Actions6
- Vector Store Actions
- Graph Database Actions
Overview
This node integrates with a Neo4j database, providing two main resource types: Vector Store and Graph Database. For the Vector Store resource, it allows operations related to managing and querying vector embeddings stored in Neo4j, such as performing similarity searches or adding new text documents to the vector index.
The Default operation for the Vector Store resource supports:
- Similarity Search: Finds vectors similar to a given query text by embedding the query and searching the vector index.
- Add Texts: Adds new text documents to the vector store, converting them into embeddings and storing them.
This node is beneficial in scenarios where you want to leverage Neo4j as a vector database for semantic search, recommendation systems, or any application requiring similarity-based retrieval of textual data.
Practical Examples
- Searching for documents semantically similar to a user query in a knowledge base.
- Adding new articles or notes to a vector index for later retrieval based on content similarity.
- Filtering search results by metadata properties (e.g., tags, categories).
Properties
| Name | Meaning |
|---|---|
| Index Name | The name of the vector index to use in Neo4j. Defaults to "vector". |
Note: The provided property is optional and can be used to specify which vector index to target.
Output
The node outputs JSON arrays containing the results of the executed operation:
For Similarity Search, the output is an array of objects representing matched documents, each including:
content: The text content of the matched document.score: Similarity score indicating how close the match is.- Additional metadata fields from the stored document.
For Add Texts, the output is a simple success confirmation object:
[{ "success": true }]
If binary data were involved (e.g., embeddings), it would be handled internally but not exposed directly in the output.
Dependencies
- Requires a configured Neo4j database accessible via credentials including URI, username, password, and database name.
- Uses external libraries from the LangChain community packages for Neo4j graph and vector store integration.
- Needs an API key credential or equivalent authentication token for connecting to the Neo4j instance.
Troubleshooting
Common Issues:
- Incorrect Neo4j credentials or unreachable database endpoint will cause connection failures.
- Malformed JSON in metadata filters or node properties may cause parsing errors.
- Using unsupported operations (e.g., "Create Relationship" in Graph DB) will throw explicit errors.
Error Messages:
- Errors thrown during execution are wrapped and reported as node operation errors with descriptive messages.
- If the similarity search fails, ensure the embedding input is connected and valid.
- JSON parsing errors typically indicate invalid JSON strings in input parameters like metadata filters.
Resolutions:
- Verify Neo4j credentials and network connectivity.
- Validate JSON inputs using online validators before usage.
- Confirm that the required input connections (like embeddings) are properly set up.