NATS_Expert/CLAUDE.md

NATS Expert

This project deploys a NATS server behind Coolify, reachable over WebSockets only. NATS' native TCP port (4222) is not exposed to the outside — the WS endpoint is the sole entry point for both applications and the managing agent.

Deployment target

• Coolify project: Volcanic MFEs (p4fts5a0kv3tahzjtbtfequk)
• Server: localhost (d0c4owww4kwo0ow08wws4ss8)
• Build pack: dockercompose
• Compose file: /docker-compose.yml
• Public domain: https://nats.tes.gd → routed by Traefik to nats:8080 (WebSocket listener)
• TLS: terminated at Traefik. NATS itself runs with no_tls: true on the WS listener.

Coolify's non-secret app config (project/server UUIDs, name, domain) lives in deploy.json at the repo root and is gitignored — it is regenerated locally when needed by the deploying-to-coolify-via-api skill flow.

Connection details (for agents managing this NATS)

There are two users, both admin (unrestricted publish/subscribe):

• admin — env vars NATS_USER / NATS_PASSWORD on the Coolify application.
• volcanic-agents — env var NATS_VOLCANIC_AGENTS_TOKEN on the Coolify application. Semantically a bearer token; on the wire it goes through NATS's password field (NATS has no per-user token field — that's only valid at the top of authorization as a single global token).

Local credentials cache

A gitignored credentials.local.json at the repo root mirrors the live values for quick local access. Shape:

{
  "url": "wss://nats.tes.gd",
  "users": {
    "admin":            { "user": "…", "password": "…", "note": "…" },
    "volcanic-agents":  { "user": "volcanic-agents", "password": "…", "note": "…" }
  }
}

Read from it with jq, e.g. jq -r '.users["volcanic-agents"].password' credentials.local.json. It is not the source of truth — Coolify env vars are. Regenerate it from Coolify with the snippet below if it's stale or missing.

The current live values are printed by:

: "${COOLIFY_URL:?}" "${COOLIFY_KEY:?}"
APP_UUID=$(jq -r '.app.uuid' deploy.json)   # from local deploy.json
curl -sS -H "Authorization: Bearer $COOLIFY_KEY" \
  "$COOLIFY_URL/api/v1/applications/$APP_UUID/envs" \
  | jq -r '.[] | select(.is_preview==false) | "\(.key)=\(.value)"'

If deploy.json is not present, list applications under the Volcanic MFEs project and find the one named nats:

curl -sS -H "Authorization: Bearer $COOLIFY_KEY" \
  "$COOLIFY_URL/api/v1/projects/p4fts5a0kv3tahzjtbtfequk" \
  | jq '.applications[] | {uuid, name}'

WebSocket URL

wss://nats.tes.gd

The NATS server accepts standard NATS-over-WebSocket framing (RFC 6455 with the nats subprotocol). No custom path; connect to the root URL.

With nats CLI

nats \
  --server=wss://nats.tes.gd \
  --user="$NATS_USER" --password="$NATS_PASSWORD" \
  server info

With nats.js / nats.ws (Node/browser)

import { connect } from 'nats.ws'; // browser
// import { connect } from 'nats';  // node with ws

const nc = await connect({
  servers: 'wss://nats.tes.gd',
  user: process.env.NATS_USER,
  pass: process.env.NATS_PASSWORD,
});

With nats-py

import asyncio, os
from nats.aio.client import Client as NATS

async def main():
    nc = NATS()
    await nc.connect(
        servers=['wss://nats.tes.gd'],
        user=os.environ['NATS_USER'],
        password=os.environ['NATS_PASSWORD'],
    )
    await nc.publish('hello', b'world')
    await nc.drain()

asyncio.run(main())

What's inside the server

• Listeners (NATS binds all three inside the container; only 8080 is declared in expose: so Traefik picks it as the routing target)
  • 4222 — native NATS protocol, reachable only from other containers on the compose network. Not routed publicly.
  • 8080 — WebSocket listener, no_tls: true. The only port in expose:. Traefik terminates TLS and forwards ws://nats:8080 from wss://nats.tes.gd.
  • 8222 — HTTP monitoring (/healthz, /varz, /jsz), used by the compose healthcheck; not routed publicly.
• JetStream: enabled, persisted to the named volume nats-data mounted at /data. Limits: 256MB memory / 4GB file. Bump max_file_store in nats-server.conf if you need more.
• Auth: two users in authorization.users — admin (password from NATS_PASSWORD) and volcanic-agents (password from NATS_VOLCANIC_AGENTS_TOKEN, semantically a bearer token). Both are unrestricted (no permissions block → admin). Substitution happens at Docker Compose parse time (the config lives inline in docker-compose.yml under configs.nats-conf.content), so NATS itself sees a static config. No accounts, no operator/JWT mode.

Managing / redeploying

Config or compose changes → commit and push to main, then trigger a Coolify deploy (fire-and-forget):

: "${COOLIFY_URL:?}" "${COOLIFY_KEY:?}"
APP_UUID=$(jq -r '.app.uuid' deploy.json)   # or look it up as shown above
curl -sS -X POST "$COOLIFY_URL/api/v1/deploy?uuid=$APP_UUID" \
  -H "Authorization: Bearer $COOLIFY_KEY"
# → { "deployments": [ { "deployment_uuid": "…" } ] }   done, don't poll

Rotating the password: PATCH the NATS_PASSWORD env on the Coolify app, then redeploy. The value is substituted into configs.nats-conf.content at compose parse time.

To add a second user (e.g. a scoped app account), edit the authorization.users array in the configs.nats-conf.content block in docker-compose.yml, and set any new credential env vars in Coolify.

Gotchas learned the hard way

• No per-user token field in NATS config. authorization.users = [...] entries only accept user+password, nkey, or JWT — a token key inside a users entry makes nats-server refuse to start with unknown field "token". The token key is only valid at the top level of authorization, and it defines a single server-wide token (mutually exclusive with users). To give a specific user a bearer-token-style credential, put the random token in the password field. On the wire it's the standard NATS user/password auth mechanism; only the semantics (opaque random string vs. human-picked password) differ.

• Coolify's deploy status ≠ container health. A deploy returning status: "finished" from /api/v1/deployments/… just means the build + docker compose up -d step ran to completion. If the container then crash-loops (e.g. bad nats-server.conf), the deploy is still "finished" but the app's /api/v1/applications/{uuid} shows status: "restarting:unknown" and /api/v1/applications/{uuid}/logs returns Application is not running. — no runtime logs via API. Fall back to docker logs <container-name> on the Coolify host to see why the process is dying. Container name pattern: nats-<app-uuid>-<random>.

Files in this repo

File	Purpose
docker-compose.yml	Single nats service, exposes 4222/8080/8222 internally; only 8080 is routed publicly via SERVICE_FQDN_NATS_8080 + Coolify's docker_compose_domains. The NATS config lives inline under configs.nats-conf.content (delivered to the container as /etc/nats/nats-server.conf) — no separate config file, no bind mount (Coolify's compose executor rewrites relative host paths into a persistent app dir and can't materialise a source file for them).
.gitignore	Keeps deploy.json and credentials.local.json out of git.
deploy.json (local only)	Coolify app config used by the deploying-to-coolify-via-api skill.
credentials.local.json (local only)	Cached NATS connection details for both users (WS URL, admin user/password, volcanic-agents user/password-that-is-a-token). Mirrors the Coolify env vars; regenerate from the API if stale.