TradingAgents是一个多智能体大语言模型金融交易框架，已正式开源发布1。该框架通过部署多个由LLM驱动的专业智能体（包括基本面分析师、情绪分析师、新闻分析师、技术分析师、交易员、风险管理团队和投资组合经理）来模拟真实交易公司的运营模式，各个智能体协同评估市场条件并作出交易决策1。
框架支持来自OpenAI、Google、Anthropic、xAI、DeepSeek、Qwen、GLM、MiniMax、Ollama和Azure OpenAI等多个LLM提供商1。最新的v0.4.0版本（2026年8月）新增了look-ahead/point-in-time修复、GPT-5.6和GLM-5.3模型支持1。用户可以通过CLI、Docker等多种方式部署该框架1，并可分析美国、香港、日本、伦敦、印度、加拿大、澳大利亚、中国A股及加密货币等全球多个市场的股票1。
框架提供决策日志持久化和检查点恢复功能，允许用户从失败步骤恢复1。值得注意的是，该框架仅供研究用途，其交易业绩可能因模型选择、温度参数、交易期间等多种因素而变化1。
TradingAgents, a multi-agent large language model financial trading framework, has been officially open-sourced.1 The framework simulates the operational structure of real trading firms by deploying specialized LLM-driven agents, including fundamental analysts, sentiment analysts, technical analysts, and others, to collaboratively assess market conditions and make trading decisions.1
The framework supports multiple LLM providers, encompassing OpenAI, Google, Anthropic, xAI, DeepSeek, Qwen, GLM, MiniMax, Ollama, and Azure OpenAI.1 It comprises seven primary roles: fundamental analyst, sentiment analyst, news analyst, technical analyst, trader, risk management team, and portfolio manager.1 The system is capable of analyzing stocks across global markets, including the United States, Hong Kong, Japan, London, India, Canada, Australia, China's A-shares, and cryptocurrencies.1
Version 0.4.0, released in August 2026, introduced look-ahead and point-in-time fixes alongside support for GPT-5.6 and GLM-5.3 models.1 The framework provides decision log persistence and checkpoint recovery functionality, enabling users to resume operations from failed steps.1 The developers emphasize that the framework is intended for research purposes only, and trading performance may vary significantly depending on model selection, temperature settings, trading period, and other factors.1
评论
还没有评论，欢迎留下第一条。