Wax: The SQLite for AI Memory

Wax simplifies the process of integrating memory into on-device AI applications, providing a streamlined alternative to complex RAG (retrieval-augmented generation) architectures. With a single file format, Wax eliminates the need for extensive infrastructure, allowing developers to focus on building intelligent applications without the overhead of managing multiple services.

Overview

Wax delivers a fully functional memory layer encapsulated in a single .mv2s file — enabling fast, reliable, and private memory management for AI agents. Forget the hassle of Docker, complex database configurations, and network dependencies; Wax is designed to enhance the efficiency of AI interactions dramatically.

Key Features

• Fast Performance: Achieve near-instantaneous vector searches, with latencies as low as 0.84ms on Apple Silicon M1 Pro.
• Durability: Built to withstand abrupt shutdowns and power losses, ensuring data integrity through robust crash recovery mechanisms.
• Deterministic Results: Consistently retrieve the same context for identical queries, providing reliability in AI responses.
• Portability: A single .mv2s file structure makes it easy to manage, backup, and distribute.
• Privacy-focused: Operates entirely on-device, eliminating network calls and potential data exposure.

Usage Example

Integrate Wax with just a few lines of code:

import Wax

// Create a memory file
let brain = try await MemoryOrchestrator(
    at: URL(fileURLWithPath: "brain.mv2s")
)

// Store user preferences
try await brain.remember(
    "User prefers dark mode and gets headaches from bright screens",
    metadata: ["source": "onboarding"]
)

// Retrieve context about user preferences
let context = try await brain.recall(query: "user preferences")
// → "User prefers dark mode and gets headaches from bright screens"

How Does Wax Work?

Wax compacts all memory into a single file, which includes:

• Your raw documents
• Embeddings and indices for efficient searching
• Metadata and entity graphs for context retrieval

This unique approach enables users to avoid fragmentation and maintain efficient access to their stored memories.

Perfect For:

• AI Assistants: Enabling memory retention across app launches.
• Offline-first Applications: Ensuring robust functionality even without internet access.
• Privacy-Critical Solutions: Keeping user data secure and localized.
• Research Tools: Facilitating reproducible results and data retrieval.
• Game Development: Supporting stateful interactions for immersive experiences.

Performance Comparison

Wax outperforms traditional memory solutions, offering single-file simplicity and offline capabilities compared to heavier architectures like Chroma and Pinecone, which require server setups and external dependencies.

Conclusion

Wax represents a significant evolution in managing AI memory, providing an efficient, user-friendly solution that prioritizes privacy and performance. By leveraging the power of a compact, single-file system, developers can create advanced AI applications without the burden of complex infrastructure. Explore more and elevate your AI projects with Wax.