Installation

How to install proto-tools

Installation

Step 0: Clone the repository

bash

git clone https://github.com/proto-bio/proto-tools.git
cd proto-tools

In the future, we plan to enable a direct PyPI install (pip install proto-tools), but prior to the public release we will be using this local install approach.

Step 1: Install the package

All you need is Python 3.10+ and pip:

bash

pip install .

System tools that standalone tool environments need (git, curl, gcc) are automatically provisioned on first use via a shared foundation environment — no manual setup required.

If you are developing or contributing to this project, follow the setup instructions in CONTRIBUTING.md instead.

Step 2: Configure storage (optional)

All persistent data (model weights, tool environments, micromamba) is stored under PROTO_HOME (defaults to ~/.proto/).

To customize the storage location (recommended for labs/HPC):

bash

# Add to your ~/.bashrc:
export PROTO_HOME=/path/to/your/proto_home

To override just model weights separately: export PROTO_MODEL_CACHE=/path/to/shared/weights. See Storage guide for all options.

Step 3: Gated model access (optional)

Some tools use gated models that require accepting a license / terms-of-use. Two access flows depending on how the upstream author publishes weights:

Model	Source	Access
ESM3	HuggingFace: EvolutionaryScale/esm3-sm-open-v1	Accept EvolutionaryScale license, then authenticate with HF (see below)
AlphaGenome	HuggingFace: google/alphagenome-all-folds	Accept Google DeepMind terms, then authenticate with HF (see below)
AlphaFold3	DeepMind request form: google-deepmind/alphafold3#obtaining-model-parameters	Submit DeepMind’s form; after approval (2–3 business days) download the weights archive and place at $PROTO_HOME/proto_model_cache/alphafold3/ (or set PROTO_ALPHAFOLD3_WEIGHTS_DIR). Not hosted on HuggingFace.

For HuggingFace-gated models:

1. Create a HuggingFace account
2. Visit each model page above and accept the license/terms
3. Install the HuggingFace CLI and log in:
   curl -LsSf https://hf.co/cli/install.sh | bash
   hf auth login
   
   Or set the token directly in your environment:
   export HF_TOKEN=hf_...
   

For AlphaFold3: follow DeepMind’s linked form (no HF token involved). See proto_tools/tools/structure_prediction/alphafold3/README.md for the full weights-setup flow.

You’re all set up! To learn what features are available in the library, check out the tutorial series — four short notebooks covering the universal tool pattern, persistent execution, device management, and parallel multi-GPU runs.

---

You’re all set! Head over to the Quickstart to run your first tool, or browse the tool catalog to see what’s available.