← 返回
未分类 Key 中文

Novyx Memory

Enterprise-grade persistent memory for AI agents — rollback, audit trails, knowledge graph, governed actions, time-travel debugging, and 60+ commands coverin...
面向 AI Agent 的企业级持久化内存——支持回滚、审计追踪、知识图谱、受控操作、时光旅行调试及 60+ 命令覆盖……
novyxlabs novyxlabs 来源
未分类 clawhub v3.0.0 1 版本 99823.3 Key: 需要
★ 0
Stars
📥 565
下载
💾 0
安装
1
版本
#audit#governed-actions#knowledge-graph#latest#memory#multi-agent#persistence#rollback#semantic-search#time-travel#tracing

概述

Novyx Memory

The only memory skill with rollback, cryptographic audit trails, and governed actions. 60+ commands covering 100% of Novyx Core — memory, drafts, knowledge graph, context spaces, time-travel replay, eval scoring, execution tracing, and policy-controlled actions.

Why Novyx Memory

Every other memory skill gives you save and search. Novyx gives you the full stack:

CapabilityNovyxBuilt-inSuperMemoryMem0memory-tools
------------------
Semantic search
Time-travel rollback
Cryptographic audit trail
Draft review workflow
Knowledge graph
Context spaces (multi-agent)
Governed actions
Execution tracing
Memory health eval
Confidence & decay

When to Use This Skill

  • Agent needs memory that persists across sessions
  • You need to undo, roll back, or audit what the agent remembered
  • Multiple agents need shared context spaces
  • Agent actions need policy approval before execution
  • You want a knowledge graph built from agent observations
  • You need time-travel debugging ("what did the agent know at 3pm?")
  • CI pipeline needs a memory health gate

Automatic Behavior

When active, this skill:

  • Auto-recalls relevant memories before each response and injects them as context
  • Auto-saves each user message and agent response to persistent storage
  • Skips trivial messages (under 15 characters) to conserve API calls
  • Filters out previously injected context to prevent feedback loops

Commands (60+)

Memory Basics

CommandWhat it does
----------------------
!remember [--tags t1,t2]Save a specific fact or note with optional tags
!search Semantic search with relevance scores
!list [--tag tag] [--limit N]List stored memories with optional filters
!forget Find and delete all memories matching a topic
!undo [N]Delete the last N saved memories (default: 1)
!supersede Replace an old memory with a corrected version

Draft Review Workflow

Stage memory changes before committing — like git for agent memory.

CommandWhat it does
----------------------
!draft [--branch name]Create a reviewable draft
!draftsList pending drafts
!draft-diff Show field-level diff for a draft
!merge-draft Commit draft to canonical memory
!reject-draft Reject a draft without saving
!branch View all drafts in a branch
!merge-branch Merge all drafts in a branch at once
!reject-branch Reject all drafts in a branch

Memory Links

CommandWhat it does
----------------------
!link [relation]Create a directed link between memories
!unlink Remove a link
!links Show all links for a memory

Knowledge Graph

Build structured knowledge from unstructured observations.

CommandWhat it does
----------------------
`!triple \ \`Add a knowledge triple
!triples [--subject x] [--predicate y]Query triples with filters
!entities [--type t]List all entities
!entity Get entity with all connected triples
!del-triple Delete a triple
!del-entity Delete entity and cascade triples
!graph [--relation r]View memory graph edges

Context Spaces (Multi-Agent Collaboration)

Isolated memory namespaces that can be shared between agents.

CommandWhat it does
----------------------
!space [--description d]Create a context space
!spacesList accessible spaces
!space-memories [query]List or search within a space
!update-space [--name n]Update space metadata
!del-space Delete a space (owner only)
!share Share a space with another user
!sharedList spaces shared with you
!accept Accept a share invitation
!revoke Revoke a share

Replay (Time-Travel Debugging)

Scrub through the full history of your agent's memory.

