Build lightweight AI agent admin

2026-06-08 18:14:59 +08:00
commit e164840f43
2530 changed files with 435693 additions and 0 deletions
@@ -0,0 +1,6 @@
+"""
+知识库 Schema
+"""
+from .knowledge_base_schema import *
+from .document_schema import *
+from .segment_schema import *
@@ -0,0 +1,37 @@
+"""
+知识库标注 Schema
+"""
+from typing import Optional, List
+from datetime import datetime
+
+from pydantic import BaseModel, Field, ConfigDict
+
+from app.base_schema import CSTDatetime
+
+
+class AnnotationCreateInput(BaseModel):
+    """创建标注"""
+    question: str = Field(..., min_length=1, description="问题")
+    answer: str = Field(..., min_length=1, description="答案")
+
+
+class AnnotationUpdateInput(BaseModel):
+    """更新标注"""
+    question: Optional[str] = Field(None, min_length=1, description="问题")
+    answer: Optional[str] = Field(None, min_length=1, description="答案")
+    enabled: Optional[bool] = Field(None, description="是否启用")
+
+
+class AnnotationResponse(BaseModel):
+    """标注输出"""
+    id: str
+    knowledge_base_id: str
+    question: str
+    answer: str
+    embedding_status: str = "pending"
+    enabled: bool = True
+    hit_count: int = 0
+    sys_create_datetime: Optional[CSTDatetime] = None
+    sys_update_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True)
@@ -0,0 +1,62 @@
+"""
+知识库文档 Schema
+"""
+from typing import Optional, List
+from datetime import datetime
+
+from pydantic import BaseModel, Field, ConfigDict
+
+from app.base_schema import CSTDatetime
+
+
+class DocumentUploadInput(BaseModel):
+    """文档上传输入（通过文件管理系统上传后传入 file_id）"""
+    file_id: str = Field(..., description="文件ID（来自文件管理系统）")
+    name: Optional[str] = Field(None, description="文档名称（不传则使用文件名）")
+
+
+class DocumentBatchUploadInput(BaseModel):
+    """批量文档上传"""
+    file_ids: List[str] = Field(..., min_length=1, description="文件ID列表")
+
+
+class DocumentResponse(BaseModel):
+    """文档输出"""
+    id: str
+    knowledge_base_id: str
+    file_id: Optional[str] = None
+    name: str
+    file_type: str = ""
+    file_size: int = 0
+    content_hash: str = ""
+    segment_count: int = 0
+    token_count: int = 0
+    char_count: int = 0
+    status: str = "pending"
+    error_message: str = ""
+    duplicate_warning: Optional[str] = None
+    enabled: bool = True
+    indexing_started_at: Optional[CSTDatetime] = None
+    indexing_completed_at: Optional[CSTDatetime] = None
+    sys_create_datetime: Optional[CSTDatetime] = None
+    sys_update_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True)
+
+
+class DocumentListResponse(BaseModel):
+    """文档列表输出"""
+    id: str
+    knowledge_base_id: str
+    file_id: Optional[str] = None
+    name: str
+    file_type: str = ""
+    file_size: int = 0
+    segment_count: int = 0
+    token_count: int = 0
+    status: str = "pending"
+    duplicate_warning: Optional[str] = None
+    enabled: bool = True
+    sys_create_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True)
@@ -0,0 +1,115 @@
+"""
+知识库 Schema
+"""
+from typing import Optional, List, Dict, Any
+from datetime import datetime
+
+from pydantic import BaseModel, Field, ConfigDict
+
+from app.base_schema import CSTDatetime
+
+
+class KnowledgeBaseCreate(BaseModel):
+    """创建知识库"""
+    model_config = ConfigDict(protected_namespaces=())
+    
+    application_id: Optional[str] = Field(None, description="所属应用ID")
+    is_global: bool = Field(default=False, description="是否在子应用中可见")
+    name: str = Field(..., max_length=100, description="知识库名称")
+    code: str = Field(..., max_length=100, description="知识库编码")
+    description: Optional[str] = Field(None, description="描述")
+    icon: str = Field(default="", description="图标")
+    embedding_model_id: Optional[str] = Field(None, description="Embedding 模型ID")
+    embedding_dimensions: int = Field(default=1536, description="向量维度")
+    chunk_strategy: str = Field(default="recursive", description="分块策略")
+    chunk_size: int = Field(default=500, ge=100, le=4000, description="分块大小")
+    chunk_overlap: int = Field(default=50, ge=0, le=500, description="分块重叠")
+    separator: Optional[str] = Field(None, description="自定义分隔符")
+    retrieval_mode: str = Field(default="hybrid", description="检索模式")
