mcp-memory-server 0.2.0__tar.gz

mcp_memory_server-0.2.0/.gitignore

@@ -0,0 +1,38 @@
+# Python
+__pycache__/
+*.py[cod]
+*.pyo
+*.egg-info/
+*.egg
+dist/
+build/
+
+# Virtual environment
+.venv/
+
+# Testing
+.pytest_cache/
+htmlcov/
+.coverage
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db
+
+# AIMemory runtime data
+memory_db/
+checkpoints/
+
+# Data artifacts (keep raw source files, ignore generated)
+data/embeddings/
+data/plots/
+data/splits/
+
+# Claude Code
+.claude/plans/

mcp_memory_server-0.2.0/CHANGELOG.md

@@ -0,0 +1,34 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.2.0] - 2026-02-28
+
+First public release.
+
+### Added
+
+- **MCP Server** with 12 tools: `auto_search`, `memory_save`, `memory_search`, `memory_update`, `memory_delete`, `memory_get_related`, `memory_pin`, `memory_unpin`, `memory_stats`, `sleep_cycle_run`, `policy_status`, `policy_decide`
+- **RL Memory Policy** — Rule-based importance scoring + MLP contextual bandit (SAVE/SKIP/RETRIEVE)
+- **Enhanced Policy** (opt-in) — 778d state encoder (SentenceTransformer 768d + 10d hand-crafted), experience replay buffer, progressive autonomy
+- **Semantic Search** — ChromaDB vector store with `intfloat/multilingual-e5-small` embeddings
+- **Knowledge Graph** — NetworkX-based entity-relation graph with multi-hop traversal
+- **GraphRAG Hybrid Retrieval** (opt-in) — Vector similarity + graph traversal with RL re-ranker (11d features)
+- **Multi-Resolution Text** — 3 levels (full text, keyword summary, entity triples) with token-budget-aware composition
+- **Forgetting Pipeline** — Decay-based aging with compress → deactivate → delete stages
+- **Sleep Cycle** — Periodic maintenance: consolidation, resolution regeneration, forgetting, checkpoint saving
+- **Immutable Memories** — Protected from modification/deletion with SHA-256 rule hash verification
+- **Pin/Unpin** — User-controlled forgetting protection
+- **Multilingual Support** — Korean (ko) and English (en) i18n patterns
+- **P2P Federated Learning** — Gossip protocol with differential privacy and Krum aggregation
+- **A/B Comparison Framework** — Baseline vs re-ranked retrieval comparison
+- **CI/CD** — GitHub Actions with Python 3.11/3.12/3.13 matrix
+- **611 tests** passing across 31 test files
+
+### Removed
+
+- Deprecated DualHeadDQN offline training pipeline (replaced by online MLP bandit)
+- `ExtractorConfig` (unused after DQN removal)

mcp_memory_server-0.2.0/CLAUDE.md

@@ -0,0 +1,8 @@
+# AIMemory Project
+
+## Overview
+This project uses Claude Code with team agents for collaborative AI-powered tasks.
+
+## Conventions
+- Use the `.claude/agents/` directory for custom agent definitions
+- Agents are defined as markdown files describing their role, tools, and instructions

mcp_memory_server-0.2.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 AIMemory Contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

