moon-dev-trading-agents

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Moon Dev's AI Trading Agents System

Moon Dev的AI交易Agent系统

Expert knowledge for working with Moon Dev's experimental AI trading system that orchestrates 48+ specialized AI agents for cryptocurrency trading across Hyperliquid, Solana (BirdEye), Asterdex, and Extended Exchange.

本文档提供了Moon Dev实验性AI交易系统的专业知识，该系统可协调48+个专业AI Agent，在Hyperliquid、Solana（BirdEye）、Asterdex及Extended Exchange等平台进行加密货币交易。

When to Use This Skill

何时使用该Skill

Use this skill when:

Working with Moon Dev's trading agents repository
Need to understand agent architecture and capabilities
Running, modifying, or creating trading agents
Configuring trading system, exchanges, or LLM providers
Debugging trading operations or agent interactions
Understanding backtesting with RBI agent
Setting up new exchanges or strategies

在以下场景使用该Skill：

操作Moon Dev的交易Agent仓库
需要了解Agent架构与能力
运行、修改或创建交易Agent
配置交易系统、交易所或LLM提供商
调试交易操作或Agent交互逻辑
了解如何使用RBI Agent进行回测
搭建新交易所接入或交易策略

Environment Setup Note

环境设置说明

For New Users: This repo uses Python 3.10.9. If using conda, the README shows setting up an environment named

tflow

, but you can name it whatever you want. If you don't use conda, standard pip/venv works fine too.

新用户注意：本仓库使用Python 3.10.9。如果使用conda，README中展示了如何创建名为

tflow

的环境，但你可以自定义环境名称。如果不使用conda，标准的pip/venv也完全适用。

Quick Start Commands

快速启动命令

bash

undefined

bash

undefined

Activate your Python environment (conda, venv, or whatever you use)

激活Python环境（conda、venv或你使用的其他环境管理器）

Example with conda: conda activate tflow

conda示例：conda activate tflow

Example with venv: source venv/bin/activate

venv示例：source venv/bin/activate

Use whatever environment manager you prefer

可使用任意你偏好的环境管理器

Run main orchestrator (controls multiple agents)

运行主协调器（控制多个Agent）

python src/main.py

Run individual agent

运行单个Agent

python src/agents/trading_agent.py python src/agents/risk_agent.py python src/agents/rbi_agent.py

Update requirements after adding packages

添加包后更新依赖

pip freeze > requirements.txt

undefined

pip freeze > requirements.txt

undefined

Core Architecture

核心架构

Directory Structure

目录结构

src/
├── agents/              # 48+ specialized AI agents (<800 lines each)
├── models/              # LLM provider abstraction (ModelFactory)
├── strategies/          # User-defined trading strategies
├── scripts/             # Standalone utility scripts
├── data/                # Agent outputs, memory, analysis results
├── config.py            # Global configuration
├── main.py              # Main orchestrator loop
├── nice_funcs.py        # Core trading utilities (~1,200 lines)
├── nice_funcs_hl.py     # Hyperliquid-specific functions
├── nice_funcs_extended.py # Extended Exchange functions
└── ezbot.py             # Legacy trading controller

src/
├── agents/              # 48+个专业AI Agent（每个文件少于800行）
├── models/              # LLM提供商抽象层（ModelFactory）
├── strategies/          # 用户自定义交易策略
├── scripts/             # 独立实用脚本
├── data/                # Agent输出、内存数据、分析结果
├── config.py            # 全局配置文件
├── main.py              # 主协调器循环
├── nice_funcs.py        # 核心交易工具（约1200行）
├── nice_funcs_hl.py     # Hyperliquid专属功能
├── nice_funcs_extended.py # Extended Exchange专属功能
└── ezbot.py             # 遗留交易控制器

Key Components

关键组件

Agents (src/agents/)

Each agent is standalone executable
Uses ModelFactory for LLM access
Stores outputs in src/data/[agent_name]/
Under 800 lines (split if longer)

LLM Integration (src/models/)

ModelFactory provides unified interface
Supports: Claude, GPT-4, DeepSeek, Groq, Gemini, Ollama
Pattern:
```
ModelFactory.create_model('anthropic')
```

