Interface Monitoring Playbook
Your dashboard says green.
Three patients in the ED are still waiting.
Your infrastructure monitoring tool reports HTTP 200 OK. The server is up. The network is fine. Meanwhile, zero ORU messages have arrived in the last 45 minutes and three STAT lab orders are pending. The interface is not down. It is failing at the message level, in a way your dashboard was never built to see.
Infrastructure monitoring does not see message-level failures.
You learned about the last interface failure from a physician. Not from a dashboard. Not from an alert. From a phone call at 7 AM asking why the lab results are not in the chart.
Your current monitoring checks whether the server is running, whether the port is open, whether the service responds to a health check. It does not check whether the HL7 messages flowing through that interface are structurally valid, whether the vendor changed a field format in their last patch, or whether the volume dropped from 200 messages per hour to zero at 3 AM.
The gap between “server is up” and “messages are arriving correctly” is where patient care delays hide.
That gap is invisible to every infrastructure monitoring tool in your stack.
The Loft monitoring progression.
Three tiers of interface intelligence — each one building on the last.
See what is happening.
Point Loft at your interface directories — Mirth pickup folders, SFTP drops, TCP listeners. It validates every message as it arrives, in real time, using the same validation engine that powers Post testing. When a message fails structurally, you know in 60 seconds. When volume drops below threshold, you know before the first clinician notices.
This is not sampling. This is not log aggregation. Every message is parsed and validated against vendor-specific rules. A missing OBR.16 is flagged. A PID.3 format change is caught. A new Z-segment variant is surfaced.
Real-time message-level validation
Every message parsed and validated across every monitored interface as it arrives.
Volume anomaly detection
"Lab interface usually sends 200 msgs/hr, now sending 0" — caught before the ED calls.
Multi-channel alerting
Alert delivery via email, Slack, Teams, or webhook. Route to the right team, not just a mailbox.
Per-interface metrics
Throughput, error rate, and queue depth. SQLite-backed 90-day retention.
Understand why it broke.
Watch tells you something failed. Insight tells you what changed, why it likely happened, and how to fix it. AI-guided remediation analyzes the deviation pattern, references the vendor profile history, and generates a plain-English explanation.
Insight does not fix the problem. It gives your engineer everything they need to fix it in minutes instead of hours.
“Lab interface error rate increased 40% after the vendor patch on Monday — OBR.2 shifted from placer-only to placer+filler format.”
Deviation classification
Field format change, segment reorder, new field, value set change — classified automatically.
Severity assessment
Based on downstream impact, not just structural severity. Critical lab routing failures rank above cosmetic issues.
Trend dashboards
Historical analytics across every interface. Catch drift before it becomes failure.
Natural language summaries
Plain-English explanations of what changed, when, and which vendor patch likely caused it.
Fix it before anyone calls.
Command generates micro-ETL rules — atomic transformations that normalize the deviation while you investigate the root cause. Two approval modes: autonomous (heal first, review after) or 1-click (hold the message, wait for your approval). Temporary rules auto-expire. Vendor communication drafts are pre-written.
The interface keeps flowing. Your team reviews the fix in the morning. The physician never calls.
Autonomous self-healing
Per-interface playbooks with auto-healing and a review queue for morning approval.
Message buffering and replay
1-click approval mode holds messages and replays them once the fix is confirmed.
TTL-managed rules
Temporary rules auto-escalate if unresolved. No silent tech debt accumulating in your pipeline.
SLA-backed 60-second alerts
Vendor communication drafts pre-written. Escalation chains configured per interface.
Loft from the command line.
The same engine that powers Post testing. Pointed at production traffic.
# Point Loft at a Mirth pickup folder
pidgeon loft watch \
--path /mirth/pickup \
--profile epic_er
# Loft watches every message as it arrives
# Alerts fire within 60 seconds of a failure
# Volume anomalies surface before the first call# Check current interface health
pidgeon loft status
# Review recent alert history
pidgeon loft alerts
# 24-hour summary report
pidgeon loft report --since 24h
# Output: per-channel throughput,
# error rates, queue depth, MTTR[LOFT INSIGHT] 2026-02-24 03:14:22
Interface: Lab_ORU_Outbound
Status: DEGRADED — error rate 38% (baseline: 0.4%)
Deviation detected: OBR.2 field format change
Before: placer order number only
After: placer + filler concatenated (OBR.2.1^OBR.2.2)
Probable cause: Lab vendor patch deployed 2026-02-24 02:00
Reference: HL7 v2.5.1 spec § 4.5.1 — OBR-2 Placer Order Number
Downstream impact: CRITICAL — lab results not routing to attending
Suggested fix: Update OBR.2 parser to accept composite format
Vendor communication draft: /loft/drafts/lab-vendor-2026-02-24.txt200+ interfaces. Six hospitals. The CIO nods.
You open your laptop to 200+ interfaces across 6 hospitals. Last month, you found out about a failure from an angry physician. This month, you open a dashboard that shows what happened overnight, why, and that the system already applied a temporary fix.
The CIO pulls up the interface reliability slide. You open Loft. Uptime by interface. MTTR by vendor. The CIO nods.
MTTR avg: 4 min 22 sec / Vendor-reported drift: 2 events
Who Loft is built for.
IT Directors
The dashboard you have been building in Excel.
Loft gives you interface health by channel, error rates by vendor, alert history with resolution timestamps. Deployed in your network. PHI never leaves.
Integration Consultants
Assurance that does not end at go-live.
Recommend Loft Watch to every client after a Post engagement. The integration team that tested with Post now monitors with Loft. Same engine, same validation rules. You become the consultant who delivers assurance that does not end at go-live.
The vendor profiles you built during go-live are the profiles Loft uses in production.
Loft uses the same validation engine that powers Post testing. The rules that caught failures in your test messages catch them in your live traffic. One engine. Unbroken assurance from pre-go-live through long-term production monitoring.
There is no retraining. There is no re-configuration. The vendor profile you built and validated during the go-live sprint is the ground truth Loft monitors against.
Post (testing)
Generate + validate test messages against vendor profiles
Loft (production)
Same vendor profiles. Same validation rules. Live traffic.
Pidgeon.Core validation engine
Shared across Post, Flock, and Loft
Loft is coming. Get in early.
Founding members shape the product, get priority access, and lock in launch pricing before general availability. Download the Post CLI now — same engine, testing mode, free.