Interact with the Orion visual AI agent via CLI to process images, videos, and documents using natural language commands.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install vlmrun-cli-skill
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 vlmrun-cli-skill 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 VLM Run CLI Agent?
The VLM Run CLI skill provides a robust interface for the Orion visual AI agent, allowing developers to perform complex visual tasks directly from their terminal. By integrating this tool with Openclaw Skills, users can automate workflows involving image generation, object detection, and video summarization. It serves as a bridge between raw visual data and actionable insights, supporting a wide range of media formats including PDF documents and high-resolution videos.
This skill is designed for high performance, featuring concurrent file uploads and session management to maintain context across multiple interactions. Whether you are building automated media pipelines or performing ad-hoc visual analysis, this component of Openclaw Skills ensures the Orion AI's capabilities are just a command away.
VLM Run CLI Agent Use Cases
- Automated OCR and text extraction from complex document images or PDF invoices.
- Generating high-quality photorealistic images and short video clips from text prompts.
- Summarizing meeting recordings or lecture videos with timestamped transcriptions.
- Performing object detection and scene analysis across multiple image inputs simultaneously.
- Comparing visual differences between two or more images for quality control or monitoring.
How VLM Run CLI Agent Works
- The user provides a natural language prompt and optional input files (images, videos, or PDFs) to the CLI.
- The VLM Run CLI authenticates with the Orion API using a pre-configured API key.
- Input files are uploaded concurrently to the VLM Run infrastructure for processing.
- The Orion visual AI agent analyzes the media based on the prompt's context.
- The skill returns a text response or generates artifacts (like new images/videos) to a specified local directory.
- For multi-step tasks, session IDs allow the agent to maintain context from previous interactions within Openclaw Skills workflows.
VLM Run CLI Agent Setup
To get started with this skill in your environment, follow these steps:
# Create and activate a virtual environment
uv venv && source .venv/bin/activate
# Install the VLM Run CLI package
uv pip install "vlmrun[cli]"
You must also configure your environment variables. It is recommended to add these to your .env file:
export VLMRUN_API_KEY="your_api_key_here"
export VLMRUN_BASE_URL="https://agent.vlm.run/v1"
VLM Run CLI Agent Data Schema & Taxonomy
The skill manages data through environment variables and local file storage. Here is how the metadata and artifacts are organized:
| Attribute | Description |
|---|---|
| VLMRUN_API_KEY | Mandatory authentication token for Orion AI access. |
| Session ID | A unique identifier used to persist conversation and artifact context. |
| Artifacts | Generated media files saved to ~/.vlmrun/cache/artifacts/ by default. |
| Input Files | Supports .jpg, .png, .mp4, and .pdf formats for analysis. |
| JSON Output | Raw structured data available via the --json flag for integration. |
VLM Run CLI Agent Advanced Features
- Session Persistence: Use the
-sflag to continue previous conversations while retaining character and scene context. - Prompt Versatility: Feed prompts directly, via file paths, or through standard input (stdin) for pipeline automation.
- Concurrent Processing: Multiple input files are uploaded and processed simultaneously to reduce latency.
- Artifact Management: Custom output directories and options to skip downloads for metadata-only queries.
- Model Selection: Switch between fast, auto, and pro Orion models to balance speed and accuracy within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Workout Track
A sophisticated logging utility that parses natural language workout data and inserts it into a structured database.
PinchBench Leaderboard
A specialized skill to fetch and analyze real-time LLM benchmark data from the PinchBench leaderboard for agentic coding tasks.
SM Saver
A powerful tool to extract, summarize, and log resources from social media posts and web URLs directly into a structured markdown file.
Skill Builder
A meta-skill designed to guide developers through the end-to-end process of creating, optimizing, and auditing high-quality Openclaw Skills.
Gmail Bridge
A comprehensive Google Workspace integration for managing emails, files, spreadsheets, and calendar events via a local API.
Base Blockchain Interface
A lightweight interface for querying live onchain data from the Base blockchain without requiring an API key.