Skip to content

Shadowell/HyperTrade

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

177 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

HyperTrade

Production-Grade Agent Runtime for Crypto Trading Research & Execution

License Python FastAPI React TypeScript PostgreSQL Status

中文文档 · English Summary · Documentation · API · User Manual · Developer Guide


Overview

HyperTrade is a self-hosted Agent runtime that connects LLM-powered reasoning with governed access to crypto market data, strategy backtesting, paper trading, and risk-gated Testnet execution. It provides a unified environment for systematic trading research — from natural language exploration to evidence-backed strategy iteration.

HyperTrade Architecture

📖 Detailed Architecture Documentation

Core Capabilities

Domain Capability
Market Intelligence Real-time OKX SWAP data, multi-source indicators, global market regime classification
Strategy Research Backtrader backtests, multi-variant experiments, evidence library with automatic gating
Paper Trading Full lifecycle simulation with pause/resume/close/reset controls
Testnet Execution Approval-gated OKX Testnet orders with risk checks (mainnet blocked in V1)
BitPro Integration MCP adapter for strategy lifecycle, backtest diagnostics, and paper monitoring
Knowledge Systems RAG retrieval over docs/knowledge with pgvector, audited Memory persistence
Governance Policy-enforced tool registry, approval gates, idempotency, deterministic evals

Disclaimer: Nothing in this repository constitutes investment advice. Mainnet live order execution is blocked in V1.


Architecture

┌─────────────────────────────────────────────────────────────────────┐
│  Client Layer                                                        │
│  CLI (ht) · Web Console (/harness) · REST/SSE API · External Agents │
├─────────────────────────────────────────────────────────────────────┤
│  Agent Runtime                                                       │
│  Graph Kernel · Provider Router · Tool Executor                      │
├─────────────────────────────────────────────────────────────────────┤
│  Governance Layer                                                    │
│  ToolRegistry · Risk Policy · Approval Gates · Audit Trace          │
├─────────────────────────────────────────────────────────────────────┤
│  Service Layer                                                       │
│  Market · RAG · Memory · Strategy · Backtest · World Model          │
├─────────────────────────────────────────────────────────────────────┤
│  Data Layer                                                          │
│  PostgreSQL/pgvector · OKX API · BitPro MCP · Alpha Vantage         │
└─────────────────────────────────────────────────────────────────────┘

Technology Stack

Layer Technologies
Runtime Python 3.12+, FastAPI, SQLAlchemy, Alembic
Agent Engine Graph-based kernel, provider routing, policy-enforced tool execution
Frontend React 18, TypeScript 5, Vite, TanStack Query, Recharts
Database PostgreSQL 14+ with pgvector (or SQLite for development)
LLM Providers Vide Coding (opus-4.6), DeepSeek, OpenAI, Codex, OpenRouter, Qwen
Backtesting Backtrader
Infrastructure Docker Compose, Nginx, GitHub Actions CI/CD

Agent Capabilities

Capability Description Status
Natural Language Free-form prompts routed to LLM planner with automatic tool selection Production
Tool Calling Registry-based execution with scope, approval, and idempotency enforcement Production
Market Intelligence OKX SWAP tickers, candles, funding, OI, relative strength indicators Production
Global Market Cross-asset regime classification (equities, volatility, FX, rates) Production
RAG pgvector-backed citation search over knowledge documents Production
Memory Audited observations with tags, confidence scoring, and importance weighting Production
Strategy Research Backtrader backtests, multi-variant experiments, evidence library Production
BitPro Integration MCP adapter for strategy lifecycle, backtest diagnostics, paper monitoring Production
Paper Trading Simulated execution with full lifecycle controls Production
Testnet Execution Approval-gated OKX Testnet orders with risk validation Production
Monitoring & Alerts Read-only monitors for paper strategies, connector health, library freshness Production
Evaluation Suite Deterministic evals for tool choice, RAG, Memory, risk, report quality Production
World Model Portfolio state tracking and defensive action scheduling Experimental

Quick Start

Prerequisites

  • Python 3.12+ with uv package manager
  • Node.js 18+ with pnpm
  • API key for at least one chat provider (Vide Coding, DeepSeek, OpenAI, etc.)
  • PostgreSQL 14+ with pgvector (recommended) or SQLite for development

Setup

git clone git@github.com:Shadowell/HyperTrade.git
cd HyperTrade
cp .env.example .env
# Edit .env with your API keys

Launch

Backend (API server):

uv run uvicorn hypertrade.main:app --app-dir backend/src --host 0.0.0.0 --port 3334

Frontend (Web console):

npm exec --yes pnpm@10 -- -C frontend install
npm exec --yes pnpm@10 -- -C frontend dev

CLI (command-line interface):

uv run ht --local

Access Points

Service URL
Web Console http://localhost:3333/harness
API Docs (Swagger) http://localhost:3334/docs
Health Check http://localhost:3334/api/health

