1. Home
  2. Install
  3. Windows
Windows install guide · updated June 2026

Install Odysseus AI on Windows in four steps.

A complete, beginner-safe guide to installing Odysseus AI on Windows 10 or Windows 11. PowerShell commands, Python and Git prerequisites, Ollama connection, and fixes for every common Windows error — including the script execution policy, port 7000 conflicts and missing C++ build tools.

Time · ~15 min Difficulty · beginner Tested · Win 10 / 11 Cost · free
Section I

What you need before you start.

A clean Windows install with these four prerequisites in place takes the Odysseus AI install from 15 minutes down to 8 minutes. Skip a step and you'll likely hit one of the errors covered later on this page.

Windows 10 or Windows 11 (build 19044+)

Odysseus AI runs on both Windows 10 and Windows 11 with no functional differences. Older Windows 10 builds before 19044 (the 21H2 update) lack the modern PowerShell features used by the launcher script. Press Win + R, type winver, hit Enter, and confirm your build number before proceeding.

Python 3.11 or newer with PATH

Download the latest Python 3.11+ installer from python.org/downloads. Critically: tick the "Add Python to PATH" checkbox on the very first screen of the installer. If you miss it, every PowerShell command that calls python will silently fall back to the Microsoft Store stub and break the Odysseus AI install.

Verify after install:

py -3.11 --version # expected: Python 3.11.x or newer

Git for Windows

Download from git-scm.com/download/win. Accept all defaults during install — the default editor and PATH options are fine. Git for Windows ships with Git Bash, which can substitute for PowerShell if you prefer Unix-style commands.

About 8 GB free disk space

Odysseus AI itself is small (~200 MB), but the Python virtual environment, dependencies, and your first downloaded model can easily push past 5 GB. If you plan to run multiple models, budget 20 GB. Keep the install on a fast SSD if you have one — model load times scale directly with disk speed.

Section II

The four-step Odysseus AI install for Windows.

Open PowerShell — not Command Prompt — in a folder where you want the project to live. Right-click the folder in Explorer and pick "Open in Terminal" (Windows 11) or "Open PowerShell window here" (Windows 10 with Shift+Right-click).

I
Clone the official Odysseus AI repository

The official Odysseus AI repo is hosted on GitHub under pewdiepie-archdaemon/odysseus. Always clone from this exact path — there are mirror and impostor repositories that look identical at a glance.

# clone the official Odysseus AI repository git clone https://github.com/pewdiepie-archdaemon/odysseus.git cd odysseus

If you see git: The term 'git' is not recognized, Git for Windows isn't installed or isn't on your PATH — close PowerShell, reinstall Git, and reopen.

II
Create the Python virtual environment

A virtual environment isolates Odysseus AI's Python dependencies from anything else on your system. The launcher script will create one automatically — but if you want to do it manually for control:

py -3.11 -m venv venv .\venv\Scripts\Activate.ps1

Your PowerShell prompt should now show (venv) at the start of the line. If activation fails with a script-execution error, jump to Common Windows errors below.

III
Run the Windows launcher

The included PowerShell script handles dependency install, database setup, and starting the server. It's safe to re-run later if anything goes wrong.

powershell -ExecutionPolicy Bypass -File .\launch-windows.ps1

First run takes 3–8 minutes while pip downloads dependencies. Watch the terminal output for the admin password — it's printed exactly once, in plain text, around the time you see "Starting Odysseus AI on port 7000". Copy it before scrolling away.

IV
Open Odysseus AI in your browser

Once the launcher prints Uvicorn running on http://127.0.0.1:7000, switch to your browser:

http://localhost:7000

Log in with username admin and the password from the terminal. You're inside the Odysseus AI workspace. The model dropdown will be empty — that's expected. The next section connects it to Ollama.

Section III

Connecting Ollama to Odysseus AI on Windows.

Odysseus AI is a workspace, not a model — it needs a backend to do anything useful. Ollama is the simplest backend for Windows: a free, native installer that runs in your system tray and serves models on localhost:11434.

Step A · Install Ollama for Windows

Download from ollama.com/download/windows and run the installer. After install, Ollama runs as a background service — you'll see its icon in the system tray. No additional configuration is needed on Windows.

