Your private local AI server

VenaX is a small Debian-based Linux image. Boot it from a USB stick, join your network, and use Ollama models from any browser on that network. No cloud account is involved.

Download the ISO Read the setup guide
x86_6464-bit PCs
Live USBruns from the stick
Experimentalsee project status
Ollama itself is not exposed to the network. Only the VenaX web interface is.

Running local AI usually means a weekend of setup

You install Linux, install Ollama, set up Wi-Fi, write service files, open the right port, and find the machine's IP address. Then you build or choose a front end so other devices can use it.

VenaX bundles those steps into one image. The build script installs Debian 12 (bookworm), Ollama 0.5.4, NetworkManager, a firewall, and a browser interface. On boot, systemd starts the services in order and the console prints the address to open.

You still choose and download your own models, and you still need hardware that can run them.

What is in the image

Everything below is present in the build script.

A browser control panel on port 8080

Pages for chat, models, network, hardware, storage, services, logs and settings. Chat streams replies from your selected model. You can pull and delete models, scan and join Wi-Fi, restart services, and read logs without a keyboard on the server.

Wi-Fi, Ethernet or its own hotspot

NetworkManager handles all links. The console offers a Wi-Fi menu when no connection exists. VenaX can also broadcast a hotspot (default network name VenaX, address 10.42.0.1) if the Wi-Fi adapter supports access-point mode.

Works with no internet

Only model downloads need internet. Installed models keep working offline, and the UI reports internet and local status separately.

Optional persistent models

Label a partition VENAXDATA and models plus saved Wi-Fi profiles go there. Without it, models live in RAM and are lost at reboot.

Built-in diagnostics

Console tools: venax-self-test, venax-status, venax-hardware, venax-diagnostics. Reports redact Wi-Fi secrets.

How it works

Three services run on the VenaX machine, each with a narrow job.

Boot order: venax-data, then Ollama and venax-agent, then venax-web.

Boot

GRUB loads a Linux kernel and a squashfs root filesystem (live-boot), for BIOS and UEFI. The console user logs in automatically on tty1 and shows the status screen.

Storage

venax-data looks for a partition labeled VENAXDATA and tells Ollama where to store models.

Control and inference are separate

The web interface controls the machine and relays chat. Ollama does the inference. The ISO contains Ollama but no models.

Technical overview

From build-iso.sh.

BaseDebian 12 (bookworm), minbase, amd64
Bootlive-boot, squashfs (zstd), GRUB with BIOS and UEFI support. Menu entries: normal, safe graphics (nomodeset), verbose
Ollama0.5.4 from the official release archive. Listens on 127.0.0.1:11434. Keep-alive 10m, one parallel request, one loaded model
Web UI and APIPython standard library server on 0.0.0.0:8080. Host header and cross-origin checks. No login or TLS
NetworkNetworkManager only. Ethernet DHCP with link-local fallback. Wi-Fi WPA2, WPA3 and open. mDNS name venax.local
Firewallnftables, default drop on input. Allows 8080, DHCP, DNS, mDNS, ICMP. 11434 not opened
Model storagePartition labeled VENAXDATA mounted at /mnt/venax-data, otherwise RAM at /var/lib/venax-ephemeral-models
Build host needsroot, debootstrap, grub-mkrescue, mksquashfs, xorriso, mtools, python3, curl, zstd

Before you start

  • A 64-bit (x86_64) PC you can boot from USB, and a USB drive you can erase.
  • A USB writing tool such as Rufus (Windows) or balenaEtcher.
  • Ethernet, or Wi-Fi that uses WPA2, WPA3 or no password. Enterprise Wi-Fi is not supported.
  • Internet during setup only to download models.

Hardware

The project defines no minimum or recommended specs. Whether a model runs well depends on RAM, CPU, storage and model size. A PC that boots VenaX may still be too slow for a given model. GPU acceleration is not configured or verified: the build installs no GPU drivers, and the Hardware page reports whether Ollama is actually using a GPU.

Installation guide

VenaX runs live from USB. Your existing system is untouched unless you erase the wrong drive.

  1. Download the ISO

    A bootable disk image, not a normal program. The download button goes straight to the release file.

    Download VenaX ISO

  2. Write it to a USB drive

    1. Plug in the USB drive.
    2. Open Rufus or balenaEtcher.
    3. Pick the VenaX ISO and then the USB device.
    4. Start and wait for it to finish.
    This erases the drive.Check the selected device twice. Do not just copy the ISO file onto the stick. If your tool asks for a write mode, follow its recommendation for the image.

  3. Boot the target PC from USB

    Insert the stick, power on, and open the boot menu (the key varies by manufacturer, often shown briefly at startup). Choose the USB drive. If the screen stays blank, pick the nomodeset entry from the VenaX menu.

