llm-output-quality-monitor
Cloudflare Workers MCP server: llm-output-quality-monitor
我该使用它吗
质量与安全性
A
发现（1）
- LOWTool 'hallucination_scorer' description lacks action verb在 hallucination_scorer 中
基于对工具定义和协议合规性的自动分析。
上下文开销
~421token 数（工具定义）
~756 B典型响应大小
对注意力的影响极小（占 128k 上下文窗口的 0.33%）
这是每次将服务器的工具加载到模型上下文窗口时所消耗的大致 token 数。数值越高，可用于其他任务的注意力就越少。
安装
一键安装
将以下内容添加到你的 `claude_desktop_config.json` 文件中：
{
"mcpServers": {
"llm-output-quality-monitor": {
"url": "https://api.lazy-mac.com/llm-output-quality-monitor/mcp"
}
}
}远程端点
https://api.lazy-mac.com/llm-output-quality-monitor/mcpstreamable-http它能做什么
工具清单
工具（5）
🟢 只读🟡 写入🔴 删除⚪ 未知
⚪quality_validator(response, minLength, maxLength, strictFormat)
Validate LLM response quality based on length, format, and structure
输入模式
{
"type": "object",
"properties": {
"response": {
"type": "string",
"description": "LLM response to validate"
},
"minLength": {
"type": "number",
"description": "Minimum response length (default: 10)"
},
"maxLength": {
"type": "number",
"description": "Maximum response length (default: 10000)"
},
"strictFormat": {
"type": "boolean",
"description": "Enforce punctuation and capitalization"
}
},
"required": [
"response"
]
}⚪drift_detector(currentResponse, previousResponse, threshold)
Detect quality drift between current and previous LLM responses
输入模式
{
"type": "object",
"properties": {
"currentResponse": {
"type": "string",
"description": "Current LLM response"
},
"previousResponse": {
"type": "string",
"description": "Previous LLM response"
},
"threshold": {
"type": "number",
"description": "Drift threshold (0-1, default: 0.15)"
}
},
"required": [
"currentResponse",
"previousResponse"
]
}⚪hallucination_scorer(response, context)
Pattern-based heuristic risk scoring for LLM responses (0-100). Detects linguistic signals such as contradictory assertions, unsourced claims, and uncertainty markers. Not a semantic hallucination detector.
输入模式
{
"type": "object",
"properties": {
"response": {
"type": "string",
"description": "LLM response to analyze"
},
"context": {
"type": "string",
"description": "Reference context for grounding"
}
},
"required": [
"response"
]
}⚪schema_enforcer(response, schema)
Validate JSON response against schema
输入模式
{
"type": "object",
"properties": {
"response": {
"type": "string",
"description": "JSON response to validate"
},
"schema": {
"type": "object",
"description": "JSON schema definition"
}
},
"required": [
"response",
"schema"
]
}🟢consistency_check(responses)
Check consistency across multiple LLM responses
输入模式
{
"type": "object",
"properties": {
"responses": {
"type": "array",
"items": {
"type": "string"
},
"description": "Array of responses to compare"
}
},
"required": [
"responses"
]
}社区
证据
最近观测
已验证未记录版本5 个工具
已验证未记录版本5 个工具