RAG tools

nae.tools.make_retriever_tool

make_retriever_tool(retriever: object, name: str = 'search_documents', description: str = 'Search the knowledge base for relevant context.') -> StructuredTool

Wrap a LangChain retriever as a StructuredTool bindable to an AgentNode.

Parameters:

Name	Type	Description	Default
retriever	object	Any object with .invoke(query) returning a list of documents.	required
name	str	Tool name exposed to the model.	'search_documents'
description	str	Tool description exposed to the model.	'Search the knowledge base for relevant context.'

nae.tools.ChromaRAG

ChromaRAG(embeddings, collection_name: str = 'nae', persist_directory=None)

Convenience wrapper around a Chroma vector store.

Requires the optional rag extra: pip install -e .[rag].