Documentation

YieldBoost AI Docs Center

Product context, honest integration notes, user walkthroughs, and developer reference for the current state of YieldBoost AI.

Current Status

0G Mainnet is the active default workspace path, while the judge review can still switch networks on demand.

Runtime Truth

UI-ready deterministic scoring with streamed narrative fallback

Product Fundamentals

Why YieldBoost AI

Why the project exists, which problem it tackles, and what trust model it assumes.

Problem

Idle DeFi capital and fragmented proof trails

Target Users

Retail users, judges, and contributors

Trust Model

Transparent fallbacks over fake certainty

Problem statement

Many DeFi dashboards stop at recommendation cards. They show a higher APY route, but they do not explain how the route was produced, what part is simulated, where proof is stored, or how a reviewer should validate the claim.

YieldBoost AI exists to reduce that trust gap. The interface is opinionated, the optimization flow is guided, and the proof surfaces are always close to the main action paths.

Why the product is shaped as a dashboard first

  • A user should be able to understand the pitch in under a minute.
  • A judge should be able to see action, result, and proof without switching tools repeatedly.
  • A developer should be able to inspect the actual routes that produce the optimization snapshot, streamed reasoning, storage write, and history ledger.

Who the product is for

Audience
What they need
Where to start
Best proof surface
DeFi user
Simple action path and low-friction CTA
Dashboard
Proof modal + latest tx
Hackathon judge
Fast narrative and verifiable artifacts
Docs overview
History and explorer link
Developer
Route-level understanding
Architecture
API & Data Flow
Contributor
Honest current-state map
Roadmap
Troubleshooting + architecture

Trust model

YieldBoost AI is not a custody layer. Wallet access stays in the browser wallet or a manually entered watch-only address. The app reads wallet state, prepares optimization output, and stores proof records.

The product is strongest when it is explicit about uncertainty. Optimization numbers can still come from deterministic or sandbox-assisted logic, but proof storage, history surfacing, and verifier-friendly metadata remain visible either way.

  • Connected wallets can switch networks and broadcast the selected address into the app state.
  • Watch mode allows a valid address to be tracked without an injected wallet session.
  • Proof records are stored in KV when configured, or in the local runtime artifact file when running locally.
  • The app defaults to testnet-first behavior and should be presented as testnet unless the active environment says otherwise.