docs/integrations/retrievers/self_query/pgvector_self_query/

[giscus[bot]]

docs/integrations/retrievers/self_query/pgvector_self_query/

PGVector is a vector similarity search package for Postgres data base.

https://python.langchain.com/docs/integrations/retrievers/self_query/pgvector_self_query/

[alchemine]

You can customize available operators like below:

retriever = SelfQueryRetriever.from_llm(
    llm,
    vectorstore,
    document_content_description,
    metadata_field_info,
    verbose=True,
    chain_kwargs={
        "allowed_comparators": [
            "eq",
            "ne",
            "not",
            "gt",
            "gte",
            "lt",
            "lte",
            "between",
            "in",
            "nin",
            "like",
            "ilike",
            "exists",
        ]
    },
)

The extraction prompt can be found at these links:

1. enable_limit=True
   

   langchain/libs/langchain/langchain/chains/query_constructor/prompt.py

   Line 150 in 81d1653

   SCHEMA_WITH_LIMIT = """\

2. enable_limit=False
   

   langchain/libs/langchain/langchain/chains/query_constructor/prompt.py

   Line 117 in 81d1653

   DEFAULT_SCHEMA = """\