std.serve — Serving and guardrails_

import std.serve as serve

std.serve provides local serving queues plus guardrails for AI serving baselines:

• server_set_input_policy(server: int, min: int, max: int) -> bool
• server_set_output_policy(server: int, min: int, max: int) -> bool
• server_set_rate_limit(server: int, limit: int) -> bool
• server_set_fallback(server: int, value: int) -> bool
• server_last_diagnostic(server: int) -> string
• server_audit_log(server: int) -> string

Guardrail failures complete the request with the configured fallback value and record structured diagnostic/audit JSON.

import std.serve as serve

let server = serve.server_new(3)
serve.server_set_fallback(server, -999)
serve.server_set_input_policy(server, 0, 100)
serve.server_set_output_policy(server, 0, 200)
serve.server_warmup(server)

Monitoring and drift

• server_set_model_version(server: int, version: string) -> bool
• server_monitoring_snapshot(server: int) -> string
• server_distribution_summary(server: int) -> string
• drift_check(reference: string, live: string, threshold_per_mille: int) -> string
• export_monitoring(server: int, path: string, distribution: string, drift: string, audit: string) -> string

These APIs emit versioned JSON for request metrics, latency/error/throughput, input/output distribution summaries, drift checks, and observability export.

---

Próximo / Next: 06 — Referência Rápida / Quick Reference
Anterior / Previous: 04 — Avançado / Advanced