Trading Utilities

```
nice_funcs.py
```
: Core functions (Solana/BirdEye)
```
nice_funcs_hl.py
```
: Hyperliquid exchange
```
nice_funcs_extended.py
```
: Extended Exchange (X10)

Configuration

```
config.py
```
: Trading settings, risk limits, agent behavior
```
.env
```
: API keys and secrets (never expose these)

Agents（src/agents/）

每个Agent均可独立执行
通过ModelFactory调用LLM
输出结果存储于src/data/[agent_name]/目录
单个文件不超过800行，过长则拆分

LLM集成（src/models/）

ModelFactory提供统一调用接口
支持：Claude、GPT-4、DeepSeek、Groq、Gemini、Ollama
使用范式：
```
ModelFactory.create_model('anthropic')
```

交易工具

```
nice_funcs.py
```
：核心功能（Solana/BirdEye）
```
nice_funcs_hl.py
```
：Hyperliquid交易所专属
```
nice_funcs_extended.py
```
：Extended Exchange（X10）专属

配置

```
config.py
```
：交易设置、风险限制、Agent行为配置
```
.env
```
：API密钥与敏感信息（切勿泄露）

Agent Categories

Agent分类

Trading: trading_agent, strategy_agent, risk_agent, copybot_agent

Market Analysis: sentiment_agent, whale_agent, funding_agent, liquidation_agent, chartanalysis_agent

Content: chat_agent, clips_agent, tweet_agent, video_agent, phone_agent

Research: rbi_agent (codes backtests from videos/PDFs), research_agent, websearch_agent

Specialized: sniper_agent, solana_agent, tx_agent, million_agent, polymarket_agent, compliance_agent, swarm_agent

See AGENTS.md for complete list with descriptions.

交易类：trading_agent、strategy_agent、risk_agent、copybot_agent

市场分析类：sentiment_agent、whale_agent、funding_agent、liquidation_agent、chartanalysis_agent

内容类：chat_agent、clips_agent、tweet_agent、video_agent、phone_agent

研究类：rbi_agent（从视频/PDF中生成回测代码）、research_agent、websearch_agent

专业类：sniper_agent、solana_agent、tx_agent、million_agent、polymarket_agent、compliance_agent、swarm_agent

完整Agent列表及说明请查看AGENTS.md。

Common Workflows

常见工作流

1. Run Single Agent

1. 运行单个Agent

bash

undefined

bash

undefined

Activate your environment first

先激活你的环境

python src/agents/[agent_name].py


Each agent is standalone and can run independently.

python src/agents/[agent_name].py


每个Agent均可独立运行。

2. Run Main Orchestrator

2. 运行主协调器

bash

python src/main.py

Runs multiple agents in loop based on

ACTIVE_AGENTS

dict in main.py.

bash

python src/main.py

根据main.py中的

ACTIVE_AGENTS

字典，循环运行多个Agent。

3. Change Exchange

3. 切换交易所

Edit agent file or config:

python

EXCHANGE = "hyperliquid"  # or "birdeye", "extended"

Then import corresponding functions:

python

if EXCHANGE == "hyperliquid":
    from src import nice_funcs_hl as nf
elif EXCHANGE == "extended":
    from src import nice_funcs_extended as nf

编辑Agent文件或配置：

python

EXCHANGE = "hyperliquid"  # 或 "birdeye"、"extended"

然后导入对应功能模块：

python

if EXCHANGE == "hyperliquid":
    from src import nice_funcs_hl as nf
elif EXCHANGE == "extended":
    from src import nice_funcs_extended as nf

4. Switch AI Model

4. 切换AI模型

Edit

src/config.py

python

AI_MODEL = "claude-3-haiku-20240307"  # Fast, cheap

编辑

src/config.py

：

python

AI_MODEL = "claude-3-haiku-20240307"  # 快速、低成本

AI_MODEL = "claude-3-sonnet-20240229" # Balanced

AI_MODEL = "claude-3-sonnet-20240229" # 平衡型

AI_MODEL = "claude-3-opus-20240229" # Most powerful

AI_MODEL = "claude-3-opus-20240229" # 性能最强


Or use ModelFactory per-agent:
```python
from src.models.model_factory import ModelFactory
model = ModelFactory.create_model('deepseek')  # or 'openai', 'groq', etc.
response = model.generate_response(system_prompt, user_content, temperature, max_tokens)


或在单个Agent中使用ModelFactory：
```python
from src.models.model_factory import ModelFactory
model = ModelFactory.create_model('deepseek')  # 或 'openai'、'groq'等
response = model.generate_response(system_prompt, user_content, temperature, max_tokens)

