Changeset 0.33.0 (#207)

Co-authored-by: Jeffrey Smith <jasafpro@gmail.com>
Co-committed-by: Jeffrey Smith <jasafpro@gmail.com>

CHANGELOG.md

@@ -1,5 +1,68 @@
 # Changelog
 
+## [0.33.0] — 2026-03-19
+
+### Summary
+
+Observability: structured logging, Prometheus metrics, OpenAPI docs,
+Grafana dashboards, alerting rules, and a built-in admin monitoring
+dashboard.  Operate the platform without reading Go source code.
+
+### New
+
+- **Structured logging (`slog`)** — `LOG_FORMAT=json|text`,
+  `LOG_LEVEL=debug|info|warn|error`.  JSON handler for production,
+  text handler for development.  Request ID, method, path, status,
+  latency, client IP, and user ID on every request log line.
+- **Request ID middleware** — `X-Request-Id` UUID header generated per
+  request (or forwarded from upstream).  Propagated through completion
+  chain as correlation ID.
+- **Prometheus `/metrics` endpoint** — `switchboard_http_requests_total`,
+  `switchboard_http_request_duration_seconds`, `switchboard_websocket_connections`,
+  `switchboard_completion_tokens_total`, `switchboard_completions_total`,
+  `switchboard_completion_duration_seconds`, `switchboard_provider_status`,
+  `switchboard_db_open_connections` (+ `_in_use`, `_idle`, `_wait_count`,
+  `_wait_duration`), `switchboard_task_executions_total`,
+  `switchboard_sandbox_executions_total`.  Uses `c.FullPath()` for
+  path patterns (no cardinality explosion).
+- **OpenAPI 3.0.3 spec** — hand-curated `openapi.yaml` covering auth,
+  channels, completions, health, and admin API groups.  Served at
+  `/api/docs/openapi.yaml`.
+- **Swagger UI** — `/api/docs` renders interactive API browser via CDN.
+  System-respecting light/dark mode with WCAG AA 4.5:1+ contrast ratios.
+- **Grafana dashboard** — `switchboard-overview.json` with request rate,
+  latency percentiles, provider health, token usage, DB pool, Go runtime.
+- **PrometheusRule alerting** — OOM recovery, provider down, pool
+  exhaustion, high error rate, task failure spike.
+- **ServiceMonitor** — Helm template for Prometheus Operator discovery
+  (gated: `monitoring.serviceMonitor.enabled`).
+- **Admin monitoring dashboard** — `GET /api/v1/admin/dashboard`
+  aggregates provider health, 24h usage totals, DB pool stats, WS
+  connections, Go runtime (goroutines, heap, sys, GC count, Go version),
+  storage status, and recent errors.  Auto-refreshes every 30s.
+- **ICD observability tests** — 6 tests covering `/metrics`,
+  `/api/docs`, `/api/docs/openapi.yaml`, `X-Request-Id` generation,
+  `X-Request-Id` passthrough, and admin dashboard.
+
+### Changed
+
+- **`middleware/logging.go`** — rewritten from `gin.Logger` wrapper to
+  `slog.Info` with structured fields.  `SkipPaths` still honored.
+- **`config/config.go`** — added `LogFormat`, `LogLevel` fields with
+  `LOG_FORMAT`, `LOG_LEVEL` env var loading.
+- **`events/ws.go`** — WebSocket connect/disconnect now updates
+  `switchboard_websocket_connections` gauge.
+- **`health/accumulator.go`** — flush cycle updates
+  `switchboard_provider_status` gauge per provider config.
+- **`handlers/completion.go`** — instrumented with token counters,
+  completion latency histogram, and status counter.
+- **Dockerfile** — Go 1.22 → Go 1.23 (required by `prometheus/client_golang`).
+- **`nginx.conf`** — added `location = /metrics` proxy rule.
+- **Admin monitoring tab** — landing page changed from Usage to Dashboard.
+- **Helm `values.yaml`** — added `monitoring` section (serviceMonitor,
+  grafanaDashboard, prometheusRule — all `enabled: false` by default)
+  and `logging.format`, `logging.level`.
+
 ## [0.32.0] — 2026-03-19
 
 ### Summary