Local LLM Setup

RTILA X can run AI models 100% locally on your computer. No internet connection is needed. No data leaves your machine.

Supported Local Models

Model	Size	Description
RTILA Lite 1.5 Assistant 9B	~5.5 GB	Custom-tuned model optimized for web automation tasks
Custom Model	Varies	Any GGUF model from Hugging Face

Configuring Local LLM

1. Go to Preferences → AI Assistant tab.
2. Set AI Provider to Local LLM.
3. Choose a model from the dropdown.

Server Configuration

Setting	Description	Default
Context Size	How much memory the AI has for a single conversation. Higher = remembers more, uses more RAM.	32768
GPU Layers	Number of model layers to offload to GPU. 99 = all layers. Lower this if you experience crashes.	99
Additional Flags	Advanced command-line options for the local AI engine.	-fa on

Context Size Options

Option	Value	Best For
4K	4096	Low-RAM machines
8K	8192	Basic tasks
16K	16384	Standard use
32K	32768	Recommended default
64K	65536	Complex multi-page tasks
128K	131072	Very complex tasks
256K	262144	Large context needs
512K	524288	Maximum context
1M	1048576	Extreme context

GPU Layers

Value	Description
99	Offload all layers to GPU (best performance)
0	Run entirely on CPU (use if no GPU or GPU crashes)
1–98	Partial offload (adjust based on available VRAM)

Tip: If the AI engine crashes or fails to start, try lowering GPU Layers to 0 (CPU-only) or reducing the Context Size.

Using a Custom Model

1. In the model dropdown, select Custom Model (Hugging Face GGUF).
2. In the Custom Model Path / Repo field, enter a Hugging Face repository path:

   bartowski/DeepSeek-R1-GGUF/DeepSeek-R1.gguf

3. The model is downloaded automatically on first use.

Hardware Requirements for Local AI

Model Size	Minimum RAM	Recommended RAM
7B–8B parameters	4 GB	8 GB
13B parameters	8 GB	16 GB
30B+ parameters	16 GB	32 GB+

Tip: For the best experience, use a machine with a dedicated GPU and at least 16 GB RAM.

Monitoring the Local AI Engine

The sidebar shows the Local AI Engine status:

Status	Description
Stopped	Engine is not running. Click Pre-load Engine to start it.
Starting…	Engine is loading the model into memory.
Downloading…	Model is being downloaded.
Ready	Engine is ready. Shows the active model name and a Stop button.

Switching Models

If you change the model while the engine is running:

1. The current engine is stopped.
2. The new model is downloaded (if needed).
3. The engine restarts with the new model.

Tip: You can also click the Stop button next to the model name to manually unload the engine and free up RAM.