> **RETIRED + DROPPED (2026-07-15).** Central has been replaced by [[conduit]]. Its app [[services]] were stopped and disabled 2026-07-14 (zero live consumers remained); on 2026-07-15 its database was archived to pi-nas (sha256-verified) and dropped (`DROP DATABASE central`, ~41 GB reclaimed), and the shared Postgres instance was cleaned back to plain (TimescaleDB removed). Central is recoverable only from the pi-nas archive. **Everything below this point is historical** — it describes how central worked while it was live, and is kept for reference only.
central is a multi-domain real-time data-hub spine. Adapters normalize upstream sources, publish CloudEvents to **NATS/JetStream**, and archive to **TimescaleDB/PostGIS** for historical and geospatial query. It is the live data backbone for [[navi]] traffic tiles and related situational-awareness feeds.
The data flow is: upstream APIs → adapters (central-supervisor) → NATS/JetStream :4222 → central-archive (TimescaleDB/PostGIS :5432). The central-gui (FastAPI + HTMX, :8000) exposes the API consumed by [[navi]] via recon-vm's nginx (port 8440, `^~ /api/traffic/` → `central.echo6.mesh:8000`).
navi-traffic (navi's in-VM :8421 extraction service) was **retired on 2026-05-26** and cut over to central. recon-vm's nginx (`/etc/nginx/sites-available/navi.echo6.co`, port 8440) proxied `^~ /api/traffic/` → `central.echo6.mesh:8000` with a 120s tile cache. navi-traffic:8421 is confirmed dead and disabled.
**As of 2026-07-14, navi's `/api/traffic/` tiles were repointed to [[conduit]]:** recon-vm's nginx now rewrites `^~ /api/traffic/` → `/up/tomtom_flow_tiles/...` and proxies to `central.echo6.mesh:8010` ([[conduit]], co-resident on the same CT 104 host). Central's own tile endpoint (`:8000/api/traffic/...`) still exists and still works, but is no longer on navi's hot path — it remains the rollback target if the Conduit cutover needs to be reverted. See [[conduit-operations]] for the cutover and rollback procedure.
- **v0.15.0 (deployed 2026-07-02, PRs #120–124):** the GUI can now add public REST/GeoJSON data sources **no-code** — a generic adapter kind created and configured entirely through adapter management, no repo change. None enabled in this [[deployment]] yet; all 23 current adapters remain bespoke per-source kinds (each `config.adapters.kind` is unique to its source).
- **Deploy mechanism:** manual, tag-based (`git checkout <tag>` → `uv sync` → `central-migrate` → restart). One-command `scripts/deploy.sh` ships in-repo; full procedure in the [[central-deploy-cutover]] runbook. No CI/CD; no automated DB backup (pre-flight `pg_dump` is the only migration safety net; migrations are forward-only).
- **`central-migrate` env gotcha:** manual calls need `cd /opt/central && set -a && . /etc/central/central.env && set +a && …` (the units supply this via WorkingDirectory + EnvironmentFile).
- **EONET:** now **global** as of v0.14.6 (2026-06-28) — `bypass_bbox_filter=True` + the adapter `region` key removed. NOTE: its `events`-table row count is a misleading health signal (dedup + partition expiry); use the CENTRAL_DISASTER stream as the EONET flow signal. `avalanche_org` zero-events is expected (off-season gate).
- **Supervisor CPU** ~17%→~1.4% and **cursors.db** 1.3 GB→496 MB after the v0.14.6 dedup/WAL fix + a VACUUM.
