vision-reader

name	vision-reader
description	Read and understand image content using vision-capable AI models or OCR. Use when the user asks to read/analyze an image, extract text from a screenshot, understand what's in a photo, do OCR on a document, or needs visual content interpreted. Supports local images and URLs.
triggers	null
metadata	null
openclaw	null
emoji	👁️
requires	null
anyBins	["python3"]
version	1.0.0
status	stable
updated	"2026-03-11T00:00:00.000Z"
provides	["vision_reader"]
os	["darwin","linux","win32"]
clawdbot	{"emoji":"🛠️","category":"tools","priority":"medium"}

Vision Reader

图片识别助手。触发后必须先询问用户确认，再执行分析。

触发场景	询问确认内容
OCR 提取	确认图片路径、识别语言 (中文/英文)、输出格式
图片分析	确认分析深度 (简述/详细)、关注重点
截图理解	确认截图来源、需要提取的信息类型
文档识别	确认文档类型、是否保留格式

执行流程: 触发 → 询问用户 → 用户确认 → 执行识别 → 返回结果

Full docs: read SKILL-REFERENCE.md

Analyze images using AI vision models or local OCR.

Capabilities

1. AI Vision Analysis — send image to vision model (qwen3.5-plus, kimi-k2.5) for understanding
2. OCR Text Extraction — extract text from images using macOS Vision framework
3. Screenshot Analysis — combine with screenshot-sender for screen understanding
4. Document Reading — extract text from scanned documents, receipts, etc.

Method 1: AI Vision Model (Recommended)

Use vision-capable models directly. These models accept image input:

• bailian/qwen3.5-plus — 1M context, supports ["text", "image"]
• bailian/kimi-k2.5 — 262K context, supports ["text", "image"]
• qwen-portal/vision-model — 128K context, supports ["text", "image"]

When analyzing images, switch to one of these models if the current model doesn't support image input.

Workflow

1. User provides image path or URL
2. Agent reads the image file
3. Agent sends to vision model with analysis prompt
4. Return structured analysis

Method 2: Local OCR (macOS Vision Framework)

For pure text extraction without needing an AI model:

python3 {baseDir}/scripts/ocr_extract.py <image_path> [--lang zh-Hans,en] [--json]

Parameters

Flag	Description	Default
<image_path>	Path to image file (PNG, JPG, etc.)	required
--lang	OCR languages (comma-separated)	zh-Hans,en
--json	Output as JSON with bounding boxes	plain text
--confidence	Min confidence threshold (0-1)	0.5

Output (plain text mode)

第一行文字
第二行文字
...

Output (JSON mode)

{
  "text": "完整提取文本",
  "blocks": [
    {"text": "第一行", "confidence": 0.98, "bbox": [x, y, w, h]},
    ...
  ],
  "language": "zh-Hans"
}

Method 3: Tesseract OCR (Fallback)

If macOS Vision Framework is unavailable:

brew install tesseract tesseract-lang
tesseract <image_path> stdout -l chi_sim+eng

Workflow Examples

User says: "帮我看看这张图片上写了什么"

1. Get image path
2. Run OCR: python3 {baseDir}/scripts/ocr_extract.py /path/to/image.png
3. Return extracted text

User says: "分析一下这个截图的内容"

1. Get screenshot path (or take new one with screenshot-sender)
2. Send to vision model (qwen3.5-plus) with prompt: "请详细描述这张截图的内容"
3. Return model analysis

User says: "这个价格截图里哪个最便宜？"

1. Run OCR to extract prices
2. Send to vision model for comparison analysis
3. Return structured comparison

Notes

• macOS Vision Framework requires macOS 10.15+
• For Chinese text, always include zh-Hans in language list
• Vision models are more accurate for complex analysis; OCR is faster for pure text
• Large images may be resized before sending to vision models