road-trip-planner-x-6

原始内容

---
name: road-trip-planner
description: >
  基于高德地图 personal-map skill 规划多日自驾游路线，生成连续路线个人地图小程序二维码，并输出每日行程详情。当用户提到自驾、公路旅行、路线规划、旅游行程、生成地图、规划环线、控制驾驶时间或查询沿途里程时使用。支持闭合环线设计和跨天里程补全。
  Plan multi-day road trip routes using Amap APIs. Generate continuous-route personal map QR codes and daily itinerary details. Triggered when users mention road trips, route planning, travel itineraries, map generation, loop routes, driving time control, or mileage queries. Supports closed-loop design and cross-day mileage completion. (China only — requires Amap API Key and @lbs-amap/personal-map skill)
---

# Road Trip Planner / 自驾游路线规划

> China only — 本 skill 基于高德地图 API，仅适用于中国大陆地区。  
> Dependency: [@lbs-amap/personal-map](https://clawhub.ai/lbs-amap/personal-map) (ClawHub) / [GitHub](https://github.com/amap-demo/amap-sdk-skills)

## Trigger Conditions / 触发条件

Trigger when the user has any of these intents:

当用户出现以下任意意图时，主动调用本 skill：

- Plan a road trip / road trip route / loop route (规划自驾游路线 / 公路旅行 / 环线)
- Design a travel itinerary involving multi-city driving (设计涉及多地驾驶的旅游行程)
- Generate a personal map / travel map / route QR code (生成个人地图 / 旅游地图 / 路线二维码)
- Control daily driving time / avoid fatigued driving (控制每日驾驶时间 / 避免疲劳驾驶)
- Query mileage, gas stations, hospitals along the route (查询沿途里程、加油站、医院等补给点)

## Dependencies / 前置依赖

| Dependency | Source | Description |
|-----------|--------|-------------|
| `lbs-amap/personal-map` | [ClawHub](https://clawhub.ai/lbs-amap/personal-map) | Amap Web Service API wrapper / 高德 Web 服务 API 封装。安装命令：`openclaw skills install personal-map`。提供 `AMapPersonalMapClient` 类（import 路径 `scripts.amap_personal_map_client`），本仓库自身不含该脚本。 |
| Amap API Key | [Amap Open Platform](https://lbs.amap.com/) | Required env var: `AMAP_API_KEY` / 需配置环境变量 |

### Dependency Availability Check / 依赖可用性检查

Before any API call, verify that the `personal-map` skill is installed and the API key is configured. If either is missing, stop and tell the user what's needed — do not attempt to proceed.

在任何 API 调用之前，先检查 `personal-map` skill 是否已安装以及 API Key 是否已配置。任一缺失则停止并告知用户缺失项，不要尝试继续。

```
Checklist / 检查清单:
1. personal-map skill installed? → No → 提示: "请先安装 personal-map skill，运行: openclaw skills install personal-map"
2. AMAP_API_KEY set? → No → 提示: "请先配置高德 API Key，获取地址 https://lbs.amap.com/"
3. AMapPersonalMapClient importable? → No → 提示: "personal-map skill 版本可能不兼容，请确认其提供 AMapPersonalMapClient 类"
```

Set the environment variable in the active shell; never put a real key in the skill or source code:

```bash
# macOS / Linux
export AMAP_API_KEY='your_key'

# PowerShell
$env:AMAP_API_KEY = 'your_key'

# Windows cmd
set "AMAP_API_KEY=your_key"
```

## Core Principles / 核心原则

1. **No fatigued driving / 不疲劳驾驶**：Keep daily actual driving time under 6 hours; 4–5 hours is a planning target, while short sightseeing or transfer days may be shorter. For high-altitude, mountainous, or winding roads, further reduce driving time.（每日实际驾驶时间控制在 6 小时以内；4-5 小时是规划目标，纯玩或短途换乘日可更短。高原、山区、弯道多的路段进一步压缩。）

2. **Continuous route / 连续路线**：Merge all daily waypoints into one continuous route, either a closed loop or with a clear endpoint.（所有天的节点合并为一条连续路线，首尾闭合或明确终点。）

3. **Closed loop / 闭合环线**：If start and end are the same city, ensure the endpoint is preserved after deduplication.（如果起点和终点是同一城市，确保终点节点在去重后仍然保留在路线末尾。）

4. **Daily flexibility / 每日弹性**：Leave buffer room in the itinerary; avoid over-scheduling.（行程中保留机动空间，避免把时间排得太满。）

## Workflow / 工作流程

### Step 1: Confirm itinerary framework / 确认行程框架

Confirm with the user:
- Start point, end point (loop or not)
- Total number of days
- Must-visit attractions / cities
- Travel season (affects road conditions and accommodation)
- Exact departure dates, vehicle type, and any driving/health constraints when safety or road choice depends on them

与用户确认以下信息：起点、终点（是否环线）、总天数、必去景点/途经城市、出发季节、准确出发日期、车型及会影响路线选择的驾驶/健康限制。

If the date is unspecified, provide a provisional seasonal plan and explicitly mark the real-time checks as pending.