Step B · Pull your first model

Open a fresh PowerShell window (Ollama needs to be running in the background). Pull a small model that fits 8–16 GB of RAM:

# 8 GB RAM: stick to ~1-3B models ollama pull llama3.2:3b # 16 GB RAM: 7-8B models work comfortably ollama pull llama3.1:8b # 32 GB RAM with GPU: 13B models ollama pull qwen2.5:14b

Step C · Tell Odysseus AI where Ollama lives

Inside the Odysseus AI workspace, click the settings gear → Providers → Add Provider. Pick "OpenAI-compatible" and use these values:

# Provider type OpenAI-compatible # Base URL http://localhost:11434/v1 # API key (Ollama doesn't check it, but a value is required) ollama

Save, refresh the page, and your pulled model should appear in the model dropdown at the top of the chat panel. Send a message to confirm Odysseus AI is talking to Ollama.

Section IV · tool

Generate your exact install command.

Answer four questions about your Windows setup. The generator outputs a clean, copy-ready command block tailored to your situation — Windows version, Python state, GPU plan, and whether you'd rather run native or in WSL2.

Windows install command generator

Builds the right command sequence for your machine so you skip the guesswork.

Your install plan

Section V

Six Windows-specific errors and how to fix them.

These six cover roughly 90% of failed Odysseus AI installs on Windows. Match your terminal output against the error text in each card and apply the fix below.

PowerShell execution policy

running scripts is disabled on this system

By default, Windows blocks PowerShell scripts for security. Bypass it for the current PowerShell session only — this is the safe scope and reverts when you close the window.

# run once per PowerShell session Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass # then re-run the Odysseus AI launcher .\launch-windows.ps1

Python not recognized

'python' is not recognized as the name of a cmdlet

Python isn't on your PATH. Either reinstall Python with the "Add to PATH" checkbox ticked, or use the py launcher which Windows installs automatically.

# use the py launcher with explicit version py -3.11 -m venv venv .\venv\Scripts\Activate.ps1 py -3.11 -m pip install -r requirements.txt

Port 7000 already in use

[WinError 10048] Only one usage of each socket address

Another process — usually a stuck Odysseus AI from a previous run — is holding port 7000. Find the PID and kill it.

# find the process holding port 7000 netstat -ano | findstr :7000 # kill it (replace 12345 with the PID from above) taskkill /F /PID 12345 # then relaunch .\launch-windows.ps1

Missing C++ build tools

Microsoft Visual C++ 14.0 or greater is required

Some Python packages compile native code on Windows and need the Visual C++ build tools. Free download, ~3 GB, one-time install.

# 1. Download Build Tools for Visual Studio # https://visualstudio.microsoft.com/downloads/ # 2. Install "Desktop development with C++" workload # 3. Reboot # 4. Re-run pip install -r requirements.txt

WSL2 kernel update required

WSL 2 requires an update to its kernel component

If you chose the WSL2 path, Windows needs the latest WSL kernel. Microsoft ships it via Windows Update, but it sometimes lags.

# open elevated PowerShell (Run as administrator) wsl --update wsl --shutdown wsl # verify WSL2 is the default wsl --set-default-version 2

ModuleNotFoundError

ModuleNotFoundError: No module named 'fastapi'

Your virtual environment isn't activated, or you installed dependencies into a different Python. Always activate the venv before running anything.

# activate the venv (PowerShell) .\venv\Scripts\Activate.ps1 # confirm you're in it — prompt should show (venv) # then reinstall just in case pip install -r requirements.txt
Section VI · tool

Paste your error log, get the fix.

The diagnostic pattern-matches your error text against 12 known Windows-specific Odysseus AI install failures and surfaces the exact command sequence to fix it. Runs entirely in your browser — your log text never leaves your machine.

Windows error diagnostic

Copy the relevant error lines from PowerShell, paste below, and click match. The longer the snippet the better.

Diagnosis

Section VII

Native, WSL2, or Docker — pick the right path.

There are three legitimate ways to install Odysseus AI on Windows. None is universally best. Pick by your existing workflow, not by what sounds most impressive on Reddit.