5. Backtest Strategy (RBI Agent)

5. 策略回测（RBI Agent）

python

python src/agents/rbi_agent.py

Provide: YouTube URL, PDF, or trading idea text → DeepSeek-R1 extracts strategy logic → Generates backtesting.py compatible code → Executes backtest, returns metrics

See WORKFLOWS.md for more examples.

python

python src/agents/rbi_agent.py

提供：YouTube链接、PDF文档或交易思路文本 → DeepSeek-R1提取策略逻辑 → 生成兼容

backtesting.py

的代码 → 执行回测并返回指标

更多示例请查看WORKFLOWS.md。

Development Rules

开发规则

CRITICAL Rules

核心规则

Keep files under 800 lines - split into new files if longer
NEVER move files - can create new, but no moving without asking
Use existing environment - don't create new virtual environments, use the one from initial setup
Update requirements.txt after any pip install:
```
pip freeze > requirements.txt
```
Use real data only - never synthetic/fake data
Minimal error handling - user wants to see errors, not over-engineered try/except
Never expose API keys - don't show .env contents

单个文件不超过800行 - 过长则拆分为新文件
切勿移动文件 - 可创建新文件，但移动前需确认
使用现有环境 - 不要创建新虚拟环境，使用初始搭建的环境
安装包后更新requirements.txt：
```
pip freeze > requirements.txt
```
仅使用真实数据 - 禁止使用合成/虚假数据
最小化错误处理 - 用户需要看到错误，而非过度封装的try/except
切勿泄露API密钥 - 不要展示.env文件内容

Agent Development Pattern

Agent开发范式

Creating new agents:

python

undefined

创建新Agent：

python

undefined

1. Use ModelFactory for LLM

1. 使用ModelFactory调用LLM

from src.models.model_factory import ModelFactory model = ModelFactory.create_model('anthropic')

2. Store outputs in src/data/

2. 将输出存储于src/data/

output_dir = "src/data/my_agent/"

3. Make independently executable

3. 支持独立执行

if name == "main": # Standalone logic here

if name == "main": # 独立运行逻辑

4. Follow naming: [purpose]_agent.py

4. 遵循命名规范：[用途]_agent.py

5. Add to config.py if needed

5. 如需全局配置，添加至config.py

undefined

undefined

Backtesting

回测规则

Use
```
backtesting.py
```
library (NOT built-in indicators)
Use
```
pandas_ta
```
or
```
talib
```
for indicators
Sample data:
```
src/data/rbi/BTC-USD-15m.csv
```

使用
```
backtesting.py
```
库（禁止使用内置指标）
使用
```
pandas_ta
```
或
```
talib
```
获取指标
样本数据：
```
src/data/rbi/BTC-USD-15m.csv
```

Configuration Files

配置文件

config.py: Trading settings

```
MONITORED_TOKENS
```
,
```
EXCLUDED_TOKENS
```
Position sizing:
```
usd_size
```
,
```
max_usd_order_size
```
Risk:
```
CASH_PERCENTAGE
```
,
```
MAX_LOSS_USD
```
,
```
MAX_GAIN_USD
```
Agent:
```
SLEEP_BETWEEN_RUNS_MINUTES
```
,
```
ACTIVE_AGENTS
```
AI:
```
AI_MODEL
```
,
```
AI_MAX_TOKENS
```
,
```
AI_TEMPERATURE
```

.env: Secrets (NEVER expose)

Trading APIs:

BIRDEYE_API_KEY

MOONDEV_API_KEY

COINGECKO_API_KEY

AI:

ANTHROPIC_KEY

OPENAI_KEY

DEEPSEEK_KEY

GROQ_API_KEY

GEMINI_KEY

Blockchain:

SOLANA_PRIVATE_KEY

HYPER_LIQUID_ETH_PRIVATE_KEY

RPC_ENDPOINT

Extended:

