modlens

Permanently free. Proudly temporary.

永久免费。也故意短暂。

Give a text-only model sight.

为纯文本模型补上视觉能力。

The flagship DeepSeek and GLM chat models are text-only. ModLens is a CLI they can call. An image goes in. Structured evidence comes out.

DeepSeek 和 GLM 的主力对话模型是纯文本的，无法读图。ModLens 是它们可以调用的 CLI。图片进去，结构化证据出来。

npx -y @deepseek-ai/dsh plugin --profile web add @liustack/modlens@3.24.2
Docs文档GitHub
modlens · stdout

$ modlens -i screenshot.png

{
  "image": "/path/to/screenshot.png",
  "provider": "gemini-api",
  "result": {
    "summary": "A workflow diagram with four nodes connected by labeled arrows.",
    "ocr": {
      "full_text": "/shaping\nBEFORE YOU BUILD\n...",
      "lines": []
    },
    "layout": {
      "regions": [
        { "reading_order": 1, "type": "title", "text": "/shaping" }
      ]
    },
    "semantics": {
      "scene": "workflow diagram",
      "entities": [],
      "relations": []
    },
    "visual": {
      "dominant_colors": ["white", "black"],
      "style": "flat",
      "notes": []
    },
    "uncertainty": []
  },
  "meta": {
    "generatedAt": "2026-08-06T12:00:00.000Z",
    "model": "gemini-3.6-flash",
    "conversationId": null,
    "durationSeconds": 6.4,
    "usage": {
      "promptTokenCount": 1234,
      "candidatesTokenCount": 567
    },
    "attempts": [
      { "provider": "gemini-api", "ok": true, "durationSeconds": 6.4 }
    ],
    "warnings": []
  }
}

How it works

怎么工作

An image goes in. Structured evidence comes out.

图片进去，结构化证据出来。

01

An image goes in.

图片进去。

It reads images pasted straight into the chat. You do not have to save a file and pass a path first. Or run modlens -i screenshot.png.

支持直接粘贴图片识别，不必先存成文件再提供路径。也可以运行 modlens -i screenshot.png。

02

A CLI they can call.

它们可以调用的 CLI。

Six built-in providers. Any one is enough. Four local CLIs can join the same chain after you grant them.

六个内置 provider，配好任意一个就能用。四家本机 CLI 在你授权后进入同一条故障转移链。

03

Structured evidence comes out.

结构化证据出来。

Output is a fixed JSON shape. summary, ocr, layout, semantics, visual.

输出是固定的 JSON。summary、ocr、layout、semantics、visual。

Other harnesses

其他宿主

On Claude Code, Codex, Pi, or OpenCode, send this to your agent:

在 Claude Code、Codex、Pi 或 OpenCode 里，把这句话发给你的 AI：

Install and configure the modlens skill following https://github.com/liustack/modlens/blob/main/INSTALL.md, then run the health check and tell me the result.
按 https://github.com/liustack/modlens 的 INSTALL.md 安装并配置 modlens skill，完成后运行体检并把结果告诉我。

Providers

引擎

Vision engines

视觉引擎

Six built-in providers. Any one is enough. Four local CLIs can join the same chain after you grant them.

六个内置 provider，配好任意一个就能用。四家本机 CLI 在你授权后进入同一条故障转移链。

ProviderNeeds需要什么Per read单次耗时
gemini-apifree Gemini API key免费 Gemini key5-10s
openaiany OpenAI-compatible endpoint任意 OpenAI 兼容端点5-10s
anthropicAnthropic API key5-10s
antigravity-clifree agy CLI, browser sign-in免费 agy CLI，浏览器登录15-45s
claude-clisigned-in Claude Code已登录的 Claude Code20-45s
kimi-clisigned-in Kimi Code, named explicitly已登录的 Kimi Code，需显式点名20-45s

Provider setup and failover引擎配置与故障转移

Audience

给谁

Who

给谁用

People running text-only DeepSeek, GLM, or MiMo Pro inside Claude Code, Codex, Pi, OpenCode, or DeepSeek Harness.

在 Claude Code、Codex、Pi、OpenCode 或 DeepSeek Harness 里跑纯文本 DeepSeek、GLM 或 MiMo Pro 的人。

Sunset

谢幕

Permanently free. Proudly temporary.

永久免费。也故意短暂。

No signup fee, no paid tier, not now, not later. ModLens exists because some of the best coding models still cannot see. Multimodal is coming for everyone. When text-only models leave the stage, we will too.

不收费，没有付费档，现在没有，以后也不会有。modlens 活着，是因为最好用的编程模型里，还有一批看不见图。多模态很快会变成标配。纯文本模型谢幕那天，modlens 一起谢幕。