An enterprise-grade Retrieval-Augmented Generation system offering semantic embeddings, hybrid search, and intelligent reranking for high-performance document retrieval.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install rag-retriever-v3
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 rag-retriever-v3 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 RAG Retriever V3.0?
RAG Retriever V3.0 is a sophisticated Retrieval-Augmented Generation framework designed to provide AI agents with high-fidelity document intelligence. By integrating this module into your Openclaw Skills workflow, you gain access to a multi-layered retrieval pipeline that combines semantic vector similarity with traditional keyword search. This version introduces significant upgrades over its predecessor, including support for both local Transformers.js models and cloud-based OpenAI embeddings, ensuring flexibility for various deployment environments.
The system is engineered for developers who need more than just simple vector lookups. It provides a robust architecture for managing complex knowledge bases, featuring enterprise-ready tools like Cross-Encoder reranking and automatic source citation. Whether you are building a local research assistant or a cloud-scale support bot, RAG Retriever V3.0 delivers the technical foundation required for accurate, context-aware AI interactions.
RAG Retriever V3.0 Use Cases
- Building intelligent enterprise knowledge bases for internal technical documentation.
- Enhancing AI chatbot accuracy by providing context-aware answers with verified source citations.
- Performing semantic searches across multilingual datasets with specialized Chinese language optimizations.
- Creating high-performance research tools that require hybrid search capabilities to balance precision and recall.
How RAG Retriever V3.0 Works
- Initialization: The system sets up the LanceDB vector store and initializes the chosen embedding provider such as Xenova or OpenAI.
- Document Ingestion: Text files are processed using a recursive chunking strategy to maintain semantic coherence.
- Embedding and Indexing: Content chunks are converted into vectors and indexed alongside BM25 metadata for hybrid search.
- Hybrid Retrieval: The skill executes simultaneous vector and keyword searches, merging results via the Reciprocal Rank Fusion (RRF) algorithm.
- Reranking: A Cross-Encoder model re-evaluates the top candidates to ensure the most relevant context is prioritized.
- Citation Management: The system generates a final prompt for the LLM, complete with numbered citations and a detailed source list.
RAG Retriever V3.0 Setup
First, navigate to the skill directory and install the necessary dependencies:
cd skills/rag-retriever-v3
npm install
Initialize the database and configuration:
node src/cli.js init
Add your first document to the system to start utilizing Openclaw Skills capabilities:
node src/cli.js add ./document.txt '{"source":"Internal Wiki","title":"Onboarding Guide"}'
RAG Retriever V3.0 Data Schema & Taxonomy
The skill organizes data using LanceDB, ensuring high-speed vector and metadata management. Below is the primary configuration schema:
| Field | Description | Type |
|---|---|---|
| dbPath | The local file path for the LanceDB instance | String |
| collectionName | The specific name of the document collection | String |
| chunking | Configuration for chunk size, overlap, and recursive strategy | Object |
| embeddingOptions | Model definitions for Xenova or OpenAI providers | Object |
| rerank | Settings for enabled status and Cross-Encoder model type | Object |
| citation | Formatting rules for source attribution (numbered/footnote) | Object |
RAG Retriever V3.0 Advanced Features
- Multi-Model Embedding: Toggle between local Xenova models (BGE, MiniLM) and OpenAI cloud embeddings for Openclaw Skills users.
- Hybrid Search Logic: Seamlessly combines vector similarity and BM25 keyword matching using the RRF algorithm.
- Intelligent Reranking: Uses Cross-Encoders to perform deep semantic re-scoring, significantly improving retrieval precision.
- Comprehensive Source Management: Automatically tracks document metadata for complete provenance and verifiability.
- Semantic Chunking: Beyond fixed-size splits, this skill supports recursive strategies to preserve paragraph and sentence integrity.
SKILL.md
Loading
Related Openclaw Skills
BlockBeats API Skill
A comprehensive crypto market intelligence skill providing real-time news, institutional ETF flows, on-chain analytics, and global macro indicators.
zzheng Legal Article Generator
A specialized automation tool for legal professionals to generate trending WeChat Official Account articles from real-time news sources.
Web Intel
A sophisticated routing layer that orchestrates multiple extraction engines to provide high-fidelity web search and content retrieval.
Facebook Manager v2
A comprehensive Facebook management tool for Openclaw that automates Fanpage engagement and personal account data retrieval using Meta Graph APIs.
Azure Speech TTS
A comprehensive text-to-speech integration that converts plain text or SSML into high-fidelity audio files using Azure's neural voices.
PPT-with-Nano-Banana
A specialized skill for generating sketchy, whiteboard-style PowerPoint presentations featuring Chinese-first content and strategic color accents.
Featured*
