An AI assistant to generate information and auto-tune from documents
Copy and paste the following script to your terminal to get started:
curl -fsSL https://smartloop.ai/install | shOptionally, on macos install using Homebrew:
brew tap smartloop-ai/smartloop
brew install smartloopTip
To upgrade: brew update && brew upgrade smartloop
From source:
Note
Requires Python 3.11. For NVIDIA GPU acceleration on Linux/Windows, install CUDA 12.4 before proceeding. The build step auto-creates and activates a virtual environment, then detects your GPU backend — Metal on macOS, CUDA on Linux/Windows with NVIDIA.
git clone https://github.com/smartloop-ai/smartloop.git
cd smartloop
make build # creates and activates .venv, installs dependencies, and configures GPU backend
make test # verifies CLI and GPU offload support# If installed via curl
curl -fsSL https://smartloop.ai/uninstall | shIf install using homebrew , pase the folllowing in your terminal:
brew uninstall smartloop
brew untap smartloop-ai/smartloop# View available commands
slp --help
# Initialize the workspace with a different model:
slp init -t <developer_token> --model=<llama3-1b>
# start the tui
slp Tip
You can generate a developer token from the Console.
In order ensure that your service is running correctly, type the following command in your terminal:
slp statusThis will print details like current GPU being detected, loaded project and context available based on the size of your VRAM or system memeory in case macos based systems
| Property | Value |
|---|---|
| Server | http://127.0.0.1:63838 |
| PID | 17320 |
| Model loaded | True |
| Model | gemma3-1b |
| Quantization | Q8_0 |
| Context window | 31232 |
| Flash attention | False |
| Model size | 1020 MB |
| Memory usage | 8% |
| GPU | Apple Silicon (MPS) |
| Active project | Personal (id=71db0b23-6d6d-401a-b04e-cbb64e7e9636) |
| Project model | gemma3-1b |
| Requirement | Description | Required |
|---|---|---|
| OS | macOS (Apple Silicon) or Linux (x86_64) or WSL | Yes |
| Python | 3.11+ | Yes |
| CUDA | 12.6+ (NVIDIA GPU acceleration) Metal | No (defaults to CPU) |
| Metal | Bespoke on mac | yes |
If the app falls back to CPU on a GPU-enabled system:
-
Enable persistence mode:
sudo nvidia-smi -pm ENABLED
-
Verify GPU detection:
nvidia-smi
If issues persist, ensure NVIDIA drivers are properly installed.
© 2016 Smartloop Inc.
All code is licensed under the GPL, v3 or later. See LICENSE file for details.