+    top_k: int = Field(default=5, ge=1, le=20, description="检索数量")
+    score_threshold: float = Field(default=0.5, ge=0, le=1, description="相似度阈值")
+    rerank_enabled: bool = Field(default=False, description="是否启用重排序")
+    rerank_model_id: Optional[str] = Field(None, description="重排序模型ID")
+    retrieval_weight: float = Field(default=1.0, ge=0.1, le=10.0, description="检索权重")
+    process_rules: Optional[Dict[str, Any]] = Field(None, description="预处理规则")
+    indexing_technique: str = Field(default="high_quality", description="索引模式: high_quality/economy")
+
+
+class KnowledgeBaseUpdate(BaseModel):
+    """更新知识库"""
+    model_config = ConfigDict(protected_namespaces=())
+    
+    name: Optional[str] = None
+    description: Optional[str] = None
+    icon: Optional[str] = None
+    embedding_model_id: Optional[str] = None
+    embedding_dimensions: Optional[int] = None
+    chunk_strategy: Optional[str] = None
+    chunk_size: Optional[int] = Field(None, ge=100, le=4000)
+    chunk_overlap: Optional[int] = Field(None, ge=0, le=500)
+    separator: Optional[str] = None
+    retrieval_mode: Optional[str] = None
+    top_k: Optional[int] = Field(None, ge=1, le=20)
+    score_threshold: Optional[float] = Field(None, ge=0, le=1)
+    rerank_enabled: Optional[bool] = None
+    rerank_model_id: Optional[str] = None
+    retrieval_weight: Optional[float] = Field(None, ge=0.1, le=10.0)
+    process_rules: Optional[Dict[str, Any]] = None
+    indexing_technique: Optional[str] = None
+    status: Optional[str] = None
+    is_global: Optional[bool] = None
+
+
+class KnowledgeBaseResponse(BaseModel):
+    """知识库详情输出"""
+    id: str
+    application_id: Optional[str] = None
+    is_global: bool = False
+    name: str
+    code: str
+    description: str = ""
+    icon: str = ""
+    embedding_model_id: Optional[str] = None
+    embedding_model_name: str = ""
+    embedding_dimensions: int = 1536
+    chunk_strategy: str = "recursive"
+    chunk_size: int = 500
+    chunk_overlap: int = 50
+    separator: Optional[str] = None
+    retrieval_mode: str = "hybrid"
+    top_k: int = 5
+    score_threshold: float = 0.5
+    rerank_enabled: bool = False
+    rerank_model_id: Optional[str] = None
+    retrieval_weight: float = 1.0
+    process_rules: Optional[Dict[str, Any]] = None
+    indexing_technique: str = "high_quality"
+    document_count: int = 0
+    segment_count: int = 0
+    total_token_count: int = 0
+    total_char_count: int = 0
+    status: str = "active"
+    sort: int = 0
+    sys_create_datetime: Optional[CSTDatetime] = None
+    sys_update_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True, protected_namespaces=())
+
+
+class KnowledgeBaseListResponse(BaseModel):
+    """知识库列表输出"""
+    id: str
+    application_id: Optional[str] = None
+    application_name: str = ""
+    is_global: bool = False
+    name: str
+    code: str
+    description: str = ""
+    icon: str = ""
+    embedding_model_name: str = ""
+    document_count: int = 0
+    segment_count: int = 0
+    status: str = "active"
+    sys_create_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True, protected_namespaces=())
@@ -0,0 +1,28 @@
+"""
+检索日志 Schema
+"""
+from typing import Optional, List, Dict, Any
+from datetime import datetime
+
+from pydantic import BaseModel, Field, ConfigDict
+
+from app.base_schema import CSTDatetime
+
+
+class RetrievalLogResponse(BaseModel):
+    """检索日志输出"""
+    id: str
+    query: str
+    knowledge_base_ids: List[str] = []
+    retrieval_mode: str = "hybrid"
+    top_k: int = 5
+    score_threshold: float = 0.5
+    result_count: int = 0
+    results: Optional[List[Dict[str, Any]]] = None
+    rerank_applied: str = "false"
+    elapsed_time: int = 0
+    source: Optional[str] = None
+    user_id: Optional[str] = None
+    sys_create_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True)
@@ -0,0 +1,140 @@
+"""
+知识库分段 Schema
+"""
+from typing import Optional, List, Dict, Any
+from datetime import datetime
+
+from pydantic import BaseModel, Field, ConfigDict
+
+from app.base_schema import CSTDatetime
+
+
+class SegmentResponse(BaseModel):
+    """分段输出"""
+    id: str
+    knowledge_base_id: str
+    document_id: str
