存储（Storage）¶

配置嵌入模型与向量数据库，用于表结构/样例数据、文档与指标的嵌入与检索。

结构¶

storage:
  base_path: data                # RAG 向量存储根目录
  embedding_device_type: cpu     # cpu/cuda/mps/auto

  database:
    registry_name: openai
    model_name: text-embedding-v3-small
    dim_size: 1024
    batch_size: 10
    target_model: openai

  document:
    model_name: all-MiniLM-L6-v2
    dim_size: 384

  metric:
    model_name: all-MiniLM-L6-v2
    dim_size: 384

路径与设备¶

storage:
  base_path: data
  embedding_device_type: auto

路径示例：data/datus_db_<namespace>（如 data/datus_db_snowflake）
设备选项：cpu、cuda、mps、auto

嵌入模型¶

数据库嵌入（表结构/样例）¶

database:
  registry_name: openai                # openai 或 sentence-transformers
  model_name: text-embedding-v3-small
  dim_size: 1024
  batch_size: 10
  target_model: openai                 # 关联 agent.models

参数：registry_name、model_name、dim_size、batch_size、target_model

文档嵌入¶

document:
  model_name: all-MiniLM-L6-v2
  dim_size: 384

指标嵌入¶

metric:
  model_name: all-MiniLM-L6-v2
  dim_size: 384

提供方选项¶

OpenAI（云）¶

database:
  registry_name: openai
  model_name: text-embedding-v3-small   # 或 v3-large
  dim_size: 1536                        # small=1536, large=3072
  batch_size: 10
  target_model: openai

Sentence-Transformers（本地）¶

database:
  registry_name: sentence-transformers
  model_name: all-MiniLM-L6-v2
  dim_size: 384

其它本地模型

intfloat/multilingual-e5-large-instruct（~1.2GB，1024 维，多语种）
BAAI/bge-large-en-v1.5 / BAAI/bge-large-zh-v1.5（~1.2GB，1024 维）

方案建议¶

轻量本地

storage:
  base_path: data
  embedding_device_type: auto
  database:
    registry_name: sentence-transformers
    model_name: all-MiniLM-L6-v2
    dim_size: 384
  document:
    model_name: all-MiniLM-L6-v2
    dim_size: 384
  metric:
    model_name: all-MiniLM-L6-v2
    dim_size: 384

混合云本地

storage:
  base_path: data
  embedding_device_type: cpu
  database:
    registry_name: openai
    model_name: text-embedding-v3-small
    dim_size: 1536
    batch_size: 10
    target_model: openai
  document:
    model_name: intfloat/multilingual-e5-large-instruct
    dim_size: 1024
  metric:
    model_name: intfloat/multilingual-e5-large-instruct
    dim_size: 1024

企业高质

storage:
  base_path: /opt/datus/embeddings
  embedding_device_type: cuda
  database:
    registry_name: openai
    model_name: text-embedding-v3-large
    dim_size: 3072
    batch_size: 5
    target_model: openai
  document:
    model_name: BAAI/bge-large-en-v1.5
    dim_size: 1024
  metric:
    model_name: BAAI/bge-large-en-v1.5
    dim_size: 1024

与其它组件集成¶

metrics:
  duckdb:
    domain: sale
    layer1: layer1
    layer2: layer2
    ext_knowledge: ""

storage:
  metric:
    model_name: all-MiniLM-L6-v2
    dim_size: 384