中文版

ai-goofish-monitor（GitHub：https://github.com/Usagi-org/ai-goofish-monitor ，2025-07-16 创建，今晨抓取时已 16,366★）是基于 Playwright 与 AI 的闲鱼多任务实时/定时监控系统：帮你从闲鱼海量商品里蹲到心仪的那一件，配有完整的 Web 管理界面。

按 README 的说法，核心是多任务并发监控：每个任务可独立配关键词、价格范围、筛选条件与 AI Prompt；AI 模式可用自然语言描述需求、由多模态模型深度分析商品详情与图片，关键词模式则不经过 AI、直接按规则创建。高级筛选包括包邮、新发布时间范围、省/市/区三级区域；命中后即时通知，渠道包括 ntfy.sh、企业微信、Bark、Telegram、Webhook 等；定时调度支持 Cron；多账号管理、任务绑定账号、代理池轮换与失败重试都已内置。部署推荐 Docker 一键起（镜像已内置 Chromium，默认 Web UI 在 http://127.0.0.1:8000），数据存在本地 SQLite，MIT 协议开源。

为什么值得了解

• 蹲二手好价最怕“刷到了但没看到”：把“定时刷闲鱼 + 人工判断成色价格”交给自动监控 + AI 初筛，省下的时间与错过的漏远比手动刷划算
• 对中文用户尤其友好：闲鱼场景、中文界面、三级区域筛选、微信/Bark 通知，都是按国内使用习惯做的
• 但要按 README 的提醒来用：AI 模式需要你自己提供 OpenAI 兼容接口的 API Key 与支持图片输入的模型（会产生模型调用费用），只想零成本用的话可选关键词模式；且请遵守闲鱼用户协议与 robots 规则、不要过于频繁请求，作者也声明本项目仅供学习与技术研究

English Version

ai-goofish-monitor (GitHub: https://github.com/Usagi-org/ai-goofish-monitor , created 2025-07-16, 16,366 stars at this morning’s fetch) is a Playwright- and AI-based multi-task real-time and scheduled monitoring system for Xianyu (Goofish): it helps you catch the item you want from Xianyu’s huge listings, with a complete web admin UI.

Per the README, the core is concurrent multi-task monitoring: each task can independently set keywords, price ranges, filters and an AI prompt; AI mode takes a natural-language description and has a multimodal model analyse listing details and images in depth, while keyword mode skips AI and creates tasks straight from rules. Advanced filters cover free shipping, newly-published time ranges and province/city/district selection; hits trigger instant notifications over ntfy.sh, WeCom, Bark, Telegram, webhooks and more; scheduling supports cron; multi-account management, per-task account binding, proxy pool rotation and failure retries are built in. Docker deployment is recommended (the image already bundles Chromium, with the default web UI at http://127.0.0.1:8000), data lives in local SQLite, and it is MIT licensed.

Why it’s worth knowing

• Hunting second-hand bargains fails most often on “I saw it too late”: handing scheduled Xianyu scanning plus first-pass AI judgement to automation beats manual refreshing on both time saved and deals missed
• Especially friendly to Chinese users: a Xianyu scenario, a Chinese UI, three-level region filters and WeCom/Bark notifications are all built around local usage habits
• Use it with the README’s warnings attached: AI mode needs your own OpenAI-compatible API key and an image-capable model (which incurs model call costs) — keyword mode is the zero-model-cost option; and respect Xianyu’s user agreement and robots rules without overly frequent requests, as the authors state the project is for study and technical research only