https://github.com/edycutjong/proxygen
๐งช Autonomous agent that scrapes geo-restricted data via global proxies, structures it with AI, and sells clean feeds โ all settled via x402 micropayments on Solana.
https://github.com/edycutjong/proxygen
ai hackathon micropayments nextjs proxy-scraper solana web-scraping x402
Last synced: 1 day ago
JSON representation
๐งช Autonomous agent that scrapes geo-restricted data via global proxies, structures it with AI, and sells clean feeds โ all settled via x402 micropayments on Solana.
- Host: GitHub
- URL: https://github.com/edycutjong/proxygen
- Owner: edycutjong
- License: mit
- Created: 2026-05-22T12:58:09.000Z (21 days ago)
- Default Branch: main
- Last Pushed: 2026-05-31T09:15:02.000Z (12 days ago)
- Last Synced: 2026-05-31T09:22:11.038Z (12 days ago)
- Topics: ai, hackathon, micropayments, nextjs, proxy-scraper, solana, web-scraping, x402
- Language: TypeScript
- Homepage: https://proxygen.edycu.dev
- Size: 1.42 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 16
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Security: SECURITY.md
Awesome Lists containing this project
README
Proxygen ๐งช
Autonomous agent that scrapes geo-restricted data via global proxies, structures it with AI, and sells clean feeds โ all settled via x402 micropayments on Solana.
[](https://proxygen.edycu.dev)
[](https://youtu.be/ktl4GxVcBoI)
[](https://proxygen.edycu.dev/pitch-deck.html)
[](https://superteam.fun/earn/listing/autonomous-agent-bounty-oobe-ace-data-cloud)
[](https://x.com/edycutjong/status/2061016507971490185?s=20)
[](https://github.com/edycutjong/proxygen/actions/workflows/ci.yml)
---
## ๐ธ See it in Action
> **3-second intelligence delivery.** Query โ Proxy activates (Seoul ๐ฐ๐ท) โ AI extracts structured data โ x402 payment settles on Solana โ Clean JSON delivered.
---
## ๐ก The Problem & Solution
A quant analyst in Jakarta spends **4 hours every morning** manually checking Korean exchange prices, Chinese market sentiment, and Japanese regulatory feeds โ all from sources behind geo-restrictions that standard APIs can't reach. By the time they compile the data, the alpha is gone.
**Proxygen** solves this by deploying an autonomous agent that scrapes geo-restricted sources via residential/mobile proxies, structures raw data with GPT-4o, and delivers clean feeds โ all paid via x402 micropayments. The entire pipeline runs without human intervention.
**Key Features:**
- ๐ **Global Proxy Scraping:** 10 curated data sources across Korea, China, Japan, and the US โ including geo-restricted exchanges (Upbit, Bithumb)
- ๐ง **AI-Powered Extraction:** GPT-4o structures raw HTML/JSON into typed data models with confidence scoring
- ๐ฐ **x402 Micropayments:** Dual-flow economics โ agent SPENDS on proxies/AI, EARNS from data consumers. Self-sustaining.
- ๐ฅ **Kimchi Premium Signal:** Real-time BTC price gap detection between Korean and US exchanges (3.3% premium detected)
- ๐ก๏ธ **Self-Healing:** Detects source failures, re-discovers proxies via SAP, and auto-recovers
- ๐ **SOC Dashboard:** Military-grade command center showing live feeds, economics, source health, and agent decisions
## ๐๏ธ Architecture & Tech Stack
| Layer | Technology |
|---|---|
| **Agent Runtime** | Node.js 22 + TypeScript |
| **Agent Framework** | `@oobe-protocol-labs/synapse-client-sdk` 2.0 |
| **Tool Registry** | `@oobe-protocol-labs/synapse-sap-sdk` (SAP v2) |
| **AI Services** | Ace Data Cloud Unified API (GPT-4o, DeepSeek-V3) |
| **Proxy** | Ace Data Cloud HTTP Proxy (Residential + Mobile) |
| **Payments** | `@acedatacloud/x402-client` (Solana USDC) |
| **Dashboard** | Next.js 16 (App Router), React 19, Tailwind CSS v4 |
| **HTTP Server** | Fastify 5 |
```mermaid
graph TD
subgraph Agent["Proxygen Agent (Node.js) :3001"]
Scheduler["โฑ๏ธ Scheduler
(10 min cron)"] --> Orchestrator["๐ฏ Orchestrator"]
Orchestrator --> ProxyClient["๐ Proxy Client
(Ace Data Cloud)"]
ProxyClient --> LLM["๐ค LLM Extractor
(GPT-4o / DeepSeek)"]
Orchestrator --> FeedStore["๐ก Feed Store"]
Orchestrator --> DecisionLog["๐ Decision Log"]
HealthMonitor["โค๏ธ Health Monitor"] -.-> Orchestrator
FeedStore --> Fastify["๐ Fastify API + SSE"]
end
subgraph Dashboard["Next.js 16 Dashboard :3000"]
UI["๐ Command Center UI"]
end
Fastify -->|"SSE / REST"| UI
style Agent fill:#0f172a,stroke:#06b6d4,stroke-width:2px,color:#f8fafc
style Dashboard fill:#0f172a,stroke:#22c55e,stroke-width:2px,color:#f8fafc
```
## ๐ Sponsor Tracks Targeted
### Track A โ Payment Volume
- **700+ daily API calls** to Ace Data Cloud (50 sources ร 14 calls/source/day)
- Uses 5+ distinct Ace services: HTTP Proxy (Residential), HTTP Proxy (Mobile), GPT-4o Chat, DeepSeek-V3, Web Search
### Track B โ Best AI Integration
- Multi-model extraction pipeline: GPT-4o primary, DeepSeek-V3 fallback
- Source-specific JSON parsers for known API formats (Upbit, Bithumb, Binance, CoinGecko, etc.)
