Settings

Configure appearance, manage GPU detection, and control all three tool environments

Settings (07)

Configure appearance, detect GPU hardware, and manage all three tool environments — llama.cpp, Training, and Convert — from one place.

Appearance

Theme

Theme	Description
Dark (default)	Dark background (#0a0a0a) with monochrome + amber accents
Light	Light background (#fafafa) variant

Click the theme toggle in the header to switch. Changes apply immediately.

Font Settings

Setting	Options
Font Family	System monospace fonts (JetBrains Mono preferred)
Font Size	Adjustable base font size

GPU Detection

ForgeAI auto-detects your hardware:

Field	Description
OS	Operating system
ARCH	CPU architecture
NVIDIA	GPU name, VRAM, CUDA version
VULKAN	Vulkan support status
METAL	Metal support (macOS)
RECOMMENDED	Suggested llama.cpp variant

Environment Management

All three tool environments are managed from Settings:

llama.cpp Tools

Required for GGUF quantization (Compress) and GGUF inference (Test).

State	Actions
Not installed	Select variant (CPU/CUDA/Vulkan) → DOWNLOAD & INSTALL
Installed	View version, path → REINSTALL or REMOVE

Training Environment

Python virtual environment for fine-tuning (PyTorch + PEFT + TRL + BitsAndBytes).

State	Display
Not installed	”Install from Training page”
Installed	Python version, venv status, package list, CUDA availability
Partial	Missing packages highlighted

Actions: CLEAN/DELETE to remove the environment and start fresh.

Convert Environment

Python virtual environment for SafeTensors → GGUF conversion.

State	Display
Not installed	”Install from Convert page”
Installed	Python version, venv status, script, packages
Partial	Missing packages highlighted

Actions: CLEAN/DELETE to remove the environment and start fresh.

Color Semantics

Color	Role
Amber	Idle, accent, interactive elements
Blue	Working, in-progress operations
Green	Success, loaded, complete
Red	Error, failure, destructive actions
Gray	Paused, disabled, inactive