extract-entities-multi

Agent Skills › zerogpu/zerogpu-router › extract-entities-multi

extract-entities-multi

GitHub

用于多语言文本的自定义标签命名实体识别，支持提取用户指定类型的实体。

agents/claude/skills/extract-entities-multi/SKILL.md zerogpu/zerogpu-router

Trigger Scenarios

从非英文段落中提取实体使用自定义标签进行多语言 NER

Install

npx skills add zerogpu/zerogpu-router --skill extract-entities-multi -g -y

More Options

Non-standard path

npx skills add https://github.com/zerogpu/zerogpu-router/tree/main/agents/claude/skills/extract-entities-multi -g -y

Use without installing

npx skills use zerogpu/zerogpu-router@extract-entities-multi

指定 Agent (Claude Code)

npx skills add zerogpu/zerogpu-router --skill extract-entities-multi -a claude-code -g -y

安装 repo 全部 skill

npx skills add zerogpu/zerogpu-router --all -g -y

预览 repo 内 skill

npx skills add zerogpu/zerogpu-router --list

SKILL.md

Frontmatter

{
    "name": "extract-entities-multi",
    "description": "Custom-label named-entity recognition across multiple languages (gliner2.5-multi-v1, 287M). Use when the user wants to extract entities with their own labels — people, organizations, locations, attributes, or any caller-defined entity type — from a passage that is not in English. Use extract-entities for English text; the two cost the same.",
    "allowed-tools": "Bash(zerogpu chat_completions *)",
    "argument-hint": "<text> (-l <label>... | --labels a,b,c) [-t <0..1>]"
}

Run multilingual custom-label NER. The request below holds the source text and the entity labels, given as flags (-l person -l company, --labels person,company, -t 0.4) or in plain words.

Run this with the Bash tool, filling in the labels and pasting the source text into the heredoc verbatim — no escaping, the quoted heredoc handles every shell metacharacter, newline, quote, and paren:

zerogpu chat_completions -m gliner2.5-multi-v1 --metadata '{"usecase":"ner","labels":["person","company"],"threshold":0.3}' <<'ZGPU_END_OF_INPUT'
<the source text, verbatim>
ZGPU_END_OF_INPUT

"usecase":"ner" is required exactly as written; the API rejects the request without it.
labels needs at least one label. If the request names none, ask the user which entity types to extract.
threshold is the minimum confidence in [0, 1]. Use 0.3 unless the request gives one.
--metadata is single-quoted JSON, so write any ' inside a label as '\''.

Output is a JSON object whose entities maps each label to the spans found.

At $0.02 / $0.05 per 1M input/output tokens this matches /zerogpu-router:extract-entities exactly, so the choice is about language, not cost: this model is built on mDeBERTa-v3-base for multilingual text, where extract-entities is English-only. It also supports "usecase":"json" and "usecase":"classification", the shapes /zerogpu-router:extract-json and /zerogpu-router:classify-structured use.

Savings note: only if the command output literally contains a line starting with 💰 ZeroGPU savings, append that exact line, unchanged, as the last line of your reply. If no such line is present, say nothing about savings and do not mention or suggest /zerogpu-router:cost-savings — this note is intentionally occasional, not shown every time.

Request: $ARGUMENTS