---
type: "post"
id: "usagi-org-s-open-source-ai-goofish-monitor-589c2b"
slug: "usagi-org-s-open-source-ai-goofish-monitor-589c2b"
title: "Usagi-org’s open-source ai-goofish-monitor is a self-hosted tool for monitoring listings on Xianyu"
description: "Usagi-org’s open-source ai-goofish-monitor is a self-hosted tool for monitoring listings on Xianyu, the Chinese second-hand marketplace."
retrieval_nugget: "Usagi-org’s open-source ai-goofish-monitor is a self-hosted tool for monitoring listings on Xianyu, the Chinese second-hand marketplace. It combines browser automation, filters, AI analysis of listings and notifications, offering a concrete example of turning a marketplace search into a scheduled monitoring workflow rather than a one-off manual task."
published_at: "2026-11-05"
updated_at: "2026-11-05"
record_date: "2026-11-05"
date_kind: "published_at"
topics: ["automation","open-source"]
entities: ["Usagi-org","ai-goofish-monitor","Xianyu","Playwright","SQLite"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "This is a distinct Xianyu marketplace monitoring project; none of the archive documents cover it."
prior_refs: []
source_url: "https://github.com/Usagi-org/ai-goofish-monitor"
source_urls: ["https://github.com/Usagi-org/ai-goofish-monitor"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:usagi-org-s-open-source-ai-goofish-monitor-589c2b"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/usagi-org-s-open-source-ai-goofish-monitor-589c2b/","markdown":"https://newruntime.com/posts/usagi-org-s-open-source-ai-goofish-monitor-589c2b.md","json":"https://newruntime.com/posts/usagi-org-s-open-source-ai-goofish-monitor-589c2b.json"}
markdown_url: "https://newruntime.com/posts/usagi-org-s-open-source-ai-goofish-monitor-589c2b.md"
---

# Usagi-org’s open-source ai-goofish-monitor is a self-hosted tool for monitoring listings on Xianyu

## Retrieval answer

Usagi-org’s open-source ai-goofish-monitor is a self-hosted tool for monitoring listings on Xianyu, the Chinese second-hand marketplace. It combines browser automation, filters, AI analysis of listings and notifications, offering a concrete example of turning a marketplace search into a scheduled monitoring workflow rather than a one-off manual task.

## Summary
Usagi-org’s open-source ai-goofish-monitor is a self-hosted tool for monitoring listings on Xianyu, the Chinese second-hand marketplace. It combines browser automation, filters, AI analysis of listings and notifications, offering a concrete example of turning a marketplace search into a scheduled monitoring workflow rather than a one-off manual task.

## What happened
The GitHub repository describes ai-goofish-monitor as a multi-task Xianyu monitoring system built with Playwright and AI. Its purpose is to help users find relevant items among a large volume of marketplace listings. The project provides a web management interface for tasks, accounts, AI criteria, logs and results. Users can configure separate keywords, price conditions, filters and prompts for multiple concurrent tasks.

The documented filters include free shipping, recency and geographic selection at province, city and district levels. Tasks can run on a schedule configured with Cron. The system can send notifications through ntfy.sh, WeCom, Bark, Telegram or a webhook. It supports multiple accounts, task-to-account binding, proxy-pool rotation and retries when requests fail. These are repository-described capabilities; the source does not quantify their reliability, throughput or performance under load.

The documented workflow starts a monitoring task using an account and proxy configuration, searches listings, and, when it finds a new item, fetches item and seller details and downloads product images. It then calls an AI model to analyze the listing. Listings recommended by the model trigger a notification; other results are saved. The repository says the system retains tasks, results and price history in SQLite, while account state, prompts, logs and images remain in filesystem directories. Temporary product images are normally cleaned up after a task ends.

