← 返回 Skills 市场
Rag Retriever
作者
yuyonghao-123
· GitHub ↗
· v0.1.1
· MIT-0
158
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install yuyonghao-rag-retriever
功能描述
RAG 2.0 检索系统，支持中文分词，混合向量与关键词搜索，智能文档分块与加权重排，自动生成 RAG 格式输出。
安全使用建议
This skill appears to implement a legitimate RAG retriever, but exercise caution because: (1) the code will call the OpenAI Embeddings API if an OPENAI_API_KEY is available — that is not declared in the registry metadata or SKILL.md; providing a key will cause your texts to be sent to OpenAI. (2) npm install pulls heavy native deps (transformers, onnxruntime, sharp), so review resource/compatibility and run in a sandbox if possible. Before installing, either: (A) review src/embeddings.js and other source files to confirm what data is sent externally, (B) avoid setting OPENAI_API_KEY to force use of the local SimpleEmbedding fallback, or (C) run the skill in an isolated environment and inspect network traffic. If you plan to provide an API key, make sure you understand what documents/queries will be transmitted and consider data-sanitization or policy constraints. If you want to be extra safe, run tests locally and audit the code paths that call external APIs and write to ./data/.
功能分析
Type: OpenClaw Skill
Name: yuyonghao-rag-retriever
Version: 0.1.1
The skill bundle implements a legitimate RAG (Retrieval-Augmented Generation) system featuring hybrid search (BM25 and vector), document chunking, and support for both OpenAI and local Transformers.js embeddings. The code (src/rag2.js, src/retriever.js) is well-structured, includes extensive tests, and follows standard practices for document retrieval systems. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; the use of a Hugging Face mirror (hf-mirror.com) and standard environment variables for API keys are consistent with the tool's stated purpose.
能力评估
Purpose & Capability
The libraries and files (LanceDB, jieba, HuggingFace transformers, BM25 implementation) are appropriate for a RAG retriever and align with the description. The included local model/tokenizer artifacts also make sense for local embedding support. No obvious unrelated dependencies are present.
Instruction Scope
SKILL.md instructs npm install and running the CLI and JS APIs and documents using external embedding providers, but it does not mention that the code will call the OpenAI Embeddings API. The runtime instructions do not declare or warn about network requests or the need to supply an OpenAI API key, which the code will use if present.
Install Mechanism
There is no formal install spec in the registry entry, but SKILL.md/README instruct users to run 'npm install'. package.json and package-lock.json will pull many native and heavy dependencies (transformers, onnxruntime, sharp, etc.), which is expected for local transformer support but can be large and may require native build/runtime dependencies. This is expected for the skill's purpose but worth knowing.
Credentials
Registry metadata declares no required environment variables, but src/embeddings.js reads process.env.OPENAI_API_KEY and will POST user text to https://api.openai.com/v1/embeddings if used. That means sensitive data (documents or queries) can be sent to an external API if a key is provided — this external credential access is not declared in the skill metadata.
Persistence & Privilege
The skill does write local caches and model/tokenizer artifacts under ./data/ (embedding cache, lancedb, model-cache). It does not request permanent 'always: true' privileges and does not attempt to modify other skills or global agent settings. Local file writes are expected for this functionality.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install yuyonghao-rag-retriever - 安装完成后，直接呼叫该 Skill 的名称或使用
/yuyonghao-rag-retriever触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v0.1.1
- Updated package version to 0.1.1 in package.json.
- No other functional or documentation changes in this release.
v0.1.0
Initial release of RAG 2.0 document retrieval system for OpenClaw.
- Implements document chunking with configurable size and overlap
- Supports simple TF-normalized vector embedding and LanceDB storage
- Provides hybrid retrieval: vector search combined with BM25 keyword search (RRF fusion)
- Features source citation, context enhancement for RAG prompts, and Chinese/English mixed search with jieba tokenizer
- Includes CLI tools, JavaScript API, and basic testing
- Lays out roadmap for embedding model integration, hybrid search, reranker, and multi-collection management
元数据
常见问题
Rag Retriever 是什么？
RAG 2.0 检索系统，支持中文分词，混合向量与关键词搜索，智能文档分块与加权重排，自动生成 RAG 格式输出。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 158 次。
如何安装 Rag Retriever？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install yuyonghao-rag-retriever」即可一键安装，无需额外配置。
Rag Retriever 是免费的吗？
是的，Rag Retriever 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Rag Retriever 支持哪些平台？
Rag Retriever 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Rag Retriever？
由 yuyonghao-123（@yuyonghao-123）开发并维护，当前版本 v0.1.1。
推荐 Skills