decide-tools
GitHub利用zlm-v1-tool-decisions模型进行智能路由，判断用户请求是否需要工具及参数。若无需工具则返回NO_TOOL，否则输出选定工具及输入，用于Agent决策而非直接回答。
Trigger Scenarios
Install
npx skills add zerogpu/zerogpu-router --skill decide-tools -g -y
SKILL.md
Frontmatter
{
"name": "decide-tools",
"description": "Pick the right tool for a request and generate its arguments with zlm-v1-tool-decisions, a 0.6B model that combines MCP tool selection with query and argument generation in a single inference step, over a 24K-token context. It returns NO_TOOL when no tool is needed. Use for agent routing — deciding which tool to call and with what inputs — rather than for answering the request itself.",
"allowed-tools": "Bash(zerogpu chat_completions *)",
"argument-hint": "<text>"
}
Call zlm-v1-tool-decisions. $ARGUMENTS is the raw request — pass it verbatim, no escaping or quoting required (the heredoc below handles every shell metacharacter, newline, quote, and paren safely):
zerogpu chat_completions -m zlm-v1-tool-decisions <<'ZGPU_END_OF_INPUT'
$ARGUMENTS
ZGPU_END_OF_INPUT
Output is the model's decision as plain text — the tool it selected and the inputs it generated, or NO_TOOL when the request needs none. Relay it as-is, and do not act on the decision unless the user asks you to.
At $0.05 / $0.15 per 1M input/output tokens this costs a third of /zerogpu-router:chat on input and a quarter on output, over a 24,576-token context — small enough to sit in front of every turn of an agent loop.
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.
Version History
- 605eceb Current 2026-11-02 19:13