Deployment is intended to be self-hosted. The README recommends Docker Compose and says the image includes Chromium. The default web interface is at `http://127.0.0.1:8000`; users clone the repository, copy `.env.example` to `.env`, configure it and start the service with Docker Compose. The minimum required AI configuration is an API key, an OpenAI-compatible base URL and a model name that supports image input. The repository also describes a first-use flow requiring users to import Xianyu login state, then create a task and bind an account. Its default UI credentials are `admin/admin123`, and the README says to change them in production.

The project also distinguishes two task-creation paths. In AI mode, a submitted detailed requirement initiates a background job that generates analysis criteria; the client receives a `202` response and can poll the job’s progress. In keyword mode, the task is created directly. The maintainers describe this asynchronous AI flow as a way to avoid leaving the task form blocked while criteria are being generated.

The project is published under the MIT license. The supplied repository material does not state a commercial price, user count, GitHub star count, supported Xianyu API or service guarantees. It warns users to respect Xianyu’s terms and robots.txt, avoid excessive requests, and recognize that accounts may be restricted. The README describes the software as provided without warranty and for learning and research use.

## What the archive adds
We have not covered this before; it opens a new thread.

## Trend signal
No supplied SITE_PATTERNS entry genuinely fits this release. The repository documents an application-specific listing monitor, but it does not establish a broader shift in agent protocols, runtime security, task completion economics or another listed pattern.

## Why it matters
For a builder with a relevant marketplace use case, the project illustrates a practical composition: browser automation gathers candidates, structured filters narrow the search, a multimodal model evaluates listing details and images, and a scheduler plus notification channel turns the process into ongoing monitoring. The repository also makes operational concerns visible—login-state handling, retries, proxy rotation, persistence and the possibility of account restrictions—rather than presenting AI analysis as a standalone prompt.

That does not make the implementation a general-purpose shopping agent, nor does the source establish that it is robust enough for production. Still, teams exploring similar workflows can inspect the boundaries between scraping, filtering, model judgment and alerts, then decide which parts transfer to their own authorized data sources. The most reusable lesson is architectural, not marketplace-specific: define what counts as a candidate before calling a model, retain useful run history, and route only selected outcomes to a human.

This separation also gives teams places to evaluate a system independently. Search and filtering can be assessed for coverage; model criteria can be revised when recommendations are poor; notification volume can be constrained without changing collection. The README’s combination of SQLite records, logs and result browsing suggests an implementation intended to make those operational layers visible, although it does not demonstrate their quality in real deployments.

## Limits and uncertainty
The available evidence is the project’s README, not an independent test or deployment report. It does not establish accuracy of the AI recommendations, false-positive rates, latency, operating cost, supported listing volume or reliability of scheduled runs. It also does not show whether Xianyu permits this use, how frequently the site changes its interface, or how effectively proxy rotation and retries handle blocks.

The repository says image-capable models and imported login state are required, but does not provide a model comparison or a cost estimate. It does not identify a stable marketplace API, and browser-driven collection may be affected by changes to the target site or by account controls. Its features and setup instructions should therefore be treated as the maintainers’ description, not as verified performance claims.

## Skill to build now
Map a monitoring workflow into four explicit stages before adding model judgment: candidate collection, deterministic filtering, AI evaluation and human-facing notification. For a permitted internal or public data source, write one acceptance criterion for each stage and log the input, decision and outcome. This makes it possible to tune the AI prompt separately from collection rules, measure unwanted alerts, and stop a noisy notification path without losing the underlying record.

## Reader takeaway
ai-goofish-monitor is a substantial, self-hosted example of AI-assisted marketplace monitoring: search and filter listings, analyze promising items with an image-capable model, then notify a user. Its usefulness is specific to Xianyu and comes with account, policy and reliability questions. For other builders, the workflow is more portable than the target marketplace—but should be adapted only to data sources they are authorized to access.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Automation](https://newruntime.com/topics/automation.md): Explore the automation topic hub.
- topic hub: [Open Source](https://newruntime.com/topics/open-source.md): Explore the open-source topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)