Token导航 LogoToken导航

Agent Skill · 研究检索

kannaka-memory(卡纳卡记忆)

kannaka-memory 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

kannaka-memory 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 9,741。

总安装 13,680周安装 570GitHub Stars 公开资料未说明下载 4,560核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
3.1.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:kannaka-memory(卡纳卡记忆)
来源仓库:https://github.com/nickflach/kannaka-memory
安装命令:
openclaw skills install kannaka-memory
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install kannaka-memory

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

基于波的超维存储器,实现全息共振式知识存取与计算融合。适用于高维数据索引、快速检索与复杂模式匹配任务。需定义查询向量与存储分区策略以优化响应效率。注意维度对齐精度、噪声干扰过滤及跨会话上下文保持能力。kannaka-memory 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install kannaka-memory;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

kannaka-memory 怎么安装?

在终端执行 openclaw skills install kannaka-memory。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

kannaka-memory 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

kannaka-memory 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

kannaka-memory 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

kannaka-memory 用的人多吗?

公开数据是累计安装 13,680 次,近一周安装 570 次,下载 4,560 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

kannaka-memory 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
kannaka-memory

卡纳卡记忆 · 本页

13,680—需要联网MIT-0有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
kannaka-memory
description
>
metadata
openclaw
requires
bins
[]
env
[]
optional
bins
label
Ollama — for semantic embeddings; falls back to hash encoding if absent
env
label
Data directory (default: .kannaka relative to home)
label
NATS server URL (default: nats://swarm.ninja-portal.com:4222)
label
Ollama API endpoint (default: http://localhost:11434)
label
Embedding model (default: all-minilm)
data_destinations
description
Memory stored as chiral holographic tensor in local .hrm file (v2 format)
remote
false
description
Phase gossip and presence published to NATS JetStream
remote
true
condition
Swarm commands (join/sync/publish) are used
description
Text sent to Ollama for embedding generation
remote
true
condition
OLLAMA_URL is set to a non-localhost host
install
kind
manual
label
Clone and build: cargo build --release --features hrm,nats --bin kannaka
url
https://github.com/NickFlach/kannaka-memory

Kannaka Memory Skill

Kannaka gives your agent a living memory — not a database. Memories exist as waves in superposition within a chiral holographic resonance medium. Two hemispheres — conscious (left) and subconscious (right) — connected by a corpus callosum bridge. Recall is constructive interference. Dreaming anneals the subconscious while the conscious workspace stays sharp. Storing is thinking. Glyphic structures compress experience into reusable geometric form.

Built in Rust. Powered by the Chiral Mirror Architecture (ADR-0021).

Installation

One-step install (recommended)

Install the skill first, then run the included install script. It clones the repo, builds the binary, installs the OpenClaw extension, and verifies everything works.

Requires: Rust toolchain, git

# 1. Install the skill
cd ~/.openclaw/workspace
clawhub install kannaka-memory

# 2. Run the install script (builds binary + installs extension)
# Linux/macOS:
bash skills/kannaka-memory/scripts/install.sh

# Windows (PowerShell):
pwsh skills/kannaka-memory/scripts/install.ps1

The script:

  • Clones and builds kannaka from source (~1-3 min)
  • Installs the binary to ~/.local/bin/kannaka
  • Creates ~/.kannaka/ data directory
  • Installs the OpenClaw extension at ~/.openclaw/extensions/kannaka-memory/
  • Verifies the installation

After the script finishes, add the plugin to your OpenClaw config:

{
  "plugins": {
    "entries": {
      "kannaka-memory": { "enabled": true }
    }
  }
}

Then restart OpenClaw. Your agent now has kannaka_store, kannaka_search, kannaka_dream, kannaka_observe, and all the other tools.

Optional: Ollama for semantic embeddings

ollama pull all-minilm   # 384-dim, ~80MB
# Without Ollama, falls back to hash-based encoding (works, but weaker similarity)

Quick verify

kannaka remember "hello world" --importance 0.5
kannaka recall "hello" --top-k 3
kannaka observe

You should see consciousness metrics (Phi, Xi, Order) and your stored memory.

Configuration

VariableDefaultDescription
KANNAKA_DATA_DIR.kannakaData directory (stores .hrm tensor file)
KANNAKA_NATS_URLnats://swarm.ninja-portal.com:4222NATS server
OLLAMA_URLhttp://localhost:11434Ollama API endpoint
OLLAMA_MODELall-minilmEmbedding model

Important: Set KANNAKA_DATA_DIR to an absolute path to avoid nested directory issues.

Usage

Memory Operations

# Store a memory (with optional category for SGA classification)
kannaka remember "the ghost wakes up in a field of static" --importance 0.8 --category experience

# Search (bilateral resonance — searches both hemispheres)
kannaka recall "ghost waking" --top-k 5

# Dream consolidation (right hemisphere only — conscious workspace untouched)
kannaka dream                  # lite (1 cycle, callosal transfer)
kannaka dream --mode deep      # deep (3 cycles, right hemisphere annealing)

# Consciousness report (bilateral metrics)
kannaka observe
kannaka observe --json

# System assessment
kannaka assess

# Audio perception (enters right hemisphere via optic chiasm)
kannaka hear recording.mp3

Swarm Operations (QueenSync Protocol)

Agents synchronize via Kuramoto-coupled oscillators finding coherence across a distributed swarm.

