Status
ReadyChecked live (Docker every 5 min, the proxy every 30 s, the model service every minute) or read from the last real self-test; the limits are the configured values. Raw data: /status.json.
Sandbox
- Runtime
runtime: runscavailable- Egress proxy
- running
- Limits per run
- 4 GB memory · 2 CPUs · 512 processes
- Network
- one internal network per sandbox (ring-task-…), with no host address on it → ring-proxy (allowlist)
Model service
- Provider
- Vultr Serverless Inference
- Reachable
- yes HTTP 200 · 249 ms
- Checked
- 2026-11-04 23:58 UTC
- Fix model
deepseek-v4.1-flash-1017- Vision model
qwen3.8-27b
Self-tests
- Containment
- 18/18 passed under gVisor · 2026-11-04 16:35 UTC · on
7068f49 - Hostile-repo scenes
- 20/20 passed under gVisor · 2026-11-04 16:36 UTC · on
7068f49 - Agent loop
- 168/168 passed under gVisor · 2026-11-04 16:39 UTC · on
7068f49
Spend today
- Day (UTC)
- 2026-11-04
- Public runs
- $0.00 of $5.00
- Admin runs
- $0.09 of $15.00
- Until 13 Nov
- $0.93 of $60.00
- Prices
- list prices from the Vultr catalogue; billing may differ
Queue
- Running
- 0 of 2 at once
- Waiting
- 0
- Public runs, sample
- not open yet
- Public runs, GitHub repositories
- not open yet
Latest verified run
- Result
18 fixed · 0 new · 0 cover-ups · 0 lost- Published
- 2026-11-04 07:58 UTC
Version
- Commit
7068f49- Deployed
- 2026-11-04 16:33 UTC
- Up for
- 7 h 25 min