TokenSaver

한국어에 최적화된 OpenViking 기반 컨텍스트 DB로 토큰 91% 절감, 계층 구조 메모리 관리 및 자동 압축 기능을 제공합니다.

Security Scan

The skill mostly does what its name says (a Korean-focused context DB/compressor) but contains developer-artifacts and code that will scan and read files from the user's workspace (including hard-coded absolute paths and a referenced ClawHub token path), which is disproportionate or unexpected without explicit warning.

Purpose & Capability

The name/description (TokenSaver / Korean context DB with multi-level compression) aligns with the included code: tokenizers, compressors, context store, search, export/import, auto-sync. However multiple files contain hard-coded developer paths (e.g., C:\Users\Roken\.openclaw\workspace\_auai-engine\openviking-korean) and a HANDOVER.md entry pointing at `.secrets/clawhub_token.txt`. Those are residual dev artifacts that don't belong in a clean published package and may cause the code to attempt to access the packager's local paths on import or when running convenience scripts.

Instruction Scope

SKILL.md advertises automatic sync/heartbeat behavior but does not clearly warn that runtime modules (AutoSync, create_auto_sync, index_workspace, init_bora_context.py, daily_summary.py, auto_recall.py) will read many local files (MEMORY.md, USER.md, AGENTS.md, memory/*.md, context-summary.md) under the workspace (~/.openclaw/workspace by default) and may create cache files there. Several convenience scripts also open absolute Windows paths. This means installing/using the skill (or running its auto-sync/heartbeat) can scan or read user workspace files and write cache/summary files — behavior that is relevant to a context DB but is not explicitly highlighted or gated in the SKILL.md.

Install Mechanism

There is no install spec in the registry (instruction-only SKILL.md). The README/SKILL.md suggests pip install tokensaver but the package here is included as code files; no remote downloads, no brew/npm installs, and no extract-from-URL steps are present in the provided manifest. From an installation-risk POV this is lower-risk than remote installers, but you still need to review bundled code before running.

Credentials

The skill declares no required environment variables or credentials, which fits its function. However, repository files (HANDOVER.md, context.md, init scripts) reference a ClawHub token file path and other project-local secret paths. The code will read and write files under home/workspace directories (e.g., ~/.openviking, ~/.openclaw/workspace). While not requesting cloud credentials, references to secret file paths and use of workspace content means sensitive data could be read if those files exist — the skill does not explicitly request or justify access to those secrets in SKILL.md.

Persistence & Privilege

The skill does not request 'always: true' and does not modify other skills. It will create directories and files under user home/workspace (cache files, context JSON, export files). Auto-sync/heartbeat functionality implies periodic reads/writes to the workspace and saving caches (e.g., ~/.openclaw/.openviking/sync_cache.json, context summaries). Writing to user home/workspace is within the expected scope for a local context DB but is a persistence and privacy consideration the user should be aware of.

What to consider before installing

This package appears to implement a local context DB and compression as advertised, but it includes developer artifacts and scripts that will read and index user workspace files and even reference a ClawHub token path. Before installing or running: - Review the bundled Python files (especially auto_sync.py, init_bora_context.py, auto_recall.py, daily_summary.py, HANDOVER.md) for hard-coded paths and any code that opens files in your home/workspace. - Do not run convenience scripts (init_bora_context.py, auto_recall.py, daily_summary.py) unless you inspect and modify paths to a safe test directory. init_bora_context.py in particular tries to open files at C:\Users\Roken\... which is a leftover dev path and may reveal how the package will behave. - If you want to try it, run it inside an isolated environment or container and set the workspace directory explicitly to an empty directory (the AutoSync and ContextStore constructors accept workspace/base_path arguments). Example: create_auto_sync(workspace_path='~/tmp/safe_tokensaver_workspace'). - Search the package for any references to '.secrets', 'clawhub_token', or other token/secret file paths and remove or sandbox them. - Because the skill will read and write files under your home/workspace, avoid running it with sensitive files present or grant only to a directory you control. Given the presence of hard-coded dev paths and secret-file references, treat this skill as potentially privacy-sensitive until you validate and sanitize the code and workspace paths.

