Creating Vectors

Vectors are created through a multi-step process that includes configuring basic information, selecting data sources, and building knowledge graphs. This guide walks you through each step.

Required Information

• Name: Unique name for the vector (required)
• Description: Description of the vector's purpose (required)
• Model Provider: AI provider for embeddings (required)
• Embedding Model: Specific embedding model (required)
• Data Sources: At least one data source (Objects, MFlows, MReports, or Files)

Creation Process

Vector creation consists of three steps:

1. Basic Information: Configure name, description, and embedding model
2. Data Source: Select data sources (Objects, MFlows, MReports, Files)
3. Knowledge Graph: Build knowledge graph defining relationships

Step 1: Basic Information

Configure the basic information for your vector:

Enter Vector Name

Provide a unique name for the vector.

Required: Yes

Field: Name

Validation: Must be unique

Example: "Customer Data Vector", "Product Knowledge Base", "Document Repository"

Enter Description

Provide a description explaining the vector's purpose and what data it contains.

Required: Yes

Field: Description

Example: "Vector store containing customer information and order history for AI agent context"

Select Model Provider

Choose the AI provider for generating embeddings.

Required: Yes

Field: Model Provider

Available Providers:

• OpenAI: OpenAI embedding models
• Ollama: Local Ollama embedding models
• VertexAI: Google Vertex AI embedding models

How to select: Use the dropdown to select a provider. The available embedding models will update based on your selection.

Select Embedding Model

Choose the specific embedding model to use.

Required: Yes

Field: Embedding Model

How to select:

• After selecting a provider, choose from available embedding models
• Different providers offer different models
• Model selection affects embedding quality and dimensions

Note: Use the same embedding model consistently for a vector to ensure consistent embeddings.

Configure Model Settings (Optional)

Configure additional model settings:

• Temperature: Controls randomness in embeddings (default: 0.7)
• Top K: Number of top results to return in searches (default: 3)

Note: These settings can typically be adjusted after vector creation.

Navigate to Next Step

Click "Next" or "Confirm" to proceed to the Data Source step.

Step 2: Data Source

Select the data sources to include in your vector:

Select Objects

Choose Objects whose data will be included in the vector.

Required: No (but at least one data source is required)

Field: Objects

How to select:

• Use the multi-select dropdown
• Search for Objects by name
• Select multiple Objects
• Selected Objects' records will be embedded

Note: Only non-transient Objects are included in embeddings.

Select MFlows

Choose workflows (MFlows) whose definitions will be included in the vector.

Required: No

Field: MFlows

How to select:

• Use the multi-select dropdown
• Search for MFlows by name
• Select multiple MFlows
• Workflow definitions and configurations will be embedded

Select MReports

Choose reports (MReports) whose configurations will be included in the vector.

Required: No

Field: MReports

How to select:

• Use the multi-select dropdown
• Search for MReports by name
• Select multiple MReports
• Report configurations and queries will be embedded

Upload Files

Upload files (documents, PDFs, etc.) to include in the vector.

Required: No

Field: Files

How to upload:

• Click the file upload area
• Select files from your computer
• Files are uploaded and processed
• File content will be extracted and embedded

Supported Formats: Documents, PDFs, and other text-based files

Navigate to Next Step

Click "Next" or "Confirm" to proceed to the Knowledge Graph step.

Step 3: Knowledge Graph

Build a knowledge graph defining relationships between data sources:

Understanding Knowledge Graphs

Knowledge graphs visualize relationships between data sources:

• Nodes: Represent data sources (Objects, MFlows, MReports)
• Edges: Represent relationships between data sources
• Visual Interface: Drag-and-drop interface for building graphs
• Relationship Labels: Custom labels describing relationships

Building the Knowledge Graph

Add Nodes

1. Select Data Source: Choose a data source from the sidebar (Objects, MFlows, MReports)
2. Drag to Canvas: Drag the data source onto the canvas
3. Configure Node:
   • Select Columns: For Object nodes, select specific columns to include
   • Node Label: Customize the node label
   • Node Type: Node type is automatically set based on data source

Create Relationships

1. Connect Nodes: Click and drag from one node to another to create an edge
2. Label Relationship: Click on the edge to add/edit relationship label
3. Relationship Types: Define how data sources relate (e.g., "contains", "references", "belongs to")

Configure Node Details

For Object nodes:

• Select Columns: Choose which columns from the Object to include in embeddings
• Column Selection: Select specific fields that are relevant for search
• Metadata: Column selections affect what data is embedded

Knowledge Graph Best Practices

• Logical Relationships: Create relationships that make logical sense
• Relevant Connections: Connect related data sources
• Clear Labels: Use clear, descriptive labels for relationships
• Column Selection: Select relevant columns from Objects

Create Vector Store

After configuring the knowledge graph:

1. Review your configuration
2. Click "Create" or "Save" to create the vector store
3. The system will:
   • Create the vector store
   • Generate embeddings for selected data sources
   • Store embeddings in the vector database
   • Make the vector available for search and AI agent integration

Note: Embedding generation may take time depending on the amount of data.

Editing Vectors

To edit an existing vector:

1. Navigate to the Vectors page
2. Find the vector in the list
3. Click the "Edit" icon next to the vector
4. The creation dialog opens with existing configuration
5. Navigate through steps to modify:
   • Basic information
   • Data sources
   • Knowledge graph
6. Click "Save" to apply changes

Note: Changes to data sources may require regenerating embeddings.

Searching Vectors

To search a vector:

1. Navigate to the Vectors page
2. Find the vector in the list
3. Click the "Search" icon next to the vector
4. Enter your search query
5. View results showing:
   • Similarity Score: How similar the result is to your query
   • Content: The relevant content from data sources
   • Source: Which data source the result came from

Best Practices

Vector Naming

• Descriptive Names: Use clear names that indicate the vector's purpose
• Consistent Naming: Use consistent naming conventions
• Include Context: Include context in names (e.g., "Customer Data Vector")

Data Source Selection

• Relevant Data: Select only relevant data sources
• Avoid Overload: Don't include unnecessary data
• Quality over Quantity: Focus on quality, relevant data

Knowledge Graph Design

• Logical Structure: Create logical relationships
• Clear Relationships: Use clear relationship labels
• Column Selection: Select relevant columns from Objects
• Avoid Complexity: Keep knowledge graphs simple and understandable

Embedding Model Selection

• Consistent Models: Use the same model for a vector
• Appropriate Models: Choose models appropriate for your data type
• Provider Selection: Consider provider costs and capabilities

Related Introduction

• Vectors Introduction - Overview of vectors
• Objects Introduction - Understand Objects used as vector data sources
• Workflows Introduction - Learn about workflows used as vector data sources
• Reports Introduction - Understand reports used as vector data sources