AIM-OS bridges Linux system introspection with large‑language‑model reasoning. It captures real‑time data from the host and lets you administer the machine in plain English requests—no shell scripting required.
- Natural‑language administration: ask AIM‑OS to check resources, manage packages, or troubleshoot services with simple English instructions.
- Real‑time telemetry: Rust and Bash collectors stream process, network, and hardware metrics directly to the LLM.
- Secure, deterministic execution: every action is sandboxed and logged for auditability.
- Extensible pipeline: plug‑in collectors and response adapters to tailor AIM‑OS to any Linux distribution or cloud instance.
- Rust
- Python 3.9+
Python packages (install via pip
):
rich
aioconsole
python-dotenv
openai
typing_extensions
psutil
git clone https://github.com/DCV05/AIM-OS.git
cd AIM-OS
pip install -r requirements.txt
AIM‑OS communicates with an OpenAI Assistant that follows a strict JSON contract for every response.
- Open the Assistants section in your OpenAI dashboard and create a new assistant.
- Paste the system prompt provided in
prompts/assistant_system_prompt.md
into the System instructions box. - Add a Function tool and supply the response JSON schema located at
schema/response_schema.json
. The schema enforces the following structure:
{
"response_type": "text | linux_command",
"response": ["..."],
"metacommand": "",
"error": "",
"dangerous_command": 0 | 1
}
- Save the assistant and copy its
assistant_id
into your.env
file.
Create a .env
file in the project root and supply your OpenAI credentials:
OPENAI_API_KEY=sk-***************************************
ASSISTANT_ID=asst_**************************************
Start the interactive terminal:
python main.py
When prompted, type requests in natural language, for example:
Show me the top 5 processes by memory
Clean the apt cache
How much free disk space do I have on /home?
Create a tar.gz backup of /var/log and store it in /backups
AIM‑OS will translate your request into the appropriate shell commands, execute them safely, and return the results in a readable format.
- Multiple‑model back‑end support (OpenAI, local LLMs)
- Role‑based permission profiles
- Web dashboard