Retrieval Layer
The Retrieval Layer integrates vector stores with conversational AI frameworks (like LangChain and LangGraph), allowing AI agents to query database updates.
Module Summary
Create a LangChain Tool backed by the VectorStoreAdapter. |
|
Create a LangGraph-compatible node that queries the VectorStoreAdapter. |
API Reference
LangGraph integration for PyDebeziumAI vector store adapters.
- create_retriever_tool(adapter, name, description, **kwargs)[source]
Create a LangChain Tool backed by the VectorStoreAdapter.
This tool is designed to be passed to LangGraph agents or standard LangChain agent executors.
- Parameters:
adapter (VectorStoreAdapter) – The VectorStoreAdapter instance to query.
name (str) – Name of the tool.
description (str) – Description of what the tool does and when to call it.
**kwargs (Any) – Additional parameters passed to adapter.as_retriever().
- Returns:
A LangChain Tool instance.
- Return type:
BaseTool
- create_retriever_node(adapter, state_key='documents', query_key='query', query_extractor=None, **retriever_kwargs)[source]
Create a LangGraph-compatible node that queries the VectorStoreAdapter.
This node executes the retrieval query and updates the graph state with the resulting list of Document objects under the specified state_key.
- Parameters:
adapter (VectorStoreAdapter) – The VectorStoreAdapter instance.
state_key (str) – The key in the returned state dictionary where retrieved documents will be stored.
query_key (str) – The key in the input state where the search query string is stored. Only used if query_extractor is None.
query_extractor (Callable[[Any], str] | None) – An optional callable that extracts the query string from the graph state. If provided, it overrides query_key.
**retriever_kwargs (Any) – Options passed to adapter.as_retriever() (e.g. search_kwargs).
- Returns:
A callable node function for LangGraph.
- Return type: