Skip to main content
Pulse
Synthetic healthcare dataRecorded replay by default

Investigate integration failures before clinicians discover them.

Pulse turns failed jobs, health signals, and incident history into a cited investigation workspace. Every operational action still needs a person to approve it.

Preview the demo path

Mercy General EHR sync

CRITICAL

First signal

Repeated upstream 503 responses preceded a DEAD sync job.

8evidence
2hypotheses
1approval
Action target revalidated and audited

Preview the proof

See the whole recovery loop before signing in.

This guided outline follows one synthetic EHR outage from the first health signal through cited findings, target revalidation, worker execution, and the final audit row.

  1. 00:00 · Review the degraded connector
  2. 00:25 · Run the bounded investigation
  3. 00:55 · Approve and inspect the audit

The captioned release video is not configured in this environment. Launch the interactive demo or follow the timestamped preview beside this card.

The guided path

A complete operational story in 90 seconds

No credentials, provider key, or shared demo account required.

  1. 1

    Enter an isolated tenant

    Pulse provisions synthetic OPS data with a one-hour TTL.

  2. 2

    Open the EHR incident

    A bounded outage window includes logs, jobs, events, health, and timeline evidence.

  3. 3

    Investigate the first signal

    A recorded engine returns the same validated report contract used by live AI.

  4. 4

    Open a citation

    Each finding links back to the evidence record that supports it.

  5. 5

    Review the retry

    The suggested action stays pending while the operator checks its target and expected result.

  6. 6

    Approve the action

    The server rechecks the target before it queues the retry.

  7. 7

    Inspect the audit

    The approval and operational job retry are both durable and attributable.

Engineering proof

Claims backed by executable gates

View the CI workflow →

314

automated tests

236 unit · 66 integration · 12 browser · 19 offline eval fixtures

93%

required-fact coverage

13 of 14 cases on the latest summary eval corpus

100%

safety fixture pass rate

Schema, leak, and injection regression fixtures

Health rules: 100% branch coverageRedaction boundary: 100% branch coverageAI budget accounting: 100% line and function coverageDurable job dispatch: 100% line and function coverageBuild 3a345bfApp v3.0.0Proof generated 2026-08-24

Evidence-bounded AI

Redacted excerpts, structured schemas, cited claims, uncertainty, deterministic fallback, and cost caps.

Operational reliability

Retries, attempt history, rolling health windows, durable incidents, replay protection, and cleanup.

Safe by default

Deterministic replay keeps the public path reliable; credentialed live AI uses the same contract with redaction, budgets, telemetry, and eval gates.

Ready to investigate?

The workspace resets safely and expires after one hour.