mcp_memory_server-0.2.0/PKG-INFO

@@ -0,0 +1,219 @@
+Metadata-Version: 2.4
+Name: mcp-memory-server
+Version: 0.2.0
+Summary: Intelligent memory MCP server for AI assistants — auto-save, semantic search, RL-powered retrieval, and knowledge graphs
+Project-URL: Homepage, https://github.com/ihwoo/mcp-memory-server
+Project-URL: Repository, https://github.com/ihwoo/mcp-memory-server
+Project-URL: Issues, https://github.com/ihwoo/mcp-memory-server/issues
+Project-URL: Changelog, https://github.com/ihwoo/mcp-memory-server/blob/main/CHANGELOG.md
+Author: ihwoo
+License-Expression: MIT
+License-File: LICENSE
+Keywords: ai,chatbot,claude,knowledge-graph,llm,mcp,memory,rag,reinforcement-learning
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Requires-Python: <3.14,>=3.11
+Requires-Dist: chromadb>=0.5.0
+Requires-Dist: mcp[cli]>=1.2.0
+Requires-Dist: pydantic>=2.0
+Requires-Dist: sentence-transformers>=3.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.5; extra == 'dev'
+Provides-Extra: ko
+Requires-Dist: mecab-ko-dic>=1.0.0; extra == 'ko'
+Requires-Dist: mecab-python3>=1.0.9; extra == 'ko'
+Provides-Extra: train
+Requires-Dist: matplotlib>=3.8; extra == 'train'
+Requires-Dist: ollama>=0.4.0; extra == 'train'
+Requires-Dist: pandas>=2.0; extra == 'train'
+Requires-Dist: pyarrow>=15.0; extra == 'train'
+Requires-Dist: tqdm>=4.66; extra == 'train'
+Description-Content-Type: text/markdown
+
+# AIMemory
+
+**Intelligent memory for AI assistants that actually remembers.**
+
+> Drop-in MCP server that gives Claude (and any MCP client) persistent, searchable, self-organizing memory — powered by semantic search, knowledge graphs, and reinforcement learning.
+
+[![CI](https://github.com/ihwoo/mcp-memory-server/actions/workflows/ci.yml/badge.svg)](https://github.com/ihwoo/mcp-memory-server/actions)
+[![PyPI](https://img.shields.io/pypi/v/mcp-memory-server)](https://pypi.org/project/mcp-memory-server/)
+[![Python](https://img.shields.io/pypi/pyversions/mcp-memory-server)](https://pypi.org/project/mcp-memory-server/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+
+---
+
+## Why AIMemory?
+
+Current AI memory tools have two critical problems:
+
+| Problem | How AIMemory solves it |
+|---------|----------------------|
+| **Manual retrieval** — you must ask "do you remember X?" | `auto_search` runs every turn, injecting relevant memories automatically |
+| **Token waste** — entire memory dump inserted into context | Multi-resolution composer selects top-K memories within a token budget |
+
+## Key Features
+
+- **RL-powered policy** — Contextual bandit decides when to save, skip, or retrieve (not just keyword matching)
+- **Semantic search** — ChromaDB + multilingual sentence-transformer embeddings (`intfloat/multilingual-e5-small`)
+- **Knowledge graph** — Entity-relation graph (NetworkX) for multi-hop reasoning ("who likes what?")
+- **GraphRAG hybrid retrieval** — Vector similarity + graph traversal, fused and re-ranked by an RL re-ranker
+- **Multi-resolution text** — Full text → summary → entity triples, composed within token budget
+- **Forgetting pipeline** — Decay-based aging with consolidation, pinning, and immutable protection
+- **Sleep cycle** — Periodic maintenance: dedup, compress, forget, checkpoint
+- **Multilingual** — Korean and English pattern support out of the box
+
+---
+
+## Quick Start (2 minutes)
+
+### 1. Install
+
+```bash
+pip install mcp-memory-server
+```
+
+Or with [uv](https://docs.astral.sh/uv/):
+
+```bash
+uv pip install mcp-memory-server
+```
+
+<details>
+<summary>Optional: Korean NLP support</summary>
+
+```bash
+pip install mcp-memory-server[ko]
+```
+</details>
+
+### 2. Connect to Claude Desktop
+
+Add to your `claude_desktop_config.json`:
+
+```json
+{
+  "mcpServers": {
+    "aimemory": {
+      "command": "aimemory-mcp"
+    }
+  }
+}
+```
+
+That's it. Claude now has persistent memory across all conversations.
+
+<details>
+<summary>Advanced: custom data path or uv project mode</summary>
+
+```json
+{
+  "mcpServers": {
+    "aimemory": {
+      "command": "uv",
+      "args": ["run", "--project", "/path/to/AIMemory", "aimemory-mcp"],
+      "env": {
+        "AIMEMORY_DB_PATH": "/path/to/memory_db"
+      }
+    }
+  }
+}
+```
+</details>
+
+### 3. Connect to Claude Code
+
+```bash
+claude mcp add aimemory -- aimemory-mcp
+```
+
+---
+
+## MCP Tools (12)
+
+| Tool | Description |
+|------|-------------|
+| `auto_search` | Auto-retrieve relevant memories at turn start (multi-resolution context) |
+| `memory_save` | Save a new memory with keywords, category, and relations |
+| `memory_search` | Semantic similarity search |
+| `memory_update` | Update content or keywords of an existing memory |
+| `memory_delete` | Delete a memory (respects immutability) |
+| `memory_get_related` | BFS graph traversal for related memories |
+| `memory_pin` / `memory_unpin` | Protect memories from forgetting |
+| `memory_stats` | Total count and category breakdown |
+| `sleep_cycle_run` | Trigger maintenance (consolidation + forgetting + checkpoint) |
+| `policy_status` | RL policy state (epsilon, action distribution, updates) |
+| `policy_decide` | Ask the RL policy for a SAVE/SKIP/RETRIEVE decision with reasoning |
+
+---
+
+## Configuration
+
+All settings via environment variables:
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `AIMEMORY_DB_PATH` | `./memory_db` | ChromaDB persistence directory |
+| `AIMEMORY_LANGUAGE` | `ko` | Language for pattern matching (`ko` / `en`) |
+| `AIMEMORY_EMBEDDING_MODEL` | `intfloat/multilingual-e5-small` | Sentence-transformer model |
+| `AIMEMORY_LOG_LEVEL` | `INFO` | Logging level |
+| `AIMEMORY_ENHANCED_POLICY` | `0` | Enable 778d enhanced RL policy (`1` to enable) |
+| `AIMEMORY_GRAPH_RAG` | `0` | Enable GraphRAG hybrid retrieval (`1` to enable) |
+
+---
+
+## Architecture
+
+```
+┌─────────────────────────────────────────────────┐
+│                   MCP Client                     │
+│          (Claude Desktop / Claude Code)          │
+└────────────────────┬────────────────────────────┘
+                     │ stdio (JSON-RPC)
+┌────────────────────▼────────────────────────────┐
+│              FastMCP Server (12 tools)           │
+├──────────────────────────────────────────────────┤
+│              MemoryBridge (orchestrator)          │
+├──────────┬──────────┬──────────┬─────────────────┤
+│ RL Policy│ Retrieval│ Storage  │ Maintenance      │
+│          │          │          │                  │
+│ Rule-    │ ChromaDB │ Graph    │ Sleep Cycle      │
+│ Based +  │ vector + │ Memory   │ (consolidation,  │
+│ MLP      │ Knowledge│ Store    │  forgetting,     │
+│ Bandit   │ Graph    │          │  checkpoints)    │
+│          │ (GraphRAG)│         │                  │
+│ Re-ranker│          │          │                  │
+│ (11d MLP)│          │          │                  │
+└──────────┴──────────┴──────────┴─────────────────┘
+```
+
+---
+
+## Development
+
+```bash
+# Clone and install dev dependencies
+git clone https://github.com/ihwoo/mcp-memory-server.git
+cd mcp-memory-server
+uv sync --extra dev
+
+# Run tests (611 tests)
+uv run pytest tests/ -q
+
+# Lint & format
+uv run ruff check src/ tests/
+uv run ruff format src/ tests/
+```
+
+---
+
+## License
+
+MIT — see [LICENSE](LICENSE) for details.

