A local, pure-Python semantic memory engine that provides AI agents with persistent entity tracking, relationship mapping, and automated reflection.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install nova-memory
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 nova-memory 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 nova-memory?
nova-memory is a robust local memory skill designed to give AI agents a long-term, structured understanding of their environment and interactions. Built entirely in Python with no external API dependencies, it ensures maximum privacy and data security. It enables agents to remember specific facts, perform semantic searches across past interactions, and manage complex entity relationships within a local workspace.
By utilizing advanced indexing like TF-IDF and Cosine Similarity, this Openclaw Skills contribution allows for high-performance retrieval without the overhead of heavy vector databases. It integrates seamlessly into existing workflows, providing a bridge between raw logs and high-level project insights through its automated reflection and OLAP analysis modules.
nova-memory Use Cases
- Long-term fact retention for user preferences and project history.
- Building complex entity-relationship graphs for knowledge management.
- Semantic retrieval of historical events and lessons learned during development.
- Automated daily reflection to generate insights and summaries for long-term memory logs.
- Deep analytical reporting on memory trends and content diversity using local OLAP.
How nova-memory Works
- The system captures information through the
rememberfunction, assigning a unique ID and storing it as a local JSON file. - Inverted indexes map tags to memory IDs, while a vector index stores TF-IDF representations for every entry.
- When a query is initiated, the system uses Cosine Similarity to compare the query vector against stored vectors to find the most relevant context.
- Entity management allows for the storage of structured facts about specific subjects, which are then linked in a memory graph.
- Periodically, the
auto_reflectengine analyzes recent memories to generate high-level insights, which can be persisted back into the project's permanent documentation.
nova-memory Setup
To install and configure the nova-memory skill, follow these steps:
- Ensure you have
sklearnandnumpyinstalled in your Python environment. - Clone or place the skill directory into your workspace.
- Initialize the memory directory:
mkdir -p /workspace/memory/nova-memory/
- You can now use the CLI to test the memory system:
# Store a new memory
python -m nova_memory.scripts.memory_cli remember "Project Alpha uses Python 3.11" --tags dev,config
# Recall information
python -m nova_memory.scripts.memory_cli recall "What version of Python?"
nova-memory Data Schema & Taxonomy
The skill organizes data within /workspace/memory/nova-memory/ using a flat file structure for maximum portability.
| Component | Format | Description |
|---|---|---|
| Memory Entry | .json |
Individual files containing content, timestamp, tags, and entity IDs. |
| Inverted Index | JSON | Map of tag -> [memory_ids] for fast filtering. |
| Vector Index | NumPy/JSON | Stored TF-IDF vectors for semantic similarity calculations. |
| Entity Store | JSON | Key-value store of entity_name -> {attributes}. |
| OLAP Cache | SQLite | Local database for trend analysis and statistical reporting. |
nova-memory Advanced Features
- Automated Reflection: Generates synthesis reports from recent raw data to update long-term status files.
- Memory Graph: Visualizes connections between different entities and facts stored in the system.
- Semantic Search: Uses local machine learning techniques to find contextually relevant information without calling external LLMs.
- OLAP Analysis: Built-in SQLite engine to perform trend analysis, tag heatmaps, and content entropy calculations.
- Persistence Hooks: Simple
save()andload()mechanisms to ensure memory is never lost between agent restarts.
SKILL.md
Loading
Related Openclaw Skills
Video Narration
A high-precision technical skill for generating professional AI voiceovers with automated scene analysis and frame-accurate audio-video synchronization.
Iberia Insight Agent
A specialized research skill for deep-diving into Iberia brand history, product ecosystems, and global market positioning.
DataWorks Diagnoser
A specialized CLI tool for fetching and diagnosing Alibaba Cloud DataWorks task instance logs to identify and resolve execution failures.
Memory Focal System
An intelligent memory management framework that optimizes AI context by dynamically adjusting memory focal depth and applying tiered storage.
Humana Intelligence & Research Skill
An AI agent skill designed to provide structured intelligence, business maps, and market performance data for the Humana organization.
hugo-boss Background Context Skill
A comprehensive research tool designed to provide deep contextual insights into the hugo-boss brand history, products, and market strategy.
Featured*
