DeepSeek Harness plugin

dsh-vision-android

Vision + Android UI automation plugin for DeepSeek Harness: gpt-5.6-luna screenshot understanding and real-tap mobile app testing via adb

Jump to install

Source facts

Repository
superclaude1/dsh-vision-android
Latest update
Aug 14, 2026
Category
Vision & Multimodal
GitHub stars
2
Format
plugin
Catalog evidence
Upstream dsh.bundle evidence
Evidence path
package.json#dsh.bundle
Checked against
0.1.0-rc.8
Upstream check date
2026-08-20

This evidence comes from the upstream catalog. This site has not installed, run, or security-reviewed the plugin.

Install

Start with a prompt that asks an agent to review the GitHub repository and source. Switch to the command if you want to install it yourself.

Copy this prompt into DSH, Codex, or another agent and ask it to review the GitHub repository and source first.

Do not install or run any commands yet. Read this plugin's GitHub repository, README, and relevant source code. Then answer the questions below clearly and directly so I can decide whether it fits my needs:

1. What is this plugin, and what problem does it solve?
2. Who is it for, and what are its typical use cases?
3. How is it used after installation? Include one minimal example.
4. What known limitations or privacy, security, compatibility, or maintenance risks does it have?
5. Give a clear recommendation: recommend, conditionally recommend, or do not recommend, with reasons.

Distinguish statements documented by the repository, inferences from source code, and unknowns. If evidence is insufficient, say so explicitly. Do not guess or simply repeat the README.

GitHub: https://github.com/superclaude1/dsh-vision-android
Plugin: dsh-vision-android
Author: superclaude1

Check the source files

Read the README and other files from this plugin directory before installing.

File explorer3 files
README.mdSource · read only

dsh-vision-android

DeepSeek Harness 插件:多模态视觉 + Android UI 自动化

解决两个问题: 1. 主模型没有多模态输入:把任意截图/图片发给 OpenAI 兼容的视觉模型提问(默认 gpt-5.6-luna,可配任意 OpenAI 兼容端点)。 2. agent 测试手机软件:通过 adb 在模拟器/真机上做真实点击、滑动、输入、按键,形成"截图 → 视觉理解 → 真实点击 → 再截图验证"的闭环。

工具清单(注册到 ctx.tools,模型可见)

工具作用
vision_ask把本地图片文件发给视觉模型提问(通用多模态)
mobile_look截屏 + 视觉理解:返回屏幕描述 + 可交互元素 [{label, role, x, y}](像素中心坐标)
mobile_dumpuiautomator 抓取 UI 层级:精确坐标兜底(不花视觉 token)
mobile_screenshot只截屏存 PNG,返回路径
mobile_tap真实点击 (x, y)
mobile_swipe真实滑动手势
mobile_text输入文本(adb 限制:仅 ASCII,空格自动转义)
mobile_keyevent按键:back/home/enter/tab/esc/dpad_* 等
mobile_installadb install -r 安装 APK(保留数据)
mobile_launcham start 启动组件 / monkey 启动包
mobile_screen_size屏幕物理分辨率(校验视觉坐标用)

配置

插件 Config(Schemastery 校验,环境变量兜底):

字段默认说明
visionBaseUrlhttps://api.openai.com/v1OpenAI 兼容地址(env DSH_VISION_BASE_URL
visionApiKeyAPI Key(env DSH_VISION_API_KEY推荐用环境变量,不要写进配置
visionModelgpt-5.6-luna视觉模型 ID(env DSH_VISION_MODEL
visionDetailhigh图片 detail 档位
visionMaxTokens / visionTimeoutMs2000 / 120000生成上限与超时
adbPathadbadb 路径(env ADB 兜底,可写绝对路径)
adbTimeoutMs30000单条 adb 命令超时
screenshotDir.dsh-mobile-shots截图落盘目录

安装

A. overlay(开发/本机快速生效)

复制 cordis.example.ymlcordis.yml(已被 gitignore),把插件路径改成你的绝对路径,然后:

dsh web --patch /path/to/dsh-vision-android/cordis.yml

B. 安装进 profile(bundle 形态,推荐)

dsh plugin --profile web add ./dsh-vision-android     # 本地目录(link)
# 或从 GitHub 安装:dsh plugin --profile web add github:you/dsh-vision-android
dsh --profile web --dump-config   # 应看到 # == dsh-vision-android 层

机器私密配置(API Key / adb 绝对路径)建议放在 $DSH_HOME/cordis.patch.yml 或 profile 自己的 patch 里按行 id 覆盖,不要把 Key 写进仓库。

构建与演示

node scripts/build.mjs    # esbuild 打包 src/ → 自包含 lib/index.js(已提交构建产物)
node scripts/smoke.mjs    # 加载冒烟:Config 校验 + 11 个工具注册(无需 Key/设备)
node scripts/demo.mjs     # 端到端:截图→视觉→真实点击→验证(需 DSH_VISION_API_KEY + 设备在线)

模型对话中的用法示例

请打开诡秘之主这本书:先 mobile_look 看当前界面,找到这本书的坐标,
用 mobile_tap 点击,再 mobile_look 确认进入了目录页。

注意事项

  • 视觉坐标为像素值;mobile_dump 的坐标永远精确,视觉坐标偶有偏差,重要点击前可先用 dump 校验。
  • mobile_text 仅 ASCII(adb input text 限制),中文输入需设备 IME;空格已自动转义为 %s
  • 视觉模型可能把系统软键盘等非应用 UI 误读为应用元素;关键操作前用 mobile_dump 交叉验证。