rag_server/applications/utils/milvus/field.py

from pymilvus import FieldSchema, DataType

# milvus 向量数据库
fields = [
    # 主键 ID
    FieldSchema(
        name="id",
        dtype=DataType.INT64,
        is_primary=True,
        auto_id=True,
        description="自增id",
    ),
    # 文档 id 字段
    FieldSchema(
        name="doc_id", dtype=DataType.VARCHAR, max_length=64, description="文档id"
    ),
    FieldSchema(name="chunk_id", dtype=DataType.INT64, description="文档分块id"),
    # 三种向量字段
    FieldSchema(
        name="vector_text",
        dtype=DataType.FLOAT_VECTOR,
        dim=2560,
        description="chunk文本 embedding",
    ),
    FieldSchema(
        name="vector_summary",
        dtype=DataType.FLOAT_VECTOR,
        dim=2560,
        description="总结 embedding",
    ),
    FieldSchema(
        name="vector_questions",
        dtype=DataType.FLOAT_VECTOR,
        dim=2560,
        description="衍生问题 embedding",
    ),
]

# pattern fields
mode_fields = [
    FieldSchema(
        name="id",
        dtype=DataType.INT64,
        is_primary=True,
        auto_id=True,
        description="自增id",
    ),
    # 文档 id 字段
    FieldSchema(
        name="mode_id", dtype=DataType.VARCHAR, max_length=64, description="模式id"
    ),
    FieldSchema(
            name="mode_vector",
            dtype=DataType.FLOAT_VECTOR,
            dim=2560,
            description="chunk文本 embedding",
        ),
]


__all__ = ["fields", "mode_fields"]