Sign in →

Ingestion

Monitor the metering pipeline in real time — pulse KPIs, source health, data quality, and performance — inspect recent usage events, and bulk-upload event files.

Updated 2026-06-15Suggest edits

Ingestion

Ingestion (Intelligence → Ingestion) is the operational view of your metering pipeline: what's flowing in, how healthy it is, and the ability to inspect or bulk-load events.

Tabs

Overview (monitoring)

A pulse KPI strip — Events/min · Success Rate · Active Sources · Avg Latency — plus per-source metrics grouped by source type:

Source typeLabel
APIAPI Gateway
AGENTAI Agent
MCPMCP Desktop
FILEFile Upload
AGENTICAgentic

Each source shows event count, success rate, last-event time, and status (HEALTHY, DEGRADED, DOWN, THROTTLED). The Overview also reports data quality (validation pass rate, schema violations, dead letters, top errors) and performance (P99/P95/avg latency, throughput, backlog depth, throttling alerts).

Recent Events

A paginated table of raw usage_events, filterable by customer, metric name, product type (API, AGENTIC_API, AI_AGENT, MCP_SERVER), and time range. Each event shows its status:

StatusMeaning
ACCEPTEDIngested and counted
DUPLICATEIdempotency-key match — not double-counted
REJECTEDFailed validation

Click a row to open the full JSON payload in a detail drawer, or Export events to S3 or email.

File Upload

Drag-drop a CSV, JSON, NDJSON, or XML file (up to 50 MB). The wizard previews the file, maps columns to customerId, metricName, quantity, occurredAt, and idempotencyKey (with default metric/customer overrides), then ingests. Upload jobs track QUEUED → PROCESSING → COMPLETED (or PARTIAL / FAILED) with accepted/failed row counts.

For programmatic ingestion patterns (gateway plugins, SDKs, webhooks, batch API), see the Protocols and Gateways docs.