A local visual analysis skill that enables AI agents to perform OCR and image description using Ollama models without cloud dependencies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ollama-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 ollama-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 Ollama Vision Skill?
The Ollama Vision Skill is a specialized integration designed for the Openclaw Skills ecosystem that brings local computer vision capabilities to AI agents. It leverages the qwen3-vl:4b model via Ollama to process images, allowing users to keep their main conversation model (like Kimi or GPT-4) while offloading visual tasks to a local instance. This ensures maximum privacy and reduces latency for image-heavy workflows.
By utilizing this skill, developers can implement high-performance image processing including text extraction and visual scene description. As a part of the Openclaw Skills collection, it provides a bridge between large language models and local vision-language models (VLMs), facilitating a multi-model approach to complex AI tasks.
Ollama Vision Skill Use Cases
- Extracting text from screenshots or scanned documents using local OCR.
- Generating descriptive metadata for image assets in a development directory.
- Automated data extraction from tables or forms within images using custom prompts.
- Integrating private visual analysis into sensitive Openclaw Skills workflows.
How Ollama Vision Skill Works
- The AI agent detects an image input or a path to a local image file.
- The analyze_image tool is triggered within the Openclaw Skills environment.
- The skill checks the image file size; if it exceeds 2MB, it automatically compresses the image using the Pillow library to optimize processing speed.
- The image data is sent to the locally running Ollama service using the qwen3-vl:4b vision model.
- The model processes the image according to the specified mode: describe (general description), ocr (text extraction), or extract (custom prompt analysis).
- The resulting text analysis is returned to the agent, and any temporary compressed files are automatically deleted.
Ollama Vision Skill Setup
To get started with this skill in your Openclaw Skills setup, ensure you have the following prerequisites:
- Install and run Ollama on your machine.
- Pull the vision model:
ollama pull qwen3-vl:4b
- Install the required Python image processing library:
pip install Pillow
- Ensure your AI agent has access to the tool functions defined in the skill configuration.
Ollama Vision Skill Data Schema & Taxonomy
The Ollama Vision Skill processes image data and returns structured text results. Below is the schema for the tool parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
| image_path | string | Yes | The full local file system path to the image. |
| mode | string | No | Analysis mode: "describe" (default), "ocr", or "extract". |
| prompt | string | No | Custom instructions used specifically when mode is set to "extract". |
Output: Returns a plain text string containing the visual analysis or extracted text.
Ollama Vision Skill Advanced Features
- Intelligent Image Compression: Prioritizes JPEG quality reduction before scaling dimensions to preserve semantic data.
- Resource Management: Automatic cleanup of temporary files to prevent disk bloat during heavy Openclaw Skills automation tasks.
- Custom Prompt Support: Allows power users to define specific extraction logic for complex visual data structures.
- Local-First Architecture: Requires zero API keys or external cloud calls, making it ideal for air-gapped or privacy-conscious environments.
SKILL.md
Loading
Related Openclaw Skills
Tencent Cloud General Accurate OCR
A high-precision optical character recognition skill that extracts text from complex layouts, images, and PDF documents using Tencent Cloud AI.
Workspace Context Linter
A diagnostic tool designed to audit, de-bloat, and organize core workspace context files for AI agents.
Scrum Master Expert
A data-driven agile coaching skill for Openclaw Skills that automates sprint forecasting, team health scoring, and retrospective analysis.
investment-portfolio
Track and analyze stock and crypto holdings with local-first security using Openclaw Skills.
Quick Proposal Skill
An automated project proposal generator that transforms client requirements into professional, structured business documents including cost and time estimates.
CLI Builder
A comprehensive automation skill for scaffolding professional command-line applications with full lifecycle support across multiple programming languages.
Featured*