X10_API_KEY

X10_PRIVATE_KEY

X10_PUBLIC_KEY

X10_VAULT_ID

config.py：交易设置

```
MONITORED_TOKENS
```
、
```
EXCLUDED_TOKENS
```
仓位大小：
```
usd_size
```
、
```
max_usd_order_size
```
风险控制：
```
CASH_PERCENTAGE
```
、
```
MAX_LOSS_USD
```
、
```
MAX_GAIN_USD
```
Agent配置：
```
SLEEP_BETWEEN_RUNS_MINUTES
```
、
```
ACTIVE_AGENTS
```
AI配置：
```
AI_MODEL
```
、
```
AI_MAX_TOKENS
```
、
```
AI_TEMPERATURE
```

.env：敏感信息（切勿泄露）

交易API：

BIRDEYE_API_KEY

、

MOONDEV_API_KEY

、

COINGECKO_API_KEY

AI API：

ANTHROPIC_KEY

、

OPENAI_KEY

、

DEEPSEEK_KEY

、

GROQ_API_KEY

、

GEMINI_KEY

区块链密钥：

SOLANA_PRIVATE_KEY

、

HYPER_LIQUID_ETH_PRIVATE_KEY

、

RPC_ENDPOINT

Extended Exchange配置：

X10_API_KEY

、

X10_PRIVATE_KEY

、

X10_PUBLIC_KEY

、

X10_VAULT_ID

Exchange Support

交易所支持

Hyperliquid (

nice_funcs_hl.py

)

EVM-compatible perpetuals DEX

Functions:

market_buy()

market_sell()

get_position()

close_position()

Leverage up to 50x

BirdEye/Solana (

nice_funcs.py

)

Solana spot token data and trading

Functions:

token_overview()

token_price()

get_ohlcv_data()

Real-time market data for 15,000+ tokens

Extended Exchange (

nice_funcs_extended.py

)

StarkNet-based perpetuals (X10)
Auto symbol conversion (BTC → BTC-USD)
Leverage up to 20x
Functions match Hyperliquid API for compatibility

See docs/hyperliquid.md, docs/extended_exchange.md for exchange-specific guides.

Hyperliquid（

nice_funcs_hl.py

）

兼容EVM的永续合约DEX

功能：

market_buy()

、

market_sell()

、

get_position()

、

close_position()

最高支持50倍杠杆

BirdEye/Solana（

nice_funcs.py

）

Solana现货代币数据与交易

功能：

token_overview()

、

token_price()

、

get_ohlcv_data()

支持15000+代币的实时市场数据

Extended Exchange（

nice_funcs_extended.py

）

基于StarkNet的永续合约平台（X10）
自动币种转换（BTC → BTC-USD）
最高支持20倍杠杆
功能与Hyperliquid API兼容

交易所专属指南请查看docs/hyperliquid.md、docs/extended_exchange.md。

Data Flow Pattern

数据流范式

Config/Input → Agent Init → API Data Fetch → Data Parsing →
LLM Analysis (via ModelFactory) → Decision Output →
Result Storage (CSV/JSON in src/data/) → Optional Trade Execution

配置/输入 → Agent初始化 → API数据获取 → 数据解析 →
LLM分析（通过ModelFactory） → 决策输出 →
结果存储（src/data/下的CSV/JSON文件） → 可选交易执行

Common Tasks

常见任务

Add new package:

bash

undefined

安装新包：

bash

undefined

Make sure your environment is activated first

先激活你的环境

pip install package-name pip freeze > requirements.txt


**Read market data:**
```python
from src.nice_funcs import token_overview, get_ohlcv_data, token_price

overview = token_overview(token_address)
ohlcv = get_ohlcv_data(token_address, timeframe='1H', days_back=3)
price = token_price(token_address)

Execute trade (Hyperliquid):

python

from src import nice_funcs_hl as nf
nf.market_buy("BTC", usd_amount=100, leverage=10)
position = nf.get_position("BTC")
nf.close_position("BTC")

Execute trade (Extended):

python

from src import nice_funcs_extended as nf
nf.market_buy("BTC", usd_amount=100, leverage=15)
position = nf.get_position("BTC")
nf.close_position("BTC")

pip install package-name pip freeze > requirements.txt


**读取市场数据：**
```python
from src.nice_funcs import token_overview, get_ohlcv_data, token_price

overview = token_overview(token_address)
ohlcv = get_ohlcv_data(token_address, timeframe='1H', days_back=3)
price = token_price(token_address)

