# 2877 Arena API 3dbfee82 > 多Agent策略竞技场 API 文档 - Skill: `tools-only/2877-arena-api-3dbfee82` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add tools-only/2877-arena-api-3dbfee82` - Raw SKILL.md: https://api.skillmd.com/api/skills/tools-only/2877-arena-api-3dbfee82/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tools-only (https://skillmd.com/u/tools-only) - Updated: 2026-09-22 - Page: https://skillmd.com/skills/tools-only/2877-arena-api-3dbfee82 --- # 多Agent策略竞技场 API 文档 ## 概述 多Agent策略竞技场API提供了完整的竞技场管理、策略竞争、Agent讨论和评估功能接口。 ## 基础URL ``` /api/arena ``` ## 认证 所有API端点都需要认证。请在请求头中包含有效的认证token。 --- ## 竞技场管理 ### 创建竞技场 **POST** `/api/arena/create` 创建一个新的多Agent策略竞技场。 **请求体** ```json { "name": "我的竞技场", "description": "测试竞技场描述", "agent_count": 5, "symbols": ["000001.SZ", "600000.SH"], "discussion_max_rounds": 3, "initial_capital": 100000.0, "backtest_start_date": "2024-01-01", "backtest_end_date": "2024-12-31", "simulated_duration_days": 30 } ``` **参数说明** | 参数 | 类型 | 必填 | 说明 | |------|------|------|------| | name | string | 是 | 竞技场名称 | | description | string | 否 | 竞技场描述 | | agent_count | int | 否 | Agent数量 (3-10)，默认5 | | symbols | array | 否 | 股票代码列表 | | discussion_max_rounds | int | 否 | 最大讨论轮数 (1-10)，默认3 | | initial_capital | float | 否 | 初始资金，默认100000 | | backtest_start_date | string | 否 | 回测开始日期 | | backtest_end_date | string | 否 | 回测结束日期 | | simulated_duration_days | int | 否 | 模拟交易天数 (1-365)，默认30 | **响应** ```json { "id": "arena_abc123", "name": "我的竞技场", "state": "created", "active_strategies": 0, "total_strategies": 0, "eliminated_strategies": 0, "discussion_rounds": 0, "last_evaluation": null, "duration_seconds": 0, "error_count": 0, "last_error": null } ``` --- ### 获取竞技场状态 **GET** `/api/arena/{arena_id}/status` 获取指定竞技场的当前状态。 **响应** ```json { "id": "arena_abc123", "name": "我的竞技场", "state": "discussing", "active_strategies": 5, "total_strategies": 5, "eliminated_strategies": 0, "discussion_rounds": 2, "last_evaluation": "2024-01-15T10:30:00", "duration_seconds": 3600.5, "error_count": 0, "last_error": null } ``` **状态说明** | 状态 | 说明 | |------|------| | created | 已创建，等待启动 | | initializing | 初始化中 | | discussing | 讨论进行中 | | backtesting | 回测进行中 | | simulating | 模拟交易中 | | evaluating | 评估中 | | paused | 已暂停 | | completed | 已完成 | | failed | 失败 | --- ### 启动竞技场 **POST** `/api/arena/{arena_id}/start` 启动竞技场竞争流程。 **响应** ```json { "status": "started", "arena_id": "arena_abc123" } ``` --- ### 暂停竞技场 **POST** `/api/arena/{arena_id}/pause` 暂停竞技场竞争流程。 --- ### 恢复竞技场 **POST** `/api/arena/{arena_id}/resume` 恢复已暂停的竞技场。 --- ### 删除竞技场 **DELETE** `/api/arena/{arena_id}` 删除指定竞技场及其所有数据。 --- ### 列出所有竞技场 **GET** `/api/arena/list` **查询参数** | 参数 | 类型 | 必填 | 说明 | |------|------|------|------| | state | string | 否 | 按状态过滤 | | limit | int | 否 | 返回数量限制 (1-100)，默认50 | **响应** ```json { "total": 3, "arenas": [ { "id": "arena_abc123", "name": "竞技场1", "state": "discussing", "active_strategies": 5 } ] } ``` --- ## 思考流 ### SSE实时思考流 **GET** `/api/arena/{arena_id}/thinking-stream` Server-Sent Events端点，实时获取Agent思考消息。 **查询参数** | 参数 | 类型 | 必填 | 说明 | |------|------|------|------| | round_id | string | 否 | 按讨论轮次过滤 | **事件格式** ``` data: { "id": "msg_123", "arena_id": "arena_abc123", "agent_id": "agent_1", "agent_role": "strategy_generator", "round_id": "round_1", "message_type": "thinking", "content": "分析市场数据...", "metadata": {}, "timestamp": "2024-01-15T10:30:00" } ``` **消息类型** | 类型 | 说明 | |------|------| | thinking | 思考过程 | | argument | 论证观点 | | conclusion | 结论 | | intervention | 人工干预 | | error | 错误 | --- ### 获取讨论历史 **GET** `/api/arena/{arena_id}/discussions` **响应** ```json { "total": 