### Step 2: Design daily waypoints / 设计每日节点

Split waypoints by day following these rules:
- Target 4–6 hours of driving per day, never exceed 6 hours; short sightseeing or transfer days may be shorter
- Each day's `points` list includes the core locations passed that day
- The end city of each day must also serve as the start point for the next day (for mileage calculation)

按天拆分节点：每天驾驶 4-6 小时，不超过 6 小时；每天的 points 列表包含当天途经的核心地点；终点城市需同时作为下一天的起点。

### Step 3: Call Amap APIs / 调用高德 API

Uses `AMapPersonalMapClient` provided by the `personal-map` skill. The class is at `scripts.amap_personal_map_client` inside that skill's directory — this is the canonical import path. This repo itself contains no scripts.

```python
import os
from scripts.amap_personal_map_client import AMapPersonalMapClient

# Check API key BEFORE creating client
api_key = os.getenv("AMAP_API_KEY")
if not api_key:
    print("AMAP_API_KEY not set. Get a key at https://lbs.amap.com/ and run:")
    print("  Set AMAP_API_KEY in your active shell (see Dependencies above).")
    # stop here — do not proceed

# Create client (auto-reads AMAP_API_KEY from env):
client = AMapPersonalMapClient()
# Or pass explicitly: client = AMapPersonalMapClient(api_key="your_key")
```

**Normalize every response before using it.** The client does not raise API failures, but a failure can be either an error dict or an error dict as the first item of a list (for example, POI search). Always check both shapes:

```python
def api_error_message(result):
    if isinstance(result, dict) and "error" in result:
        return result.get("message", str(result["error"]))
    if (
        isinstance(result, list)
        and result
        and isinstance(result[0], dict)
        and "error" in result[0]
    ):
        return result[0].get("message", str(result[0]["error"]))
    return None

geo = client.maps_geo("北京市朝阳区", "北京")
if error := api_error_message(geo):
    # Stop handling this location; never reuse another location's coordinates.
    raise ValueError(f"Cannot resolve this location: {error}")

# Convert external fields once. All later workflow steps use only this shape.
point = {
    "name": "地点名称",
    "lon": float(geo["longitude"]),
    "lat": float(geo["latitude"]),
    "poiId": "",             # fill from the verified POI-search result
    "required": False,         # True for the start, end, and user must-visit points
}
```

For each location:
1. `maps_geo(address, city)` — get precise coordinates. Pass the city name as second arg for better accuracy.
2. `maps_text_search(keywords, city, offset=5)` — get poiId. Use `offset=5` to limit results (API default is 20).
3. Verify search result coordinates are within 30km of geo coordinates to prevent cross-city mismatches
4. `time.sleep(0.3)` between API calls to avoid QPS throttling

对每个地点执行：`maps_geo` 获取坐标 → 立即标准化为 `lon`/`lat` → `maps_text_search` 获取 poiId → 验证距离 < 30km → sleep 0.3s。后续步骤不得直接混用 API 返回的 `longitude`/`latitude` 与内部 `lon`/`lat` 字段。

**Error handling / 错误处理**：
- If `AMapPersonalMapClient` is not found → check the pre-flight checklist in the Dependencies section, inform user, and **stop**
- If `AMAP_API_KEY` is empty → inform user and **stop**
- For every response → call `api_error_message()` before reading result fields
- If `maps_geo` fails → do not substitute a previous coordinate. Stop QR generation when this is a start, end, or must-visit point; otherwise list the optional point as unresolved and exclude it from routing
- If `maps_text_search` fails or has no result within 30km → retry with `{city_name} {location_name}`. For a required point, stop QR generation and ask the user to disambiguate; for an optional point, retain it only in the text itinerary as unresolved and do not give it a fake poiId
- If QPS exceeded (`CUQPS_HAS_EXCEEDED_THE_LIMIT` in error message) → wait 1s and retry once

#### 3.1 Supply point search per day / 每日补给搜索

After all waypoint coordinates are resolved, search for supply points near key nodes on **days that enter sparsely populated areas or have long uninterrupted stretches**:

对于进入人烟稀少区域或存在长距离无补给路段的日期，在关键节点附近搜索补给点：

```python
# For each day flagging a supply concern (long stretch, remote area):
# Search gas stations and hospitals near the day's midpoint or last town before uninhabited area
result = client.maps_around_search(
    keywords="加油站|医院",
    location=f"{lon},{lat}",    # "经度,纬度" 格式
    radius=5000,               # 搜索半径 5km（默认 1000m）
    offset=5
)
if error := api_error_message(result):
    print(f"补给搜索失败: {error}")
```

**When to search / 何时搜索**：
| Condition / 条件 | Action / 动作 |
|------------------|---------------|
| Day mileage > 300km / 日里程超 300km | Search gas stations at midpoint / 在中点搜索加油站 |
| Entering uninhabited area (>100km no town) / 进入无人区（>100km 无城镇） | Search all supply points at the last town before entry / 在进入前最后城镇搜索全部补给点 |
| High-altitude segment (>3500m) / 高海拔路段（>3500m） | Search hospitals at the nearest low-altitude town / 在最近的低海拔城镇搜索医院 |
| All days / 每天 | Search at least one hospital near the day's endpoint (overnight city) / 在每天终点城市至少搜索一家医院 |

### Step 4: Driving route planning with cross-day completion / 驾车路径规划（跨天补全）

When calculating daily mileage, **must** insert the previous day's endpoint as the start of the current day's points list:

```python
for i, day in enumerate(days):
    points = [p for p in day["points"] if "lon" in p]
    if i > 0:
        prev = [p for p in days[i-1]["points"] if "lon" in p]
        if prev:
            points.insert(0, prev[-1])
    # Drive segment by segment using maps_direction_driving:
    #   For each adjacent pair in points, call:
    #     result = client.maps_direction_driving(
    #         origin=f"{points[j]['lon']},{points[j]['lat']}",
    #         destination=f"{points[j+1]['lon']},{points[j+1]['lat']}"
    #     )
    #   Sum distances for the day's total.
    #   Parameter format: "longitude,latitude" (no spaces).
```

计算每日里程时，**必须**把前一天的终点插入当天 points 列表头部作为起点。
`maps_direction_driving` 的 `origin` / `destination` 参数格式为 `"经度,纬度"`（逗号分隔，无空格）。

### Step 5: Generate personal map QR code / 生成个人地图二维码

#### 5.1 Collect valid points / 收集有效点

Collect only normalized points with a non-empty verified `poiId`. Preserve insertion order (day-by-day). Before collecting, confirm that all required points (start, end, user must-visits) resolved successfully. If any did not, do not generate a QR code that claims to be the complete route.

收集所有天的有效点（已标准化且 poiId 已验证），按天顺序保持插入顺序。收集前必须确认起点、终点和用户必去点均已解析；任一失败时不得生成声称完整的二维码路线。

#### 5.2 Deduplicate by identity / 按身份去重

Rules (apply in order):

1. **Identity / 身份**：Use `poiId` as the primary identity. Only when no poiId is available for a non-QR text item, treat points as duplicates when their coordinates are within a small tolerance; never deduplicate by name alone.
2. **Closed-loop anchor / 环线锚点**：If the first and last points have the same identity, keep both, and remove only middle occurrences of that identity. This ensures the route stays closed.
3. **Intermediate duplicates / 中间重复**：For all other repeated identities, keep the first occurrence and remove later ones. Keep the day-by-day itinerary data unchanged; this rule applies only to the QR route list.
4. **Adjacent duplicates / 连续重复**：Remove one only when two adjacent QR points have the same identity. Do not infer a duplicate solely because their display names match.

```
Deduplication flow:
  identities = [p["poiId"] for p in points]
  if len(identities) < 2:
      → stop; do not generate a route QR code
  elif identities[0] == identities[-1]:
      → keep points[0] and points[-1], remove middle occurrences of this identity
  else:
      → for all repeated identities, keep first occurrence and remove rest
```

#### 5.3 Trim to ≤ 16 nodes / 精简至 ≤ 16 节点

If the deduplicated list exceeds 16 nodes, trim using this priority order:

| Priority | Keep | Drop |
|----------|------|------|
| 1 (highest) | Start and end points (环路锚点) | — |
| 2 | User-specified must-visit attractions | — |
| 3 | Major scenic spots / landmarks (景点/地标) | Rest stops, service areas (服务区) |
| 4 | Geographically isolated nodes (唯一覆盖该区域的点) | Dense urban clusters (keep 1 per city) |
| 5 | POI with verified poiId | Optional unresolved POI (text itinerary only; never add to QR route) |

**Geographic distribution principle / 地理分布原则**：When choosing between nodes of equal priority, prefer nodes that are farther apart (ensure the route visually represents the full journey). Avoid dropping the only node in a sparse region.

**Minimum trim / 最小精简**：Only remove as many nodes as needed to reach 16. Never remove the start or end anchor.

#### 5.4 Data integrity before calling API / 调用 API 前的数据完整性

```python
# Build lineList structure
point_info_list = [
    {"name": p["name"], "lon": p["lon"], "lat": p["lat"], "poiId": p["poiId"]}
    for p in deduplicated_points
]
line_list = [{
    "title": "Route Name / 路线名称",
    "pointInfoList": point_info_list
}]

# Validate before calling maps_schema_personal_map
if len(point_info_list) > 16:
    raise ValueError(f"Still {len(point_info_list)} nodes, must be ≤ 16")
if len(point_info_list) < 2:
    raise ValueError("A route QR code needs at least a resolved start and end point")
if any(not p["poiId"] for p in point_info_list):
    raise ValueError("Empty poiId found")
```

#### 5.5 Call API and download QR / 调用 API 并下载二维码

```python
result = client.maps_schema_personal_map(
    orgName="路线名称",
    lineList=line_list,
    sceneType=3          # 3 = route planning mode (仅创建路线)
)

# Check for API errors before reading result fields
if error := api_error_message(result):
    print(f"地图生成失败: {error}")
    # fallback: output the coordinates as text so user can manually create route
else:
    qr_url = result["qr_code_url"]
    # Download QR image:
    import urllib.request
    qr_path = "/tmp/road_trip_qr.png"
    urllib.request.urlretrieve(qr_url, qr_path)
    # Display inline using the local absolute path:
    # ![路线地图](/tmp/road_trip_qr.png)
    # Fallback link: [在 Amap 中打开](qr_url)
```

1. Call `maps_schema_personal_map` with `sceneType=3` (route planning mode / 仅创建路线)
2. Check for error in result dict (no exceptions thrown)
3. Download QR code image and display inline
4. Always include the fallback URL link

收集已验证节点 → 按 poiId/坐标身份去重（保留闭环首尾）→ 构建 lineList → 按优先级精简至 ≤ 16 节点 → 数据完整性校验 → 调用 `maps_schema_personal_map`(sceneType=3) → 检查错误 → 下载二维码 → 显示备用链接。

### Step 6: Output Markdown itinerary / 输出 Markdown 行程表

Output directly in the conversation as Markdown, including:
- Itinerary overview table (day, route, mileage, duration, road type, attractions) with total row
- Daily detailed itinerary (waypoints table with coordinates and altitude, supply points)
- Practical tips (altitude sickness, road conditions, tickets, temperature, seasonal notes)
- QR code image and fallback link

在对话中直接输出 Markdown：行程概览表格（含总里程合计行）、每日详细行程（含海拔和补给点）、实用提示、二维码图片和备用链接。

### Step 6.1: Verify date-sensitive facts / 核验时效信息

Before presenting a final itinerary, verify the route against the actual departure dates. Static seasonal guidance and route-planning API results are not proof that a road, attraction, or service is currently available.

出最终行程前，必须按准确出发日期核验。静态季节提示和路径规划 API 结果均不能证明道路、景区或服务设施当前可用。

- Check official traffic/road-authority notices for closures, controls, and weather-related restrictions
- Check current weather warnings for high-altitude, mountain, desert, or remote segments
- Check attraction opening status, ticket/reservation rules, and holiday calendar/traffic-freeway policy for that year
- Mark each unavailable source as `未核验` rather than inferring a current status; advise the user to recheck immediately before departure

## Key Constraints / 关键约束

| Constraint / 约束 | Description / 说明 |
|-------------------|-------------------|
| Dependency check / 依赖检查 | Before any API call, verify `personal-map` skill is installed AND `AMAP_API_KEY` is set. Fail fast with clear user-facing messages if missing. |
| Max nodes per route / 单条路线节点上限 | Amap `lineList` max 16 nodes per `pointInfoList` |
| poiId required / poiId 必填 | `maps_schema_personal_map` requires non-empty poiId |
| poiId strategy / poiId 获取策略 | `maps_text_search` first → retry with city name. Required points must resolve; optional unresolved points stay in text only and never receive a fake poiId. |
| API error pattern / API 错误模式 | API failures do not throw. Check an error dict or an error dict in the first list item before reading result fields. |
| QPS limit / QPS 限制 | ≥ 0.3s interval between calls, or `CUQPS_HAS_EXCEEDED_THE_LIMIT` |
| API Key / API Key | Set `AMAP_API_KEY` env var (auto-read by `AMapPersonalMapClient()` with no args) or pass explicitly |
| Region / 地域限制 | China only — Amap (高德) covers mainland China |
| External dependency / 外部依赖 | This repo contains no scripts. `AMapPersonalMapClient` lives in the `personal-map` skill at `scripts.amap_personal_map_client`. Install via `openclaw skills install personal-map`. |

## Output Template / 输出模板

```markdown
# [Route Name] Road Trip Guide / 自驾行程指南

> Start/End: [City]　> Total: ~X km　> Days: X　> Max Altitude: ~Xm

## Itinerary Overview / 行程概览

| Day | Route | Est. Mileage | Est. Duration | Road Type | Key Attractions |
|-----|-------|-------------|--------------|-----------|-----------------|
| Day 1 | ... | ... | ... | 高速/国道/山路 | ... |
| **Total** | — | **X km** | **~X h** | — | — |

## Daily Itinerary / 每日详细行程

### Day 1: [Title]
- **Driving mileage**: X km (高速 Xkm / 国道 Xkm / 山路 Xkm)
- **Driving duration**: ~X hours
- **Max altitude**: ~Xm (仅高原线路标注 / only for high-altitude routes)

**Waypoints / 途经点**：
| # | Name | Coordinates (lon, lat) | Altitude | Note |
|---|------|------------|----------|------|
| 1 | ... | (lon, lat) | ~Xm | ... |

**Supply points / 沿途补给**：
- ⛽ Gas stations / 加油站: [name] at [location]
- 🏥 Hospitals / 医院: [name] at [location]

## Practical Tips / 实用提示
1. **Altitude sickness / 高原反应**：...（高海拔线路必须包含 / required for routes >3000m）
2. **Road conditions / 路况**：...
3. **Tickets & booking / 门票与预订**：...
4. **Temperature & clothing / 温差与衣物**：...

## Personal Map QR Code / 个人地图二维码
![Route Name](/absolute/path/to/road_trip_qr.png)
> Fallback link / 备用链接: [Open in Amap](url)
```

## Advanced Capabilities / 进阶能力

- **Supply point search / 沿途补给搜索**：Integrated into Step 3.1 — conditions and API calls are defined there. / 已集成到 Step 3.1，按条件触发。
- **Mileage anomaly troubleshooting / 里程异常排查**：If a day's mileage is 0 or obviously wrong, check whether the previous day's endpoint was correctly inserted (Step 4 cross-day completion), and whether coordinates are duplicated.
- **Seasonal adjustments / 季节性调整**：For routes affected by holidays, winter closures, rainy seasons, high altitude, desert heat, or peak tourism windows, read [references/seasonal.md](references/seasonal.md) before finalizing the itinerary, then complete Step 6.1 with current official sources.

## References / 参考

- Full examples: [examples.md](examples.md)
- Seasonal and regional risk checklist: [references/seasonal.md](references/seasonal.md)
- Underlying API: [@lbs-amap/personal-map](https://clawhub.ai/lbs-amap/personal-map) skill

name: road-trip-planner description: > 基于高德地图 personal-map skill 规划多日自驾游路线，生成连续路线个人地图小程序二维码，并输出每日行程详情。当用户提到自驾、公路旅行、路线规划、旅游行程、生成地图、规划环线、控制驾驶时间或查询沿途里程时使用。支持闭合环线设计和跨天里程补全。 Plan multi-day road trip routes using Amap APIs. Generate continuous-route personal map QR codes and daily itinerary details. Triggered when users mention road trips, route planning, travel itineraries, map generation, loop routes, driving time control, or mileage queries. Supports closed-loop design and cross-day mileage completion. (China only — requires Amap API Key and @lbs-amap/personal-map skill)

Road Trip Planner / 自驾游路线规划

China only — 本 skill 基于高德地图 API，仅适用于中国大陆地区。
Dependency: @lbs-amap/personal-map (ClawHub) / GitHub

Trigger Conditions / 触发条件

Trigger when the user has any of these intents:

当用户出现以下任意意图时，主动调用本 skill：

Plan a road trip / road trip route / loop route (规划自驾游路线 / 公路旅行 / 环线)
Design a travel itinerary involving multi-city driving (设计涉及多地驾驶的旅游行程)
Generate a personal map / travel map / route QR code (生成个人地图 / 旅游地图 / 路线二维码)
Control daily driving time / avoid fatigued driving (控制每日驾驶时间 / 避免疲劳驾驶)
Query mileage, gas stations, hospitals along the route (查询沿途里程、加油站、医院等补给点)

Dependencies / 前置依赖

Dependency	Source	Description
`lbs-amap/personal-map`	ClawHub	Amap Web Service API wrapper / 高德 Web 服务 API 封装。安装命令：`openclaw skills install personal-map`。提供 `AMapPersonalMapClient` 类（import 路径 `scripts.amap_personal_map_client`），本仓库自身不含该脚本。
Amap API Key	Amap Open Platform	Required env var: `AMAP_API_KEY` / 需配置环境变量

Dependency Availability Check / 依赖可用性检查

Before any API call, verify that the personal-map skill is installed and the API key is configured. If either is missing, stop and tell the user what's needed — do not attempt to proceed.

在任何 API 调用之前，先检查 personal-map skill 是否已安装以及 API Key 是否已配置。任一缺失则停止并告知用户缺失项，不要尝试继续。

Checklist / 检查清单:
1. personal-map skill installed? → No → 提示: "请先安装 personal-map skill，运行: openclaw skills install personal-map"
2. AMAP_API_KEY set? → No → 提示: "请先配置高德 API Key，获取地址 https://lbs.amap.com/"
3. AMapPersonalMapClient importable? → No → 提示: "personal-map skill 版本可能不兼容，请确认其提供 AMapPersonalMapClient 类"

Set the environment variable in the active shell; never put a real key in the skill or source code:

# macOS / Linux
export AMAP_API_KEY='your_key'

# PowerShell
$env:AMAP_API_KEY = 'your_key'

# Windows cmd
set "AMAP_API_KEY=your_key"

Core Principles / 核心原则

No fatigued driving / 不疲劳驾驶：Keep daily actual driving time under 6 hours; 4–5 hours is a planning target, while short sightseeing or transfer days may be shorter. For high-altitude, mountainous, or winding roads, further reduce driving time.（每日实际驾驶时间控制在 6 小时以内；4-5 小时是规划目标，纯玩或短途换乘日可更短。高原、山区、弯道多的路段进一步压缩。）
Continuous route / 连续路线：Merge all daily waypoints into one continuous route, either a closed loop or with a clear endpoint.（所有天的节点合并为一条连续路线，首尾闭合或明确终点。）
Closed loop / 闭合环线：If start and end are the same city, ensure the endpoint is preserved after deduplication.（如果起点和终点是同一城市，确保终点节点在去重后仍然保留在路线末尾。）
Daily flexibility / 每日弹性：Leave buffer room in the itinerary; avoid over-scheduling.（行程中保留机动空间，避免把时间排得太满。）

Workflow / 工作流程

Step 1: Confirm itinerary framework / 确认行程框架

Confirm with the user:

Start point, end point (loop or not)
Total number of days
Must-visit attractions / cities
Travel season (affects road conditions and accommodation)
Exact departure dates, vehicle type, and any driving/health constraints when safety or road choice depends on them

与用户确认以下信息：起点、终点（是否环线）、总天数、必去景点/途经城市、出发季节、准确出发日期、车型及会影响路线选择的驾驶/健康限制。

If the date is unspecified, provide a provisional seasonal plan and explicitly mark the real-time checks as pending.

Step 2: Design daily waypoints / 设计每日节点

Split waypoints by day following these rules:

Target 4–6 hours of driving per day, never exceed 6 hours; short sightseeing or transfer days may be shorter
Each day's points list includes the core locations passed that day
The end city of each day must also serve as the start point for the next day (for mileage calculation)

按天拆分节点：每天驾驶 4-6 小时，不超过 6 小时；每天的 points 列表包含当天途经的核心地点；终点城市需同时作为下一天的起点。

Step 3: Call Amap APIs / 调用高德 API

Uses AMapPersonalMapClient provided by the personal-map skill. The class is at scripts.amap_personal_map_client inside that skill's directory — this is the canonical import path. This repo itself contains no scripts.

import os
from scripts.amap_personal_map_client import AMapPersonalMapClient

# Check API key BEFORE creating client
api_key = os.getenv("AMAP_API_KEY")
if not api_key:
    print("AMAP_API_KEY not set. Get a key at https://lbs.amap.com/ and run:")
    print("  Set AMAP_API_KEY in your active shell (see Dependencies above).")
    # stop here — do not proceed

# Create client (auto-reads AMAP_API_KEY from env):
client = AMapPersonalMapClient()
# Or pass explicitly: client = AMapPersonalMapClient(api_key="your_key")

Normalize every response before using it. The client does not raise API failures, but a failure can be either an error dict or an error dict as the first item of a list (for example, POI search). Always check both shapes:

def api_error_message(result):
    if isinstance(result, dict) and "error" in result:
        return result.get("message", str(result["error"]))
    if (
        isinstance(result, list)
        and result
        and isinstance(result[0], dict)
        and "error" in result[0]
    ):
        return result[0].get("message", str(result[0]["error"]))
    return None

geo = client.maps_geo("北京市朝阳区", "北京")
if error := api_error_message(geo):
    # Stop handling this location; never reuse another location's coordinates.
    raise ValueError(f"Cannot resolve this location: {error}")

# Convert external fields once. All later workflow steps use only this shape.
point = {
    "name": "地点名称",
    "lon": float(geo["longitude"]),
    "lat": float(geo["latitude"]),
    "poiId": "",             # fill from the verified POI-search result
    "required": False,         # True for the start, end, and user must-visit points
}

For each location:

maps_geo(address, city) — get precise coordinates. Pass the city name as second arg for better accuracy.
maps_text_search(keywords, city, offset=5) — get poiId. Use offset=5 to limit results (API default is 20).
Verify search result coordinates are within 30km of geo coordinates to prevent cross-city mismatches
time.sleep(0.3) between API calls to avoid QPS throttling

对每个地点执行：maps_geo 获取坐标 → 立即标准化为 lon/lat → maps_text_search 获取 poiId → 验证距离 < 30km → sleep 0.3s。后续步骤不得直接混用 API 返回的 longitude/latitude 与内部 lon/lat 字段。

Error handling / 错误处理：

If AMapPersonalMapClient is not found → check the pre-flight checklist in the Dependencies section, inform user, and stop
If AMAP_API_KEY is empty → inform user and stop
For every response → call api_error_message() before reading result fields
If maps_geo fails → do not substitute a previous coordinate. Stop QR generation when this is a start, end, or must-visit point; otherwise list the optional point as unresolved and exclude it from routing
If maps_text_search fails or has no result within 30km → retry with {city_name} {location_name}. For a required point, stop QR generation and ask the user to disambiguate; for an optional point, retain it only in the text itinerary as unresolved and do not give it a fake poiId
If QPS exceeded (CUQPS_HAS_EXCEEDED_THE_LIMIT in error message) → wait 1s and retry once

3.1 Supply point search per day / 每日补给搜索

After all waypoint coordinates are resolved, search for supply points near key nodes on days that enter sparsely populated areas or have long uninterrupted stretches:

对于进入人烟稀少区域或存在长距离无补给路段的日期，在关键节点附近搜索补给点：

# For each day flagging a supply concern (long stretch, remote area):
# Search gas stations and hospitals near the day's midpoint or last town before uninhabited area
result = client.maps_around_search(
    keywords="加油站|医院",
    location=f"{lon},{lat}",    # "经度,纬度" 格式
    radius=5000,               # 搜索半径 5km（默认 1000m）
    offset=5
)
if error := api_error_message(result):
    print(f"补给搜索失败: {error}")

When to search / 何时搜索：

Condition / 条件	Action / 动作
Day mileage > 300km / 日里程超 300km	Search gas stations at midpoint / 在中点搜索加油站
Entering uninhabited area (>100km no town) / 进入无人区（>100km 无城镇）	Search all supply points at the last town before entry / 在进入前最后城镇搜索全部补给点
High-altitude segment (>3500m) / 高海拔路段（>3500m）	Search hospitals at the nearest low-altitude town / 在最近的低海拔城镇搜索医院
All days / 每天	Search at least one hospital near the day's endpoint (overnight city) / 在每天终点城市至少搜索一家医院

Step 4: Driving route planning with cross-day completion / 驾车路径规划（跨天补全）

When calculating daily mileage, must insert the previous day's endpoint as the start of the current day's points list:

for i, day in enumerate(days):
    points = [p for p in day["points"] if "lon" in p]
    if i > 0:
        prev = [p for p in days[i-1]["points"] if "lon" in p]
        if prev:
            points.insert(0, prev[-1])
    # Drive segment by segment using maps_direction_driving:
    #   For each adjacent pair in points, call:
    #     result = client.maps_direction_driving(
    #         origin=f"{points[j]['lon']},{points[j]['lat']}",
    #         destination=f"{points[j+1]['lon']},{points[j+1]['lat']}"
    #     )
    #   Sum distances for the day's total.
    #   Parameter format: "longitude,latitude" (no spaces).

计算每日里程时，必须把前一天的终点插入当天 points 列表头部作为起点。 maps_direction_driving 的 origin / destination 参数格式为 "经度,纬度"（逗号分隔，无空格）。

Step 5: Generate personal map QR code / 生成个人地图二维码

5.1 Collect valid points / 收集有效点

Collect only normalized points with a non-empty verified poiId. Preserve insertion order (day-by-day). Before collecting, confirm that all required points (start, end, user must-visits) resolved successfully. If any did not, do not generate a QR code that claims to be the complete route.

收集所有天的有效点（已标准化且 poiId 已验证），按天顺序保持插入顺序。收集前必须确认起点、终点和用户必去点均已解析；任一失败时不得生成声称完整的二维码路线。

5.2 Deduplicate by identity / 按身份去重

Rules (apply in order):

Identity / 身份：Use poiId as the primary identity. Only when no poiId is available for a non-QR text item, treat points as duplicates when their coordinates are within a small tolerance; never deduplicate by name alone.
Closed-loop anchor / 环线锚点：If the first and last points have the same identity, keep both, and remove only middle occurrences of that identity. This ensures the route stays closed.
Intermediate duplicates / 中间重复：For all other repeated identities, keep the first occurrence and remove later ones. Keep the day-by-day itinerary data unchanged; this rule applies only to the QR route list.
Adjacent duplicates / 连续重复：Remove one only when two adjacent QR points have the same identity. Do not infer a duplicate solely because their display names match.

Deduplication flow:
  identities = [p["poiId"] for p in points]
  if len(identities) < 2:
      → stop; do not generate a route QR code
  elif identities[0] == identities[-1]:
      → keep points[0] and points[-1], remove middle occurrences of this identity
  else:
      → for all repeated identities, keep first occurrence and remove rest

5.3 Trim to ≤ 16 nodes / 精简至 ≤ 16 节点

If the deduplicated list exceeds 16 nodes, trim using this priority order:

Priority	Keep	Drop
1 (highest)	Start and end points (环路锚点)	—
2	User-specified must-visit attractions	—
3	Major scenic spots / landmarks (景点/地标)	Rest stops, service areas (服务区)
4	Geographically isolated nodes (唯一覆盖该区域的点)	Dense urban clusters (keep 1 per city)
5	POI with verified poiId	Optional unresolved POI (text itinerary only; never add to QR route)

Geographic distribution principle / 地理分布原则：When choosing between nodes of equal priority, prefer nodes that are farther apart (ensure the route visually represents the full journey). Avoid dropping the only node in a sparse region.

Minimum trim / 最小精简：Only remove as many nodes as needed to reach 16. Never remove the start or end anchor.

5.4 Data integrity before calling API / 调用 API 前的数据完整性

# Build lineList structure
point_info_list = [
    {"name": p["name"], "lon": p["lon"], "lat": p["lat"], "poiId": p["poiId"]}
    for p in deduplicated_points
]
line_list = [{
    "title": "Route Name / 路线名称",
    "pointInfoList": point_info_list
}]

# Validate before calling maps_schema_personal_map
if len(point_info_list) > 16:
    raise ValueError(f"Still {len(point_info_list)} nodes, must be ≤ 16")
if len(point_info_list) < 2:
    raise ValueError("A route QR code needs at least a resolved start and end point")
if any(not p["poiId"] for p in point_info_list):
    raise ValueError("Empty poiId found")

5.5 Call API and download QR / 调用 API 并下载二维码

result = client.maps_schema_personal_map(
    orgName="路线名称",
    lineList=line_list,
    sceneType=3          # 3 = route planning mode (仅创建路线)
)

# Check for API errors before reading result fields
if error := api_error_message(result):
    print(f"地图生成失败: {error}")
    # fallback: output the coordinates as text so user can manually create route
else:
    qr_url = result["qr_code_url"]
    # Download QR image:
    import urllib.request
    qr_path = "/tmp/road_trip_qr.png"
    urllib.request.urlretrieve(qr_url, qr_path)
    # Display inline using the local absolute path:
    # ![路线地图](/tmp/road_trip_qr.png)
    # Fallback link: [在 Amap 中打开](qr_url)

Call maps_schema_personal_map with sceneType=3 (route planning mode / 仅创建路线)
Check for error in result dict (no exceptions thrown)
Download QR code image and display inline
Always include the fallback URL link

