Roboflow Enterprise
Secure Gateway proxies all traffic between your inference servers and Roboflow Cloud. It caches model weights, enforces access policies, and ships audit logs to your SIEM.
$ docker run -d \
--name secure-gateway \
-p 443:443 \
-e ROBOFLOW_API_KEY=rf_... \
-e CACHE_DIR=/var/cache/gw \
-e SIEM_ENDPOINT=https://splunk:8088 \
-v gw-cache:/var/cache/gw \
roboflow/secure-gateway:latest
Container started: a3f8c...
$ curl -k https://localhost/health
{"status":"ok","cached_models":0}
One egress point. Everything else stays on your network.
First download gets cached. Every server after that pulls from the gateway's local store. Cut bandwidth costs and eliminate cold-start delays when scaling your fleet.
# Cache hit — served in 38ms
GET /my-workspace/yolov8n/3
200 — 104 MB — cache:HIT
mode: deny-all
allow:
- model: "ws/safety-model"
versions: ["3","4"]
- model: "ws/counter"
versions: ["*"]
Lock down exactly which models and endpoints your servers can reach. Rules live in a YAML file, hot-reloaded on change — no restart needed.
Every request through the gateway produces a structured JSON log entry. Ship to Splunk, Datadog, Sentinel, or any syslog target. Know who downloaded what, when, and whether it was served from cache.
{"source_ip":"10.0.1.42",
"path":"/ws/yolov8n/3",
"action":"allowed",
"cache_hit":true,
"duration_ms":38}
# Just swap the image
docker stop license-server
docker run -d \
--name secure-gateway \
-p 443:443 \
roboflow/secure-gateway
Same port, same API key, same inference server config. Replace the container image and you get caching, policy enforcement, and audit logging on top of what you already had.
Secure Gateway is available with a Roboflow Enterprise license. Talk to us about deploying in your environment, or dive into the docs.