Retrieval Documentation

Functions

create_retrieval_tool(vectorstore)

Creates a LangChain retrieval tool to query account details.

Parameters:

Name	Type	Description	Default
vectorstore		The vectorstore object.	required

Returns:

Name	Type	Description
Tool	Tool	Configured LangChain Tool instance for document retrieval.

Example

tool = create_retrieval_tool(my_vectorstore)
response = tool.run("Invoice for FedEx shipments")
print(response)

is_substring_of_any(doc, seen_texts)

Check if the content of a document is a substring of any text in the provided set.

Parameters:

Name	Type	Description	Default
doc	Document	The document to be checked.	required
seen_texts	set	Set of previously observed document texts.	required

Returns:

Name	Type	Description
bool	bool	True if doc.page_content is a substring of any text in seen_texts, False otherwise.

Example

doc = Document(page_content="Hello world!")
seen_texts = {"Hello", "Hello world! This is an example."}
result = is_substring_of_any(doc, seen_texts)
print(result)  # Output: True

retrieve_accounts(vectorstore, query, retrieved_examples=5)

Retrieves formatted account details from the vectorstore based on query relevance.

Parameters:

Name	Type	Description	Default
vectorstore		Vectorstore object for semantic retrieval.	required
query	str	The query string.	required
retrieved_examples	int	Number of documents to retrieve. Default is 5.	5

Returns:

Name	Type	Description
str	str	Formatted text containing retrieved document details.

Example

output = retrieve_accounts(vectorstore, "invoice from FedEx")
print(output)

retrieve_from_pcg(vectorstore, test_query, retrieved_examples=5)

Retrieve unique, relevant documents from a vectorstore based on semantic similarity.

Parameters:

Name	Type	Description	Default
vectorstore		Vectorstore object supporting semantic search.	required
test_query	str	The query string used for retrieval.	required
retrieved_examples	int	Number of top documents to retrieve. Default is 5.	5

Returns:

Type	Description
List[Tuple[Document, float]]	List[Tuple[Document, float]]: A list of tuples with Documents and their similarity scores.

Example

results = retrieve_from_pcg(vectorstore, "invoice number details", 3)
for doc, score in results:
    print(doc.page_content, score)