mcp_memory_server-0.2.0/PROJECT_SUMMARY.md

@@ -0,0 +1,295 @@
+# AIMemory 프로젝트 정리
+
+> 작성일: 2026-02-28 | v0.1.0
+
+## 1. 프로젝트 개요
+
+**AI Memory System** — AI 챗봇이 사용자의 정보를 자동으로 기억하고 활용하는 시스템.
+
+핵심 문제 해결:
+- **수동 검색 문제**: 사용자가 명시적으로 요청해야만 메모리를 활용 → 매 턴 자동 검색
+- **토큰 낭비 문제**: 전체 기억을 매번 context에 삽입 → top-K 선별 + 다해상도 압축
+
+## 2. 시스템 아키텍처
+
+```
+사용자 메시지
+    │
+    ▼
+┌─────────────────────────┐
+│  State Encoder (394d)    │  텍스트 임베딩(384d) + 수작업 특징(10d)
+│  → Action Selection      │  SAVE / SKIP / RETRIEVE (Contextual Bandit)
+└─────────────┬───────────┘
+              │
+    ┌─────────┼─────────┐
+    ▼         ▼         ▼
+ SAVE      SKIP    RETRIEVE
+    │                   │
+    ▼                   ▼
+┌────────┐      ┌──────────────┐
+│GraphRAG│      │Context       │
+│Store   │◄────►│Composer      │
+│(Chroma)│      │(Token Budget)│
+└────────┘      └──────────────┘
+    │
+    ▼ (주기적)
+┌─────────────┐
+│Sleep Cycle   │  망각 · 통합 · 재인코딩
+└─────────────┘
+```
+
+### 6 레이어
+
+| 레이어 | 모듈 | 역할 |
+|--------|------|------|
+| 1. Policy | `online/policy.py`, `online/enhanced_policy.py` | Feature → Action (SAVE/SKIP/RETRIEVE) |
+| 2. Rule Filter | `online/rule_verifier.py` | 불변 규칙 (보안/프라이버시) |
+| 3. Storage | `memory/graph_store.py`, `memory/knowledge_graph.py` | ChromaDB + Knowledge Graph |
+| 4. Retrieval | `memory/graph_retriever.py`, `online/reranker.py` | Vector search + RL Re-ranking |
+| 5. Composer | `memory/composer.py`, `memory/resolution.py` | 토큰 budget 내 최적 조합 + 다해상도 압축 |
+| 6. Sleep Cycle | `memory/sleep_cycle.py`, `memory/forgetting.py`, `memory/consolidation.py` | 주기적 기억 정리 |
+
+## 3. 모듈 구조
+
+```
+src/aimemory/               # 9,709 lines
+├── __init__.py
+├── config.py               # 전체 설정 (Pydantic)
+├── schemas.py              # 데이터 모델 (Turn, Episode, MemoryActionType)
+│
+├── i18n/                   # 다국어 지원
+│   ├── __init__.py         # LanguagePatterns 레지스트리
+│   ├── ko.py               # 한국어 패턴 (정규식 40+개)
+│   └── en.py               # 영어 패턴
+│
+├── mcp/                    # MCP 서버 (12 tools)
+│   ├── __init__.py
+│   ├── __main__.py         # python -m aimemory.mcp
+│   ├── server.py           # FastMCP 서버 정의
+│   └── bridge.py           # Policy ↔ MCP 연결
+│
+├── memory/                 # 기억 저장/검색/관리
+│   ├── graph_store.py      # ChromaDB + sentence-transformers
+│   ├── knowledge_graph.py  # 트리플 기반 지식 그래프
+│   ├── graph_retriever.py  # 하이브리드 검색 (Vector + Graph)
+│   ├── composer.py         # Context 조합 (토큰 budget)
+│   ├── resolution.py       # 다해상도 텍스트 (L0/L1/L2)
+│   ├── forgetting.py       # 망각 파이프라인
+│   ├── consolidation.py    # 기억 통합
+│   └── sleep_cycle.py      # 수면 사이클 오케스트레이션
+│
+├── online/                 # RL 온라인 학습
+│   ├── policy.py           # Contextual Bandit (10d → 3 actions)
+│   ├── enhanced_policy.py  # Enhanced MLP (394d → 3 actions)
+│   ├── enhanced_encoder.py # 384d embedding + 10d features
+│   ├── replay_buffer.py    # Experience replay
