vision

name	vision
description	视觉理解工具：识别人物/物体/场景/图表/UI 布局等【非文字】图像内容，或用户明确要求视觉识别时使用。文字内容（文档、证据、票据、截图文字）一律走 OCR，不要用本工具。

vision

Multi-provider vision tool. Call various vision models to describe images. Feed it a prompt + image path, get back a text description.

⚠️ 使用边界：文字内容一律走 OCR，不要用本工具

本工具是视觉理解（识人、识物、识场景、看图说话），不是 OCR。

• 文字内容（文档、证据、票据、合同页、截图里的文字等）→ 用 OCR 提取文字，不要调用本工具：
  • 本地 OCR（默认优先，不联网）：你本机的本地 OCR 工具（路径按部署替换，如 ocr.bat / ocr_run.py）
  • 线上 OCR（须经用户明确同意）：mcp__paddleocr__*
• 非文字内容（识别人物/物体/场景/图表/UI 布局等）→ 用本工具。
• 触发条件：仅在用户明确要求视觉识别（"识别这是谁""看看这张图""用视觉识别"）或内容明显为非文字时调用。用户未指明时，默认按文字内容处理、走 OCR。

聊天中粘贴的图片：先提取到临时文件，按上述规则路由——文字内容 → 本地 OCR；非文字/用户指明 → 本工具（见 pasted_image_flow.py）。

When to use this tool

If you can already see and understand the image yourself (native multimodal model), skip this tool — analyze it directly.

A SessionStart hook normally announces this session's routing status up front. If that context isn't visible (e.g. compacted out of a long conversation, or the hook isn't installed), check before calling this tool:

python vision.py --check-routing

• native → you already have native image understanding this session; don't call this tool.
• external (default) → proceed with the quick start below.

Quick start

python vision.py [--provider <name>] <image_path> <prompt>

When --provider is omitted, the provider is resolved by: --provider flag > VISION_PROVIDER env > first API key found.

Providers

doubao (Volcengine Ark)

• API key: DOUBAO_API_KEY
• Default model: doubao-seed-2-0-pro-260215
• Custom endpoint: DOUBAO_BASE_URL

qwen (DashScope)

• API key: DASHSCOPE_API_KEY
• Default model: qwen-vl-max
• Custom endpoint: DASHSCOPE_BASE_URL
• Available models: qwen-vl-max, qwen-vl-plus, qvq-max

openai (GPT-4o)

• API key: OPENAI_API_KEY
• Default model: gpt-4o
• Custom endpoint: OPENAI_BASE_URL
• Also works with any OpenAI-compatible endpoint.

anthropic (Claude)

• API key: ANTHROPIC_API_KEY
• Default model: claude-sonnet-5
• Custom endpoint: ANTHROPIC_BASE_URL
• Requires the anthropic package (pip install anthropic); it's imported lazily so other providers work without it.

any custom provider

Any --provider name outside the built-in four is resolved dynamically from environment variables named after it — no code changes needed:

Env Var	Required	Notes
{NAME}_API_KEY	yes	checked at request time, same as built-ins
{NAME}_BASE_URL	yes	no default — arbitrary endpoint
{NAME}_MODEL	yes	no default (or set global VISION_MODEL instead)
{NAME}_PROTOCOL	no	openai (default) or anthropic — picks the request shape

openai covers essentially every OpenAI-compatible endpoint (vLLM, Ollama, LiteLLM, OpenRouter, Azure OpenAI, self-hosted proxies, ...). Use {NAME}_PROTOCOL=anthropic only if the endpoint speaks the Anthropic Messages API shape.

export MYAPI_API_KEY="sk-xxx"
export MYAPI_BASE_URL="https://my-endpoint.example.com/v1"
export MYAPI_MODEL="my-vision-model"
python vision.py --provider myapi "screenshot.png" "describe this"

If {NAME}_BASE_URL or {NAME}_MODEL is missing, the tool prints exactly which variables to set instead of a generic "unknown provider" error.

Configuration

Env Var	Scope	Default
VISION_PROVIDER	Default provider (built-in or custom name)	auto-detect (built-ins only)
VISION_MODEL	Override model (all providers)	provider default
{PROVIDER}_MODEL	Override model (per provider)	—
{PROVIDER}_BASE_URL	Override/define endpoint (per provider)	built-in default, or required for custom
{PROVIDER}_PROTOCOL	Request shape for a custom provider: openai | anthropic	openai
VISION_TEMPERATURE	Response creativity 0–1	0
VISION_MAX_TOKENS	Max response tokens	4096

Note: auto-detect (no --provider / VISION_PROVIDER set) only scans the four built-in providers' API keys — a custom provider must always be named explicitly.

Examples

# Auto-detect provider from API keys
python vision.py "screenshot.png" "Describe the page layout and any visible UI issues."

# Explicit provider
python vision.py --provider qwen "mockup.png" "List all components, colors, and spacing patterns."

# Custom model
QWEN_MODEL=qvq-max python vision.py --provider qwen "diagram.png" "Explain the architecture."

# GPT-4o for visual regression
python vision.py -p openai "after.png" "Compare with app design spec, flag differences."

# Fully custom provider (self-hosted, third-party proxy, any OpenAI-compatible endpoint)
MYAPI_API_KEY=sk-xxx MYAPI_BASE_URL=https://host/v1 MYAPI_MODEL=my-model \
  python vision.py --provider myapi "ui.png" "Analyze layout issues"