Agent roles

“Agent” describes a role in this project, not a blanket claim of autonomy, safety, or live financial execution.

Yield selection

The backend includes a yield-optimizer service. Its source is organized around strategy candidates and their attributes, with a corresponding Solidity contract prototype. Treat candidate selection as an implementation surface. Do not read it as a current quote, a ranked market survey, or realized APY. The site’s selector uses letters and qualitative fields so it does not invent yield data.

Arbitrage model

The browser repository contains an arbitrage model, and the Solidity package has an arbitrage contract source. The model includes a simulated execution path. A schematic route comparison helps explain the shape of the work, but it does not establish live liquidity, executable prices, net profit, or transaction settlement. See the simulation boundary note for the evidence levels used on this site.

Risk thresholds

The risk contract source defines a warning marker at 700 and a higher marker at 900. The product-page slider moves an illustrative value across those source constants. It is not connected to an account, does not invoke the contract, and does not prove that a loss can be prevented. A threshold is only as meaningful as its inputs, model, enforcement, and review.

Portfolio rebalancing

The Python engine computes allocation deviations and forms a candidate rebalance. Its execution path explicitly labels itself simulated. The related Solidity source does not turn that local example into a transaction. The page’s allocation view contains no token balances, prices, or dollar values; it only switches a schematic pattern and announces that no account was read.

What “agent” does not mean

In these notes, an agent is not assumed to be funded, continuously connected, risk-free, audited, or entitled to act without confirmation. The code map labels orchestration, API, Python, contract, and research boundaries separately. A module can be present without an operational service, and a service can run without being safe for live funds.

When writing about these roles, name the exact module, the inputs it expects, the observable output, and the boundary that remains. Avoid performance claims until the team publishes a reproducible benchmark and defines its data, fees, time range, and review process.