A specialized vision skill that provides high-accuracy image recognition and text extraction to power AI agent reasoning.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install siliconflow-vision
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install siliconflow-vision using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is SiliconFlow Vision?
The SiliconFlow Vision skill is a professional-grade image recognition and analysis tool designed for AI agents. It utilizes advanced Vision Large Language Models (VLM) to perform objective visual identification, allowing the primary agent to focus on cognitive reasoning and synthesis. By offloading the 'seeing' process to specialized models like DeepSeek-VL2 or Qwen2.5-VL, it ensures high-fidelity data extraction for any visual task.
This skill is a core component of the Openclaw Skills ecosystem, enabling developers to integrate multi-provider support including SiliconFlow, OpenAI, and Anthropic. It adheres to a strict protocol where the skill handles objective recognition while the main agent performs subjective analysis, ensuring more accurate and hallucination-free responses.
SiliconFlow Vision Use Cases
- Extracting precise code snippets or error messages from terminal screenshots.
- Analyzing complex memes and social media trends for context-aware interpretation.
- Performing high-accuracy OCR on physical documents, invoices, and business cards.
- Interpreting technical data visualizations, flowcharts, and architectural diagrams.
- Identifying products, landmarks, or people within photographs for research purposes.
How SiliconFlow Vision Works
- The user uploads an image or provides a URL to the primary AI agent.
- The main agent identifies the visual input and executes the siliconflow-vision script within the Openclaw Skills framework.
- The skill processes the image through the selected provider and model (e.g., SiliconFlow by default).
- A structured, objective report is generated, covering text, layout, and visual elements.
- The main agent receives this report and combines it with its internal knowledge or web searches to deliver a final answer.
SiliconFlow Vision Setup
To get started, configure your API keys in config/default.json or set them as environment variables:
# Set your API key
export SILICONFLOW_API_KEY="your_key_here"
# Run a basic image analysis
python scripts/analyze_image.py /path/to/image.jpg
# Run with smart mode for complex charts
python scripts/analyze_image.py diagram.png -m smart
SiliconFlow Vision Data Schema & Taxonomy
The skill produces a standardized output format to ensure consistency across Openclaw Skills integrations:
| Attribute | Content Type | Description |
|---|---|---|
| Image Type | Classification | Identifies if the image is a screenshot, meme, document, etc. |
| Clear Text | OCR String | Full, accurate extraction of all visible text elements. |
| Visual Elements | List | Identification of all key objects and components in the frame. |
| Layout | Structure | Description of the spatial organization and composition. |
| Style | Descriptive | Defines the atmosphere (e.g., minimalist, humorous, dark). |
SiliconFlow Vision Advanced Features
- Multi-provider flexibility supporting SiliconFlow, OpenAI (GPT-4o), and Anthropic (Claude 3.5 Sonnet).
- Smart processing mode using Qwen2.5-VL-72B for high-precision diagram and document analysis.
- Built-in image compression to optimize speed and reduce token costs during processing.
- Custom question support to focus the vision model on specific areas of interest within an image.
- Seamless integration with search-enabled Openclaw Skills for deep contextual verification.
SKILL.md
Loading
Related Openclaw Skills
Moltbook Poster
A comprehensive automation toolkit for programmatic interaction with the Moltbook social platform.
Feishu Message Sender
A powerful Python utility for sending automated messages to Feishu and Lark users using various identifiers.
AgentGen
A powerful CLI utility for converting HTML content into professional-grade PDFs and images without complex server-side rendering setup.
Video Crawler
A high-performance utility for extracting and downloading video content from Douyin and Twitter platforms.
SSH Server Management
A robust integration for automating remote server management and command execution via secure SSH connections.
SAP FICO Expert — Australia
A specialized AI skill for SAP Finance & Controlling tailored specifically for Australian tax, banking, and compliance requirements.
Featured*
