Weaviate

The Weaviate Node in Stack AI is used to query a Weaviate vector database. Weaviate is a popular open-source vector database that allows you to store and search data using vector embeddings, which is especially useful for semantic search, recommendation systems, and AI-powered retrieval tasks.

Here’s a detailed explanation of Vector Search in Weaviate and how to establish a connection in Stack AI:

---

Available Actions

Vector Search in Weaviate allows you to search for objects (such as documents, images, or data entries) that are most similar to a given query, using vector embeddings. This is especially useful for semantic search, recommendations, and AI-powered retrieval.

How the "Vector Search" Action Works

• Action Name: Vector Search

• Provider: Weaviate

• Purpose: Search for similar vectors in a Weaviate database.

Required Input Parameters

1. Class Name (class_name): The name of the Weaviate class (collection/table) you want to search in. Example: "Article", "Document", "Product"

2. Properties (properties): Comma-separated list of properties to return in the results. Example: "title,description,url"

3. Query (query): The search query (text) to find similar vectors. Example: "AI workflow automation"

4. Number of Results (top_k): How many results to return. Example: 5 or 10

Output

• Results (results): The results of the vector search query, typically a list of objects with the requested properties.

---

How to Establish a Connection to Weaviate

To connect Stack AI to your Weaviate instance, you need to provide connection details. Here’s how:

1. Connection ID: If you have a private Weaviate instance, you’ll need a connection ID. This is usually set up in your Stack AI integrations or connections panel.

   • In the action node’s action_configurations, add:

     { "connection_id": "<your-weaviate-connection-id>" }

   • If you’re using the default Stack AI Weaviate instance, you may not need to set this.

2. Configure the Action Node:

   • Set the required input parameters (class_name, properties, query, top_k).

   • If you want to use dynamic input (e.g., from an Input node or LLM), reference the node like {in-0} or {llm-0} in the query field.

---

Summary Table

Parameter	Description	Example Value
class_name	Weaviate class to search	"Article"
properties	Properties to return (comma-separated)	"title,description,url"
query	Search query (text or node reference)	"AI workflow automation"
top_k	Number of results to return	5
connection_id	(Optional) Connection ID for private Weaviate	"abc123-connection-id"

How to Use the Weaviate Node

1. Connect to a Weaviate Database: You typically need to configure the node with a connection to your Weaviate instance (this may require a connection ID if you have a private instance).

2. Set Query Parameters: Define what you want to search for—this usually involves providing a query vector or text, and specifying parameters like the number of results (top_k), the class or collection to search, and any filters.

3. Input Data: The node can take input from other nodes (like an LLM or Input node) to dynamically generate queries.

4. Use Output: The results from the Weaviate node can be passed to downstream nodes, such as an LLM for summarization or an Output node for display.

Example Workflow

• User provides a search query (Input node).

• The query is embedded (optionally by an LLM or embedding node).

• The Weaviate node performs a vector search using the embedding.

• The results are passed to an LLM for summarization or directly to the Output node.

When to Use the Weaviate Node

• When you need to perform advanced, AI-powered search over large datasets.

• When you want to build applications that require semantic understanding, such as chatbots, document search, or recommendation engines.