Documentation

Getting data in.

Northwind ingests OpenTelemetry, Prometheus remote-write and syslog. If you already run an OTel collector, the integration is one exporter block.

Quickstart

Five minutes from nothing to a chart. You need an API key, which the free tier issues without a card.

Terminal shell

# 1. Install the collector
curl -fsSL https://get.northwind.example/install.sh | sh

# 2. Point it at your account
northwind configure --api-key "${NORTHWIND_API_KEY}"

# 3. Send something
northwind test --signal metrics
# → 200 OK · 1 metric accepted · view at app.northwind.example

Install the collector

The collector is the standard OpenTelemetry Collector with one extra exporter. If you have an existing deployment, add the exporter rather than replacing anything.

otel-collector.yaml yaml

receivers:
  otlp:
    protocols:
      grpc:
        endpoint: 0.0.0.0:4317

processors:
  batch:
    timeout: 5s
  redact:
    # Redaction happens here, before anything leaves your network
    blocked_key_patterns: [".*_ssn", "patient_.*", "authorization"]

exporters:
  northwind:
    endpoint: "ingest.northwind.example:443"
    api_key: "${NORTHWIND_API_KEY}"
    compression: zstd
    sending_queue:
      # Buffer to disk so our outage is not your outage
      storage: file_storage

service:
  pipelines:
    metrics:
      receivers: [otlp]
      processors: [redact, batch]
      exporters: [northwind]

On Kubernetes, the Helm chart wraps all of the above: helm install northwind northwind/collector --set apiKey=…

Your first query

NWQL reads left to right as a pipeline. Every stage takes the previous stage's result.

NWQL nwql

metric("http.server.duration")
  | where service == "checkout-api"
  | percentile(95) by endpoint
  | over(1h)

Metrics

Accepts OTLP, Prometheus remote-write and StatsD. Existing Prometheus setups need only a remote-write URL change — recording rules, alerting rules and Grafana dashboards are unaffected.

prometheus.yml yaml

remote_write:
  - url: "https://ingest.northwind.example/api/v1/write"
    authorization:
      credentials_file: /etc/northwind/api-key

Traces

OTLP over gRPC or HTTP. Sampling policy is set server-side, so you do not redeploy services to change what is kept.

sampling-policy.yaml yaml

policies:
  - name: keep-all-errors
    type: status_code
    status_codes: [ERROR]

  - name: keep-slow
    type: latency
    threshold_ms: 800

  - name: baseline
    type: probabilistic
    sampling_percentage: 5

Logs

Structured JSON is preferred but not required — unstructured lines are parsed with the rules you configure and stored either way.

NWQL nwql

log()
  | where service == "checkout-api" and level >= "warn"
  | where body ~ "timeout"
  | count() by host
  | over(15m)

NWQL reference

NWQL functions
metric(name)	Select a metric series by name
trace()	Select spans
log()	Select log records
where expr	Filter; supports ==, !=, >, <, ~ (regex) and in
percentile(n) by k	Percentile aggregate, grouped
count() by k	Count, grouped
over(window)	Bucket into a time window
above()	Reference the previous query’s result set

Limits and quotas

Platform limits
Max log line	256 KB
Max span size	2 MB
Trace assembly window	60 seconds
Metric resolution	10 seconds
Ingest rate limit	None — bursts are absorbed and billed
Query timeout	120 seconds
API rate limit	600 requests/minute/key

These docs are an excerpt written to demonstrate documentation design. The endpoints are not real — see book a demo if you want to talk about the real thing.