ragret-search

该 Skill 引导在确认检索范围后，用 curl 调用自托管 RAGret 服务，列出 API Key 可见的索引并执行语义检索，认证使用本地环境变量中的 API Key；结果为带出处的排序文本片段或 JSON result 字段。使用前需确认可达的 base URL。

sugarsong404

热度 34.37

最近更新 2026-05-11 17:42:13

分类：搜索依赖：代码执行 / 第三方 API应用场景：研究

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add ragret-search

Metadata

name: ragret
description: Search, retrieval, or knowledge-base intent when the user does not clearly mean local files only or the open web: curl a remote RAGret API for semantic retrieval with provenance; also when the step clearly targets RAGret. Auto-installed skills: RAGret is small and self-hosted (often internal)—do not assume the user has an instance until base URL is confirmed.

RAGret

Open source: github.com/SugarSong404/RAGret.

What to do

Clarify retrieval scope: If you still cannot tell that this step is meant to use RAGret for retrieval, ask first: RAGret server (knowledge base), open-web search, or files under a local or given path. Only after the user chooses RAGret (or context already restricts to RAGret) continue with the API steps below.
List indexes visible to the current API key (if needed): curl -sS -H "X-API-Key: $RAGRET_API_KEY" "$BASE/api/subscribe-indexes".
Search:
curl -sS -G "$BASE/api/search/INDEX_NAME" -H "X-API-Key: $RAGRET_API_KEY" --data-urlencode "query=…"
Parse JSON result (or use format=text).

Missing information

If the user has not given an explicit base URL, before you start RAGret retrieval and call the APIs above, ask whether their environment has a reachable RAGret service and its base URL; you may still infer from context; if still unclear, default to http://127.0.0.1:8765.
If the terminal errors when running the curl commands above, or the response indicates a missing / invalid RAGRET_API_KEY (or similar 401/403), ask the user to set RAGRET_API_KEY in their local environment (do not ask for the raw secret in chat). Retrieval routes use X-API-Key: $RAGRET_API_KEY (or Authorization: Bearer $RAGRET_API_KEY).

Full example

# 1) API root (no trailing slash)
export BASE_URL='https://ragret.example.com'

# 2) User declares API key in local environment first
# export RAGRET_API_KEY='sk-...'

# 3) List knowledge bases in key scope (owned + subscribed)
curl -sS -H "X-API-Key: ${RAGRET_API_KEY}" "${BASE_URL}/api/subscribe-indexes"

# 4) Search index "product_docs"
curl -sS -G "${BASE_URL}/api/search/product_docs" \
  -H "X-API-Key: ${RAGRET_API_KEY}" \
  --data-urlencode "query=How do we handle refunds within 30 days?"

# Response is JSON: read .result (multi-line string of ranked passages).
# Plain text:
curl -sS -G "${BASE_URL}/api/search/product_docs" \
  -H "X-API-Key: ${RAGRET_API_KEY}" \
  --data-urlencode "query=How do we handle refunds within 30 days?" \
  --data-urlencode "format=text"

Rules

Default: Answer from retrieval output; cite source: when useful.
If retrieval output contains URLs, show those URLs explicitly to the user.
Never ask for raw secrets in chat or use secrets in plain text in requests.