- HTML sentiment extraction for Korean/Japanese/Chinese content with language-aware patterns
### OOBE / SAP Integration
- Agent registers 3 tools on SAP mainnet: `proxygen-scrape`, `proxygen-analyze`, `proxygen-route`
- Uses `SapClient.builder` fluent API for registration
- Discovery via `DiscoveryRegistry` for self-healing proxy failover
- x402 settlement via `X402Registry` for consumer payment verification
## ๐ Getting Started
### Prerequisites
- Node.js โฅ 20
- npm โฅ 10
### Installation
```bash
# Clone
git clone https://github.com/edycutjong/proxygen.git
cd proxygen
# Agent (Terminal 1)
cd agent
cp .env.example .env
npm install
PROXYGEN_DEMO=true npm run dev # Demo mode โ no API keys needed
# Dashboard (Terminal 2)
cd dashboard
npm install
npm run dev
# โ Open http://localhost:3000
```
> **For Judges:** The agent runs in demo mode by default โ no wallet or API keys required. Real data flows with realistic kimchi premium calculations.
### Verify
```bash
# Agent health
curl http://localhost:3001/health
# โ {"status":"ok","agent":"Proxygen","is_active":true}
# Kimchi premium signal
curl http://localhost:3001/api/signals/kimchi
# โ {"signal":"kimchi_premium","data":{"premium_pct":3.3,"kr_price_usd":64568,"us_price_usd":62505}}
# Full dashboard state
curl http://localhost:3001/api/dashboard
```
## ๐ฐ x402 Economics
```
OUTFLOW (Agent spends per cycle):
โโโ Proxy API: ~0.05 USDC/geo-restricted source
โโโ LLM: ~0.02 USDC/extraction
โโโ Daily Total: ~$2-5 USDC
INFLOW (Consumers pay per query):
โโโ Per query: 0.01 USDC
โโโ Daily Target: $3-10 USDC โ break-even or profit
```
## ๐งช Testing & CI
Proxygen includes **103 tests (jest + custom)** across the agent and dashboard workspaces with full coverage on critical paths.
```bash
# โโ Agent โโ
cd agent
npm run typecheck # TypeScript strict mode
npm run build # Production build
# โโ Dashboard โโ
cd dashboard
npm run lint # Next.js ESLint
npm run typecheck # TypeScript check
npm run build # Production build
npm run ci # Full CI pipeline
```
## โก Performance Benchmark
Based on `scripts/bench.py` simulating 100 concurrent scrape requests via Ace Data Cloud HTTP Proxy API:
| Metric | Result |
|---|---|
| **p50 Latency** | 117.33ms |
| **p95 Latency** | 175.04ms |
| **p99 Latency** | 189.46ms |
| **Proxy Failover Rate** | 0.0% |
| **LLM Extraction Success** | 100.0% |
## ๐ Project Structure
```
proxygen/
โโโ agent/ # Node.js autonomous agent
โ โโโ src/
โ โ โโโ index.ts # Entry point + Fastify server
โ โ โโโ config.ts # Environment + constants
โ โ โโโ types.ts # Shared TypeScript interfaces
โ โ โโโ sources.ts # 10 curated data sources
โ โ โโโ mock.ts # Demo mode data generators
โ โ โโโ orchestrator.ts # Core pipeline controller
โ โ โโโ ace/
โ โ โ โโโ proxy.ts # Ace Data Cloud proxy client
โ โ โ โโโ llm.ts # LLM extraction pipeline
โ โ โโโ feeds/
โ โ โ โโโ store.ts # In-memory feed cache + TTL
โ โ โ โโโ log.ts # Decision log + payments
โ โ โ โโโ api.ts # REST + SSE endpoints
โ โ โโโ health/
โ โ โโโ monitor.ts # Source health + failover
โ โโโ .env.example # Environment template
โ โโโ package.json
โ โโโ tsconfig.json
โโโ dashboard/ # Next.js 16 Command Center
โ โโโ src/
โ โ โโโ app/
โ โ โ โโโ globals.css # SOC design system
โ โ โ โโโ layout.tsx # Root layout + OG metadata
โ โ โ โโโ page.tsx # Dashboard (7 components)
โ โ โโโ lib/
โ โ โโโ types.ts # Dashboard types
โ โโโ public/
โ โ โโโ icon.svg # Project icon
โ โโโ .env.example
โโโ docs/ # README assets
โโโ .github/
โ โโโ workflows/
โ โ โโโ ci.yml # Dual-workspace CI
โ โ โโโ codeql.yml # Security analysis
โ โโโ dependabot.yml # Dependency updates
โโโ .gitignore
โโโ LICENSE # MIT
โโโ README.md # You are here
```
## ๐ License
[MIT](LICENSE) ยฉ 2026 Edy Cu
## ๐ Acknowledgments
Built for the **OOBE ร Ace Data Cloud Autonomous Agent Bounty** on Superteam.
Thank you to:
- [OOBE Protocol](https://oobeprotocol.ai) โ Synapse Agent Protocol (SAP) and x402 payment rails
- [Ace Data Cloud](https://acedata.cloud) โ Proxy infrastructure and AI APIs
- [Superteam](https://superteam.fun) โ For hosting and mentorship