CAT

CAT is a local Windows tuning workbench for NVIDIA GPUs. It coordinates MSI Afterburner V/F curve changes, RTSS OSD output, and 3DMark validation from an Electron UI.

Detailed architecture, control flow, safety boundaries, UI rationale, current capabilities, limitations, and audit checkpoints are documented in CAT 软件说明书与代码审计指南.

Safety Model

• CAT does not require administrator privileges by default.
• Runtime data lives under .runtime/; Python is managed by conda under .runtime/conda/cat-python, and JavaScript dependencies live under electron-app/node_modules.
• CAT runs as a local Electron desktop app.
• Hardware writes are bounded, snapshotted, verified, and rolled back on apply failure.

Setup

.\scripts\manage-conda-env.ps1 ensure
cd electron-app
npm install

Python environment management:

.\scripts\manage-conda-env.ps1 doctor
.\scripts\manage-conda-env.ps1 update
.\scripts\manage-conda-env.ps1 rebuild
.\scripts\manage-conda-env.ps1 remove

Run

cd electron-app
npm run dev

Verify

.\scripts\manage-conda-env.ps1 ensure
.runtime\conda\cat-python\python.exe -m pytest
cd electron-app
npm run verify
npm run test:e2e
npm audit

Control Scripts

• scripts/abctl.py: Afterburner shared-memory status, V/F dump, profile import, and transactional V/F apply.
• scripts/rtssctl.py: RTSS shared-memory OSD slot control with dynamic telemetry-file rendering.
• scripts/tdmctl.py: 3DMark WSS automation and result identity validation.

Release Notes

This is still a local development app, not an installer build. Keep validation conservative and always confirm stability with real workloads after synthetic tests.