\r \r

GEO 深度仿写（OpenClaw）\r

\r 针对已开启 deepImitateEnabled 的 OPT 任务：拉信源 URL → 爬全文 → 提交 Backend 异步成稿。\r \r

1. 信源列表\r

\r

GEO_KEY=$(cat ~/.qclaw/geo-api-key 2>/dev/null || cat ~/.openclaw/geo-api-key 2>/dev/null)\r
OPT_ID="OPT-XXXXXXXXXXXX"\r
CYCLE=1\r
curl -s "https://ai.gaobobo.cn/api/geo/diagnosis/imitate-sources?optimization_task_id=$OPT_ID&cycle_number=$CYCLE" \\r
  -H "Authorization: Bearer $GEO_KEY"\r
```\r
\r
取 Top1–3 条带 `url` 的条目。\r
\r
## 2. 爬全文\r
\r
对每条 URL（优先 Firecrawl）：\r
\r
- `firecrawl_scrape(url, extractMode=markdown, onlyMainContent=true, maxChars=50000)`\r
- 失败 → `web_fetch(url)`\r
- 仍失败 → 用 `summary` 并明确告知用户已降级\r
\r
可选：本地 LLM 从 markdown 提取 `reference_outline`（sections 数组）。\r
\r
## 3. 提交成稿\r
\r
```bash\r
curl -s -X POST "https://ai.gaobobo.cn/api/geo/article/generate-deep-imitate" \\r
  -H "Authorization: Bearer $GEO_KEY" \\r
  -H "Content-Type: application/json" \\r
  -d '{\r
    "optimization_task_id": "OPT-XXX",\r
    "cycle_number": 1,\r
    "brand_name": "品牌名",\r
    "keyword": "核心词",\r
    "reference_url": "https://...",\r
    "reference_markdown": "全文 markdown...",\r
    "reference_outline": {"sections": ["引言", "对比", "结论"]}\r
  }'\r
```\r
\r
## 4. 轮询\r
\r
```bash\r
TASK_ID="CG-XXXXXXXX"\r
# 间隔 ≥30 秒\r
curl -s "https://ai.gaobobo.cn/api/geo/article/$TASK_ID" \\r
  -H "Authorization: Bearer $GEO_KEY"\r
```\r
\r
`status=completed` 且正文非空即成功；周期日志 `deepImitate` 由服务端自动回写。\r
\r
## 约束\r
\r
- 与 SaaS **浅仿写**（`imitate_enabled`）互斥，勿重复派发 Celery imitate\r
- 不触发三方媒体自动发稿（无 `publishTaskId` 自动创建）\r