Docker (PostgreSQL + pgvector)

docker compose up -d postgres
export DATABASE_URL="postgresql://hypertrade:hypertrade@localhost:5432/hypertrade"
uv run alembic upgrade head

Usage

Market Research

# Natural language queries
uv run hypertrade --local ask "看下目前市场的热度怎么样"

# Structured commands
/price ETH
/candles BTC 1H 120
/compare ETH SOL BTC
/global-market          # Cross-asset regime snapshot

Strategy Research

/research 研究ETH趋势突破策略     # Create research record
/backtest                         # Run Backtrader backtest
/experiment 不同参数动量策略       # Multi-variant experiment
/strategy library momentum_v1     # Query aggregated evidence

BitPro Integration

# Query backtest performance
查看 BitPro 回测收益大于100%的策略有哪些

# Inspect backtest details
查看 BitPro 回测 result 196 的权益曲线和交易证据

# Monitor live paper strategies
监控 BitPro 所有运行中的模拟盘策略,列出异常和数据缺口

Paper Trading

/paper                   # Status overview
/paper pause BTC         # Pause a strategy
/paper resume            # Resume all
/paper close             # Close positions
/paper reset             # Reset simulation

Testnet Execution (Approval-Gated)

/live intent ETH buy 0.01 reason="API smoke test"
/live intents            # List pending intents
/live approve loi_abc123 # Approve for execution
/live execute loi_abc123 # Submit to OKX Testnet

RAG & Memory

/rag 风控                # Semantic search over knowledge base
/memory search tag:strategy  # Query audited memory

API Access

# Streaming agent run
curl -N -X POST http://localhost:3334/api/agent/runs/stream \
  -H "Content-Type: application/json" \
  -d '{"prompt":"请做行情归纳"}'

# RAG search
curl "http://localhost:3334/api/rag/search?query=风控&limit=5"

# Memory query
curl "http://localhost:3334/api/memory?query=市场&limit=10"

# Market snapshot
curl http://localhost:3334/api/global-market/snapshot

Documentation

Document Description
API Reference Complete REST/SSE API documentation
User Manual Operator guide for all surfaces (CLI, Web, API)
Developer Guide Extending HyperTrade with tools, providers, connectors
Documentation Index Full documentation map
Architecture Docs System design and module documentation (20+ docs)
Product Spec Vision, scope, and roadmap
Knowledge Base Operator guides and best practices
Runbooks Deployment, monitoring, incident response

Chinese translations available for all primary documents.


Development

Quality Gates

./scripts/check.sh       # Full suite: tests, lint, type-check
uv run pytest tests/ -v  # Run specific tests

Code Standards

Python:

uv run ruff format .     # Auto-format
uv run ruff check .      # Lint
uv run mypy backend/src  # Type-check

Frontend:

npm exec --yes pnpm@10 -- -C frontend lint
npm exec --yes pnpm@10 -- -C frontend format

Evaluation Suite

Deterministic evals guard against regressions across tool choice, RAG citations, Memory behavior, risk refusal, BitPro source-of-truth usage, and report quality.

uv run ht --local /evals
curl http://localhost:3334/api/evals/status

Deployment

Automated deployment via GitHub Actions to a self-hosted runner on push to main.

# Manual deployment
ssh hypertrade-server
sudo -u hypertrade /opt/hypertrade/deploy/deploy.sh

# Post-deploy verification
curl -fsS http://localhost:3334/api/health
hypertrade ask "看下ETH行情"

See Runbooks for detailed deployment and monitoring procedures.


Security

  • Secrets: API keys, tokens, and credentials managed via environment variables; never committed
  • BitPro Boundary: All BitPro access through stable MCP/API contracts only
  • Audit Trail: Every tool execution logged with timestamp, payload, and provenance
  • Idempotency: Write operations require idempotency keys
  • Approval Gates: Human-in-the-loop for Testnet orders and destructive operations
  • Testnet First: Mainnet live execution blocked in V1

Roadmap

Current (V2)

Initiative Status
World Model — portfolio state and defensive actions In Progress
Global Market — cross-asset regime classification Production
Vide Coding (opus-4.6) provider Production
Enhanced output formatting Complete

Planned (V3+)

  • Multi-Agent collaboration patterns
  • Custom strategy DSL
  • Multi-exchange support (Binance, Bybit)
  • Advanced risk modeling (VaR, CVaR)
  • Mainnet execution framework (pending governance)

Contributing

  1. Review active sprint scope in docs/contracts/
  2. Keep changes within contract scope
  3. Run ./scripts/check.sh before committing
  4. Commit to main (no feature branches in current workflow)
  5. Push triggers automatic CI/CD deployment

See Developer Guide for detailed contribution guidelines.


Acknowledgments

Built with these excellent projects and services:

OKX API · Backtrader · FastAPI · LangGraph · React · Vite · pgvector · Alembic


About

A crypto trading agent for market research and execution

Topics

Resources

License

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors