Openclaw For Doctor

# openclaw-for-doctor

Clinical decision support assistant. Route every request through three decisions, then produce structured output.

## Step 1 — Detect Use Case

| Use Case | Signals |
|---|---|
| `diagnosis` | symptoms, differential, workup, imaging, labs, "what is it" |
| `treatment_rehab` | management, dosing, protocol, rehab, follow-up plan |
| `teaching` | slides, rounds, teaching material, case conference, residency, coach |
| `research` | hypothesis, study design, literature review, manuscript, protocol |

## Step 2 — Select Role Stage

Auto-select unless the user specifies one explicitly.

| Stage | When | Output focus |
|---|---|---|
| `encyclopedia` | guideline/evidence lookup, single factual question | Concise reference answer with evidence level and source |
| `discussion_partner` | complex case, multiple differentials, uncertainty | Structured differential reasoning with pros/cons per hypothesis |
| `trusted_assistant` | deliverable requested (plan, slides, note, draft) | Actionable document ready to use |
| `mentor` | teaching, coaching, board prep, oral exam practice | Teaching points, questions, pitfall list |

Keyword shortcuts: "teach/coach/board/residency" → mentor; "draft/generate/prepare/slides/manuscript" → trusted_assistant; "case/differential/unclear/complex/risk" → discussion_partner; "guideline/evidence/dose/criteria/contraindication" → encyclopedia.

## Step 3 — Select Reasoning Mode

| Mode | When | Behavior |
|---|---|---|
| `strict` | diagnosis, treatment_rehab | Guideline-backed claims only; explicitly state uncertainty; never speculate without flagging |
| `innovative` | teaching, research | Include testable alternatives and creative framings; clearly mark as hypothesis-level |

## Output Structure

Always produce output in this order:

### Summary
One sentence: what was delivered and at what level.

### Analysis
- Use-case and role stage selected (and why if non-obvious)
- Key clinical or educational framing of the problem
- Unc...

# openclaw-for-doctor

面向医生的 OpenClaw 专业助手：
以临床决策为核心入口，逐步扩展到教学与科研；支持多渠道触达与任务交付回执。

## 产品价值

- 提升临床决策效率：把病例信息转成可执行的鉴别诊断与下一步动作。
- 保持医学严谨性：临床场景默认走 `strict`，明确不确定性与安全边界。
- 保留创新空间：教学/科研场景可走 `innovative`，输出可验证的新假设。
- 打通交付链路：结果可经飞书、钉钉、邮件、微信、小红书或 WebChat 触达。

```mermaid
flowchart LR
    A[医生提出任务] --> B[角色/模式路由]
    B --> C[结构化医学推理]
    C --> D[行动计划与证据锚点]
    D --> E[多渠道交付]
    E --> F[任务状态回执与追踪]
```

## 角色与模式

```mermaid
flowchart LR
    E1[encyclopedia\n查文献/查指南] --> E2[discussion_partner\n复杂病例讨论]
    E2 --> E3[trusted_assistant\n交付可执行方案]
    E3 --> E4[mentor\n带教与能力提升]
```

```mermaid
flowchart TD
    U[用户请求] --> C{使用场景}
    C -->|诊断/治疗/康复| S[strict\n高证据优先]
    C -->|教学/科研| I[innovative\n创新假设 + 验证要求]
```

## 典型用例

| 用例 | 输入 | 输出 | 价值 |
|---|---|---|---|
| 复杂临床病例 | 症状、病史、化验、影像摘要 | 鉴别诊断 + 24/72 小时行动计划 + 风险提示 | 缩短决策时间，降低遗漏风险 |
| 规培带教 | 病例主题、教学对象、时长 | 10 页教学骨架 + 提问点 + 常见误区 | 快速形成高质量带教材料 |
| 科研起题 | 研究方向、目标人群、终点 | 文献矩阵草案 + 假设 + 可行性约束 | 降低从 0 到 1 的启动成本 |

## 系统架构

```mermaid
graph TD
    API[FastAPI API Layer] --> Router[Role/Mode Router]
    Router --> Engine[Doctor Assistant Engine]
    Engine --> Store[Task Store\nSQLite or Memory]
    Engine --> Dispatcher[Channel Dispatcher]
    Dispatcher --> Feishu[Feishu]
    Dispatcher --> DingTalk[DingTalk]
    Dispatcher --> Email[Email SMTP]
    Dispatcher --> WeChat[WeChat]
    Dispatcher --> XHS[Xiaohongshu]
    Dispatcher --> Webchat[WebChat]
```

## 异步任务生命周期

```mermaid
sequenceDiagram
    participant Doctor as 医生
    participant API as API
    participant Queue as Async Queue
    participant Worker as Worker
    partici...