LLM-Box Base — Private AI (Qwen3-8B) in Your Cloud

LLM-Box Base is a preconfigured private AI server. The image bundles the Qwen3-8B model, the vLLM inference engine with an OpenAI-compatible API, and the Open WebUI chat interface behind an nginx reverse proxy.

Why it matters. Access to foreign AI services is limited, and businesses and the public sector need their data to stay inside their perimeter. LLM-Box runs in your own cloud — requests and data never leave it.

Features:
— OpenAI-compatible API: point your app to your own server by changing a single URL;
— web chat for non-technical staff;
— one-click deployment, password and key generated on first boot;
— works fully offline: the model is baked into the image, no external calls required.

Suitable for chatbots, customer support, document processing, RAG and app integration.

The Qwen3-8B model is distributed under the Apache 2.0 license.

1. Create a VM from this image on a platform with NVIDIA T4 (min 8 vCPU, 32 GB RAM, 60 GB disk) and assign a public IP.
2. In the network security group, open inbound ports 80/TCP (web chat and API) and 22/TCP (SSH).
3. After boot, services start automatically (2–5 minutes to load the model).
4. Open http://<public IP>/ for the web chat. The first login creates the administrator.
5. The API is available at http://<public IP>/v1 (OpenAI-compatible). The address and API key are shown in the MOTD over SSH.
6. For integration, set this address as the OpenAI base URL and use the key from the MOTD.

— Chatbots and assistants on your own model without data leakage.
— Customer support: answers to common questions within your perimeter.
— Document processing and RAG: search and answers over your internal knowledge base.
— Analytics and summarization of texts and reports.
— AI integration into apps via an OpenAI-compatible API.

Technical support by email: info@papadata.ru. We assist with deployment and operation, responding within 1–2 business days.