Your guide to
the tradeoff.

1. Explore the public showcase

Open the workspace, choose a scenario, then select Replay LinuxONE run. Inspect the round allocations, acceptance feedback, true-cost benchmark, and messages. Export the complete public run as JSON.

The five scenario records were generated by the real LinuxONE deployment. Replaying loads a recorded result; it does not submit a new optimization job. Timestamps and solver times belong to that recorded execution.

Try Delivery risk with new inputs. On the showcase, this computes a fresh prediction in your browser using the same learned logistic regression coefficients deployed on LinuxONE.

2. Read the outcomes

• Baseline shortage: two rounds, A560 / B300 / alternate60. Accepted true cost $15,000; oracle $14,700.
• Primary capacity loss: A480 / B300 / alternate300. Accepted cost $45,000; oracle $44,900.
• Combined disruption: available supply cannot meet both local minimums. It stays unresolved.
• No scarcity: accepted, but coarse requests draw stock and incur $3,300 cost against a zero-cost oracle. A useful limitation.
• No alternate supply: explores agreement when alternate sourcing is unavailable.

“Surrogate cost” uses disclosed priorities. “True cost” comes from the separate privileged synthetic evaluator. They measure different objectives.

3. Run the full product locally

Install Python 3.10 or newer, then run:

git clone https://github.com/SushantBotla/TradeOff.git
cd TradeOff
cp .env.example .env
python -m tradeoff dev

Open the address printed by the launcher. Open the workspace and paste the dashboard token from the generated .tradeoff/coordinator/dashboard-token.txt file. Local development runs separate processes under your user; the LinuxONE installer adds OS user isolation.

To choose the same real model mode as the deployment, copy .env.example to .env before starting. The learned model is included; no dataset download or external AI key is required for inference.

4. Access the live LinuxONE runtime

The deployed coordinator binds to loopback. An authorized operator connects through SSH, then uses the dashboard token. Substitute your own VM host, account and SSH key:

ssh -N -L 8080:127.0.0.1:8000 USER@VM_HOST

Open http://127.0.0.1:8080/. Select Open workspace, authenticate, choose a scenario, and select Run coordination. These are new runs executed by the VM. The launcher/deployment administrator provides the token; never commit it or paste it into a public issue.

What this prototype proves

A two-factory, one-supplier coordination loop with real integer optimization, bounded feedback, local learned-model inference, separate private data stores, identity checks, and an auditable record. LinuxONE verification includes actual s390x execution and permission-denied tests.

It does not prove strategic truthfulness, formal privacy, production readiness, prospective manufacturing-risk accuracy, or hardware cryptography acceleration. The synthetic evaluator intentionally has more information than the coordinator.