LatentMAS
Qwen3‑VL Console

Cloudflare-hosted UI for your local RTX 5090 inference API. Run normal vision inference, a single latent continuation, or the full 5-agent Visual Perception → Planner → Critic → Refiner → Judger chain.

Qwen3‑VL • Latent cache chain

Connection + request

API base URL

Upload an image

Outside users should upload an image here. The server-side image path below is optional/fallback for files already on the API machine.

Fallback image path on the API machine Question / prompt

Latent steps m

Max new tokens

If the Cloudflare HTTPS page cannot call an HTTP LAN API due browser mixed-content rules, open this UI over LAN/local HTTP later or expose the API through an HTTPS tunnel. The API itself is currently live at 192.168.1.5:8898.

Not checked model: unknown gpu: unknown

Run a request to see output here.

{}