Before the storm, the tools must be forged.
| Requirement | Version |
|---|---|
| Python | ≥ 3.9 |
| pip | any recent |
| An LLM server | See Supported Servers |
Typhon does not require a GPU to install or scan — but a GPU is required to face the full trial and get meaningful benchmark data.
=== “From GitHub (recommended)”
```bash
git clone https://github.com/IAcriolla/typhon-stress-test.git
cd typhon-stress-test
pip install -e .
```
The `-e` flag installs in editable mode and registers all `typhon-*` commands in your shell. No manual PATH changes needed.
=== “In a virtual environment”
```bash
git clone https://github.com/IAcriolla/typhon-stress-test.git
cd typhon-stress-test
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e .
```
Recommended for keeping Typhon's dependencies isolated from your system Python.
typhon-scan --help
You should see the Typhon banner and the help text. If you get command not found, make sure the Python bin directory is on your PATH (it always is inside a virtual environment).
All dependencies are declared in pyproject.toml and installed automatically by pip install -e .:
| Package | Purpose |
|---|---|
requests |
HTTP — speaks to LLM servers and feeds the scanner |
psutil |
CPU and RAM detection |
openai |
OpenAI-compatible client for typhon-ask — works with any local LLM |
fastapi |
The herald API server |
uvicorn |
ASGI server that runs FastAPI |
GPU monitoring — VRAM, temperature, power draw, utilization — requires nvidia-smi, which ships with the NVIDIA driver. Typhon detects it automatically. Without it, those fields will be absent from benchmark results. The trial still runs; it just fights partially blind.
Verify:
nvidia-smi --query-gpu=name --format=csv,noheader
# NVIDIA GeForce RTX 3090
cd typhon-stress-test
git pull
pip install -e . # re-registers entry points if they changed
Your data/ directory is gitignored and will not be touched by a pull. The chronicle survives.