一个简单的 AI 辩论小工具:接入 API 后,让 AI-A 与 AI-B 围绕同一议题进行观点生成、矛盾判定与多回合辩论,最后由裁判模型给出总结。
- 支持 AI-A / AI-B / 裁判 三角色独立配置
- 支持 OpenAI 兼容接口与 Claude 接口
- 支持在线拉取模型列表,也支持手动填写模型名称
- 支持可选 Tavily 联网搜索(仅用于首次观点生成,且由 AI 自主决定是否搜索及搜索词)
- 自动注入当前系统时间上下文,减少“今天/今年/最近”时间理解偏差
- 自动判定双方观点是否存在矛盾
- 多回合攻防辩论
- 裁判最终总结
- 一键导出辩论文本
- 打开
index.html - 分别填写 AI-A、AI-B、裁判的 API Key
- 选择或手动填写模型,并输入议题
- (可选)填写 Tavily API Key,用于首次观点生成时联网参考
- 点击“开始”
A lightweight AI debate playground: connect APIs, let AI-A and AI-B generate initial views, have a judge model detect conflicts, run multi-round rebuttals, and produce a final summary.
- Independent configuration for AI-A / AI-B / Judge
- Supports OpenAI-compatible APIs and Claude APIs
- Fetch model lists online, with manual model name input fallback
- Optional Tavily web search (used only for the initial view generation, with AI deciding whether and what to search)
- Injects current system-time context to reduce ambiguity for terms like "today/this year/recent"
- Automatic conflict detection between initial views
- Multi-round debate flow
- Final judge summary
- One-click export of debate transcript
- Open
index.html - Fill API keys for AI-A, AI-B, and Judge
- Select or manually enter models, then enter a topic
- (Optional) Fill Tavily API Key for web references in initial views
- Click "Start"
