Overview
A Space refers to the process of transforming your defined Corvic Tables into embeddings. Here, you specify the “how” - the method of embedding. This involves selecting an algorithm, setting parameters, and initiating the space. Once the space is complete, your embeddings are ready for analysis and export.A Space details how to use input sources in your Corvic Table to generate embeddings for each key defined by the output entities within that Table.
Creating a Space
Step 1: Select Corvic Table
- Navigate to Spaces in your data room
- Click “Create Space”
- Select the Corvic Table you want to embed
- Review the entities that will be embedded
Step 2: Choose Algorithm
Select an embedding algorithm based on your data type:- Graph Structural Encoding: For graph-structured data with relationships
- Text Embedding: For text-based entities
- Tabular Embedding: For structured tabular data
- Multi-Modal: For mixed data types
Step 3: Configure Parameters
Set algorithm-specific parameters:- Embedding Dimension: Size of embedding vectors
- Algorithm Parameters: Algorithm-specific settings
- Quality Thresholds: Minimum quality requirements
Step 4: Run Space Generation
- Review your configuration
- Click “Generate Space”
- Monitor the generation progress
- Wait for completion
Space Generation Process
Processing Stages
- Validation: Verify Corvic Table and parameters
- Data Loading: Load entities and relationships
- Embedding Generation: Apply selected algorithm
- Quality Analysis: Compute quality metrics
- Storage: Store embeddings for access
Monitoring
Track space generation progress:- Status: Pending, Processing, Completed, Failed
- Progress: Percentage complete
- Estimated Time: Remaining time
- Logs: Detailed processing logs
Quality Metrics
Once a space is generated, you can analyze its quality using:Stable Rank
Measures the stability and dimensionality of the embedding space.Reciprocal Condition Number
Indicates the numerical stability of the embeddings.NE Sum
Normalized entropy sum, measuring the information content.Visualization
- UMAP Plots: Visualize embedding distributions
- Quality Dashboards: Monitor quality metrics
- Comparison Tools: Compare different spaces
Using Spaces
Querying Embeddings
Once generated, you can:- Query embeddings via API
- Use vector search
- Retrieve similar entities
- Export embeddings
Agent Integration
Spaces can be used by agents:- Configure agents to use specific spaces
- Multi-space traversal
- Dynamic policy generation
Agents
Learn about using spaces with agents.
Managing Spaces
Viewing Spaces
- List all spaces in a room
- View space details and metrics
- Preview embedding samples
- Check dependencies
Updating Spaces
- Regenerate with new parameters
- Update algorithm settings
- Refresh from updated Corvic Tables
Deleting Spaces
- Remove spaces that are no longer needed
- Free up storage
- Clean up unused resources
Best Practices
Algorithm Selection
- Choose algorithms appropriate for your data type
- Consider relationship structures
- Test different algorithms for best results
Parameter Tuning
- Start with default parameters
- Experiment with different settings
- Monitor quality metrics
- Optimize for your use case
Quality Monitoring
- Regularly check quality metrics
- Monitor embedding distributions
- Compare spaces over time
- Address quality issues promptly
Performance Considerations
Generation Time
- Depends on data size and algorithm
- Use distributed processing for large datasets
- Monitor and optimize generation time
Storage
- Embeddings require storage space
- Plan for storage needs
- Archive unused spaces