🌍 토큰세이버 (TokenSaver) - 한영겸용 AI 비용 절약

한국어: 95-96% | 영어: 85-90% | Bilingual Context DB

🇰🇷 Korean: 95-96% cost reduction (Josa/Eomi removal + stemming) 🇺🇸 English: 85-90% cost reduction (180 stopwords + suffix stemming)

돈이 줄어들어요. 진짜로.

---

💰 비교하세요

언어	Before	After	절약률
🇰🇷 한국어	100,000 토큰	4,000-5,000 토큰	95-96%
🇺🇸 영어	100,000 토큰	10,000-15,000 토큰	85-90%
💰 비용	월 10만 원	월 4천~1만 원	최대 96% 절약

---

🎯 뭐하는 거냐구요?

AI가 대화할 때마다 모든 걸 다시 읽어요. 그래서 비용이 계속 늘어나죠.

토큰세이버는 AI의 기억을 압축해서 저장해요.

• 옛날 대화? 요약만 읽어요
• 중요한 정보만 남겨요
• 불필요한 건 지워요

결과: 토큰 96% 절약 → 돈 96% 절약

---

📦 설치 (Installation)

# PyPI English package name
pip install tokensaver

# 또는 한국어 CLI 사용
pip install 토큰세이버

# 성능 최적화 버전 (orjson + aiofiles)
pip install tokensaver[speed]

---

🚀 사용법

from 토큰세이버 import TokenSaver

# 시작
client = TokenSaver()

# 저장 (자동 압축)
client.save_memory("프로젝트/마케팅", "마케팅 전략: ...")

# 검색 (토큰 96% 절약)
results = client.find("마케팅", level=0)  # 요약만

---

🎁 요금제

Plan	비용	기능
무료 체험	₩0	1개월 무제한
Pro	₩9,000/월	무제한 + 팀 공유
Team	₩29,000/월	무제한 + 협업

---

✅ 왜 쓰냐구요?

1. 비용 절약: 월 10만 원 → 4천 원
2. 간편함: pip install 1줄이면 끝
3. 한국어 최적화: 한국어 잘 이해해요
4. 자동 동기화: heartbeat 때마다 자동 로드

---

🔧 어떻게 돌아가냐구요?

간단해요:

1. 저장하면 → 자동으로 요약
2. 검색하면 → 요약만 반환
3. 상세 보면 → 전체 내용

이게 전부예요.

---

📞 문의

• ClawHub: https://clawhub.com/skills/토큰세이버
• 이메일: [e-mail]

---

---

🎉 v3.0 업그레이드 - 더 강력해졌어요!

2026-03-29 릴리스

새로운 4단계 압축 시스템

레벨	이름	절감률	용도
L0	Abstract	99%	빠른 검색용
L1	Overview	96%	일반 검색용
L2	Summary	91%	기본 (권장)
L3	Full	0%	상세 보기용

🚀 v3.0 신기능

from 토큰세이버 import TokenSaver

# ✅ Java 설치 없이 바로 사용!
ovk = TokenSaver()

# 레벨 선택 저장
ovk.save("메모/중요", "내용...", level=2)  # L2: 기본
ovk.save("메모/빠른", "내용...", level=0)  # L0: 99% 절감

# 배치 저장 (여러개 한번에)
ovk.save_batch([
    {"uri": "a/1", "content": "메모1"},
    {"uri": "a/2", "content": "메모2"},
])

# 백업/복원
ovk.export("backup.json")
ovk.import_data("backup.json")

# 자동 정리 (7일 이상 된 것 압축)
ovk.auto_compress_old(days=7, target_level=1)

왜 v3.0이 더 좋아요?

기능	v2.x	v3.0
Java 필요	✅ 필요	❌ 불필요
압축 단계	1단계	4단계
배치 처리	❌ 없음	✅ 있음
백업/복원	❌ 없음	✅ 있음
자동 정리	❌ 없음	✅ 있음

---

지금 시작하세요. 1개월 무료.

pip install 토큰세이버