security/injection lenses).

Testing gaps: timeout-omission path and dedup cache in attachMetricNamePreviews are unverified; all-informational adoption-check edge is untested.

[github-actions]

E2E Test Results

✅ All tests passed • 315 passed • 1 skipped • 1145s

Status	Count
✅ Passed	315
❌ Failed	0
⚠️ Flaky	4
⏭️ Skipped	1

Tests ran across 4 shards in parallel.

View full report →

[pulpdrew]

I see my comments in packages/api/src/mcp/tools/sources/metricNames.ts are on code that was just relocated. It'd be nice to fix them but we can treat them as non-blocking.

[karl-power]

Thanks @pulpdrew all of your points addressed in fc770da

[pulpdrew]

LGTM