{
  "user_id": "swarm_project",
  "version": 1,
  "entries": [
    {
      "entry_id": "e9ebd05737a4",
      "user_message": "虫群项目什么时候开始的",
      "ai_response": "2026年5月9日，基于Hermes智能体框架开始开发虫群项目",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.668591",
      "created_at": 1778432517.6686053,
      "access_count": 0,
      "keywords": [
        "虫群项目什么时候开始的"
      ]
    },
    {
      "entry_id": "23b83fa4041e",
      "user_message": "虫群的架构是什么",
      "ai_response": "五层虫族体系：虫皇(Overmind)→虫巢(Hive)→脑虫(Cerebrate)→虫后(Queen)→子虫(Drone)",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.669479",
      "created_at": 1778432517.669485,
      "access_count": 0,
      "keywords": [
        "虫群的架构是什么"
      ]
    },
    {
      "entry_id": "60ef627c7fd4",
      "user_message": "训练了多少个模型",
      "ai_response": "3个本地模型：tiny(3.83M)、small(11.8M)、medium(26.8M)，词表5526",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.669942",
      "created_at": 1778432517.6699476,
      "access_count": 0,
      "keywords": [
        "训练了多少个模型"
      ]
    },
    {
      "entry_id": "2baa4d369d38",
      "user_message": "MOA是什么",
      "ai_response": "Mixtures of Agents，多模型聚合引擎。路由→并行执行→聚合，不同于MOE的模型内门控",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.670543",
      "created_at": 1778432517.6705499,
      "access_count": 0,
      "keywords": [
        "MOA是什么"
      ]
    },
    {
      "entry_id": "7ce113e8089a",
      "user_message": "MOA和MOE的区别",
      "ai_response": "MOA是模型间聚合(Mixtures of Agents)，多个独立模型投票/加权；MOE是模型内分工(Mixture of Experts)，门控网络选择专家。虫群用MOA",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.671046",
      "created_at": 1778432517.6710517,
      "access_count": 0,
      "keywords": [
        "MOA和MOE的区别"
      ]
    },
    {
      "entry_id": "35aa5aaff82b",
      "user_message": "种子模型是什么",
      "ai_response": "虫皇模型是种子，喂养不同素材长成不同模型。体量大=虫巢/脑虫，体量小=虫后个人端",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.671607",
      "created_at": 1778432517.6716137,
      "access_count": 0,
      "keywords": [
        "种子模型是什么"
      ]
    },
    {
      "entry_id": "3bb266c68b04",
      "user_message": "记忆模型有什么功能",
      "ai_response": "v7.1增强：encode_realtime实时训练、fork_for_task任务分叉、MemoryMatrix记忆矩阵、超长对话支持",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.672503",
      "created_at": 1778432517.6725132,
      "access_count": 0,
      "keywords": [
        "记忆模型有什么功能"
      ]
    },
    {
      "entry_id": "a4fad3521533",
      "user_message": "自动训练机制",
      "ai_response": "AutoTrainer：should_train判断触发、auto_train增量训练、grow_param参数扩展(d_model逐级增大)",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.673484",
      "created_at": 1778432517.6734908,
      "access_count": 0,
      "keywords": [
        "自动训练机制"
      ]
    },
    {
      "entry_id": "3439af625a45",
      "user_message": "数据采集器功能",
      "ai_response": "DataCollector：对话自动采集、公开数据集下载、质量过滤、去重、格式统一",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.674044",
      "created_at": 1778432517.6740494,
      "access_count": 0,
      "keywords": [
        "数据采集器功能"
      ]
    },
    {
      "entry_id": "9c2c27827c01",
      "user_message": "1B模型能拆成几百个小模型吗",
      "ai_response": "理论上可以。1B参数≈1000个1M小模型。但不是参数拆分，而是多小模型协作覆盖1B能力(MOA)",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.674744",
      "created_at": 1778432517.674753,
      "access_count": 0,
      "keywords": [
        "1B模型能拆成几百个小模型吗"
      ]
    },
    {
      "entry_id": "dd6bc68fb04a",
      "user_message": "MOA是模型间聚合",
      "ai_response": "多个独立模型投票/加权合并结果",
      "intent": "",
      "route": "",
      "timestamp": "2026-05-11T01:01:57.676965",
      "created_at": 1778432517.6769743,
      "access_count": 0,
      "keywords": [
        "MOA是模型间聚合"
      ]
    }
  ]
}