Merge #9 into develop from fix/memory_reflection

新增反思功能（功能配置接口+反思celery后台检测反思的迭代周期） * fix/memory_reflection: (24 commits squashed) - 新增反思功能（功能配置接口+反思celery后台检测反思的迭代周期） - 新增反思功能（功能配置接口+反思celery后台检测反思的迭代周期） - 新增反思功能（检测代码/规范化程序） - 新增反思功能（检测代码/规范化程序） - 新增反思功能（检测代码/规范化程序） - 新增反思功能（检测代码/规范化程序） - 新增反思功能（检测代码/规范化程序） - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 - 反思优化 Signed-off-by: aliyun8644380055 <accounts_68c0f5d519f260d93ee2997e@mail.teambition.com> Commented-by: aliyun8644380055 <accounts_68c0f5d519f260d93ee2997e@mail.teambition.com> Commented-by: aliyun6762716068 <accounts_68cb7c6b61f5dcc4200d6251@mail.teambition.com> Reviewed-by: aliyun6762716068 <accounts_68cb7c6b61f5dcc4200d6251@mail.teambition.com> Merged-by: aliyun6762716068 <accounts_68cb7c6b61f5dcc4200d6251@mail.teambition.com> CR-link: https://codeup.aliyun.com/redbearai/python/redbear-mem-open/change/9
2025-12-19 08:04:12 +00:00
parent 8d810af1d0
commit 5c0d8b42f3
21 changed files with 2384 additions and 337 deletions
--- a/api/app/schemas/end_user_schema.py
+++ b/api/app/schemas/end_user_schema.py
@@ -13,5 +13,6 @@ class EndUser(BaseModel):
    other_id: Optional[str] = Field(description="第三方ID", default=None)
    other_name: Optional[str] = Field(description="其他名称", default="")
    other_address: Optional[str] = Field(description="其他地址", default="")
+    reflection_time: Optional[datetime.datetime] = Field(description="反思时间", default_factory=datetime.datetime.now)
    created_at: datetime.datetime = Field(description="创建时间", default_factory=datetime.datetime.now)
    updated_at: datetime.datetime = Field(description="更新时间", default_factory=datetime.datetime.now)
--- a/api/app/schemas/memory_reflection_schemas.py
+++ b/api/app/schemas/memory_reflection_schemas.py
@@ -0,0 +1,54 @@
+from pydantic import BaseModel, Field
+from typing import Optional
+from enum import Enum
+
+
+class OptimizationStrategy(str, Enum):
+    """优化策略枚举"""
+    SPEED_FIRST = "speed_first"
+    ACCURACY_FIRST = "accuracy_first"
+    BALANCED = "balanced"
+
+
+class Memory_Reflection(BaseModel):
+    config_id: Optional[int] = None
+    reflectionenabled: bool
+    reflection_period_in_hours: str
+    reflexion_range: str
+    baseline: str
+    reflection_model_id: str
+    memory_verify: bool
+    quality_assessment: bool
+    
+    # 新增快速引擎优化参数
+    optimization_strategy: Optional[OptimizationStrategy] = OptimizationStrategy.BALANCED
+    use_fast_model: Optional[bool] = True
+    enable_caching: Optional[bool] = True
+    enable_streaming: Optional[bool] = True
+    batch_size: Optional[int] = Field(default=3, ge=1, le=10)
+    max_concurrent: Optional[int] = Field(default=5, ge=1, le=20)
+    
+    class Config:
+        use_enum_values = True
+
+
+class FastReflectionRequest(BaseModel):
+    """快速反思请求模型"""
+    reflection: Memory_Reflection
+    host_id: Optional[str] = "88a459f5_text02"
+    optimization_strategy: Optional[OptimizationStrategy] = OptimizationStrategy.BALANCED
+    
+    class Config:
+        use_enum_values = True
+
+
+class ReflectionBenchmarkRequest(BaseModel):
+    """反思基准测试请求模型"""
+    reflection: Memory_Reflection
+    host_id: Optional[str] = "88a459f5_text02"
+    iterations: Optional[int] = Field(default=3, ge=1, le=10)
+    
+    class Config:
+        use_enum_values = True
+
+
--- a/api/app/schemas/memory_storage_schema.py
+++ b/api/app/schemas/memory_storage_schema.py
@@ -2,7 +2,7 @@
 所有的内容是放错误地方了，应该放在models
 """

