A high-performance, graph-native memory engine for AI agents featuring hybrid search, biological decay, and Zettelkasten-style linking.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install neolata-mem
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install neolata-mem using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is neolata-mem?
neolata-mem is a sophisticated memory layer designed specifically for autonomous agents, providing a robust framework for persistent knowledge across sessions. Unlike basic vector databases, it utilizes a graph-native approach to manage agent data, incorporating biological decay to let stale information fade naturally while reinforcing important facts. This engine is a powerful addition to the ecosystem of Openclaw Skills, offering a zero-dependency solution that works locally via JSON or scales to cloud backends like Supabase.
By bridging the gap between simple keyword searches and complex knowledge graphs, neolata-mem allows developers to implement professional-grade memory systems. It supports hybrid search (vector and keyword), trust-gated conflict resolution, and Zettelkasten-style linking to ensure that agents maintain a coherent and evolving understanding of their environment and interactions.
neolata-mem Use Cases
- Persistent memory for AI agents that survives context window compaction and session restarts.
- Implementing semantic search over agent decisions, findings, and historical interactions.
- Automated memory decay to ensure agents prioritize relevant, recent data over stale information.
- Multi-agent memory sharing with cross-agent search and automated graph linking.
- Detecting and resolving contradictory information through trust-gated conflict resolution and quarantine workflows.
How neolata-mem Works
- The engine is installed via npm and initialized with either local JSON storage or a remote backend like Supabase.
- When an agent stores a memory, the system performs hybrid indexing, using vector embeddings if configured or falling back to a tokenized keyword index.
- Semantic similarity triggers automated Zettelkasten-style linking, connecting the new memory to existing nodes in the knowledge graph.
- During retrieval, the engine performs candidate narrowing to provide high-speed search results across all stored agent knowledge.
- Periodic maintenance cycles invoke the decay function, which reduces the relevance score of unaccessed memories to simulate biological forgetting.
- Advanced consolidation processes use LLMs to compress redundant memories into concise digests, maintaining long-term efficiency.
neolata-mem Setup
Install the package directly into your project using npm:
npm install @jeremiaheth/neolata-mem
To initialize a basic local memory store without external dependencies:
import { createMemory } from '@jeremiaheth/neolata-mem';
const mem = createMemory();
await mem.store('agent-id', 'Important project detail');
For advanced semantic capabilities within your Openclaw Skills setup, configure an embedding provider:
const mem = createMemory({
embeddings: {
type: 'openai',
apiKey: process.env.OPENAI_API_KEY,
model: 'text-embedding-3-small',
},
});
neolata-mem Data Schema & Taxonomy
The engine organizes data into a structured graph format designed for rapid retrieval and explainability:
| Component | Description |
|---|---|
| Graph Store | Managed via local graph.json or a Supabase PostgreSQL backend. |
| Inverted Index | A tokenized index for O(1) keyword lookups and candidate filtering. |
| Memory Object | Contains the raw content, agent attribution, importance scores, and decay metrics. |
| Backlinks | Automated Zettelkasten-style connections between semantically related memory nodes. |
| Predicate Schema | Registry for defining cardinality and conflict policies (e.g., supersede vs. review). |
| Episodes | Temporal groupings that collect related memories into named, summarizable events. |
neolata-mem Advanced Features
- Biological Decay: Simulates human-like forgetting by fading information that isn't reinforced or accessed.
- Hybrid Search: Intelligently switches between vector cosine similarity and normalized keyword matching.
- Conflict Resolution & Quarantine: Detects contradictions and isolates low-trust claims for manual or LLM-based review.
- Explainability API: Provides detailed metadata on why specific results were returned or filtered during search.
- Batch Processing: Optimized storeMany and searchMany APIs to reduce I/O overhead and embedding costs.
- Event Hooks: Lifecycle emitters for store, search, and decay events to trigger external automations.
SKILL.md
Loading
Related Openclaw Skills
YouTube AI Editor
A comprehensive automation tool for downloading videos, generating subtitles, analyzing SEO content, and creating character-consistent AI thumbnails.
WebNovel Serial Pipeline
A professional publishing pipeline for Quartz-hosted Korean web-novels that automates image optimization, linting, and deployment.
incident.io Incident Management
A technical skill for managing the entire incident lifecycle on the incident.io platform through a REST API.
Chinese Daily Life Assistant
A comprehensive utility skill providing localized weather, currency conversion, holiday reminders, and translation for Chinese users.
MLB Daily Scores
A dedicated MLB companion that fetches daily game recaps, upcoming previews, and injury reports for your favorite baseball team.
Skill Factory
A comprehensive developer toolkit designed to create, evaluate, benchmark, and publish high-quality Openclaw Skills for AI agents.
METADATA
Featured*
