A privacy-first vector memory solution that uses Ollama embeddings and Qdrant local storage to provide semantic search without cloud dependencies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install local-vector-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 local-vector-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 Local Vector Memory?
Local Vector Memory is a specialized skill designed for developers who prioritize privacy and performance by keeping all vector embeddings and search operations on-premises. By integrating Ollama's powerful embedding models with Qdrant's high-performance vector database, this skill allows users to build a robust memory layer for their AI agents. It is a critical addition for Openclaw Skills users who need to handle sensitive data or work in air-gapped environments.
This skill excels at processing both English and Chinese text, making it a versatile choice for global applications. It replaces the need for costly and privacy-invasive cloud-based embedding APIs, providing a local alternative for semantic search, memory storage, and Retrieval-Augmented Generation (RAG) workflows.
Local Vector Memory Use Cases
- Creating a persistent, local memory layer for personal AI assistants within Openclaw Skills.
- Implementing semantic search across large repositories of local Markdown notes or technical documentation.
- Developing privacy-compliant RAG applications that never send data to third-party embedding providers.
- Optimizing AI responses by providing relevant context retrieved from a local knowledge base.
How Local Vector Memory Works
- The user installs the local-vector-memory package and pulls a supported embedding model via Ollama.
- The CLI or Python API is used to initialize a local Qdrant database instance on the user's machine.
- Text data is ingested, and the local Ollama service generates high-dimensional vector embeddings for each chunk.
- These vectors, along with original text and metadata, are stored securely in a local directory.
- Upon a search query, the system converts the query to a vector and performs a similarity search to return the most relevant results.
Local Vector Memory Setup
To get started with this skill, ensure you have Ollama installed and then follow these steps:
# Pull the recommended embedding model
ollama pull qwen3-embedding:4b
# Install the core package
pip install local-vector-memory
# Initialize your local vector database
lvm init
Local Vector Memory Data Schema & Taxonomy
The skill organizes its data using a structured local storage approach optimized for vector retrieval:
| Attribute | Specification |
|---|---|
| Embedding Models | Supports qwen3-embedding, bge-m3, and nomic-embed-text |
| Vector Storage | Qdrant binary storage located at ~/.local-vector-memory/qdrant |
| Text Chunking | Configurable (default 400 chars) with 50 char overlap |
| Data Metadata | Stores text content, source file paths, and similarity scores |
| Network Usage | 0% cloud dependency; restricted to localhost communication |
Local Vector Memory Advanced Features
- Direct integration with Openclaw Skills workflows via HEARTBEAT.md for automated reindexing.
- Support for large-scale directory reindexing using glob patterns with built-in path traversal protection.
- SSRF-protected Ollama communication, ensuring the embedding service remains strictly local.
- Customizable vector dimensions and chunking strategies to balance performance and accuracy.
- Structured JSON output for CLI searches, enabling seamless piping into other automation tools.
SKILL.md
Loading
Related Openclaw Skills
CN Content Factory
A comprehensive Chinese content production suite that transforms single ideas into platform-optimized, human-like social media copy.
Dream Journal
An intelligent dream recording and analysis tool that learns your unique subconscious language through dynamic imagery extraction.
Elon Review
A high-signal decision-testing skill that applies first-principles logic to product, strategy, and operations to eliminate bureaucracy and accelerate execution.
AI Resume Generator
A smart CLI-based resume builder designed for developers and engineers to transform raw data into professional, job-matched resumes.
No-News: International Tech News Aggregator
A high-performance news aggregator that fetches and summarizes the latest headlines from eight world-leading technology publications.
Bidding-Writing Professional Skill
A professional AI-driven bid writing system for end-to-end tender document preparation and quality control.
Featured*
