Knowledge Base with Faiss and Bailian (embedding and rerank)

使用 Python、FAISS、BM25、阿里云百炼 text-embedding-v4 与可选的 qwen3-rerank，维护基于文件目录的本地知识库；适用于在 OpenClaw 预先抽取文本后，遍历 chunks 与 T2Q 建立索引，以及对指定知识库或全部知识库做综合、语义或关键词查询。

This skill appears to do what it says: it builds and queries a local FAISS + BM25 knowledge base and sends text only to the Bailian embedding/rerank endpoints using the BAILIAN_SK key. Before installing: (1) confirm the registry metadata is corrected so the required environment variable (BAILIAN_SK) is clearly declared; (2) only provide a Bailian key you trust—embeddings and rerank calls send text to Bailian over HTTPS; (3) run the script in a controlled environment because it will read and write all files under the KB root (default /var/openclaw-kb); (4) review the included Python script (scripts/bailian_faiss_kb.py) yourself if you need to ensure no additional telemetry or unexpected network calls; and (5) if you cannot trust the Bailian endpoint or do not want external embeddings, do not provide the API key or avoid invoking the index/query operations that require network calls.

Type: OpenClaw Skill Name: rag-kb Version: 1.0.1 The skill bundle implements a local RAG (Retrieval-Augmented Generation) system using FAISS for semantic search and BM25 for keyword search. The Python script `scripts/bailian_faiss_kb.py` handles indexing, querying, and maintenance of a knowledge base stored in a local directory (defaulting to `/var/openclaw-kb`). It communicates exclusively with Alibaba Cloud's DashScope API (`dashscope.aliyuncs.com`) for text embeddings and reranking, using the `BAILIAN_SK` environment variable for authentication. The documentation in `SKILL.md` and the `references/` directory provides clear, non-malicious instructions for the agent to manage file layouts and content formatting. No evidence of data exfiltration, unauthorized execution, or prompt injection was found.