Setup:

Install langchain-snowflake and configure Snowflake connection.

.. code-block:: bash

pip install -U langchain-snowflake

Key init args:

service_name: str Fully qualified name of the Cortex Search service session: Optional[Session] Active Snowflake session k: int Number of documents to retrieve (default: 4) search_columns: Optional[List[str]] Columns to return in search results filter_dict: Optional[Dict[str, Any]] Filter criteria for search results content_field: str Metadata field containing the actual content (default: "TRANSCRIPT_TEXT") join_separator: str String to join multiple documents (default: "\n\n") fallback_to_page_content: bool Fall back to page_content when metadata field is empty (default: True)

Instantiate:

.. code-block:: python

from . import SnowflakeCortexSearchRetriever

Using existing session (recommended)

retriever = SnowflakeCortexSearchRetriever( service_name="mydb.myschema.my_search_service", session=session, k=5 )

Using connection parameters

retriever = SnowflakeCortexSearchRetriever( service_name="mydb.myschema.my_search_service", account="your-account", user="your-user", password="your-password", warehouse="your-warehouse", k=3 )

Using custom content field (e.g., for datasets that store content in "CHUNK")

retriever_custom = SnowflakeCortexSearchRetriever( service_name="mydb.myschema.my_search_service", session=session, k=5, content_field="CHUNK", # Extract from metadata["CHUNK"] instead of "TRANSCRIPT_TEXT" join_separator="\n---\n", # Custom separator fallback_to_page_content=True )

Usage:

.. code-block:: python

query = "What is machine learning?" docs = retriever.invoke(query) for doc in docs: print(doc.page_content)

Use within a chain:

.. code-block:: python

from langchain_core.output_parsers import StrOutputParser from langchain_core.prompts import ChatPromptTemplate from langchain_core.runnables import RunnablePassthrough from . import ChatSnowflake

prompt = ChatPromptTemplate.from_template( """Answer the question based only on the context provided.

Context: {context}

Question: {question}""" )

llm = ChatSnowflake(model="llama3.1-70b", session=session)

With auto_format_for_rag=True (default), no format_docs needed!

chain = ( {"context": retriever, "question": RunnablePassthrough()} | prompt | llm | StrOutputParser() )

Or with manual control:

from .formatters import format_cortex_search_documents

retriever_manual = SnowflakeCortexSearchRetriever(..., auto_format_for_rag=False)

chain = (

{"context": retriever_manual | format_cortex_search_documents, "question": RunnablePassthrough()}

| prompt | llm | StrOutputParser()

)

response = chain.invoke("What is the capital of France?")