feat: 工具系统迁移 + 重型插件骨架 + 前端交互增强

- 工具系统从 kilostar/plugin/tool_plugin/ 迁移到 data/toolset/（manifest.json 声明式）
- 新增 plugin_runtime 模块：BaseOrganization / GlobalPluginManager / loader / tool_bridge
- 新增 org_task + org_task_event 表及 DAO（alembic 0009）
- 新增 /api/v1/plugin 路由（submit/status/stream/install/reload）
- 新增 data/plugin/example_dept 示例重型插件
- regulatory_node 支持聊天历史上下文注入
- send_file 改为 artifact 存盘 + SSE 推送下载链接
- 前端 WorkflowFileCard 组件 + ToolSettings README 渲染
- utils 整理：合并 access/role_check、standalone_proxy→ray_compat、删除废弃模块
- 项目结构文档移至 docs/STRUCTURE.md 并详细展开

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

kilostar/api/chat.py

@@ -26,6 +26,40 @@ from kilostar.core.individual.regulatory_node.template import (
 
 chat_router = APIRouter(prefix="/api/v1/chat", tags=["chat"])
 
+# 单次注入历史的最大轮数（user+assistant 算一轮），防止 token 爆炸。
+_HISTORY_MAX_TURNS = 20
+
+
+def _build_message_history(rows) -> list:
+    """把 DB 中的 ChatHistoryMessage 列表转成 pydantic-ai message_history 格式。
+
+    历史按时间升序，截取末尾最多 _HISTORY_MAX_TURNS*2 条；user 消息映射为
+    ``ModelRequest(parts=[UserPromptPart])``，assistant（``regulatory_node``）映射为
+    ``ModelResponse(parts=[TextPart])``。其它 owner 跳过。
+    """
+    from pydantic_ai.messages import (
+        ModelRequest, ModelResponse, UserPromptPart, TextPart,
+    )
+
+    trimmed = rows[-(_HISTORY_MAX_TURNS * 2):]
+    history: list = []
+    for row in trimmed:
+        owner = row.message_owner
+        text = row.message
+        if not text:
+            continue
+        if owner == "user":
+            history.append(ModelRequest(parts=[UserPromptPart(content=text)]))
+        elif owner == "regulatory_node":
+            history.append(ModelResponse(parts=[TextPart(content=text)]))
+    return history
+
+
+async def _load_message_history(chat_id: str) -> list:
+    postgres_database = ray_actor_hook("postgres_database").postgres_database
+    rows = await postgres_database.list_chat_messages.remote(chat_id=chat_id)
+    return _build_message_history(rows or [])
+
 
 def _extract_reply(resp: MessageResponse | None) -> str | None:
     """从 RegulatoryNode.working 的输出里取出对用户的回复文本。
@@ -39,7 +73,7 @@ def _extract_reply(resp: MessageResponse | None) -> str | None:
 
 
 async def _ask_regulatory(
-    *, user_id: str, chat_id: str, message: str
+    *, user_id: str, chat_id: str, message: str, message_history: list | None = None
 ) -> str | None:
     """统一封装 chat 入口对 RegulatoryNode 的调用。"""
     regulatory_node = ray_actor_hook("regulatory_node").regulatory_node
@@ -49,7 +83,9 @@ async def _ask_regulatory(
         platform_id=chat_id,
         message=message,
     )
-    resp: MessageResponse | None = await regulatory_node.working.remote(payload)
+    resp: MessageResponse | None = await regulatory_node.working.remote(
+        payload, message_history
+    )
     return _extract_reply(resp)
 
 
@@ -120,7 +156,8 @@ async def send_chat_message(
     token_data: TokenData = Depends(Accessor.get_current_user),
 ):
     postgres_database = ray_actor_hook("postgres_database").postgres_database
-    # 存用户消息
+    # 先取历史（不含当前输入），再写入用户消息，避免历史里出现重复
+    message_history = await _load_message_history(chat_id)
     await postgres_database.add_chat_message.remote(
         chat_id=chat_id, message=request.message, message_owner="user"
     )
@@ -130,6 +167,7 @@ async def send_chat_message(
         user_id=token_data.user_id,
         chat_id=chat_id,
         message=request.message,
+        message_history=message_history,
     )
 
     # 存回复
@@ -164,10 +202,12 @@ async def stream_chat_message(
     token_data: TokenData = Depends(Accessor.get_current_user),
 ):
     """SSE 流式聊天端点：standalone 模式下逐 token 流式输出；distributed 模式 fallback 到整段回复。"""
-    from kilostar.utils.standalone_proxy import _STANDALONE
+    from kilostar.utils.ray_compat import _STANDALONE
 
     postgres_database = ray_actor_hook("postgres_database").postgres_database
 
+    message_history = await _load_message_history(chat_id)
+
     await postgres_database.add_chat_message.remote(
         chat_id=chat_id, message=request_body.message, message_owner="user"
     )
@@ -183,7 +223,7 @@ async def stream_chat_message(
 
     if not _STANDALONE:
         async def fallback_generator():
-            resp = await regulatory_node.working.remote(payload)
+            resp = await regulatory_node.working.remote(payload, message_history)
             full_response = resp.reply_message if resp else ""
             if full_response:
                 await postgres_database.add_chat_message.remote(
@@ -195,7 +235,7 @@ async def stream_chat_message(
         return StreamingResponse(fallback_generator(), media_type="text/event-stream")
 
     token_queue = asyncio.Queue()
-    stream_task = regulatory_node.stream_working.remote(payload, token_queue)
+    stream_task = regulatory_node.stream_working.remote(payload, token_queue, message_history)
 
     async def event_generator():
         full_response = ""