Lib.rs

Cargo Features

reasonkit-mem has no features set by default.

[dependencies]
reasonkit-mem = { version = "0.1.7", features = ["python", "docset", "local-embeddings", "compression", "vdreamteam", "http-server", "full", "fast-embed", "rag", "graph", "caching", "fast-hash", "research-phase2", "columnar", "pdf-parsing", "document-extraction", "hybrid-search", "research-phase3", "persistent-db", "zero-copy", "content-addressed", "data-integrity", "research-phase4", "research-full"] }

============================================================================
FEATURES
============================================================================

default

Default: minimal feature set for library use

python full? = pyo3

Python bindings via PyO3

Affects reasonkit-mem::PyRaptorTree…

docset full?

Docset ingestion for documentation URLs (Cursor-like @Docs)

Enables once_cell, regex, scraper ^0.21, and url

url:

Docset ingestion (Cursor-like @Docs) (optional)

Affects reasonkit-mem::docset…

Required by rk-mem, rk-mem-clap and rk-mem-docs-mcp binaries

local-embeddings full? = ndarray, ort, tokenizers

Local BGE-M3 ONNX embeddings (no API calls needed)

Affects embedding::local…

compression full? = lz4_flex

LZ4 compression for dual-layer memory architecture

Affects serde_utils::compress, serde_utils::decompress…

vdreamteam full? = serde_yaml

vDreamTeam AI Agent Memory System
Provides structured memory for AI agent teams with:
- Constitutional layer (shared identity, constraints)
- Role-specific memory (per-agent decisions, lessons, PxP logs)
- Cross-agent coordination logging

Affects reasonkit-mem::vdreamteam…

Required by vdream_cli and vdream_mcp binaries

http-server full? = axum, tower-http, tracing-subscriber

HTTP API Server for external RAG integration

Required by the rkmem_server binary

full = compression, docset, http-server, local-embeddings, python, vdreamteam

Full feature set

fast-embed research-phase2? = fastembed

Fast embedding generation (fastembed 5.8+ - now compatible with ort 2.0.0-rc.11)

rag = rag-toolchain

RAG toolchain integration

graph research-phase2? = graphalgs, petgraph

Knowledge graph algorithms

Affects reasonkit-mem::graph…

caching research-phase2? = moka, quick_cache

High-performance caching layer

fast-hash research-phase2? = blake3

Fast BLAKE3 content hashing

research-phase2 research-full? = caching, fast-embed, fast-hash, graph

All Phase 2 research features (fastembed re-enabled with ort 2.0.0-rc.11)

columnar research-phase3? = lancedb

============================================================================
PHASE 3 FEATURES: ARCHITECTURAL BLUEPRINT (2026-01-18)
============================================================================

Columnar storage for ML workloads (lancedb)

pdf-parsing research-phase3? = lopdf, pdf-extract

High-fidelity PDF parsing

Affects reasonkit-mem::pdf…

document-extraction research-phase3? = readability

Advanced document extraction

hybrid-search research-phase3? = bm25

Hybrid BM25 + vector search

research-phase3 research-full? = columnar, document-extraction, hybrid-search, pdf-parsing

Local embedding generation - now available via fast-embed feature (Phase 2)

All Phase 3 blueprint features

persistent-db research-phase4? = redb

============================================================================
PHASE 4 FEATURES: COMPLETE INTEGRATION GUIDE (2026-01-18)
============================================================================

Persistent ACID-compliant storage (alternative to sled)

zero-copy research-phase4? = rkyv

Zero-copy serialization for hot paths

content-addressed research-phase4? = cacache

Content-addressable storage for deduplication

data-integrity research-phase4? = blake3, rs_merkle

Data integrity with Merkle trees

research-phase4 research-full? = content-addressed, data-integrity, persistent-db, zero-copy

All Phase 4 integration guide features

research-full = research-phase2, research-phase3, research-phase4

Complete research feature set

Features from optional dependencies

In crates that don't use the dep: syntax, optional dependencies automatically become Cargo features. These features may have been created by mistake, and this functionality may be removed in the future.

============================================================================
DEPENDENCIES
============================================================================

pyo3 python?

Enables pyo3 ^0.24.1

Python Bindings (optional)

axum http-server?

Enables axum ^0.7

HTTP Server (for API sidecar mode)

tower-http http-server?

Enables tower-http ^0.5

tracing-subscriber http-server?

ort local-embeddings?

Enables ort

ONNX Runtime for local embeddings (BGE-M3)

tokenizers local-embeddings?

Enables tokenizers ^0.21

HuggingFace tokenizers

ndarray local-embeddings?

Enables ndarray

NDArray for tensor operations (must match ort's ndarray version)

lz4_flex compression?

Enables lz4_flex ^0.11

LZ4 compression (optional, for dual-layer memory)

serde_yaml vdreamteam?

Enables serde_yaml

YAML serialization (optional, for vDreamTeam memory system)

fastembed fast-embed?

Enables fastembed

============================================================================
PHASE 2: HYPER DEEP RESEARCH ADDITIONS (2026-01-18)
============================================================================

Fast Embeddings - fastembed 5.8+ now uses ort 2.0.0-rc.11 (aligned with our ort version)

rag-toolchain rag?

Enables rag-toolchain

RAG Framework

petgraph graph?

Enables petgraph ^0.6

Graph Algorithms

graphalgs graph?

moka caching?

Enables moka

High-Performance Caching

quick_cache caching?

blake3 data-integrity? fast-hash?

Enables blake3

BLAKE3 Hashing

lancedb columnar?

Enables lancedb ^0.23

============================================================================
PHASE 3: ARCHITECTURAL BLUEPRINT ADDITIONS (2026-01-18)
============================================================================

Columnar Storage - High-performance ML data format (100x faster than Parquet)
lancedb 0.23+ resolved chrono/arrow-arith conflict (issue #7196)

pdf-extract pdf-parsing?

Enables pdf-extract ^0.7

PDF Parsing - High-fidelity document extraction with layout preservation

lopdf pdf-parsing?

Enables lopdf ^0.34

readability document-extraction?

Enables readability

Document Processing - Advanced text extraction

bm25 hybrid-search?

Enables bm25 ^0.1

Hybrid Search - BM25 + Vector fusion algorithms

redb persistent-db?

Enables redb ^2.2

Embedding Models - fastembed re-enabled above in Phase 2 section
============================================================================
PHASE 4: COMPLETE INTEGRATION GUIDE ADDITIONS (2026-01-18)
============================================================================

Persistent Key-Value Storage (alternative to sled)
redb - Fast, ACID-compliant embedded database

rkyv zero-copy?

Enables rkyv

Zero-Copy Serialization
rkyv - 1.2ns access time for hot data paths

cacache content-addressed?

Enables cacache

Content-Addressable Storage
cacache - npm-style content-addressed cache

rs_merkle data-integrity?

Enables rs_merkle

Merkle Trees for Data Integrity
rs_merkle - Multi-proof support for audit trails