Observability & Telemetry Audits

Sub-millisecond routing telemetry, automated circuit breakers, and audited CSV/JSON log exports without storing user data.

Live EWMA Latency & Health Tracking

Nicrron computes real-time Exponentially Weighted Moving Average (EWMA) latency and Time-to-First-Token (TTFT) metrics for every provider endpoint. When upstream providers experience degraded performance, the router automatically steers traffic to healthier, faster endpoints.

EWMA Latency Tracker

Continuously samples round-trip response times to select lowest-latency routes.

Circuit Breakers

Automatically isolates failing endpoints (5xx spikes) to prevent cascade failures.

BullMQ Async Queue

Sub-millisecond credit settlement decoupled from the streaming response critical path.

CSV & JSON Telemetry Log Exports

From the /dashboard/logs console, export audited request traces directly to RFC 4180 CSV or structured JSON:

  • CSV Export: Includes Request ID, Model Slug, Resolved Provider, Prompt/Completion Tokens, Latency, TTFT, Cost, Status, and Timestamps for financial accounting in Excel/Sheets.
  • JSON Export: Structured telemetry array ready for ingestion into Datadog, Splunk, or internal data warehouses.

Zero Data Retention Guarantee

Every exported log record complies with our strict Zero Data Retention commitment. Only cryptographic hashes, token counts, and performance metrics are logged — prompt text and completion tokens are never retained.

Observability & Telemetry | Nicrron Documentation