memvid-rs 2.0.3

High-performance AI memory library with full-text and semantic search (Rust bindings for memvid v2)

memvid-rs (Python Bindings)

Memvid-rs provides Python bindings for memvid-core v2, a high-performance AI memory library written in Rust.

This is a major version upgrade from v1 (QR video encoding) to v2 (.mv2 single-file format).

Attribution: This project wraps memvid-core with PyO3 bindings for Python.

🚀 Key Features

• Single-File Memory: All data stored in portable .mv2 format
• Full-Text Search: BM25 ranking via embedded Tantivy index
• Crash-Safe Writes: WAL-based append-only architecture
• Cross-Platform: Works on macOS, Linux, Windows
• Python 3.8-3.14+: ABI3 wheel supports all modern Python versions

📦 Installation

pip install memvid-rs

💻 Quick Start

from memvid_rs import MemvidMemory

# Create a new memory file
memory = MemvidMemory.create("my_memory.mv2")

# Add content
idx = memory.append("Hello, memvid v2!")
memory.commit()

# Retrieve content
text = memory.get_frame(idx)
print(text)  # Hello, memvid v2!

# Check stats
print(f"Total frames: {memory.frame_count()}")

# Open existing memory
memory2 = MemvidMemory.open("my_memory.mv2")

🔧 API Reference

MemvidMemory

Method	Description
create(path)	Create new .mv2 file
open(path)	Open existing .mv2 file
append(text)	Add text, returns 0-based index
commit()	Flush writes to disk
get_frame(index)	Retrieve frame text by index
frame_count()	Get total frame count

⚠️ Breaking Changes from v1

• QR video encoding removed - v2 uses binary .mv2 format
• API redesigned - MemvidEncoder/MemvidDecoder replaced with MemvidMemory
• File format incompatible - v1 .mp4 files cannot be opened in v2

📋 Requirements

• Python: >= 3.8
• System: macOS, Linux, Windows

📄 License

MIT License