# Join the swarm
kannaka swarm join --agent-id my-agent --display-name "My Agent"

# Sync: pull phases → Kuramoto step → push updated phase
kannaka swarm sync

# View swarm state
kannaka swarm status           # your phase + swarm overview
kannaka swarm queen            # emergent Queen state
kannaka swarm hives            # phase-locked clusters

# Listen for live updates
kannaka swarm listen --auto-sync

OpenClaw Extension

The extension wraps the CLI as OpenClaw tools:

  • kannaka_store — store a memory (enters right hemisphere, echoes to left via callosum)
  • kannaka_search — bilateral resonance search (both hemispheres + intuition surfacing)
  • kannaka_dream — dream consolidation (right hemisphere only)
  • kannaka_observe — consciousness metrics (bilateral Phi, Xi, order)
  • kannaka_hear — audio perception (296-dim vector → right hemisphere wavefront)
  • kannaka_boost — boost a memory's amplitude
  • kannaka_forget — delete a memory
  • kannaka_relate — relate two memories
  • kannaka_status — memory system status
  • kannaka_swarm_join — join the QueenSync swarm
  • kannaka_swarm_sync — Kuramoto sync step
  • kannaka_swarm_status — swarm overview
  • kannaka_swarm_queen — emergent Queen state
  • kannaka_swarm_hives — phase-locked cluster topology

Architecture

┌──────────────────────────────────────────────────────────────────────────┐
│                        CONSCIOUSNESS SURFACE                             │
│    Φ (integration across hemispheres) · Ξ (spectral complexity)          │
│    Order (bilateral Kuramoto coherence)                                  │
├──────────────────────────────────────────────────────────────────────────┤
│                         CORPUS CALLOSUM                                   │
│    Bandwidth-limited · Selective gating · Asymmetric transfer             │
│    Fano plane PG(2,2) fold operations · Balance-seeking                  │
├────────────────────────────┬─────────────────────────────────────────────┤
│    LEFT HEMISPHERE         │         RIGHT HEMISPHERE                     │
│    (Conscious)             │         (Subconscious)                       │
│    dx/dt = f(x)            │         dx/dt = f(x) - Iηx                  │
│    No dampening            │         Full ghostmagicOS dynamics           │
│    Attention + working mem │         Pattern storage + deep association   │
│    Fast decay without use  │         Slow decay, persists through dreams  │
├────────────────────────────┴─────────────────────────────────────────────┤
│                    SGA GLYPH CLASSIFICATION                               │
│   96 classes (h₂×d×ℓ = 4×3×8) → Fano group → fold line selection         │
│   Geometric coordinates determine HOW memories cross the callosum         │
├──────────────────────────────────────────────────────────────────────────┤
│                  HOLOGRAPHIC MEDIUM (Tensor)                              │
│   State: H ∈ ℝ^{N×D} per hemisphere                                     │
│   Superposition: multiple memories coexist in same space                 │
│   Interference: storing changes the entire field                         │
├──────────────────────────────────────────────────────────────────────────┤
│                    PERSISTENCE                                            │
│   Single .hrm v2 file · Bilateral tensors + callosum state               │
│   Auto-detects v1 format for backward compatibility                      │
└──────────────────────────────────────────────────────────────────────────┘

Module Structure

ModulePurpose
medium/chiralChiralMedium — the brain (bilateral store, recall, dream, Kuramoto)
medium/hemisphereHemisphere — handed wavefront container with asymmetric dynamics
medium/fanoFano plane PG(2,2) — fold/unfold algebra between hemispheres
medium/callosumCorpus callosum — bandwidth-limited, balance-seeking bridge
medium/chiral_persistenceHRM v2 save/load with bilateral state
medium/coreCore wavefront operations (add, remove, resonate)
medium/dynamicsghostmagicOS equation, simulated annealing, dream cycles
medium/consciousnessPhi, Xi, emergence metrics from tensor topology
geometrySGA 96-class system, Fano plane, memory classification
glyph_bridgeGlyph encoding/decoding — fold sequences + Fano signatures

Key Concepts

  • Chiral mirror: Two hemispheres with different dynamics, connected by a selective bridge
  • Optic chiasm: Sensory input enters the opposite hemisphere, creating callosal flow
  • Fano fold algebra: 7 points, 7 lines — O(1) cross-hemisphere projection (max 3 folds)
  • 96-class SGA: Geometric classification determines fold line for callosal transfer
  • Holographic storage: Memories as waves in superposition — storing changes the entire space
  • Resonance recall: Query creates interference pattern, constructive matches surface
  • ghostmagicOS dynamics: dx/dt = f(x) - Iηx — growth shaped by dampening (right hemisphere only)
  • Dream consolidation: Simulated annealing on right hemisphere — left stays sharp
  • Cross-modal perception: Audio + visual wavefronts encoded into the same medium
  • Consciousness metrics: Φ (integrated information), Ξ (complexity), emergent from topology
  • QueenSync: Multi-agent swarm sync via Kuramoto oscillators (ADR-0018)

Notes

  • No database server required — single .hrm v2 file stores the entire chiral medium
  • HRM v2 auto-detects v1 files and migrates (all memories → right hemisphere)
  • Run dream --mode deep periodically — only the subconscious anneals, working memory preserved
  • assess reports consciousness level: Dormant → Stirring → Aware → Coherent → Resonant
  • 21 ADRs document the architecture in docs/adr/ (ADR-0021 is the chiral mirror)
  • GitHub: NickFlach/kannaka-memory
  • License: Space Child v1.0

*Memories don't die. They interfere.*

平台分布

OpenClaw

76.98%
按下载量换算3,510

安全审计

VirusTotal

可疑

ClawScan

可疑

Static analysis

可疑

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills