N8N Tools - GraphRAG
Actions9
Overview
The Vector Search - Search operation in this node enables semantic search over documents or data using vector embeddings and optionally enhanced with large language model (LLM) analysis. It supports querying external or local vector databases combined with graph databases to retrieve relevant information based on the input query or document text.
This node is beneficial when you want to perform advanced semantic searches that go beyond keyword matching, such as finding conceptually related documents, knowledge graph queries, or hybrid searches combining vector similarity and graph relationships. For example, it can be used to:
- Search a collection of documents for semantically similar content.
- Query a knowledge graph enriched with vector embeddings.
- Perform hybrid searches leveraging both vector and graph databases.
- Enhance search results with LLM-powered insights or summarization.
Properties
| Name | Meaning |
|---|---|
| Input Data | Document text or data to process and use as the basis for the vector search. |
| Additional Options | Collection of optional parameters: • Language: Processing language (English, Portuguese, Spanish, French). • Result Limit: Maximum number of results to return. • Chunk Size: Size of text chunks for processing. • Chunk Overlap: Overlap between text chunks. |
| Database Configuration | Configure external or local vector and graph databases: • Vector Database: Choose provider (Local FAISS/ChromaDB, Pinecone, Weaviate, Qdrant, Milvus). • Vector DB Connection URL & API Key: For external vector DBs. • Graph Database: Choose provider (Local NetworkX, Neo4j, ArangoDB, Amazon Neptune). • Graph DB connection details: URL, username, password, database name. • AWS credentials and region for Amazon Neptune. |
| 🤖 LLM Enhancement | Enable and configure Large Language Model enhancement for deeper analysis: • Enable LLM Enhancement (boolean). • LLM Provider: N8N Tools internal, OpenAI GPT models, or Anthropic Claude models. • LLM Model: Specific model selection depending on provider (e.g., GPT-4o Mini, GPT-4o, Claude Haiku, Claude Sonnet). |
Output
The node outputs a JSON object containing the results from the GraphRAG API call. This typically includes:
- Search results relevant to the input query or document.
- Metadata about the retrieved items.
- Possibly enhanced insights if LLM enhancement is enabled.
If binary data were involved (not indicated here), it would represent files or media related to the search results, but this node focuses on JSON structured data.
Dependencies
- Requires an API key credential for the external GraphRAG service (
n8nToolsApi). - Supports integration with various external vector databases (Pinecone, Weaviate, Qdrant, Milvus) and graph databases (Neo4j, ArangoDB, Amazon Neptune) which may require additional credentials and connection URLs.
- Optional LLM providers require appropriate API keys or access tokens configured in n8n.
- The node makes HTTP POST requests to
https://graphrag.n8ntools.io/api/v1/graphrag.
Troubleshooting
Common issues:
- Missing or invalid API key for the GraphRAG service will cause authentication errors.
- Incorrect or incomplete database configuration (URLs, credentials) can lead to connection failures.
- Exceeding result limits or chunk size constraints might cause performance degradation or timeouts.
- Enabling LLM enhancement without proper API credentials or quota may result in errors.
Error messages:
"GraphRAG Tool error: <message>"indicates an issue during the API request or processing.- Authentication errors suggest checking the API key credential.
- Connection errors to vector or graph databases indicate verifying URLs and credentials.
- Validation errors on input properties mean required fields are missing or incorrectly formatted.
Resolution tips:
- Ensure all required credentials and URLs are correctly set up in n8n.
- Validate input data format and size.
- Check network connectivity to external services.
- Review API usage limits and quotas for LLM providers.