  4. Connect to a network

    With Ethernet, VenaX asks for an address by DHCP. For Wi-Fi, about 25 seconds after boot the console opens a Wi-Fi menu if no link exists. You can reopen it with w.

    • Type a network number, then the password.
    • r rescans, h adds a hidden network, a starts the VenaX hotspot, q quits.

    No internet is needed for clients to reach VenaX. Model downloads do need it.

  5. Wait for services

    systemd starts storage detection, Ollama, the agent and the web UI. The console heading changes to VENA X READY when everything is up. No model is included.

  6. Find the address

    The console lists OPEN FROM CLIENT URLs of the form below. The IP depends on your network and may change.

    http://192.168.1.50:8080

    Example only. http://venax.local:8080 may work as a convenience but the IP is authoritative. In hotspot mode use http://10.42.0.1:8080.

  7. Open it from another device

    On a phone or laptop on the same network, open that URL. You get the VenaX control panel (port 8080), not the raw Ollama API. If it does not load, see troubleshooting below.

  8. Install a model

    Open Models, pick one from the list or type any Ollama model name, and press Install. Downloads need internet, disk or RAM space, and time. Larger models need much more memory. Then use Chat.

    PersistenceWithout a VENAXDATA partition, models are stored in RAM and disappear at reboot. The Storage page shows the current state.

  9. Verify

    Optional checks. Press s on the console for a shell, then:

    venax-self-test
venax-self-test --chat
curl http://127.0.0.1:8080/healthz
ollama ps

    --chat sends a short test message to the first installed model. Self-test uses sudo, which the console user can do without a password.

Troubleshooting

The PC will not boot from the USB

Likely cause: wrong boot entry or firmware mode. Check that the stick appears in the boot menu, rewrite it, or try the other BIOS or UEFI entry. Try the nomodeset menu item if graphics fail.

Wi-Fi will not connect

Common causes: wrong password, unsupported security (enterprise, WEP), missing firmware for the adapter, or a hardware Wi-Fi switch. Run venax-diagnostics or the Network diagnostics button. Ethernet or a USB adapter is a safe alternative.

No IP address, or my phone cannot open the page

Confirm both devices are on the same network, then check the console URL. Routers with client isolation or guest networks block device-to-device traffic. A VenaX hotspot avoids that if your adapter supports it.

The page loads but chat fails or no model appears

Install a model first. Then check the Services page: Ollama must be active. Logs has the Ollama journal. Models may fail if memory is too low for their size.

Download fails or space runs out

Downloads need internet to reach registry.ollama.ai. In RAM-only mode, space is limited by memory. Use a VENAXDATA partition or a smaller model.

Security and privacy

Prompts go to the Ollama process on your VenaX machine. VenaX does not call a cloud AI provider. Ollama only reaches the internet to download models.

The web UI has no login and no encryption.Anyone who can reach port 8080 can chat, pull or delete models, change Wi-Fi, and restart services. Use VenaX only on a network you trust and never forward 8080 to the internet.

The console user logs in automatically and has passwordless sudo, so treat physical access as full access. Ollama is localhost-only, the firewall drops other inbound ports, and the web process is unprivileged.

Project status

Treat VenaX as experimental. The build script runs checks, but there is no published compatibility list, benchmark data or security audit.

  • Version mismatch: the build script describes VenaX 2.1.0, while the linked release file is named 2.0.0. Features such as the hotspot are described from the script and may differ in the downloaded ISO.
  • Hotspot needs an adapter that supports AP mode. Some Wi-Fi chips need firmware that is not in the image.
  • Live USB only. No installer to disk is provided.
  • Models are not bundled, and RAM-only storage is the default.

FAQ

Does it replace my current operating system?

No. It boots from the USB drive. There is no disk installer in the source files.

Does it need internet?

Only to download models. Chat with installed models works offline.

Does it include AI models?

No. Install them from the Models page.

Can several devices use it at once?

Any device that can reach port 8080 can open the UI. Ollama is set to one parallel request, so simultaneous chats may queue and are not tested here.

Does it support GPU acceleration?

Not verified. No GPU drivers are installed by the build. The Hardware page shows whether Ollama is using a GPU.

Will it run on older hardware?

It may boot, but model speed depends on RAM and CPU. Small models are the realistic start.

Where do I report problems?

Use the VenaX repository.

Try VenaX on a spare PC

Clicking download starts the ISO file from GitHub Releases. Write it to USB, boot, and open the address shown on screen.

Download VenaX ISOSetup guide

https://github.com/ahsanraza1238999/VenaX/releases/download/VenaX/VenaX-2.0.0-x86_64.iso