MosswartOverlord/nginx/overlord.conf

# Nginx site config for overlord.snakedesert.se
#
# Lives on the host (not in the Docker stack) at:
#   /etc/nginx/sites-enabled/overlord
#
# This file is the source-of-truth copy committed to git. To deploy a change:
#   1. Edit this file in the repo
#   2. SSH to the host
#   3. sudo cp /home/erik/MosswartOverlord/nginx/overlord.conf /etc/nginx/sites-enabled/overlord
#   4. sudo nginx -t && sudo nginx -s reload
#
# Critical settings:
#   - proxy_read_timeout / proxy_send_timeout 1d on /websocket/ and /
#     WebSockets are long-lived; nginx's default 60s timeout drops idle clients.
#     Removing these timeouts caused all plugin connections to drop every
#     ~60s when no data flowed from backend to client (April 2026 incident).
#   - /grafana/ panel embeds rely on Grafana's anonymous Viewer auth
#     (GF_AUTH_ANONYMOUS_ENABLED=true in docker-compose.yml) — no credentials
#     in this file. Do NOT hardcode tokens here: this file is committed to a
#     public repo, and Grafana's state DB is ephemeral container storage, so
#     service-account tokens get orphaned on every container recreate. A
#     previously committed (long-dead) token was removed in June 2026.

server {
    listen 443 ssl;
    server_name overlord.snakedesert.se;

    # Security hardening
    server_tokens off;
    add_header X-Frame-Options SAMEORIGIN always;
    add_header X-Content-Type-Options nosniff always;
    add_header Strict-Transport-Security "max-age=31536000; includeSubDomains" always;
    add_header Referrer-Policy strict-origin-when-cross-origin always;
    add_header Permissions-Policy "geolocation=(), camera=(), microphone=()" always;

    # SSL certificates
    ssl_certificate     /etc/letsencrypt/live/overlord.snakedesert.se/fullchain.pem;
    ssl_certificate_key /etc/letsencrypt/live/overlord.snakedesert.se/privkey.pem;
    include             /etc/letsencrypt/options-ssl-nginx.conf;
    ssl_dhparam         /etc/letsencrypt/ssl-dhparams.pem;

    # Plugin WebSocket ingest — `/ws/position` upstream
    location /websocket/ {
        proxy_pass             http://tracker/ws/position;
        proxy_http_version     1.1;
        proxy_set_header       Host              $host;
        proxy_set_header       X-Real-IP         $remote_addr;
        proxy_set_header       X-Forwarded-For   $proxy_add_x_forwarded_for;
        proxy_set_header       X-Forwarded-Proto $scheme;
        proxy_set_header       Upgrade           $http_upgrade;
        proxy_set_header       Connection        "upgrade";
        proxy_set_header       X-Plugin-Secret   $http_x_plugin_secret;
        proxy_cache_bypass     $http_upgrade;
        # Long-lived WebSocket: don't time out the proxy
        proxy_read_timeout     1d;
        proxy_send_timeout     1d;
    }

    # Overlord Agent — host-side service running OUTSIDE the Docker stack
    # because it shells out to `claude` which depends on host-side
    # ~/.claude credentials. Long timeout because agent calls can spin
    # while Claude Code chains tool invocations.
    location /api/agent/ {
        proxy_pass           http://127.0.0.1:8767/agent/;
        proxy_http_version   1.1;
        proxy_set_header     Host              $host;
        proxy_set_header     X-Real-IP         $remote_addr;
        proxy_set_header     X-Forwarded-For   $proxy_add_x_forwarded_for;
        proxy_set_header     X-Forwarded-Proto $scheme;
        proxy_pass_request_headers on;
        # Heavy tool calls (cross-char search, suitbuilder) can take a while;
        # the python wrapper caps each turn at 240s, so 300s gives some
        # headroom for the round trip.
        proxy_read_timeout   300s;
        proxy_send_timeout   300s;
    }

    # API endpoints (live, trails, history, stats) — short-lived HTTP
    location /api/ {
        proxy_pass           http://tracker/;
        proxy_http_version   1.1;
        proxy_set_header     Host              $host;
        proxy_set_header     X-Real-IP         $remote_addr;
        proxy_set_header     X-Forwarded-For   $proxy_add_x_forwarded_for;
        proxy_set_header     X-Forwarded-Proto $scheme;
        proxy_set_header     Upgrade           $http_upgrade;
        proxy_set_header     Connection        "upgrade";
        proxy_cache_bypass   $http_upgrade;
    }

    # Frontend UI and browser WebSocket (`/ws/live` upstream)
    location / {
        proxy_pass           http://tracker/;
        proxy_http_version   1.1;
        proxy_set_header     Host              $host;
        proxy_set_header     X-Real-IP         $remote_addr;
        proxy_set_header     X-Forwarded-For   $proxy_add_x_forwarded_for;
        proxy_set_header     X-Forwarded-Proto $scheme;
        proxy_set_header     Upgrade           $http_upgrade;
        proxy_set_header     Connection        "upgrade";
        proxy_cache_bypass   $http_upgrade;
        # Long-lived browser WebSocket (/ws/live): don't time out
        proxy_read_timeout   1d;
        proxy_send_timeout   1d;
    }

    # Grafana Dashboard UI (served under /grafana)
    location /grafana/ {
        proxy_pass             http://grafana;
        proxy_http_version     1.1;
        proxy_set_header       Host              $host;
        proxy_set_header       X-Real-IP         $remote_addr;
        proxy_set_header       X-Forwarded-For   $proxy_add_x_forwarded_for;
        proxy_set_header       X-Forwarded-Proto $scheme;
        proxy_set_header       Upgrade           $http_upgrade;
        proxy_set_header       Connection        "upgrade";
        proxy_cache_bypass     $http_upgrade;
    }
}