Observability

Dashboard, OpenTelemetry tracing, Prometheus metrics, and SLA.

Eridian provides full observability across every inference request - tracing, metrics, cost attribution, and SLA monitoring. Platform teams use these signals to tune routing policies, justify cache investments, and satisfy regulatory examination requirements without reconstructing logs manually.

Dashboard

The observability dashboard shows real-time request volume, latency percentiles (p50, p95, p99), cache hit rate, cost breakdown by model and project, and PII redaction statistics.

Access via the Eridian app at /app after you sign in.

Dashboard Panels

Panel	Metrics
Overview	Requests/min, error rate, active projects
Latency	p50/p95/p99 by model and routing policy
Cost	USD spend by model, feature, and API key
Cache	Hit rate, similarity distribution, TTL remaining
Compliance	PII entities redacted, audit export status

Tracing (OpenTelemetry)

Eridian emits OpenTelemetry traces for every request span:

Schema normalization
Cache lookup
RAG retrieval
PII redaction
Model inference (per provider hop)
Response processing and structured output validation

Export traces to Datadog, Honeycomb, or any OTLP-compatible backend:

yaml

otel:
  endpoint: https://otel.example.com/v1/traces
  headers:
    authorization: Bearer ${OTEL_TOKEN}
  sample_rate: 1.0

otel:
  endpoint: https://otel.example.com/v1/traces
  headers:
    authorization: Bearer ${OTEL_TOKEN}
  sample_rate: 1.0

Span attributes include eridian.project_id, eridian.routing_policy, eridian.cache_hit, and eridian.cost_usd for correlation with billing exports.

Metrics (Prometheus)

Prometheus-compatible metrics are available at /v1/metrics for Enterprise tier customers:

Metric	Type	Description
`eridian_requests_total`	counter	Total inference requests by status
`eridian_latency_seconds`	histogram	End-to-end latency
`eridian_cache_hit_ratio`	gauge	Rolling cache hit rate
`eridian_cost_usd_total`	counter	Cumulative cost attribution
`eridian_pii_redacted_total`	counter	Entities redacted

Scrape interval recommendations: 15s for interactive dashboards, 60s for long-term storage.

Audit Export

Export immutable audit logs to your SIEM via:

http

GET /v1/audit/events?start=2025-06-01&end=2025-06-30
Authorization: Bearer eridian_sk_...

GET /v1/audit/events?start=2025-06-01&end=2025-06-30
Authorization: Bearer eridian_sk_...

Events include inference completions, key rotations, policy changes, and budget threshold crossings. Required scope: audit:read.

SLA

Eridian commits to 99.95% monthly uptime for the Inference API. See the SLA for credit schedules and latency commitments. Status updates appear on the status page.

See Usage for token and cost attribution, Audit Log for event query, and Runbooks for on-call when inference is in the path.

Production API credentials are issued with an institution workspace. Contact sales if you need access.