Operating system
Windows 11 (x64)
64-bit Windows 11 only. Server 2022 is supported for headless paper-trading. Not built or tested against Windows 10, Apple Silicon, or Linux.
Everything past the front door: system requirements, the Docker runtime, the six-step install, post-install verification, troubleshooting, and the complete artifact set with checksums. For the product story, head back to the terminal.
Aureis is Windows-first; its engine runs inside a Docker container — a pinned Python 3.12 runtime, a CCXT market layer, and an optional local LLM stack. Docker Desktop is required. Plan capacity around Docker and the engine, not the UI.
Operating system
64-bit Windows 11 only. Server 2022 is supported for headless paper-trading. Not built or tested against Windows 10, Apple Silicon, or Linux.
Container runtime · required
The engine runs in a Docker container, so Docker Desktop must be installed and running before launch. On Windows it uses the WSL2 backend, enabled during Docker setup.
Hardware baseline
For the local LLM and full bot fleet, 32 GB RAM and an 8-core CPU are recommended. NVIDIA RTX 30-series or newer (CUDA 12+) accelerates inference; CPU-only works but is slower for chat.
Disk space
The app is small, but the engine image (~3.3 GB) plus Docker Desktop’s WSL2 virtual disk consume real space, and WSL2 grows on demand. Reserve headroom for models, OHLCV cache, and trade history.
Network
First launch downloads the engine image once (~680 MB compressed) unless bundled. Live trading needs reachability to your configured exchange via CCXT. The local LLM runs offline; cloud fallback only fires if explicitly configured.
The desktop app is a thin control surface. The trading engine, Neural Brain Matrix memory, bot fleet, and local LLM stack run inside a Docker container Aureis supervises for you. Docker Desktop must be installed and running before you start Aureis — otherwise the engine cannot boot and the app waits at startup.
Why a container
Aureis depends on a precise Python 3.12 runtime, native trading/ML libraries, and a market-data stack that is brittle to install per-machine. A prebuilt image means every operator runs the exact same tested environment — no missing DLLs, no compiler mismatches.
Why a container
The engine is pinned byte-for-byte and SHA-256-verified before it loads. Trades, keys, and memory live in mounted host folders while the runtime stays sealed — so an upgrade can never half-patch your machine and a bad release rolls back cleanly.
Why a container
The runtime lives in the container, not scattered across your system, so Aureis does not pollute your global Python, PATH, or services. Removing Aureis is just deleting the app and its image.
Why a container
The default image is CPU/cloud-first so it boots fast anywhere. With an NVIDIA GPU, the same model opts into local GPU inference and RL training (Docker Desktop + WSL2 GPU + NVIDIA Container Toolkit) without changing the rest of the app.
Aureis-Setup-1.1.7.exe).certutil -hashfile Aureis-Setup-1.1.7.exe SHA256. Match the digest to the value in SHA256SUMS.txt.%LOCALAPPDATA%\Aureis). This lays down the desktop app; the engine loads as a Docker container on first launch.%APPDATA%\Aureis for the Neural Brain Matrix, bot configs, and trade history.A successful install is not yet a successful trading deployment. Walk this checklist before enabling live trading on real capital.
With Docker Desktop running, open Neural Chat. The status pill should read "Backend ready" within ~60 seconds once the engine container is healthy. If it stays on "Booting", confirm Docker is running, then check %APPDATA%\Aureis\backend.log.
Type a short message in chat. The first response confirms the local LLM (Ollama / llama.cpp) inside the engine container is running. Cloud fallback is optional and disabled by default.
Open the Bots panel. You should see the default fleet (volatility_scalper, breakout_tracker, smc_reversal, grid_integrator, tdi_mfi_reversal) all in paper_trading mode.
Configure exchange API keys under Settings → Exchanges. The connectivity probe must pass before flipping any bot to live mode. Live trading without a successful probe is blocked.
Most common: Windows blocked the launch
Double-clicking the shortcut produces no window, but launching C:\Program Files\Aureis\Aureis.exe from a terminal works. SmartScreen, Defender, or third-party AV has placed a "Mark-of-the-Web" hold on the files.
One-click recovery: download the launch helper, right-click Help-Aureis-Launch.cmd and choose Run as administrator. It unblocks every file in C:\Program Files\Aureis\, optionally adds a Defender exclusion, then launches Aureis.
Save both into the same folder. Double-click the .cmd only — the .ps1 is invoked automatically.
SmartScreen warning during install
Click More info then Run anyway. The installer is Authenticode-signed, but SmartScreen reputation accrues with download volume. If in doubt, verify the SHA-256 against SHA256SUMS.txt: certutil -hashfile Aureis-Setup-1.1.7.exe SHA256.
Two folders appeared on your Desktop
If you see $PLUGINSDIR and an Aureis folder, the installer was extracted instead of run. NSIS installers are technically archives. Delete those folders, re-download Aureis-Setup-1.1.7.exe, and double-click it (or right-click → Run as administrator) — do not extract.
Antivirus quarantined Aureis
Add C:\Program Files\Aureis as an excluded path in your AV product (usually under "Settings → Exclusions / Allowed Items / Trusted Locations"). On Windows Defender, the launch helper above does this for you with admin elevation. Then re-launch via the shortcut.
Engine won’t start
The engine runs in Docker, so Docker Desktop must be running before launch. Open Docker Desktop, wait until the tray whale reports "running", then relaunch Aureis. On a true first launch the engine image downloads once before the app is ready.
Backend never readies
The UI is up but the FastAPI backend at http://127.0.0.1:8000 is not responding. Read the tail of %LOCALAPPDATA%\Aureis\backend.log for the python exception. Common causes: port 8000 occupied, AV locking python.exe, or a corrupted SQLite schema (auto-repaired since 1.1.0).
Local LLM very slow
Aureis defaults to local-first execution with a 5s timeout before cloud fallback (if configured). On CPU-only systems, expect slower local generation — consider enabling cloud fallback for chat-heavy workflows.
Failed update
Re-download over a clean network and re-verify SHA-256 with certutil. A persistent mismatch indicates a corrupted download or CDN propagation lag — wait a few minutes and retry.
Direct artifacts, update metadata, SHA-256 checksums, and a machine-readable release manifest — so installs are inspectable instead of opaque.
Single-file Windows build with the full runtime bundled inside. No first-launch runtime download required.
Cryptographic checksums for every published binary in this release. Verify locally before installing.
Compact JSON metadata for in-app update prompts and release checks.
Published update metadata for release channels.
Machine-readable artifact inventory for this public release.
https://aureisai.com.