Introduction

The plugin ecosystem of DeepSeek Harness (DSH) already covers visual processing scenarios such as OCR and UI restoration. However, for reverse image search, there is currently no aggregation tool that can invoke multiple mainstream engines in a one-stop workflow. This plugin solves that problem by unifying the search entry points of nine engines, including Google Lens and Baidu Image Search, into a single interface.

What Is This

This is a DSH plugin developed by zimai233. Its core function is to directly construct a public image URL into search links for Google Lens, Baidu, Yandex, TinEye, Bing Visual Search, Sogou Image Search, SauceNAO, IQDB, and Ascii2d. It does not initiate any actual network scraping requests and only returns results by constructing URLs.

Core Capabilities

This plugin provides three main functions that return JSON or Markdown formatted data:

  1. image_search_urls
    Returns a JSON array [{ id, name, group, url }], with one search link per engine. Supports filtering by engines (comma-separated ids) or group (通用 / 插画/动漫).

  2. image_search_best
    Returns a human-readable Markdown summary: a one-line overview plus one clickable link per engine.

  3. image_search_engines
    Returns information for all nine engines grouped by category, including templates and encoding flags, used for discovering engine ids.

In its technical implementation, it uses a pure ESM architecture with zero build steps. It is published as runtime-ready code, has no prepare script, can run offline, and has zero dependencies.

Installation and Use

Install it using the official command:

dsh plugin --profile myprofile add dsh-image-search

Typical Usage

An Agent can trigger an image search through natural language. For example, ask the Agent to perform: “Help me search for the source of this image: https://example.com/img.png — first list the search links for all engines.”

The Agent will call image_search_urls with the following parameter:

{
  "imageUrl": "https://example.com/img.png"
}

Example response:

[
  { "id": "google", "name": "Google Lens", "group": "通用", "url": "https://lens.google.com/uploadbyurl?url=https%3A%2F%2Fexample.com%2Fimg.png" },
  { "id": "saucenao", "name": "SauceNAO", "group": "插画/动漫", "url": "https://saucenao.com/search.php?url=https://example.com/img.png" }
]

You can also limit engines or groups, for example:

{
  "imageUrl": "https://example.com/img.png",
  "engines": "google,saucenao",
  "group": "插画/动漫"
}

Engine Reference

The plugin supports nine search engines, divided into the 通用 and 插画/动漫 groups.

id name group encode
google Google Lens 通用 ✓
baidu Baidu Image Search 通用 ✓
yandex Yandex 通用 ✓
tineye TinEye 通用 ✓
bing Bing Visual Search 通用 ✓
sogou Sogou Image Search 通用 ✓
saucenao SauceNAO 插画/动漫 ✗
iqdb IQDB 插画/动漫 ✗
ascii2d Ascii2d 插画/动漫 ✗

Note: Engines with encode set to true use query-string parameters (the image URL must be encoded with encodeURIComponent); engines with false use path-based parameters (the URL should be used as-is).

Use Cases and Notes

This plugin is suitable for scenarios where multi-platform image source tracing needs to be integrated into an Agent. The plugin runs with the permissions of the current DSH process; it is recommended to review the source code and license (MIT) before installing.