Quick start

Brainbrew needs Python 3.12 or 3.13. A GPU is only needed for a local model server or LoRA training.

Install

git clone https://github.com/Yog-Sotho/Brainbrew.git
cd Brainbrew
uv sync                         # or: bash install.sh
cp .env.sample .env             # then add your keys

Optional extras:

Extra	Adds
uv sync --extra train	LoRA training with TRL + PEFT
uv sync --extra pii	Presidio name detection (then python -m spacy download en_core_web_lg)

Without uv: pip install --require-hashes -r requirements.txt (or requirements-train.txt for core + LoRA training).

A local GPU model server is not an extra: install vLLM in its own environment with uv tool install --python 3.12 vllm==0.31.0 (Linux, NVIDIA), let install.sh do it, or run the official image with docker compose (see Docker).

Run the app

uv run streamlit run app.py

Open http://localhost:8501, upload a document, pick an endpoint and a model, choose a target size and click Generate Dataset.

No API key? Start a local model server and pick Local vLLM server in the sidebar:

vllm serve Qwen/Qwen3-4B-Instruct-2507    # or: bash run_vllm.sh (install.sh, 24 GB+ VRAM)

Run from the command line

export OPENAI_API_KEY=sk-...
uv run brainbrew run notes.pdf --size 100 --format sharegpt -o dataset.jsonl

See Command line for config files and run inspection.