YAMS — Yet Another Memory System

Persistent memory for LLMs and apps. Content-addressed storage with dedupe, compression, full-text and vector search.

Warning

Experimental Software - Not Production Ready

YAMS is under active development and should be considered experimental. The software may contain bugs, incomplete features, and breaking changes between versions. Use at your own risk.

For production workloads, please wait for a stable 1.0 release.

Features

• SHA-256 content-addressed storage with block-level dedupe (Rabin chunking)
• Full-text search (SQLite FTS5) + semantic vector search (embeddings)
• Symbol extraction from source code (tree-sitter: C, C++, Python, JS, TS, Rust, Go, Java, C#, PHP, Kotlin, Dart, SQL, Solidity)
• Snapshot management with Merkle tree diffs and rename detection
• WAL-backed durability, high-throughput I/O, thread-safe
• Portable CLI, MCP server, and plugin architecture (ONNX, S3, PDF extraction)

Links

• SourceHut: https://sr.ht/~trvon/yams/
• GitHub mirror: https://github.com/trvon/yams
• Docs: https://yamsmemory.ai
• Discord: https://discord.gg/rTBmRHdTEc
• License: GPL-3.0-or-later

Install

Supported platforms: Linux x86_64/ARM64, macOS x86_64/ARM64, Windows x86_64

macOS (Homebrew)

# Stable release (recommended)
brew install trvon/yams/yams

# Verify installation
yams --version

Docker

# Pull and run
docker pull ghcr.io/trvon/yams:latest
docker run --rm -v yams-data:/home/yams/.local/share/yams ghcr.io/trvon/yams:latest --version

# MCP server
docker run -i --rm -v yams-data:/home/yams/.local/share/yams ghcr.io/trvon/yams:latest serve

Build from Source

Linux/macOS:

# Quick build (auto-detects Clang/GCC, configures Conan + Meson)
./setup.sh Release

# Build
meson compile -C build/release

# Optional: Install system-wide
meson install -C build/release

Windows:

./setup.ps1 Release
meson compile -C build/release

Prerequisites: GCC 13+, Clang 16+, or MSVC 2022+ (C++20); meson, ninja-build, cmake, pkg-config, conan

See docs/BUILD.md for detailed instructions.

Quick Start

# Initialize (interactive - prompts for grammar downloads)
yams init .

# Auto mode for containers/headless
yams init --auto

# Add content
yams add ./README.md --tags docs
yams add src/ --recursive --include="*.cpp,*.h" --tags code

# Search
yams search "config file" --limit 5
yams grep "TODO" --include="*.cpp"

# List and retrieve
yams list --limit 20
yams get <hash> -o ./output.bin

MCP Server

yams serve  # stdio transport (JSON-RPC)

{
  "mcpServers": {
    "yams": {
      "command": "yams",
      "args": ["serve"]
    }
  }
}

Plugins

YAMS supports plugins for embeddings (ONNX), storage (S3), and content extraction (PDF):

yams plugin list                    # List loaded plugins
yams plugin trust add ~/.local/lib/yams/plugins
yams plugin health                  # Check plugin status
yams doctor plugin onnx             # Diagnose specific plugin

Recommended model: all-MiniLM-L6-v2 (384-dim) — best speed/quality tradeoff.

GPU Acceleration

The ONNX plugin supports GPU acceleration for faster embedding generation:

Platform	Provider	Hardware
macOS	CoreML	Apple Silicon Neural Engine + GPU
Linux	CUDA	NVIDIA GPUs
Windows	DirectML	Any DirectX 12 GPU (NVIDIA, AMD, Intel)

GPU support is auto-detected during build. To manually control:

# Disable GPU
YAMS_ONNX_GPU=none ./setup.sh Release

# Force specific provider
YAMS_ONNX_GPU=cuda ./setup.sh Release    # Linux
YAMS_ONNX_GPU=coreml ./setup.sh Release  # macOS
YAMS_ONNX_GPU=directml ./setup.ps1       # Windows

See plugins/onnx/README.md for detailed configuration.

Troubleshooting

yams doctor              # Full diagnostics
yams stats --verbose     # Storage statistics
yams repair --all        # Repair common issues

Build issues: See docs/BUILD.md

Plugin discovery: yams plugin list empty? Add trust path: yams plugin trust add ~/.local/lib/yams/plugins

Cite

@misc{yams,
  author = {Trevon Williams},
  title = {YAMS: Content-addressable storage with semantic search},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/trvon/yams}
}