Changelog
RSSv1.0.1 — Capability discovery + honest 501s
Endpoints that aren't yet wired now return a clear HTTP 501 with a roadmap pointer instead of a confusing upstream error. New /public/capabilities endpoint lets SDKs introspect what's live.
v1.0 — Public Launch
NemoRouter v1.0 is live — managed LLM gateway with 18 models, full dashboard, credit-based billing, AI guardrails, prompt management, A/B testing, team-scoped budgets, and observability — every feature on every tier.
v1.8 — Observability: logs, callbacks, alerts, data policy
Per-org logging callbacks (Langfuse, Datadog, S3, Slack), eight alert types, full request-log search, and four data-policy modes so you decide what request content we retain.
v1.7 — Model catalog with leaderboard rankings
Browse every registered model on one page. Sort by latency, cost, or quality. Switch models with a one-line env var change — your code stays exactly the same.
v1.6 — GDPR data subject requests
Self-service GDPR access, erasure, withdraw-consent, and consent records. Every administrative action lands in the audit trail. SOC 2 controls are active. HIPAA-eligible with a BAA on Enterprise.
v1.5 — Credit safety: reserve + settle
Every LLM request now reserves credits before the call, settles against the real cost after, and releases the hold on every failure path. Negative balances are blocked at the database — not by application logic.
v1.4 — Team management
Invite teammates, assign roles (owner / admin / member / viewer), and see per-key spend by member. Single-org-per-user enforced at the database; one team per user within that org, by design.
v1.3 — Usage analytics dashboard
Per-model, per-key, per-team spend charts on a single page. Filter by date range, export CSV, drill into outliers. Numbers render in Geist Mono — easy on the eyes during the monthly close.
v1.2 — Prompt management with A/B testing
Server-side prompt templates with versioning and Jinja2 variables, plus deterministic traffic-split A/B tests. Same hash, same variant — every time.
v1.1 — Guardrails: pre/post-call enforcement
Five guardrails in the request path on every plan — PII redaction, prompt injection detection, secret scanning, abuse blocking, and response scanning — with org-default + per-key override semantics.