Oracle-aware treasury risk dashboard for multichain DAOs, built on Pyth price feeds.
Guards turns a treasury's risk mandate into a policy engine: it reads oracle data, measures drawdown against absolute fiat floors, and moves the treasury through a risk ladder of bounded, auditable protective swaps.
Live demo (simulated data)
Guards was built for Pythathon, the Pyth Network hackathon held in Buenos Aires in March 2026.
Built by Nico Fernandez (@f0x1777) and Kevin Anrique (@kevan1).
About this repository. This is the frontend-only showcase of Guards: the Next.js operator dashboard and landing page, extracted so it runs on its own. The full system (shared policy engine packages, backend collector / risk engine / keeper, Cardano off-chain tooling and Aiken contract scaffold) lives in a private repository.
Everything in this UI runs on simulated data. Balances, executions, audit events and the replay datasets are bundled demo fixtures. No wallet, backend or API key is required, and no transactions are ever sent.
DAO treasuries usually write their mandates as percentages, but the real failure mode is breaching an absolute, fiat-denominated floor. When markets move fast, coordinating a multisig by hand is too slow to protect value. Most treasury tooling either tracks prices passively or needs a human at every step.
Guards turns those rules into executable policy:
- Ingest oracle snapshots from Pyth.
- Evaluate treasury health using drawdown, liquid value, oracle confidence and freshness.
- Authorize a bounded execution intent that carries the oracle evidence.
- Execute an approved route from a capped execution wallet (never from the governance multisig).
- Record an audit trail with the intent, the result and the oracle context.
Open the landing page, then Open Demo to reach the operator dashboard (/dashboard).
| Section | What you see |
|---|---|
| Overview | Company/vault profile, liquid value after haircut, stable ratio, drawdown vs EMA price, oracle freshness and the current Pyth feed snapshot (price, emaPrice, confidence). |
| Treasury | Risk bucket and stable reserve accounts with balances, fiat value and weights. |
| Policy Builder | A browser-side vault bootstrap lab: pick a company profile, reference asset and custody mode, tune thresholds and floors, and run scenarios against the policy engine. |
| Runtime | Switch between mock replay datasets and a preprod snapshot view. |
| Risk Ladder | The risk ladder plus a mock replay lab: 7 days of 15-minute price points replayed through the strategy engine (Guards ladder, Floor defender, XAU target) over datasets such as ADA treasury base, ADA flash crash, stable depeg and XAU rotation, with a backtest summary and per-step explanations. |
| History / Swap / Audit Trail | Simulated protective swaps (de-risk, re-entry), a swap panel, and audit cards for each execution. |
The dashboard opens with a "Cardano preprod only" notice carried over from the original product: mainnet execution was intentionally disabled in the hackathon build.
| Stage | Trigger | Action |
|---|---|---|
| Normal | Drawdown below the watch threshold | Hold allocation, monitor feeds |
| Watch | Drawdown > watchDrawdownBps |
Increase monitoring, prepare routes |
| Partial de-risk | Drawdown > partialDrawdownBps |
Swap part of the risk asset into the approved stable |
| Full stable exit | Drawdown > fullExitDrawdownBps or floor breached |
Exit the remaining risk bucket into stable |
| Frozen | Oracle stale or confidence too wide | Block execution until data quality recovers |
| Auto re-entry | Recovery past reentryDrawdownBps and cooldown elapsed |
Restore risk allocation gradually |
The engine measures both percentage drawdown and absolute protected value:
liquid_value = amount × pyth_price × (1 − haircut_bps / 10000)
So it reacts not only to "the asset dropped X%", but also to "the treasury fell below the fiat floor we promised to defend".
In the Guards design, Pyth is a core dependency of the policy engine, not a cosmetic price ticker:
| Pyth signal | How Guards uses it |
|---|---|
price |
Spot valuation for liquid-value calculations |
emaPrice |
Baseline for drawdown measurement |
confidence |
A widening confidence interval can freeze execution |
| freshness (update timestamp) | Stale data blocks execution |
| snapshot IDs | Carried into the audit trail for verification |
In this frontend repository specifically:
- The dashboard, replay lab and policy views run on bundled demo snapshots shaped like Pyth updates (
price,emaPrice,confidence,publisherCount, update time). - Two optional server routes,
/api/oracle/quotes(ADA, XAU, BTC, SOL, EUR vs USD) and/api/oracle/primary(ADA/USD by default), use the@pythnetwork/pyth-lazer-sdkto fetch the latest Pyth Lazer prices only if you setPYTH_API_KEY. Without it they return503and the UI stays on demo data (the top bar showsPyth: demo fallback).
The signed-update collector, the Cardano witness wiring and the keeper that acted on these signals are part of the private repository and are not included here.
| Chain | Status |
|---|---|
| Cardano | MVP in progress: policy engine, PolicyVault simulator, Aiken scaffold, DexHunter / Minswap execution path in progress |
| Solana (SVM) | Connector scaffold, Pyth-native target |
| Ethereum (EVM) | Connector scaffold, Pyth EVM target |
The policy engine is shared across chains; execution stays local to each connected treasury.
- Next.js 16 (App Router, Turbopack) and React 19
- TypeScript 5
- Tailwind CSS 4
- Framer Motion, lucide-react, hls.js (landing background video)
@pythnetwork/pyth-lazer-sdk(optional live quotes)- Vitest for the
lib/unit tests
Requires Node.js 24 or newer (see .nvmrc and engines).
git clone https://github.com/kevan1/guards-ui.git
cd guards-ui
npm install
npm run dev # http://localhost:3000Other scripts:
npm run build # production build
npm start # serve the production build
npm run typecheck # tsc --noEmit
npm test # vitest (policy engine / replay / price helpers)No environment variables are needed. To try the live quote routes, copy .env.example to .env.local and set PYTH_API_KEY to a Pyth Lazer access token. It is read only on the server and is never sent to the browser.
The repository root is the Next.js app, so importing it needs no special configuration:
| Setting | Value |
|---|---|
| Framework preset | Next.js |
| Root directory | ./ (repository root) |
| Install command | npm ci (from vercel.json) |
| Build command | npm run build (from vercel.json) |
| Output directory | default (.next) |
| Node.js version | 24.x |
| Environment variables | none required; PYTH_API_KEY optional |
app/ Next.js routes: landing (/), dashboard (/dashboard), api/oracle/*
components/ Dashboard panels and landing sections
lib/ Demo data, policy/vault lab, mock backtest engine, types
lib/core-types.ts Domain types vendored from the private core package
public/ Logos and images
docs/ Screenshots
The original project is "All rights reserved" and this repository does not include an open-source license. It is published as a hackathon showcase. Contact the authors for licensing questions.