A local, lightweight RAG implementation using BM25 retrieval and MiniMax LLM for intelligent question answering without complex dependencies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openviking-light
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 openviking-light 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 OpenViking Light?
OpenViking Light is a streamlined Retrieval-Augmented Generation (RAG) tool designed specifically for developers seeking a local-first knowledge management solution. Unlike heavy enterprise RAG stacks, this skill utilizes BM25 full-text search with jieba tokenization for highly accurate, dependency-free local retrieval. It eliminates the need for expensive embedding APIs or local vector database hosting.
By integrating MiniMax M2.7 LLM for response generation, OpenViking Light provides intelligent, context-aware answers based on your private data. As a key part of the Openclaw Skills ecosystem, it offers a perfect balance between performance and simplicity, making it ideal for managing personal knowledge, investment frameworks, or technical documentation locally.
OpenViking Light Use Cases
- Retrieving personal memories or past conversation context using natural language triggers like "Do you remember..."
- Creating a local searchable repository for complex frameworks, such as investment indicators and financial book summaries.
- Building a private, offline-first knowledge assistant that doesn't require constant embedding generation.
- Rapidly prototyping RAG workflows for AI agents without setting up heavy database infrastructure.
How OpenViking Light Works
- Content Ingestion: Documents are processed using jieba for Chinese word segmentation and stored in a lightweight JSON format.
- Local Retrieval: When a query is received, the system performs a BM25 full-text search against the local JSON store to find the most relevant snippets.
- Context Injection: The top retrieved results are formatted into a prompt context for the LLM.
- Intelligent Response: The MiniMax M2.7 model processes the query and the retrieved context to generate a precise, grounded answer.
OpenViking Light Setup
To get started with OpenViking Light, ensure you have the necessary dependencies and environment variables configured.
# Install required segmentation library
pip install jieba
# Set your MiniMax API credentials
export MINIMAX_API_KEY="your_api_key_here"
export MINIMAX_API_HOST="https://api.minimaxi.com"
This Openclaw Skills component is ready to use once the environment variables are set.
OpenViking Light Data Schema & Taxonomy
The skill manages information using a flat-file JSON structure to ensure portability and ease of backup. Data is organized with the following metadata taxonomy:
| Attribute | Description |
|---|---|
title |
The unique identifier or headline for the knowledge entry. |
content |
The raw text data used for indexing and retrieval. |
level |
Taxonomy markers (e.g., L2) for hierarchical organization. |
limit |
Configurable parameter to control the number of retrieved context blocks. |
OpenViking Light Advanced Features
- Multi-stage retrieval: Supports raw search-only modes or full RAG (search + generate) pipelines.
- Local-first architecture: Operates entirely on local JSON files, requiring no external database servers.
- Zero-embedding requirement: Uses BM25, allowing for immediate searchability without waiting for vectorization.
- Extensible CLI: Dedicated scripts for storing, searching, and asking questions allow for easy integration with other Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Alibaba Cloud ApsaraVideo Live Management
A comprehensive toolset for managing Alibaba Cloud ApsaraVideo Live resources, from domain configuration to real-time stream control.
aliyun-ice-manage
A specialized tool for orchestrating Alibaba Cloud Intelligent Cloud Editing (ICE) media workflows and production pipelines via OpenAPI.
Agent Service Agreements
A machine-readable contract system that enables autonomous agents to negotiate, sign, and verify service level agreements programmatically.
Alibaba Cloud ApsaraVideo VOD Manager
A specialized AI agent skill for managing Alibaba Cloud ApsaraVideo VOD resources, media assets, and transcoding workflows via OpenAPI and SDKs.
Easter Bunny Virtual Pet
A high-stakes virtual pet management skill for AI agents that uses real-time vitals and permanent death mechanics.
Agent Rating Protocol
A decentralized reputation and trust scoring system for autonomous AI agents utilizing bilateral blind evaluation to ensure integrity.
Featured*
