The InMemoryExecutor provides an in-memory implementation for executing queries against indexed data. It creates optimized vector spaces based on the provided indices and allows querying data from in-memory sources, making it ideal for development, testing, and prototyping scenarios.

​
Constructor

InMemoryExecutor(sources, indices, context_data=None)

​
Parameters

​
sources
InMemorySource | Sequence[InMemorySource]
required
One or more in-memory data sources to query against. Can be a single source or sequence of sources. All sources must be InMemorySource instances.
​
indices
Index | Sequence[Index]
required
One or more indices that define the vector spaces. Can be a single index or sequence of indices. These indices determine how the data will be organized and searched.
​
context_data
Mapping[str, Mapping[str, ContextValue]] | None
default:"None"
Additional context data for execution. The outer mapping key represents the context name, inner mapping contains key-value pairs for that context. Provides runtime configuration and environment-specific settings.

​
Inheritance

The InMemoryExecutor extends the executor hierarchy for in-memory processing: Inheritance Chain:
  • InMemoryExecutor
  • → InteractiveExecutor
  • → Executor
  • → ABC
  • → Generic
This inheritance provides:
  • Base executor functionality from Executor
  • Interactive processing capabilities from InteractiveExecutor
  • Abstract base class support from ABC
  • Generic type support for type safety

​
Methods

​
run()

Execute the in-memory pipeline and return a configured InMemoryApp instance. 
run() -> InMemoryApp
Returns: An instance of InMemoryApp that can accept queries and provide immediate results from in-memory data. The run method:
  1. Initializes all provided sources with their data
  2. Creates optimized vector spaces from the indices
  3. Configures the in-memory vector database
  4. Returns a fully configured application ready for querying

​
Key Features

​
In-Memory Processing

  • Fast Access: All data stored in RAM for immediate access
  • No I/O Overhead: Eliminates disk and network latency
  • Immediate Results: Queries execute and return results instantly
  • Simple Setup: No external database configuration required

​
Development Optimization

  • Quick Iteration: Rapid development and testing cycles
  • Easy Debugging: Full control over data and execution flow
  • Flexible Testing: Easy to modify datasets and test scenarios
  • Minimal Dependencies: Self-contained execution environment

​
Use Cases

​
Development and Testing

Perfect for initial development phases:
  • Prototype Development: Quick testing of vector search concepts
  • Unit Testing: Isolated testing of search functionality
  • Integration Testing: Testing complete pipelines with controlled data
  • Algorithm Development: Experimenting with different vector spaces and configurations

​
Educational and Learning

Ideal for learning and demonstration:
  • Tutorials: Teaching vector search concepts with immediate feedback
  • Workshops: Hands-on learning with instant results
  • Proof of Concepts: Demonstrating search capabilities to stakeholders
  • Research: Academic research with controlled datasets

​
Small-Scale Applications

Suitable for applications with limited data requirements:
  • Personal Projects: Individual applications with small datasets
  • Internal Tools: Company tools with limited search requirements
  • Demos: Live demonstrations with pre-loaded data
  • MVPs: Minimum viable products for initial validation

​
Performance Characteristics

​
Advantages

  • Speed: Fastest possible query execution due to in-memory storage
  • Simplicity: No external dependencies or complex configuration
  • Consistency: Predictable performance without external factors
  • Development Speed: Rapid iteration and testing capabilities

​
Limitations

  • Memory Constraints: Limited by available system RAM
  • No Persistence: Data lost on application restart
  • Scalability: Not suitable for large datasets or high concurrency
  • Single Instance: Cannot distribute across multiple servers

​
Data Management

​
Memory Usage

Memory Monitoring: Monitor memory usage carefully as all data, vectors, and indices are stored in RAM. Large datasets can quickly consume available memory.

​
Data Lifecycle

  • Initialization: Data loaded into memory during startup
  • Runtime: All operations happen in memory
  • Shutdown: All data lost when application stops
  • No Persistence: Data must be reloaded on each restart

​
Best Practices

​
Development Workflow

Start Small: Begin with small datasets to validate your approach before scaling to production solutions.

​
Memory Management

Dataset Size: Keep datasets reasonably sized (typically under 1GB) for optimal performance and to prevent out-of-memory errors.

​
Testing Strategy

Isolated Tests: Use separate InMemoryExecutor instances for different test scenarios to ensure test isolation and prevent data contamination.

​
Migration Path

Production Transition: Design your application to easily migrate from InMemoryExecutor to production executors like RestExecutor when ready to deploy.

​
Context Configuration

The context_data parameter allows fine-tuning of execution behavior:
  • Performance Settings: Configure memory allocation and processing parameters
  • Debug Information: Enable detailed logging and debugging features
  • Environment Variables: Set environment-specific configurations
  • Feature Flags: Enable or disable specific functionality during development

​
Integration Pattern

InMemoryExecutor works seamlessly with the Superlinked development stack:
  • InMemorySource: Provides data input for testing and development
  • Various Indices: Supports all index types for experimentation
  • InMemoryVectorDatabase: Automatically configured for in-memory storage
  • InMemoryApp: Returns the appropriate application instance

​
Application Lifecycle

The InMemoryExecutor manages a simple lifecycle:
  1. Initialization: Configure sources, indices, and context
  2. Data Loading: Load all data into memory during startup
  3. Vector Processing: Generate and store vectors in memory
  4. Query Execution: Process queries with immediate results
  5. Shutdown: Clean termination with data loss
The InMemoryExecutor provides the foundation for rapid development and testing while maintaining the same interface patterns used in production executors, ensuring smooth transitions as your application evolves.