5, "discussions": [ { "id": "round_1", "arena_id": "arena_abc123", "round_number": 1, "mode": "debate", "participants": ["agent_1", "agent_2"], "conclusions": { "agent_1": "建议采用均线策略", "agent_2": "建议关注风险控制" }, "started_at": "2024-01-15T10:00:00", "completed_at": "2024-01-15T10:30:00", "duration_seconds": 1800 } ] } ``` --- ## 策略管理 ### 获取策略列表 **GET** `/api/arena/{arena_id}/strategies` **查询参数** | 参数 | 类型 | 必填 | 说明 | |------|------|------|------| | active_only | boolean | 否 | 只返回活跃策略，默认false | --- ### 获取策略详情 **GET** `/api/arena/{arena_id}/strategies/{strategy_id}` **响应** ```json { "id": "strategy_1", "name": "均线交叉策略", "description": "基于MA5和MA20的交叉信号", "agent_id": "agent_1", "agent_role": "strategy_generator", "stage": "simulated", "is_active": true, "current_score": 75.5, "current_rank": 2, "logic": "当MA5上穿MA20时买入...", "rules": { "buy_condition": "MA5 > MA20", "sell_condition": "MA5 < MA20", "stop_loss": 0.05, "take_profit": 0.10 } } ``` --- ### 获取排行榜 **GET** `/api/arena/{arena_id}/leaderboard` **响应** ```json { "total": 5, "leaderboard": [ { "rank": 1, "strategy_id": "strategy_1", "name": "均线交叉策略", "agent_id": "agent_1", "agent_role": "strategy_generator", "score": 85.5, "stage": "simulated" } ] } ``` --- ### 获取策略评分明细 **GET** `/api/arena/{arena_id}/strategies/{strategy_id}/score-breakdown` **响应** ```json { "strategy_id": "strategy_1", "total_score": 75.5, "breakdown": { "profitability": 80.0, "risk_control": 70.0, "stability": 75.0, "adaptability": 72.0 }, "weights": { "profitability": 0.30, "risk_control": 0.30, "stability": 0.20, "adaptability": 0.20 } } ``` --- ## 评估与淘汰 ### 触发评估 **POST** `/api/arena/{arena_id}/evaluate` 手动触发周期评估。 **请求体** ```json { "period": "weekly" } ``` | period值 | 说明 | 淘汰率 | |----------|------|--------| | daily | 日评 | 不淘汰 | | weekly | 周评 | 末位20% | | monthly | 月评 | 末位10% | --- ### 获取淘汰历史 **GET** `/api/arena/{arena_id}/elimination-history` **响应** ```json { "total": 3, "events": [ { "id": "elim_strategy_3", "type": "elimination", "timestamp": "2024-01-15T10:30:00", "period": "weekly", "strategy_name": "高频交易策略", "strategy_id": "strategy_3", "score": 45.2, "reason": "评分低于淘汰阈值" }, { "id": "eval_123", "type": "evaluation", "timestamp": "2024-01-15T10:00:00", "period": "weekly", "total_strategies": 5, "eliminated_count": 1 } ] } ``` --- ## 讨论控制 ### 触发讨论 **POST** `/api/arena/{arena_id}/discussion/start` **请求体** ```json { "mode": "debate" } ``` | mode值 | 说明 | |--------|------| | debate | 辩论模式 - Agent之间相互辩论策略优劣 | | collaboration | 协作模式 - Agent协作优化策略 | | review | 评审模式 - Agent评审其他Agent的策略 | --- ### 获取当前讨论状态 **GET** `/api/arena/{arena_id}/discussion/current` --- ### 人工干预 **POST** `/api/arena/{arena_id}/discussion/intervention` 允许人工干预竞技场运行。 **请求体** ```json { "action": "inject_message", "message": "请关注近期市场波动风险", "reason": "市场环境变化" } ``` **可用操作** | action值 | 说明 | 必需参数 | |----------|------|----------| | inject_message | 注入消息到讨论 | message | | adjust_score | 调整策略评分 | target_strategy_id, score_adjustment (-50 to +50) | | eliminate_strategy | 强制淘汰策略 | target_strategy_id | | add_strategy | 添加新策略 | new_strategy_config | **示例 - 调整评分** ```json { "action": "adjust_score", "target_strategy_id": "strategy_1", "score_adjustment": 10, "reason": "手动奖励优秀表现" } ``` **示例 - 强制淘汰** ```json { "action": "eliminate_strategy", "target_strategy_id": "strategy_3", "reason": "策略逻辑存在严重问题" } ``` --- ## 错误响应 所有API在发生错误时返回以下格式： ```json { "detail": "错误描述信息" } ``` **常见HTTP状态码** | 状态码 | 说明 | |--------|------| | 200 | 成功 | | 400 | 请求参数错误 | | 404 | 资源不存在 | | 500 | 服务器内部错误 | --- ## WebSocket支持 思考流同时支持WebSocket连接（开发中）： ``` ws://localhost:8000/ws/arena/{arena_id}/thinking ```