Native PowerShell

when
  • You want the fastest, simplest path
  • You don't already use WSL2 or Docker
  • You're on Windows 11 with an NVIDIA GPU
Recommended for most

WSL2 Ubuntu

when
  • You already develop in WSL2 daily
  • You want commands that match the Linux docs verbatim
  • You need newer Python tooling than Windows ships
Good for developers

Docker Desktop

when
  • You want clean isolation and easy uninstall
  • You'll deploy elsewhere later (server, NAS)
  • You're comfortable with docker compose
Cleanest for servers

Your Odysseus AI install is done when —

// landfall checklist

  • http://localhost:7000 opens the Odysseus AI login screen
  • You logged in as admin with the password from the terminal
  • Ollama is running in the Windows system tray
  • At least one model appears in the Odysseus AI chat dropdown
  • You sent a test message and got a reply within 30 seconds
  • You wrote down the admin password somewhere safe
Section VIII

Performance tuning for Windows.

Odysseus AI's speed depends almost entirely on the model and the backend, not the workspace itself. These four tweaks move the needle the most on Windows.

Use the right model size for your RAM

A model that exceeds your available RAM will spill to disk and grind to a halt. Rule of thumb on Windows: pick a quantized model whose file size is ~60% of your free RAM. 16 GB RAM → 7–8B model in Q4_K_M quantization (about 4.5 GB on disk, ~9 GB in memory). 32 GB RAM → 14B model. 64 GB RAM → 30B+ models become viable.

Enable GPU offload in Ollama

Ollama auto-detects NVIDIA GPUs on Windows. If you have an RTX card with 8 GB+ VRAM, Ollama will offload as many layers as fit into VRAM, leaving the rest on CPU. Confirm offload is working by watching ollama ps while a chat is running — you should see GPU memory listed.

Close Windows Defender real-time scanning on the project folder

Defender real-time scanning can add 30–50% overhead to Python startup and model load. In Windows Security → Virus & threat protection → Manage settings → Exclusions, add the odysseus folder and your .ollama models folder. Re-enable when you uninstall.

Move models off your system SSD if it's full

Ollama stores models in C:\Users\<you>\.ollama\models by default. If your C: drive is tight, set the OLLAMA_MODELS environment variable to point at a faster external NVMe. Restart Ollama from the system tray to apply.

Appendix

Windows install FAQ.

Does Odysseus AI work on Windows 10 or only Windows 11?
Both. Windows 10 build 19044 (the 21H2 update from late 2021) or newer is required because of PowerShell features the launcher script uses. Windows 11 has slightly better Hyper-V and Docker integration, but the native PowerShell path is identical. Run winver to confirm your build number.
How do I fix "running scripts is disabled" on Windows PowerShell?
Run Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass once in the current PowerShell session. This bypasses the policy for that window only — the change reverts the moment you close PowerShell. It's the safest fix because it doesn't change machine-wide policy.
Should I use the Microsoft Store version of Python?
No. The Microsoft Store Python is sandboxed in a way that breaks several pip workflows Odysseus AI depends on. Always install Python from python.org and tick "Add Python to PATH" during install. If you already installed the Store version, uninstall it, then install from python.org.
Can I run Odysseus AI on Windows without admin rights?
Mostly yes — the install itself only needs your user folder. Two things require admin: installing Python (if not already installed), and installing Docker Desktop (if you go that route). The Odysseus AI repo, virtual environment, and runtime all work entirely in your user folder with no elevation.
Why is Odysseus AI slow on my Windows laptop?
Almost always the model is too big for your hardware. A 7B model on a CPU-only laptop with 8 GB RAM will produce 1–3 tokens per second — painfully slow. Switch to a 1B or 3B model (ollama pull llama3.2:1b) or use an API provider through Odysseus AI's OpenRouter / OpenAI settings instead.
How do I update Odysseus AI to a new version on Windows?
Inside the project folder, run git pull to fetch new code, then re-run .\launch-windows.ps1 — the launcher detects new dependencies in requirements.txt and installs them. Your chat history, settings, and Ollama models survive the update.
Can I access my Windows Odysseus AI install from my phone?
Yes, on the same Wi-Fi. Find your Windows IP with ipconfig, allow port 7000 in Windows Defender Firewall, then open http://<your-ip>:7000 from your phone. Do not expose this port to the public internet — use Tailscale or Cloudflare Access for remote access.
How do I uninstall Odysseus AI from Windows cleanly?
Delete the odysseus folder. That's it — there are no registry entries, no system services, no Windows Add/Remove Programs entry. To also remove Ollama and downloaded models, uninstall Ollama from Settings → Apps, then delete C:\Users\<you>\.ollama.
Continue

Your next step.

Odysseus AI is running on Windows. Here's where to go next based on what you want to do.