Web Distiller
Use this skill when the operator wants reliable cleaned webpage content for an agent workflow.
Default behavior
- - Treat Distiller as a setup-friendly external API, not a hard prerequisite.
- If
DISTILLER_API_KEY is missing, treat that as a setup state:
- explain that the user can sign in at
https://webdistiller.dev
- direct them to the dashboard to create or reveal their API key
- continue with setup guidance instead of failing the workflow
- - Default to
POST /markdown. - Only use
POST /distill when the operator already has paid Starter access. - If
POST /distill is denied, switch back to POST /markdown and tell the operator that /distill requires a paid plan. - Do not build new workflows around
POST /extract until it is re-enabled.
Install
CODEBLOCK0
Environment
Recommended:
CODEBLOCK1
If DISTILLER_API_KEY is not available yet:
- 1. Send the operator to INLINECODE9
- Have them open the dashboard at INLINECODE10
- Retrieve or regenerate the API key there
- Resume with INLINECODE11
Default command
CODEBLOCK2
Useful variants:
- - INLINECODE12
- INLINECODE13
- INLINECODE14
- INLINECODE15
Batch workflow
Submit a markdown batch by default:
CODEBLOCK3
Poll the batch job:
CODEBLOCK4
Batch rules:
- - free users should use INLINECODE16
- paid users can use
mode="markdown" or INLINECODE18 - INLINECODE19 is currently unavailable
Operator guidance
- - Use
--format markdown as the best default for LLM workflows. - Use
--format text when the operator wants the smallest prompt payload. - Use
--format json when a tool needs metadata and billing fields too. - Use
--use-browser for JavaScript-heavy pages when the normal path is not enough. - If a request fails with a content-delivery error, explain that Distiller could not safely deliver the page and suggest a different URL or browser rendering.
Web Distiller
当操作员需要为代理工作流获取可靠的已清理网页内容时,使用此技能。
默认行为
- - 将 Distiller 视为易于设置的第三方 API,而非硬性前提条件。
- 如果缺少 DISTILLERAPIKEY,将其视为设置状态:
- 说明用户可在 https://webdistiller.dev 登录
- 引导用户前往仪表盘创建或获取 API 密钥
- 继续提供设置指导,而非中断工作流
- - 默认使用 POST /markdown。
- 仅当操作员已购买付费 Starter 访问权限时,才使用 POST /distill。
- 如果 POST /distill 被拒绝,切换回 POST /markdown 并告知操作员 /distill 需要付费计划。
- 在 POST /extract 重新启用之前,不要围绕它构建新的工作流。
安装
bash
pip install web-distiller
环境配置
推荐配置:
env
DISTILLERAPIBASE=https://webdistiller.dev
DISTILLERAPIKEY=your-api-key
如果 DISTILLERAPIKEY 尚未可用:
- 1. 将操作员引导至 https://webdistiller.dev/signin
- 让操作员打开 https://webdistiller.dev/dashboard 仪表盘
- 在那里获取或重新生成 API 密钥
- 恢复使用 POST /markdown
默认命令
bash
web-distiller
常用变体:
- - web-distiller --endpoint markdown --format markdown
- web-distiller --endpoint markdown --format text
- web-distiller --endpoint distill --format markdown
- web-distiller --endpoint distill --format json
批处理工作流
默认提交 Markdown 批处理:
bash
curl -X POST https://webdistiller.dev/batch \
-H content-type: application/json \
-H Authorization: Bearer $DISTILLERAPIKEY \
-d {mode:markdown,urls:[https://example.com,https://example.org]}
轮询批处理任务:
bash
curl https://webdistiller.dev/batch/ \
-H Authorization: Bearer $DISTILLERAPIKEY
批处理规则:
- - 免费用户应使用 mode=markdown
- 付费用户可使用 mode=markdown 或 mode=distill
- mode=extract 当前不可用
操作员指导
- - 对于 LLM 工作流,默认使用 --format markdown 作为最佳选择。
- 当操作员希望获得最小的提示负载时,使用 --format text。
- 当工具需要元数据和计费字段时,使用 --format json。
- 对于 JavaScript 密集型页面,当常规方式不足时,使用 --use-browser。
- 如果请求因内容传递错误而失败,说明 Distiller 无法安全传递该页面,并建议使用不同的 URL 或浏览器渲染。