nx_telemetry_bus.nx

nx_telemetry.nx -- substrate telemetry bus. Sealed-enum-typed event bus for substrate-internal observability. Closes the OBD predictive-maintenance pipeline loop (commit 87dca0a0): consumer emits anomaly verdicts -> telemetry bus -> pluggable sink (stdout JSON V1; future MQTT shim / syslog / file when those sinks land). Per INTEROPERABILITY_CHARTER.md the same telemetry bus consumes events from EVERY protocol shim's diag_fn + EVERY substrate-side consumer + the silicon-feedback hot-report. One uniform format across the whole substrate. Status: SEED v0.1.0. 2026-05-27. WINNER-TIER: WINNER-A CANDIDATE INCUMBENTS: syslog (RFC 5424), Prometheus client libs, OpenTelemetry (OTLP), structlog (Python), statsd protocol, Vector.dev event router NUMBERS: V1 ships the bus + stdout-JSON sink; throughput measurement vs structlog/Prometheus pending paired bench GAP: vs OpenTelemetry the gap is feature coverage (no distributed-trace context yet); vs syslog the gap is transport (V1 is in-process; future MQTT/syslog sinks close it); vs structlog Nishi WINS on sealed-enum typing (structlog is string-keyed and loses type safety) PLAN: M-next: paired bench vs structlog on 1M-event burst; re-rate; add distributed-trace context for OpenTelemetry parity EXEMPTION REASON: n/a; provisional pending measurement V1 SCOPE: - 8 sealed event kinds (covers anomaly / perf / health / shim_diag / silicon_feedback / substrate_panic / user / debug) - 4 sealed severity levels (INFO / WARN / ERROR / CRITICAL) - NxTelemetryEvent struct (kind / severity / source-string / ts_ns / 3 value fields for structured numeric payload) - Pluggable NxTelemetrySink (caller-supplied emit function)