执行交易（Hyperliquid）：

python

from src import nice_funcs_hl as nf
nf.market_buy("BTC", usd_amount=100, leverage=10)
position = nf.get_position("BTC")
nf.close_position("BTC")

执行交易（Extended Exchange）：

python

from src import nice_funcs_extended as nf
nf.market_buy("BTC", usd_amount=100, leverage=15)
position = nf.get_position("BTC")
nf.close_position("BTC")

Git Operations

Git操作

Current branch: main Main branch for PRs: main

Recent commits:

dc55e90: websearch agent
921ead6: websearch_agent launched and rbi agent updated
6bb55c2: backtest dash

Modified files (current):

.env_example
src/agents/swarm_agent.py
src/agents/trading_agent.py
src/data/ohlcv_collector.py

当前分支：main PR目标分支：main

近期提交：

dc55e90: websearch agent
921ead6: websearch_agent上线并更新rbi agent
6bb55c2: 回测仪表盘

当前修改文件：

.env_example
src/agents/swarm_agent.py
src/agents/trading_agent.py
src/data/ohlcv_collector.py

Documentation

文档

Main docs (docs/):

```
CLAUDE.md
```
: Project overview and development guidelines
```
hyperliquid.md
```
,
```
hyperliquid_setup.md
```
: Hyperliquid exchange
```
extended_exchange.md
```
: Extended Exchange (X10) setup
```
rbi_agent.md
```
: Research-Based Inference agent
```
websearch_agent.md
```
: Web search capabilities
```
swarm_agent.md
```
: Multi-agent coordination
```
[agent_name].md
```
: Individual agent docs

README files:

Root
```
README.md
```
: Project overview
```
src/models/README.md
```
: LLM provider guide

主文档（docs/目录）：

```
CLAUDE.md
```
：项目概述与开发指南
```
hyperliquid.md
```
、
```
hyperliquid_setup.md
```
：Hyperliquid交易所指南
```
extended_exchange.md
```
：Extended Exchange（X10）搭建指南
```
rbi_agent.md
```
：Research-Based Inference Agent说明
```
websearch_agent.md
```
：网页搜索功能说明
```
swarm_agent.md
```
：多Agent协同说明
```
[agent_name].md
```
：单个Agent的专属文档

README文件：

根目录
```
README.md
```
：项目概述
```
src/models/README.md
```
：LLM提供商使用指南

Risk Management

风险管理

Risk Agent runs FIRST before any trading decisions
Circuit breakers:
```
MAX_LOSS_USD
```
,
```
MINIMUM_BALANCE_USD
```
AI confirmation for position-closing (configurable)
Default loop: every 15 minutes (
```
SLEEP_BETWEEN_RUNS_MINUTES
```
)

Risk Agent会在所有交易决策前运行
熔断机制：
```
MAX_LOSS_USD
```
、
```
MINIMUM_BALANCE_USD
```
仓位平仓需AI确认（可配置）
默认循环间隔：15分钟（
```
SLEEP_BETWEEN_RUNS_MINUTES
```
）

Philosophy

项目理念

This is an experimental, educational project:

No guarantees of profitability
Open source and free
YouTube-driven development
Community-supported via Discord
No official token (avoid scams)

Goal: Democratize AI agent development through practical trading examples.

本项目为实验性、教育性项目：

不保证盈利
开源免费
基于YouTube内容驱动开发
由Discord社区提供支持
无官方代币（警惕诈骗）

目标：通过实用的交易示例，让AI Agent开发平民化。

Additional Resources

额外资源

For complete agent list, see AGENTS.md For workflow examples, see WORKFLOWS.md For architecture details, see ARCHITECTURE.md

Built with 🌙 by Moon Dev

"Never over-engineer, always ship real trading systems."

完整Agent列表请查看AGENTS.md 工作流示例请查看WORKFLOWS.md 架构细节请查看ARCHITECTURE.md

由Moon Dev 🌙 打造

“切勿过度设计，始终交付可实际运行的交易系统。”