BioMaker develops biomedical image analysis models from a task description and local data. It prepares data, trains and improves candidate models, and exports the selected model and reports. Use the desktop app, WebUI or command line; CPU, Apple MPS and NVIDIA CUDA are supported.
BioMaker Desktop 0.2.0 — macOS Apple silicon
The Google Drive folder contains the DMG, installation instructions and SHA-256 checksums. Drag BioMaker into Applications, then follow the setup guide to install the engine and configure your model provider. Initial setup requires internet access. This test build is not Apple-notarized, so macOS may prompt you to allow it to open.
BioMaker Desktop — start a new experiment.
Requires Python 3.11 or 3.12 and uv.
On macOS, install libomp with brew install libomp first.
git clone https://github.com/uni-medical/BioMaker.git
cd BioMaker
uv sync --python 3.11Create a .env file in the repository root with your OpenAI-compatible provider:
OPENAI_BASE_URL=https://your-provider.example/v1
OPENAI_API_KEY=your-keyDescribe your task, labels and evaluation criteria in task.md, then run:
uv run biomaker data_dir=/absolute/path/to/data \
desc_file=/absolute/path/to/task.md \
model_for_all_agents=your-modelReplace your-model with a model supported by your provider. Outputs are saved in
logs/ and workspaces/. Keep your API keys private; model calls may incur charges.
To use the browser interface:
cd apps/webui
uv run python -m webui.server --host 127.0.0.1 --port 7860Open http://127.0.0.1:7860. See the WebUI guide for configuration.
Thanks to medical workers and dataset owners, and to open-source projects including MONAI and aideml.