收集已验证节点 → 按 poiId/坐标身份去重（保留闭环首尾）→ 构建 lineList → 按优先级精简至 ≤ 16 节点 → 数据完整性校验 → 调用 maps_schema_personal_map(sceneType=3) → 检查错误 → 下载二维码 → 显示备用链接。

Step 6: Output Markdown itinerary / 输出 Markdown 行程表

Output directly in the conversation as Markdown, including:

Itinerary overview table (day, route, mileage, duration, road type, attractions) with total row
Daily detailed itinerary (waypoints table with coordinates and altitude, supply points)
Practical tips (altitude sickness, road conditions, tickets, temperature, seasonal notes)
QR code image and fallback link

在对话中直接输出 Markdown：行程概览表格（含总里程合计行）、每日详细行程（含海拔和补给点）、实用提示、二维码图片和备用链接。

Step 6.1: Verify date-sensitive facts / 核验时效信息

Before presenting a final itinerary, verify the route against the actual departure dates. Static seasonal guidance and route-planning API results are not proof that a road, attraction, or service is currently available.

出最终行程前，必须按准确出发日期核验。静态季节提示和路径规划 API 结果均不能证明道路、景区或服务设施当前可用。

Check official traffic/road-authority notices for closures, controls, and weather-related restrictions
Check current weather warnings for high-altitude, mountain, desert, or remote segments
Check attraction opening status, ticket/reservation rules, and holiday calendar/traffic-freeway policy for that year
Mark each unavailable source as 未核验 rather than inferring a current status; advise the user to recheck immediately before departure

Key Constraints / 关键约束

Constraint / 约束	Description / 说明
Dependency check / 依赖检查	Before any API call, verify `personal-map` skill is installed AND `AMAP_API_KEY` is set. Fail fast with clear user-facing messages if missing.
Max nodes per route / 单条路线节点上限	Amap `lineList` max 16 nodes per `pointInfoList`
poiId required / poiId 必填	`maps_schema_personal_map` requires non-empty poiId
poiId strategy / poiId 获取策略	`maps_text_search` first → retry with city name. Required points must resolve; optional unresolved points stay in text only and never receive a fake poiId.
API error pattern / API 错误模式	API failures do not throw. Check an error dict or an error dict in the first list item before reading result fields.
QPS limit / QPS 限制	≥ 0.3s interval between calls, or `CUQPS_HAS_EXCEEDED_THE_LIMIT`
API Key / API Key	Set `AMAP_API_KEY` env var (auto-read by `AMapPersonalMapClient()` with no args) or pass explicitly
Region / 地域限制	China only — Amap (高德) covers mainland China
External dependency / 外部依赖	This repo contains no scripts. `AMapPersonalMapClient` lives in the `personal-map` skill at `scripts.amap_personal_map_client`. Install via `openclaw skills install personal-map`.

Output Template / 输出模板

# [Route Name] Road Trip Guide / 自驾行程指南

> Start/End: [City]　> Total: ~X km　> Days: X　> Max Altitude: ~Xm

## Itinerary Overview / 行程概览

| Day | Route | Est. Mileage | Est. Duration | Road Type | Key Attractions |
|-----|-------|-------------|--------------|-----------|-----------------|
| Day 1 | ... | ... | ... | 高速/国道/山路 | ... |
| **Total** | — | **X km** | **~X h** | — | — |

## Daily Itinerary / 每日详细行程

### Day 1: [Title]
- **Driving mileage**: X km (高速 Xkm / 国道 Xkm / 山路 Xkm)
- **Driving duration**: ~X hours
- **Max altitude**: ~Xm (仅高原线路标注 / only for high-altitude routes)

**Waypoints / 途经点**：
| # | Name | Coordinates (lon, lat) | Altitude | Note |
|---|------|------------|----------|------|
| 1 | ... | (lon, lat) | ~Xm | ... |

**Supply points / 沿途补给**：
- ⛽ Gas stations / 加油站: [name] at [location]
- 🏥 Hospitals / 医院: [name] at [location]

## Practical Tips / 实用提示
1. **Altitude sickness / 高原反应**：...（高海拔线路必须包含 / required for routes >3000m）
2. **Road conditions / 路况**：...
3. **Tickets & booking / 门票与预订**：...
4. **Temperature & clothing / 温差与衣物**：...

## Personal Map QR Code / 个人地图二维码
![Route Name](/absolute/path/to/road_trip_qr.png)
> Fallback link / 备用链接: [Open in Amap](url)

Advanced Capabilities / 进阶能力

Supply point search / 沿途补给搜索：Integrated into Step 3.1 — conditions and API calls are defined there. / 已集成到 Step 3.1，按条件触发。
Mileage anomaly troubleshooting / 里程异常排查：If a day's mileage is 0 or obviously wrong, check whether the previous day's endpoint was correctly inserted (Step 4 cross-day completion), and whether coordinates are duplicated.
Seasonal adjustments / 季节性调整：For routes affected by holidays, winter closures, rainy seasons, high altitude, desert heat, or peak tourism windows, read references/seasonal.md before finalizing the itinerary, then complete Step 6.1 with current official sources.

References / 参考

Full examples: examples.md
Seasonal and regional risk checklist: references/seasonal.md
Underlying API: @lbs-amap/personal-map skill