System Requirements¶
This page separates declared compatibility from capacity planning. The first is enforced by package metadata and release builds. The second is practical guidance for choosing hardware; it is not a claim that every workload was benchmarked on exactly that configuration.
Supported Software¶
| Component | Requirement | Notes |
|---|---|---|
| Python SDK | Python 3.10 through 3.14 | Python 3.15 is not supported by the current package metadata. |
| Operating system | 64-bit Windows, macOS, or Linux | The Python package is cross-platform; availability of optional scientific wheels can differ by OS and CPU architecture. |
| Linux desktop beta | x86_64 Linux | The current AppImage is built on Ubuntu 22.04 and is not an ARM AppImage. |
| Windows desktop beta | Windows x64 | Distributed as an .exe installer. |
| macOS desktop beta | macOS build from the current GitHub macos-latest runner |
The beta is not advertised as a universal binary; inspect the release notes when CPU architecture matters. |
| Internet access | Required for initial installation | Also required for public biological APIs, Hugging Face downloads, Ollama model pulls, and update checks. Simulations can run offline after dependencies and inputs are present. |
| User storage | Writable home and output directories | The SDK writes environments, caches, run bundles, reports, models, and datasets outside the source tree when configured to do so. |
No GPU is required for ordinary simulation, validation, MDMP certification, or report generation.
Capacity Planning¶
These figures are conservative starting points. Study duration, cohort size, report resolution, dataset size, model size, and parallelism can increase demand substantially.
| Workload | Practical minimum | Recommended | Free storage to reserve |
|---|---|---|---|
| CLI, small simulations, validation | 2 CPU cores, 4 GB RAM | 4 CPU cores, 8 GB RAM | 2-5 GB |
| Reports and desktop workbench | 4 CPU cores, 8 GB RAM | 4-8 CPU cores, 16 GB RAM | 5-10 GB |
| Local Ollama explanation with a small model | 4 CPU cores, 8 GB RAM | 8 CPU cores or supported GPU, 16 GB RAM | 10-30 GB |
| Glucose-model training and large study matrices | 8 CPU cores, 16 GB RAM | 8+ CPU cores, 32 GB RAM, optional supported GPU | 25-100+ GB |
| Long-running Jetson or edge research | board-specific | active cooling and monitored storage | depends on checkpoints and telemetry |
Storage estimates exclude private datasets, Ollama models, model checkpoints, exported figures, and accumulated results/ folders. Those artifacts usually dominate long-term storage.
Installation Profiles¶
| Profile | Python package | Main use |
|---|---|---|
| Standard | iints-sdk-python35[full,mdmp] |
simulation, reports, imports, and certification |
| Research | iints-sdk-python35[full,mdmp,research] |
Torch, ONNX, Parquet/HDF5, and interactive research plots |
| Edge | iints-sdk-python35[edge,mdmp] |
serial bridges and supported hardware workflows |
| Maintained desktop engine | iints-sdk-python35[tauri-engine] |
Tauri/Python bridge, reports, interactive plots, SBML, and FMI support |
| Legacy Qt/development bundle | iints-sdk-python35[desktop-all] |
compatibility testing for the former PySide interface plus ML/packaging dependencies |
The research and legacy desktop-all profiles are much larger because they include machine-learning or GUI-packaging libraries. The maintained app uses the smaller tauri-engine profile; install a training profile only on machines that actually train models.
External Tools¶
The following tools are optional and are not silently installed with the normal SDK package:
- Ollama and its model files
- COPASI
- OpenCOR
- external FMUs used through FMPy
- private or licensed datasets
- pretrained Hugging Face model weights
The desktop workbench detects missing optional tools and should keep unrelated workflows available.
Desktop-Specific Notes¶
The Linux AppImage bundles the application shell, but the scientific Python engine remains a private environment under ~/.iints-af/python-engine. On Omarchy, the supported installer prepares both layers and installs fuse2 for normal AppImage startup.
Tauri documents that AppImage compatibility depends on the GNU C Library baseline used for the build. The IINTS-AF Linux beta is built on Ubuntu 22.04, while current rolling-release Omarchy systems provide a newer userspace. See the Tauri AppImage guide for the underlying compatibility model.
Check A Machine¶
Before installing:
=== "Linux or macOS"
```bash
uname -m
python3 --version
python3 -c "import platform; print(platform.platform())"
df -h "$HOME"
```
=== "Windows PowerShell"
```powershell
py --version
Get-CimInstance Win32_OperatingSystem | Select-Object OSArchitecture, TotalVisibleMemorySize
Get-PSDrive -PSProvider FileSystem
```
After installing:
iints --version
iints version --refresh
iints doctor --smoke-run --suggest
Use iints doctor --full --suggest before a long study, edge deployment, or AI workflow.
Platform Guides¶
Basis For These Requirements¶
- Python and dependency compatibility come from
pyproject.tomlin the released SDK. - Desktop architectures and build baselines come from
.github/workflows/tauri-desktop-beta.yml. - Omarchy package and runtime handling follows the Omarchy manual, its development tools guide, and its update procedure.
- Linux desktop build dependencies follow the official Tauri prerequisites.
Last reviewed: 2026-08-22.