CommandWhat it does
----------------------
!timeline [N]Full timeline of memory operations
!snapshot Reconstruct memory state at a point in time
!lifecycle Full biography of a single memory
!replay Complete history of a memory
!replay-diff Diff memory state between two timestamps
!recall-at What would search have returned in the past?
!replay-drift Memory composition drift analysis

Audit & Rollback

CommandWhat it does
----------------------
!audit [N]Last N operations with SHA-256 hashes (default: 10)
!audit-verifyVerify cryptographic audit chain integrity
!rollback Rewind memory to a point in time ("1h", "2 days ago", ISO)
!rollback-preview Preview what rollback would do without executing
!rollback-historyList past rollback operations

Cortex (Autonomous Intelligence)

AI that watches your memory and generates insights automatically.

CommandWhat it does
----------------------
!cortexCortex status and last run
!cortex-runManually trigger a Cortex cycle
!insightsList auto-generated insights
!cortex-configView current configuration

Eval (Memory Health Scoring)

Quantify your agent's memory quality with weighted scoring.

CommandWhat it does
----------------------
!evalRun memory health evaluation (recall 40%, drift 30%, conflict 15%, staleness 15%)
!eval-gate [threshold]CI gate — pass/fail on memory quality (default: 0.7)
!eval-historyPast evaluation scores with trends
!eval-driftDetect memory drift over time
!healthQuick memory stats overview

Tracing (Execution Audit)

Track multi-step agent workflows with cryptographic verification.

CommandWhat it does
----------------------
!trace Create a new execution trace
`!trace-step [--type THOUGHT\ACTION\OBSERVATION]`Add a step
!trace-complete Finalize trace with RSA signature
!trace-verify Verify trace chain integrity

Control (Governed Actions)

Policy-evaluated agent actions with approval workflows.

CommandWhat it does
----------------------
!pendingList pending approval requests
!approve Approve a pending action
!policyView active security policies
!actions [N]Recent action history

Overview

CommandWhat it does
----------------------
!statusMemory usage, tier, and rollback count
!dashboardFull dashboard with health score and activity
!contextCurrent memory context snapshot
!statsDetailed memory statistics
!helpList all 60+ commands

Key Differentiators

  • Rollback: !rollback 1h restores memory to exactly one hour ago. No other memory skill supports this.
  • Audit trail: Every operation logged with SHA-256 hashes forming a tamper-proof chain. !audit-verify checks integrity.
  • Draft workflow: Stage memory changes in branches, review diffs, merge or reject — like git for memory.
  • Knowledge graph: Build structured triples from unstructured observations. Query relationships, not just content.
  • Context spaces: Isolated memory namespaces with sharing. Perfect for multi-agent architectures.
  • Governed actions: Policy-evaluated agent actions with Solo/Team/Enterprise approval workflows.
  • Time-travel: !recall-at 3pm "project status" shows what the agent would have found at 3pm yesterday.
  • Eval scoring: Quantified memory health with CI gate integration. Catch memory rot before it matters.
  • Execution tracing: Cryptographically signed audit trail for multi-step agent workflows.
  • Graceful degradation: If rate limits are hit or the API is unavailable, the agent continues working — it never crashes.

Tiers

TierPriceMemoriesRollbacksFeatures
--------------------------------------------
Free$05,00010/monthMemory, search, audit, rollback
Starter$12/mo50,000100/month+ Drafts, eval, conflict resolution
Pro$39/mo500,000Unlimited+ Cortex, replay, knowledge graph, tracing
Enterprise$199/moUnlimitedUnlimited+ Governed actions, team sharing, SSO

Configuration

Requires NOVYX_API_KEY environment variable. Get a free key at https://novyxlabs.com (5,000 memories, no credit card).

Options:

  • autoSave (default: true) — Automatically save messages
  • autoRecall (default: true) — Automatically recall context before each response
  • recallLimit (default: 5) — Max memories to recall per query

版本历史

共 1 个版本

  • v3.0.0 当前
    2026-03-30 18:21 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,105 📥 829,865
ai-agent

Find Skills

guipi888
场景驱动+关键词双模式技能发现工具。当用户用自然语言描述场景/需求(如"我想做一个海报""帮我分析股票"),或明确说"安装技能/find skills/找个skill"时,自动从官方内置、本地已安装、SkillHub、虾评、GitHub、C
★ 1,469 📥 534,165
ai-agent

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,398 📥 322,926