Set up Boltz-2 as a remote tool

Sharing this model for a user, not developing it? Follow the tooluniverse-share-computer guide; no ToolUniverse source checkout is needed. The model runs in its own Python environment, with its dependencies:

~/.local/bin/uv venv --python 3.12 ~/.tooluniverse/envs/boltz
curl -fsSL -o ~/.tooluniverse/envs/boltz-tooluniverse.txt https://connect.aiscientist.tools/install-requirements.txt
~/.local/bin/uv pip install --python ~/.tooluniverse/envs/boltz/bin/python -r ~/.tooluniverse/envs/boltz-tooluniverse.txt
# then this model's own dependencies, as in "Create an isolated environment" below
~/.local/bin/tu remote check boltz --python ~/.tooluniverse/envs/boltz/bin/python   # lists anything still missing
~/.local/bin/tu serve --allow boltz --python ~/.tooluniverse/envs/boltz/bin/python --share --detach --name "<computer name>" --service https://tooluniverse-backend.onrender.com

If the check reports a missing environment variable (a data folder, a model cache, a provider key), set it in the same shell before tu serve; "Obtain credentials, data, and model weights" below says what each one is. The model reads only the variables it declares.

The rest of this guide is for developers working from the ToolUniverse source, and for diagnosing problems with this model.

Validation status (2026-08-16): Boltz 2.2.1 and the official checkpoints ran on the NVIDIA GB10 in one repaired direct MCP call and three authenticated Platform calls, each returning six finite affinity values in 63.6-66.5 seconds. The upstream MSA service timed out during validation, so those successful calls explicitly used bounded single-sequence mode. Missing, oversized, malformed, or non-finite affinity artifacts now fail closed. Public publication, cross-user isolation, broad concurrency, recovery, biological accuracy, and the live MSA path remain unvalidated; keep this deployment private.