@n8n/n8n-nodes-langchain
Version:
59 lines (55 loc) • 2.11 kB
text/typescript
/**
* Simple Vector Store Node - Version 1.1
* Discriminator: mode=retrieve
*/
/** Retrieve documents from vector store to be used as vector store with AI nodes */
export type LcVectorStoreInMemoryV11RetrieveParams = {
/**
* Exposes the store as an `ai_vectorStore` subnode for another node (e.g. `toolVectorStore`). Declare with `vectorStore({...})`. Required subnodes: `embedding`. For RAG with an AI Agent directly, prefer `mode: 'retrieve-as-tool'`.
* <patterns>
* <pattern title="retrieve mode — feed another node as a subnode (generic)">
* // Substitute the type literal and provider-specific parameters — see the rest of this file.
* const store = vectorStore({
* type: '@n8n/n8n-nodes-langchain.vectorStoreXxx',
* config: {
* name: 'Knowledge Base',
* parameters: { mode: 'retrieve' /* + provider-specific parameters *\/ },
* subnodes: { embedding: embeddingsOpenAi }
* }
* });
*
* const retrieverTool = tool({
* type: '@n8n/n8n-nodes-langchain.toolVectorStore',
* config: {
* name: 'KB Retriever',
* parameters: { description: 'Search the product knowledge base' },
* subnodes: { vectorStore: store, model: openAiModel }
* }
* });
* </pattern>
* </patterns>
*/
mode: 'retrieve';
/**
* The key to use to store the vector memory in the workflow data. The key will be prefixed with the workflow ID to avoid collisions.
* @default vector_store_key
*/
memoryKey?: string | Expression<string>;
/**
* Whether or not to rerank results
* @default false
*/
useReranker?: boolean | Expression<boolean>;
};
export interface LcVectorStoreInMemoryV11RetrieveSubnodeConfig {
embedding: EmbeddingInstance | EmbeddingInstance[];
/**
* @displayOptions.show { useReranker: [true] }
*/
reranker: RerankerInstance;
}
export type LcVectorStoreInMemoryV11RetrieveNode = {
type: '@n8n/n8n-nodes-langchain.vectorStoreInMemory';
version: 1.1;
config: NodeConfig<LcVectorStoreInMemoryV11RetrieveParams> & { subnodes: LcVectorStoreInMemoryV11RetrieveSubnodeConfig };
};