setup-genie3-remote-tool

Set up Genie 3 as a remote tool

Validation status (2026-10-01): Genie 3 ran on the NVIDIA GB10 (aarch64, cc 12.1) under the optimization kit's exact mode, all nine levers. It is the only provider here where residency has been shown not to change the output: four consecutive identical requests in one held worker produced BYTE-IDENTICAL PDB files (sha256 over every file, 9.8 s then 0.8 s each), because the worker snapshots and restores four RNG generator states to emulate the kit's own "one process per request" contract. 12x warm with that contract intact. Ladder 0.31 / 0.61 / 1.28 / 2.82 GiB at 100 / 200 / 400 / 670 tokens, 4.84 at 670x2 — the cheapest generative provider here by an order of magnitude against RFdiffusion3's 12.49 and ESMFold2's 25.28. Two operational cautions. (1) Upstream can build exactly ONE sampling config per process: its default_config is a module-level mutable holding FieldReference objects, so the second distinct request raises unless the worker resets it — which this one does. Four identical requests cannot catch that, because a held engine is built once. (2) On unified memory the kit's capacity guard reads total - free from cudaMemGetInfo, which is SYSTEM-WIDE, so a co-tenant preallocating the card makes this provider refuse batches a hundredth of its size. Expect refusals rather than slowdowns when co-scheduling. Public publication, cross-user isolation, concurrency, recovery and design quality are untested; keep this deployment private.

Prerequisites

• Run from the ToolUniverse repository root on Linux with Python 3.12.3.
• A CUDA GPU is required. This provider holds its model resident between requests; CPU is not a target.
• Keep provider data, weights, caches, and credentials outside Git.
• Bind to loopback. A non-loopback bind requires TOOLUNIVERSE_API_TOKEN; never put it in arguments or results.

Run the standard-library contract check before downloading large dependencies:

python scripts/remote_validation/setup_skill_preflight.py --implementation genie3

After exporting provider resources, add --check-provider-env. After the server starts, add --live to verify the exact MCP tool set without running the model. Before sharing, add --check-connect-prereqs; this reports only whether a key is set and never prints its value.

Create the two environments

This provider is on the remote-tool standard, so it has two environments, not one: the MCP server runs in ToolUniverse's, and the model runs in its own, built from a pinned lock. They cannot import each other, and that is the point -- the model's stack is not installable beside ToolUniverse's.

python3 -m venv .venvs/genie3
. .venvs/genie3/bin/activate
python -m pip install --upgrade pip
python -m pip install -e .
python -m pip install fastmcp

The model half lives at the path TOOLUNIVERSE_GENIE3_PYTHON names, and nothing above creates it. Create an empty interpreter there first -- python3 -m venv ~/genie3-venv, or uv venv unless this provider's lock header says otherwise -- and tu remote setup --only environment fills it from src/tooluniverse/remote/genie3/environment/requirements.lock with --no-deps, so no version is silently resolved to something else.

The rest is built by tu remote setup, which reads src/tooluniverse/remote/genie3/provider.toml and runs five steps: preflight, environment, kit, weights, smoke. Each refuses by name and carries its fix.

export GENIE3_WEIGHTS=...
export TOOLUNIVERSE_GENIE3_PYTHON=~/genie3-venv/bin/python
tu remote setup genie3

Run one step at a time with --only preflight|environment|kit|weights|smoke while bringing a machine up. The smoke step is the one that matters: it holds the model, sends one request, asserts the provider produced what its manifest says a successful run produces, and records the measured peak VRAM to ~/.cache/tooluniverse/measured/genie3.json. A host schedules on that file.

Package/network-dependent commands must be rerun in a clean environment before marking this skill complete.

Obtain credentials, data, and model weights

• GENIE3_WEIGHTS points at <root>/pretrained/v1 (2 files, 0.3 GiB) — the smallest weight set of any provider here. [kit.env] GENIE3_OPT_AUTOLOAD = "0" is load-bearing: this kit's .pth hook does not patch upstream, it KILLS any in-process importer of genie3 with exit 3, and the mode is a driver script whose position the resident worker takes.
• Nothing is fetched at request time. The manifest declares every weight file with its size and sha256, and the launch refuses if one is off its pin.

Authorize once, then share with one short command

After installing dependencies, exporting the provider resources above, and installing the pinned relay SDK described under Connect below, run from the repository root. Log in only once per machine (and again after key rotation):

tu remote login
# Or import an existing protected 0600 file without sourcing it:
tu remote login --env-file /path/to/tooluniverse-service.env

Then each private share is one short command:

tu remote share genie3

By default, tu remote login requests a short-lived device code, opens the TU Platform approval page, and polls until the signed-in user approves. No key copy/paste is required. On a headless machine, add --no-browser and open the printed link elsewhere. The CLI exchanges the approval for a computer-only key, verifies /remote-servers/preflight, stores it in a local 0600 config file, and never displays it.

The share command runs environment and TU Platform preflights, starts or reuses the exact loopback endpoint, validates discovery, and keeps the relay in the foreground until Ctrl-C. It automatically uses the reviewed Python, name, and worker count. Override them only when needed:

tu remote share genie3 --name my-genie3-remote --workers 1

Use tu remote check genie3 for a non-sharing readiness check and tu remote run genie3 for a local-only foreground server.