-from typing import Any, Optional, List, Dict, Literal
+from typing import Any, Optional, List, Dict, Literal, Union
 import time
 import uuid
 from pydantic import BaseModel, Field, ConfigDict, field_validator, model_validator
@@ -28,25 +28,48 @@ class Write_UserInput(BaseModel):
 # ============================================================================
 class BaseDataSchema(BaseModel):
    """Base schema for the data"""
-    id: str = Field(..., description="The unique identifier for the data entry.")
-    statement: str = Field(..., description="The statement text.")
-    group_id: str = Field(..., description="The group identifier.")
-    chunk_id: str = Field(..., description="The chunk identifier.")
+    # 保持原有必需字段为可选，以兼容不同数据源
+    id: Optional[str] = Field(None, description="The unique identifier for the data entry.")
+    statement: Optional[str] = Field(None, description="The statement text.")
+    group_id: Optional[str] = Field(None, description="The group identifier.")
+    chunk_id: Optional[str] = Field(None, description="The chunk identifier.")
    created_at: str = Field(..., description="The creation timestamp in ISO 8601 format.")
    expired_at: Optional[str] = Field(None, description="The expiration timestamp in ISO 8601 format.")
    valid_at: Optional[str] = Field(None, description="The validation timestamp in ISO 8601 format.")
    invalid_at: Optional[str] = Field(None, description="The invalidation timestamp in ISO 8601 format.")
    entity_ids: List[str] = Field([], description="The list of entity identifiers.")
+    description: Optional[str] = Field(None, description="The description of the data entry.")
+
+    # 新增字段以匹配实际输入数据
+    entity1_name: str = Field(..., description="The first entity name.")
+    entity2_name: Optional[str] = Field(None, description="The second entity name.")
+    statement_id: str = Field(..., description="The statement identifier.")
+    relationship_type: str = Field(..., description="The relationship type.")
+    relationship: Optional[Dict[str, Any]] = Field(None, description="The relationship object.")
+    entity2: Optional[Dict[str, Any]] = Field(None, description="The second entity object.")
+
+
+class QualityAssessmentSchema(BaseModel):
+    """Schema for memory quality assessment results."""
+    score: int = Field(..., ge=0, le=100, description="Quality score percentage (0-100).")
+    summary: str = Field(..., description="Brief summary of data quality status, including main issues and strengths.")
+
+
+class MemoryVerifySchema(BaseModel):
+    """Schema for memory privacy verification results."""
+    has_privacy: bool = Field(..., description="Whether privacy information was detected.")
+    privacy_types: List[str] = Field([], description="List of detected privacy information types.")
+    summary: str = Field(..., description="Brief summary of privacy detection results.")


 class ConflictResultSchema(BaseModel):
    """Schema for the conflict result data in the reflexion_data.json file."""
-    data: List[BaseDataSchema] = Field(..., description="The conflict memory data.")
+    data: List[BaseDataSchema] = Field(..., description="The conflict memory data. Only contains conflicting records when conflict is True.")
    conflict: bool = Field(..., description="Whether the memory is in conflict.")
-    conflict_memory: Optional[BaseDataSchema] = Field(None, description="The conflict memory data.")
+    quality_assessment: Optional[QualityAssessmentSchema] = Field(None, description="The quality assessment object. Contains score and summary when quality_assessment is enabled, null otherwise.")
+    memory_verify: Optional[MemoryVerifySchema] = Field(None, description="The memory privacy verification object. Contains privacy detection results when memory_verify is enabled, null otherwise.")

    @model_validator(mode="before")
-    @classmethod
    def _normalize_data(cls, v):
        if isinstance(v, dict):
            d = v.get("data")
@@ -61,7 +84,6 @@ class ConflictSchema(BaseModel):
    conflict_memory: Optional[BaseDataSchema] = Field(None, description="The conflict memory data.")

    @model_validator(mode="before")
-    @classmethod
    def _normalize_data(cls, v):
        if isinstance(v, dict):
            d = v.get("data")
@@ -76,21 +98,30 @@ class ReflexionSchema(BaseModel):
    solution: str = Field(..., description="The solution for the reflexion.")


+class ChangeRecordSchema(BaseModel):
+    """Schema for individual change records"""
+    field: List[Dict[str, str]] = Field(..., description="List of field changes, each containing field name and new value.")
+
 class ResolvedSchema(BaseModel):
    """Schema for the resolved memory data in the reflexion_data"""
    original_memory_id: Optional[str] = Field(None, description="The original memory identifier.")
-    resolved_memory: Optional[BaseDataSchema] = Field(None, description="The resolved memory data.")
+    # resolved_memory: Optional[BaseDataSchema] = Field(None, description="The resolved memory data (only contains records that need modification).")
+    resolved_memory: Optional[Union[BaseDataSchema, List[BaseDataSchema]]] = Field(None, description="The resolved memory data (only contains records that need modification). Can be a single record or list of records.")
+    change: Optional[List[ChangeRecordSchema]] = Field(None, description="List of detailed change records with IDs and field information.")


+class SingleReflexionResultSchema(BaseModel):
+    """Schema for a single reflexion result item."""
+    conflict: ConflictResultSchema = Field(..., description="The conflict result data for this specific conflict type.")
+    reflexion: ReflexionSchema = Field(..., description="The reflexion data for this conflict.")
+    resolved: Optional[ResolvedSchema] = Field(None, description="The resolved memory data for this conflict.")
+    type: str = Field("reflexion_result", description="The type identifier.")
+
 class ReflexionResultSchema(BaseModel):
-    """Schema for the reflexion result data in the reflexion_data.json file."""
-    # 模型输出中 "conflict" 为单个冲突对象（包含 data 与 conflict_memory），而非字典映射
-    conflict: ConflictResultSchema = Field(..., description="The conflict result data.")
-    reflexion: Optional[ReflexionSchema] = Field(None, description="The reflexion data.")
-    resolved: Optional[ResolvedSchema] = Field(None, description="The resolved memory data.")
+    """Schema for the complete reflexion result data - a list of individual conflict resolutions."""
+    results: List[SingleReflexionResultSchema] = Field(..., description="List of individual conflict resolution results, grouped by conflict type.")

    @model_validator(mode="before")
-    @classmethod
    def _normalize_resolved(cls, v):
        if isinstance(v, dict):
            conflict = v.get("conflict")