+    document_name: str = ""
+    position: int = 0
+    content: str
+    answer: Optional[str] = None
+    token_count: int = 0
+    char_count: int = 0
+    word_count: int = 0
+    page_number: Optional[int] = None
+    keywords: Optional[List[str]] = None
+    metadata: Optional[Dict[str, Any]] = None
+    embedding_status: str = "pending"
+    enabled: bool = True
+    hit_count: int = 0
+    sys_create_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True)
+
+
+class SegmentListResponse(BaseModel):
+    """分段列表输出"""
+    id: str
+    document_id: str
+    document_name: str = ""
+    position: int = 0
+    content: str
+    answer: Optional[str] = None
+    token_count: int = 0
+    char_count: int = 0
+    word_count: int = 0
+    page_number: Optional[int] = None
+    keywords: Optional[List[str]] = None
+    extra_metadata: Optional[Dict[str, Any]] = None
+    enabled: bool = True
+    hit_count: int = 0
+    embedding_status: str = "pending"
+    sys_create_datetime: Optional[CSTDatetime] = None
+
+    model_config = ConfigDict(from_attributes=True)
+
+
+class SegmentUpdateInput(BaseModel):
+    """更新分段"""
+    content: Optional[str] = Field(None, description="分段内容")
+    keywords: Optional[List[str]] = Field(None, description="关键词")
+    enabled: Optional[bool] = Field(None, description="是否启用")
+    extra_metadata: Optional[Dict[str, Any]] = Field(None, description="元数据")
+
+
+class SegmentCreateInput(BaseModel):
+    """手动创建分段"""
+    content: str = Field(..., min_length=1, description="分段内容")
+    answer: Optional[str] = Field(None, description="Q&A 模式的答案")
+    keywords: Optional[List[str]] = Field(None, description="关键词")
+
+
+class ChunkPreviewInput(BaseModel):
+    """分块预览输入"""
+    file_id: str = Field(..., description="文件ID")
+    chunk_strategy: str = Field(default="recursive", description="分块策略")
+    chunk_size: int = Field(default=500, ge=100, le=4000, description="分块大小")
+    chunk_overlap: int = Field(default=50, ge=0, le=500, description="分块重叠")
+    separator: Optional[str] = Field(None, description="自定义分隔符")
+    process_rules: Optional[Dict[str, Any]] = Field(None, description="预处理规则")
+
+
+class ChunkPreviewItem(BaseModel):
+    """分块预览结果项"""
+    position: int = 0
+    content: str = ""
+    char_count: int = 0
+    token_count: int = 0
+    word_count: int = 0
+    answer: Optional[str] = None
+    metadata: Optional[Dict[str, Any]] = None
+
+
+class ChunkPreviewResponse(BaseModel):
+    """分块预览响应"""
+    chunks: List[ChunkPreviewItem] = Field(default_factory=list)
+    total: int = 0
+    strategy: str = ""
+    chunk_size: int = 0
+    chunk_overlap: int = 0
+
+
+class RetrievalInput(BaseModel):
+    """检索输入"""
+    model_config = ConfigDict(protected_namespaces=())
+    
+    query: str = Field(..., min_length=1, description="查询文本")
+    knowledge_base_ids: List[str] = Field(..., min_length=1, description="知识库ID列表")
+    top_k: int = Field(default=5, ge=1, le=20, description="返回数量")
+    score_threshold: float = Field(default=0.5, ge=0, le=1, description="相似度阈值")
+    retrieval_mode: Optional[str] = Field(None, description="检索模式（不传则使用知识库配置）")
+    rerank_enabled: Optional[bool] = Field(None, description="是否启用重排序（不传则使用知识库配置）")
+    rerank_model_id: Optional[str] = Field(None, description="重排序模型ID（不传则使用知识库配置）")
+    metadata_filter: Optional[Dict[str, Any]] = Field(None, description="元数据过滤条件")
+
+
+class RetrievalResult(BaseModel):
+    """检索结果"""
+    segment_id: str
+    document_id: str
+    document_name: str = ""
+    knowledge_base_id: str
+    knowledge_base_name: str = ""
+    content: str
+    score: float = 0.0
+    token_count: int = 0
+    page_number: Optional[int] = None
+    metadata: Optional[Dict[str, Any]] = None
+    keywords: Optional[List[str]] = None
+    match_source: Optional[str] = Field(None, description="命中来源: vector/fulltext/annotation")
+    parent_content: Optional[str] = Field(None, description="父分段内容（Small-to-Big 模式）")
+
+
+class RetrievalResponse(BaseModel):
+    """检索响应"""
+    results: List[RetrievalResult] = Field(default_factory=list)
+    total: int = 0
+    query: str = ""
+    elapsed_time: int = 0
+    retrieval_mode: str = ""
+    rerank_applied: bool = False