+│   ├── autonomy.py         # Progressive autonomy (신뢰도 기반)
+│   ├── reranker.py         # RL Re-ranker (검색 결과 재정렬)
+│   ├── ab_comparator.py    # A/B 비교 테스트
+│   ├── gossip.py           # P2P Federated Learning (Gossip)
+│   ├── rule_verifier.py    # 불변 규칙 필터
+│   └── transport.py        # P2P 통신 계층
+│
+├── reward/                 # 보상 계산
+│   ├── calculator.py       # 종합 보상 계산기
+│   ├── feedback_detector.py # 사용자 피드백 감지 (긍정/부정/교정)
+│   ├── implicit_detector.py # 암묵적 보상 감지
+│   ├── korean_patterns.py  # (legacy) 한국어 패턴
+│   └── signals.py          # 보상 신호 정의
+│
+├── dataset/                # 학습 데이터 파이프라인
+│   ├── builder.py          # Episode → 학습 데이터 변환
+│   ├── splitter.py         # Train/Val/Test 분리
+│   └── stats.py            # 데이터셋 통계
+│
+├── extractor/              # Offline RL Feature Extractor
+│   ├── encoder.py          # Feature 인코더
+│   ├── model.py            # DualHeadDQN
+│   ├── dataset.py          # DQN 학습 데이터
+│   ├── policy.py           # 추출 정책
+│   └── trainer.py          # 학습 루프
+│
+└── selfplay/               # Self-play 데이터 생성
+    ├── engine.py            # Self-play 엔진
+    ├── llm_client.py        # Ollama 클라이언트
+    ├── memory_agent.py      # 메모리 에이전트
+    └── scenarios.py         # 시나리오 관리
+```
+
+## 4. 테스트 현황
+
+```
+tests/                      # 8,491 lines, 619 tests
+├── test_i18n.py            # 다국어 패턴 (ko/en)
+├── test_mcp_server.py      # MCP 서버
+├── test_mcp_bridge.py      # MCP 브릿지
+├── test_mcp_e2e.py         # MCP E2E
+├── test_graph_store.py     # ChromaDB 저장소
+├── test_graph_retriever.py # 하이브리드 검색
+├── test_knowledge_graph.py # 지식 그래프
+├── test_online_policy.py   # Contextual Bandit
+├── test_enhanced_policy.py # Enhanced Policy
+├── test_enhanced_encoder.py # Enhanced Encoder
+├── test_replay_buffer.py   # Replay Buffer
+├── test_reranker.py        # Re-ranker
+├── test_ab_comparator.py   # A/B 비교
+├── test_gossip.py          # Gossip Protocol
+├── test_transport.py       # P2P Transport
+├── test_autonomy.py        # Progressive Autonomy
+├── test_rule_verifier.py   # Rule Verifier
+├── test_reward_calculator.py # 보상 계산
+├── test_feedback_detector.py # 피드백 감지
+├── test_implicit_detector.py # 암묵적 보상
+├── test_resolution.py      # 다해상도 텍스트
+├── test_composer.py        # Context Composer
+├── test_forgetting.py      # 망각 파이프라인
+├── test_sleep_cycle.py     # 수면 사이클
+├── test_extractor.py       # Feature Extractor
+├── test_e2e_enhanced.py    # Enhanced E2E
+├── test_dataset_builder.py # 데이터셋 빌더
+├── test_schemas.py         # 스키마
+└── test_selfplay_engine.py # Self-play (ollama 필요, skip)
+```
+
+**결과**: 619 passed, 1 skipped (14.6s)
+
+## 5. 핵심 기술
+
+### 5.1 Contextual Bandit Policy
+- **State**: 394d feature vector (384d multilingual embedding + 10d hand-crafted)
+- **Actions**: SAVE(0), SKIP(1), RETRIEVE(2)
+- **Reward**: 사용자 피드백 기반 (긍정 +1.0, 부정 -1.0, 암묵적 보상 포함)
+- **학습**: Online MLP + Experience Replay + Epsilon-greedy
+