In an interactive terminal, sharing automatically starts the same browser flow when the key is missing, expired, or revoked. A malformed or revoked explicit TOOLUNIVERSE_SERVICE_KEY fails fast instead of being silently replaced; unset or correct it, then run tu remote login. Non-interactive jobs also fail fast. Use tu remote logout to remove only the local copy. Use tu remote logout --revoke to revoke the computer-only platform connection first; the server record remains offline for owner inspection.

Start and verify locally

python -m tooluniverse.remote.genie3.genie3_tool

The Streamable HTTP endpoint is http://127.0.0.1:8048/mcp. In a second activated shell run:

python - <<'PY'
import asyncio
from fastmcp import Client

async def main():
    async with Client("http://127.0.0.1:8048/mcp") as client:
        print([tool.name for tool in await client.list_tools()])

asyncio.run(main())
PY

Confirm discovery contains genie3_design; stop on empty, duplicate, or schema-drifted discovery.

The first request after a restart pays the model build; the ones after it do not. That is the whole reason this provider is resident, and a benchmark that measures only the first call measures the wrong thing.

Connect to ToolUniverse Connect

The tuplatform-connect relay is not yet published on PyPI. Install the reviewed public wheel below; its SHA-256 is pinned. Interactive sharing uses browser device authorization, so no key copy/paste or GitHub access is required.

python -m pip install fastmcp pyyaml "tuplatform-connect @ https://connect.aiscientist.tools/downloads/tuplatform_connect-0.3.0-py3-none-any.whl#sha256=3fad5eee5ecf7887a693d93ccd1aa112dc0955617a885d1fc3daded0030f9ae0"
tu doctor --forward http://127.0.0.1:8048/mcp --json
tu serve --share --forward http://127.0.0.1:8048/mcp --name validation-genie3 --workers 1

Prefer browser device authorization. For CI or migration, supply TOOLUNIVERSE_SERVICE_KEY only through a protected environment or use tu remote login --manual-key; never put a key in shell arguments.

This implementation was added after the authenticated 2026-08-16 Platform matrix and was not part of it. That matrix imported each implementation it covered as unpublished owner drafts and invoked them through /expert-sessions/{id}/test; this provider was imported as no such draft and invoked through no such route. It has not been exercised through the Platform at all: the evidence above is local, on one card, through the loopback endpoint. Public publication, independent-caller authorization/isolation, saturation, and persistent supervision are untested.

Run a verified example

Operation: genie3_design

{"length":"60","n_designs":1,"seed":0}

Invoke the example through the live local MCP endpoint:

python - <<'PY'
import asyncio
import json
from fastmcp import Client

async def main():
    arguments = json.loads(r'''{"length":"60","n_designs":1,"seed":0}''')
    async with Client("http://127.0.0.1:8048/mcp") as client:
        result = await client.call_tool("genie3_design", arguments)
        print(result)

asyncio.run(main())
PY

Expected success shape: n_designs above zero with per-design records, plus engine_rebuilt and rng_restored. The output glob is *_[0-9]*.pdb: the only file written regardless of success is the kit's own composed request.yaml, which it cannot match, and the worker additionally filters by this request's start time so a previous pass's file is not counted. Check scientific meaning, output bounds, invalid-input behavior, and absence of paths, secrets, and traces.

Tune GPU and concurrency

• Use one worker. The resident worker serves one request at a time by design; a second in-flight request queues on its lock rather than doubling peak VRAM.
• Read vram_budget_gb from the provider manifest and the measured file before co-tenanting anything else on the card. For a model whose peak depends on input size, the manifest states the envelope the figure was measured at; a larger request exceeds it.
• Measure cold start, then at least four consecutive identical requests, then a mixed-shape sequence (small, small, large, small). Two warm calls of different shapes is not enough: openfold3 passed exactly that and still died on the second request of the same shape, because a CUDA graph captured on the first replay attempt left a failed capture behind. A repeat and a shape change are different tests and a resident provider has to survive both.
• Then parallel levels 1, 2, 4, 8, and only 16 if memory permits.
• Record successes/errors, p50/p95, peak RAM/VRAM, utilization, queueing, cancellation cleanup, and recovery.
• Increase workers only after single-flight initialization and sanitized recoverable OOM/timeout behavior are proven.

Troubleshoot and clean up

• Read the refusal. Every step of tu remote setup names what is wrong and carries the fix; exit 3 is a refusal and exit 1 is a failure.
• The worker's own transcript is at ~/.cache/tooluniverse/remote-logs/genie3-resident.log, and it holds the kit's full lever census. A lever reading state=off reason=... is the kit declining something, which is usually correct.
• Import/executable failure in the model half: re-run tu remote setup genie3 --only environment. It rebuilds from the lock with --no-deps and will not silently resolve a different version.
• Weights off their pin: the named file is left in place for inspection rather than deleted, because a truncated transfer and a corrupt file look alike. Remove it by hand and re-run --only weights.
• 401/403 on deliberate network binding: configure matching TOOLUNIVERSE_API_TOKEN bearer auth; prefer loopback plus relay.
• Stop server/relay with Ctrl-C. If installed, run tuplatform-service uninstall --name validation-genie3.
• A worker outliving its server is a known shape: it has an idle timeout and an orphan watchdog, and it exits when its parent becomes init. Check with nvidia-smi --query-compute-apps=pid,used_memory --format=csv before assuming the card is free.
• Revoke temporary keys. After confirmation, remove only .venvs/genie3 and the provider's own cache directory; never use a broad recursive target.

Use only official upstream documentation linked by the implementation README; do not substitute third-party model mirrors.