Stores

ChatBotKit Stores are specialized storage classes designed for efficient data retrieval and management. They serve as the backbone for organizing and accessing information within your datasets, ensuring optimal performance based on your specific use case requirements.

How Stores Work

When creating a dataset, you select a store type that best matches your data size and retrieval needs. Each store uses different underlying technologies optimized for specific scenarios, from rapid searches on smaller datasets to comprehensive searches across millions of records.

Available Store Types

Ada Sprout

Best for: Small to medium datasets requiring fast search speeds

• Technology: text-embedding-ada-002 + vector database
• Ideal use cases:
  • Quick lookups on smaller datasets
  • Applications where search speed is the primary concern
  • Lightweight implementations

Lingo Sprout (beta)

Best for: General-purpose semantic search across various dataset sizes

• Technology: text-embedding-3-small + vector database
• Note: Lingo Sprout is currently in beta.
• Ideal use cases:
  • Semantic search applications
  • General-purpose data retrieval
  • Balanced performance and accuracy needs

Ada Loom (Default)

Best for: Datasets where search recall and precision both matter

• Technology: hybrid search combining text-embedding-ada-002 and BM25 over a vector database
• Ideal use cases:
  • Datasets where keyword and semantic relevance both matter
  • Applications requiring reliable, consistent search results
  • General production workloads

Choosing the Right Store

Getting Started

1. Assess your needs: Consider your dataset size and whether you prioritize speed or accuracy
2. Select your store: Choose the appropriate store type when creating your dataset
3. Monitor performance: Evaluate search results and adjust if needed