+### 5.2 GraphRAG Storage
+- **Vector Store**: ChromaDB + `intfloat/multilingual-e5-small` (384d)
+- **Knowledge Graph**: 트리플 기반 (`subject → predicate → object`)
+- **Hybrid Retrieval**: Vector similarity + Graph traversal 결합
+
+### 5.3 다해상도 텍스트
+- **L0**: 원본 텍스트
+- **L1**: 핵심 정보 추출 (주어+술어)
+- **L2**: 키워드 압축
+- Context Composer가 토큰 budget에 맞게 적절한 레벨 선택
+
+### 5.4 수면 사이클
+- 주기적으로 실행: 망각 → 통합 → 재인코딩
+- **망각**: 접근 빈도 낮은 기억 삭제 (pinned/immutable 보호)
+- **통합**: 유사 기억 병합
+- **재인코딩**: 임베딩 재생성
+
+### 5.5 다국어 지원 (i18n)
+- `LanguagePatterns` 데이터 클래스로 언어별 패턴 분리
+- 현재 지원: 한국어 (`ko`), 영어 (`en`)
+- `register()` 함수로 새 언어 추가 가능
+
+## 6. MCP 서버 도구 (12개)
+
+| 도구 | 설명 |
+|------|------|
+| `auto_search` | 매 턴 자동 기억 검색 |
+| `memory_save` | 기억 저장 |
+| `memory_search` | 명시적 기억 검색 |
+| `memory_update` | 기억 수정 |
+| `memory_delete` | 기억 삭제 |
+| `memory_stats` | 기억 통계 |
+| `memory_list` | 기억 목록 |
+| `memory_get` | 기억 상세 조회 |
+| `graph_query` | 지식 그래프 쿼리 |
+| `graph_add_triple` | 트리플 추가 |
+| `sleep_cycle` | 수면 사이클 수동 실행 |
+| `policy_stats` | 정책 통계 |
+
+## 7. 의존성
+
+### Core
+```
+pydantic>=2.0
+chromadb>=0.5.0
+sentence-transformers>=3.0
+mcp[cli]>=1.2.0
+```
+
+### Optional
+```
+[ko]   mecab-python3, mecab-ko-dic     # 한국어 형태소 분석
+[dev]  pytest, ruff                      # 개발
+[train] ollama, pandas, pyarrow, tqdm, matplotlib  # 학습
+```
+
+## 8. 설치 및 실행
+
+```bash
+# 기본 설치
+uv sync
+
+# 한국어 지원 포함
+uv sync --extra ko
+
+# 개발 환경
+uv sync --extra dev --extra ko
+
+# 테스트
+uv run pytest tests/ -q
+
+# MCP 서버 실행
+uv run python -m aimemory.mcp
+
+# OpenClaw 연동
+bash scripts/install_openclaw.sh
+```
+
+## 9. 환경 변수
+
+| 변수 | 기본값 | 설명 |
+|------|--------|------|
+| `AIMEMORY_DB_PATH` | `./memory_db` | ChromaDB 저장 경로 |
+| `AIMEMORY_LANGUAGE` | `ko` | 언어 (`ko`, `en`) |
+| `AIMEMORY_EMBEDDING_MODEL` | `intfloat/multilingual-e5-small` | 임베딩 모델 |
+| `AIMEMORY_LOG_LEVEL` | `WARNING` | 로그 레벨 |
+| `AIMEMORY_TOKEN_BUDGET` | `500` | 검색 토큰 budget |
+| `AIMEMORY_EPSILON` | `0.15` | Epsilon-greedy 탐색률 |
+
+## 10. 구현 이력
+
+| Step | 내용 | 상태 |
+|------|------|------|
+| 1 | 학습 데이터 확보 (Self-play + 공개 데이터) | ✅ |
+| 2 | Rule-Based Baseline + Online MLP Bandit | ✅ |
+| 3 | 망각 파이프라인 + 다해상도 저장 | ✅ |
+| 4 | 수면 사이클 + 기억 통합 | ✅ |
+| 5 | RL Re-ranker | ✅ |
+| 6 | P2P Federated Learning (Gossip) | ✅ |
+| 7 | MCP 서버 패키징 | ✅ |
+| 8 | GraphRAG Integration + Enhanced Policy | ✅ |
+| 9 | 다국어(i18n) + 설치 파이프라인 + 배포 준비 | ✅ |
+
+## 11. 배포 방법
+
+```bash
+# 1. 빌드
+uv build
+# → dist/aimemory-0.1.0.tar.gz
+# → dist/aimemory-0.1.0-py3-none-any.whl
+
+# 2. TestPyPI에 먼저 테스트
+uv publish --publish-url https://test.pypi.org/legacy/
+
+# 3. PyPI 배포
+uv publish
+# PyPI API token 필요: https://pypi.org/manage/account/token/
+
+# 4. 설치 확인
+pip install aimemory
+```
+
+## 12. 코드 규모
+
+| 항목 | 수치 |
+|------|------|
+| 소스 코드 | 51개 파일, 9,709 줄 |
+| 테스트 코드 | 31개 파일, 8,491 줄 |
+| 테스트 수 | 619 passed, 1 skipped |
+| 스크립트 | 14개 |
+| MCP 도구 | 12개 |
+| 지원 언어 | 2 (ko, en) |