Crypto Signal Board

AI-powered crypto trading signals, verified in a Trusted Execution Environment.

Real-time technical analysis for 25 trading pairs, combined with an LLM macro-risk verdict that runs inside an OpenGradient TEE — every prediction is cryptographically signed and tamper-proof.

Live Demo

---

How It Works

| Step | What happens | Powered by | |------|-------------|------------| | 1 | Fetch 48h of 1-hour OHLC candles | Binance API | | 2 | Compute 6 technical indicators (SMA, EMA, RSI, MACD, Bollinger Bands, Volume) | Local TA engine | | 3 | Run an LLM inside a TEE to assess macro risks (FOMC, CPI, ETF decisions, regulatory news) | OpenGradient TEE via x402 | | 4 | Produce a combined Buy / Sell / Neutral verdict with confidence score | TEE-verified AI | | 5 | Predict 1-hour volatility using an ONNX neural network | OpenGradient Model Hub | | 6 | Display everything in a real-time dashboard | Next.js 16 |

Every AI prediction comes with a cryptographic proof — it cannot be altered after the fact.

---

Why TEE Matters

Traditional signal providers can edit predictions retroactively. With TEE:

• The AI model runs inside a secure enclave — even the server operator cannot see or modify the computation
• Every output is signed and immutable — the result is tied to a verifiable on-chain transaction
• Users get a trust guarantee: what the AI predicted is exactly what you see

---

Tech Stack

| Layer | Technology | |-------|-----------| | Frontend | Next.js 16, React 19, TypeScript, Tailwind CSS 4 | | Market Data | Binance API (25 pairs, 1h + 30m candles) | | AI Inference | OpenGradient TEE (LLM via x402 protocol, Base Sepolia) | | ML Predictions | ONNX model via onnxruntime-node (local inference) | | Blockchain | viem (x402 payment flow for TEE access) | | Cache | ioredis-os (signal cache + bot snapshot reads) | | Deployment | Vercel (Stockholm region) |

---

Features

• 6 Technical Indicators — SMA, EMA, RSI, MACD, Bollinger Bands, Volume analysis with bull/bear/neutral consensus
• TEE-Verified AI Verdict — Buy/Sell/Neutral with confidence %, backed by on-chain proof
• Macro Risk Assessment — AI evaluates upcoming economic events and their crypto market impact
• ML Volatility Forecast — ONNX neural network predicts 1-hour price movement magnitude
• Interactive Price Chart — 48h candlestick chart with real-time data
• 25 Trading Pairs — BTC, ETH, BNB, SOL, XRP, DOGE, ADA, and 18 more
• Auto-Refresh — continuous monitoring with configurable refresh interval
• Signal History — locally stored past signals for comparison
• Redis Cache — /api/signals responses cached via ioredis-os (memory fallback when Redis is down)
• ML Bot Integration — reads latest trade scores from intelligent-trading-bot via Redis (itb:signal:{pair}:latest)
• Mobile Responsive — optimized layout for all screen sizes

---

Architecture

                        Binance API
                            |
                   OHLC candles (1h + 30m)
                            |
              +-------------+-------------+
              |             |             |
        Local TA Engine  OpenGradient   ONNX Model
        (6 indicators)   TEE (LLM)    (volatility)
              |             |             |
              +-------------+-------------+
                            |
                    Next.js API Route
              (parallel fetch + Redis cache)
                            |
              +-------------+-------------+
              |                           |
        ioredis-os cache          itb bot snapshots
        (csb:signals:*)           (itb:signal:*:latest)
                            |
                    Dashboard UI
           (charts, verdicts, indicators)

---

Model Hub

The volatility model runs locally via onnxruntime-node:

| Field | Value | |-------|-------| | Model | og-1hr-volatility-ethusdt v0.06 | | Input | 10 x 30-min OHLC candles — tensor float32 [10, 4] | | Output | Predicted 1-hour volatility % — tensor float32 [1] | | Size | 7 KB | | Source | OpenGradient Model Hub |

---

Getting Started

# Clone
git clone https://github.com/cryptoyasenka/crypto-signal-board.git
cd crypto-signal-board

# Install dependencies
npm install

# Configure environment
cat > .env.local << 'EOF'
APP_WALLET_PRIVATE_KEY=0x...your_base_sepolia_private_key...
NODE_TLS_REJECT_UNAUTHORIZED=0
EOF

# Run development server
npm run dev

# Run full pipeline smoke test (build + API checks)
npm run test:pipeline

# Check Redis connectivity
npm run redis:health

Open http://localhost:3000.

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | APP_WALLET_PRIVATE_KEY | Yes | Base Sepolia wallet private key for x402 TEE payments | | NODE_TLS_REJECT_UNAUTHORIZED | No | Set to 0 for TEE devnet self-signed certs | | REDIS_URL or REDIS_HOST | No | Enable Redis caching (REDIS_ENABLED=false to force off) | | REDIS_KEY_PREFIX | No | Cache key prefix (default csb) | | REDIS_SIGNAL_TTL_SEC | No | Signal cache TTL in seconds (default 60) | | BOT_REDIS_KEY_PREFIX | No | Prefix for bot snapshots from intelligent-trading-bot (default itb) |

Note: You need OPG tokens on Base Sepolia to pay for TEE inference. The wallet key is used to sign x402 payment transactions.

---

Deployment

Deployed on Vercel with automatic deploys from main:

npx vercel --prod

The ONNX model file (7 KB) is included in the repo, so Vercel bundles it automatically.

---

Built with OpenGradient for the OpenGradient hackathon.