Install OpenLM AI module management on Windows

brucelee1862026-03-22 12:07

1️⃣ Install Python

OpenLM is Python-based. Make sure you have Python 3.9+ installed.
Check installation:

python --version
pip --version
If missing, install from python.org.
Optional: Add Python to PATH during installation.

2️⃣ Install OpenLM Module Manager

Open a Windows Command Prompt or PowerShell:

复制代码

pip install openlm

This installs the CLI and API server tools.

3️⃣ Initialize OpenLM Repository

Create the module storage folder:

复制代码

openlm init

This will create folders like:

C:\Users<User>.openlm
modules
configs
logs\
All AI modules you register will be stored here.

4️⃣ Register / Add AI Modules

Suppose you have an AI module or script (e.g., weather-agent):

复制代码

openlm add-module --name weather-agent --type agent --path "D:\project\weather-agent" --gpu 0

--name → module ID
--type → agent / model / tool
--path → folder with code or checkpoint
--gpu → GPU index if available

You can repeat this for all your AI modules.

5️⃣ List Registered Modules

复制代码

openlm list

Example output:

复制代码

ID      Name            Type       Status
1       weather-agent   agent      ready
2       code-reviewer   model      ready

6️⃣ Start the Module Server

Open a terminal in Windows 11:

复制代码

openlm serve

Starts a local REST API server , typically at http://localhost:8000.
Each module is exposed via endpoints like /run?module=weather-agent.

7️⃣ Use a Module via Python

Example: call a module to check last 7 days' weather:

复制代码

import requests

resp = requests.post("http://localhost:8000/run", json={
    "module": "weather-agent",
    "input": "Check last 7 days rain in New York"
})

print(resp.json())

Works for agents, models, and tools
Responses are JSON objects with module output.

8️⃣ Update or Remove Modules

复制代码

openlm update-module --name weather-agent --path "D:\project\weather-agent-new"
openlm remove-module --name old-module

Keeps your modules versioned and organized.

9️⃣ Notes for Windows Users

GPU usage: Make sure CUDA drivers are installed and PyTorch detects the GPU.
Firewall: Allow Python/Server through firewall if you want remote access.
Paths : Use double quotes "D:\folder\path" in CMD / PowerShell.
PowerShell vs CMD: Both work; PowerShell allows better scripting.

If you want, I can also make a diagram showing OpenLM on Windows + OpenClaw + GPU + modules, so you can see visually how everything interacts.

Do you want me to do that?