← Back to Skills Marketplace
unisound-llm

unisound-similar-case-retrieval

by Unisound-LLM · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
93
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install unisound-similar-case-retrieval
Description
医生端临床科研 — 相似病例语义检索与可解释排序,锚点病例对候选池做类比推理辅助。
README (SKILL.md)

相似病例检索(科研辅助)

概述

面向临床科研与教学的「锚点病例 → 候选病例池」语义对齐能力:在用户提供的一组已脱敏摘要或结构化要点上,由内部医疗大模型完成类比排序、相似维度拆解与科研启发式提示

本实现不接真实院内病历库,候选病例须由调用方预先准备(导出队列、公开数据集子集、或文献病例摘要)。价值在于统一走公司内部医疗大模型,便于与后续真实检索管线对接时替换为向量召回 + 重排。

业界脉络(写法参考,非功能承诺)

近年医疗 AI 文献中常见的组合范式包括:**病例基推理(Case-based Reasoning)**与 RAG / GraphRAG 并用,在 EHR 长程结构数据上保留时间轴与合并症模式再做相似患者检索;也有工作将「指南证据检索」与「相似患者 exemplar」双通道融合以提升可解释性。本 skill 在接口层预留 anchor_case / candidate_cases 字段,语义上对齐上述「患者级类比」叙事,当前阶段以 LLM 重排与解释 为主。

OpenClaw 中的角色

  • 输入:锚点病例自然语言摘要 + 多条候选病例(每条含 idsummary)。
  • 输出:统一 JSON(data 为结构化回显与轻量统计,text 为 Markdown 级排序解读与科研注意点)。

快速开始

python3 scripts/run.py --input input.json --output output.json --appkey YOUR_KEY

输入字段(JSON)

字段 必填 说明
anchor_case 锚点病例摘要(主诉、关键体征、诊断线索、时间轴等)
candidate_cases 数组,元素含 idsummary
top_k 期望在解读中重点讨论的靠前条数,默认 5
task_hint 科研关注点,如「预后分层」「用药方案对照」

输出约定

与仓库内其他 LLM skill 一致:

{
  "skill": "相似病例检索",
  "status": "ok",
  "data": { },
  "text": "Markdown 解读"
}

参数

  • --input PATH必填。UTF-8 JSON 文件路径。
  • --output PATH:可选。落盘路径;省略则打印到 stdout。
  • --appkey STRING必填。调用内部医疗大模型的鉴权 key,由平台分配。

模型配置

  • endpoint:https://maas-api.hivoice.cn/v1/chat/completions
  • model:u1-insuremed
  • 鉴权:Bearer {appkey}

医疗与合规边界

输出仅供科研、教学或方法学讨论参考,不构成诊疗建议;使用真实患者数据前须完成脱敏与伦理审批。

Usage Guidance
Before installing, verify the publisher/package identity, use a protected app key, and only process de-identified, ethics-approved case summaries. Treat the output as research assistance, not clinical advice.
Capability Analysis
Type: OpenClaw Skill Name: unisound-similar-case-retrieval Version: 1.0.0 The skill is a legitimate implementation for medical case retrieval and ranking using the Unisound (hivoice.cn) medical LLM. The Python script (scripts/run.py) and documentation (SKILL.md) align with the stated purpose of clinical research assistance, using standard libraries to handle JSON input/output and making authorized API calls to a known endpoint. No evidence of data exfiltration, malicious execution, or prompt injection was found.
Capability Assessment
Purpose & Capability
The purpose, SKILL.md, and code are coherent: it ranks similar clinical cases and explains them with an internal medical LLM. This is still sensitive because the inputs may contain clinical case details.
Instruction Scope
The workflow is a user-run CLI with clearly described input and output fields, a capped top_k value, and a research-only clinical disclaimer; it does not instruct hidden or destructive actions.
Install Mechanism
There is no install script or dependency installer, and the Python code uses standard libraries. However, the packaged _meta.json owner/slug/publishedAt do not match the supplied registry metadata, so provenance should be verified.
Credentials
Outbound HTTPS access to the disclosed model endpoint and an app key are proportionate to the stated LLM-based retrieval purpose, but the registry metadata does not declare a primary credential even though the CLI requires --appkey.
Persistence & Privilege
The script only reads the user-specified input file and optionally writes a user-specified output file; it does not create background workers, persistent memory, privileged changes, or local credential-store access.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install unisound-similar-case-retrieval
  3. After installation, invoke the skill by name or use /unisound-similar-case-retrieval
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of "med-doctor-similar-case-retrieval" skill. - Provides semantic retrieval and explainable ranking of similar clinical cases for doctors’ research and teaching use. - Accepts anchor case summary and user-provided candidate cases; does not connect to internal hospital records. - Outputs structured JSON with ranked results and Markdown explanations, highlighting key analogies and research points. - Designed for integration with internal medical LLM, with fields and usage compatible with future vector retrieval pipelines.
Metadata
Slug unisound-similar-case-retrieval
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is unisound-similar-case-retrieval?

医生端临床科研 — 相似病例语义检索与可解释排序,锚点病例对候选池做类比推理辅助。 It is an AI Agent Skill for Claude Code / OpenClaw, with 93 downloads so far.

How do I install unisound-similar-case-retrieval?

Run "/install unisound-similar-case-retrieval" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is unisound-similar-case-retrieval free?

Yes, unisound-similar-case-retrieval is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does unisound-similar-case-retrieval support?

unisound-similar-case-retrieval is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created unisound-similar-case-retrieval?

It is built and maintained by Unisound-LLM (@unisound-llm); the current version is v1.0.0.

💬 Comments