Skip to main content
The vlmrun chat command enables visual AI chat with Orion directly from your terminal. Process images, videos, and documents with natural language prompts.

​
Basic Usage

# Describe an image
vlmrun chat "Describe this image" -i photo.jpg

# Analyze a document
vlmrun chat "Extract the key information" -i document.pdf

# Process a video
vlmrun chat "Summarize this video" -i video.mp4

# Compare multiple files
vlmrun chat "Compare these images" -i image1.jpg -i image2.jpg

​
Prompt Sources

Prompts can be provided in three ways (in precedence order): 
# 1. Direct argument
vlmrun chat "Your prompt here" -i file.jpg

# 2. Using -p option (text, file path, or stdin)
vlmrun chat -p "Your prompt" -i file.jpg
vlmrun chat -p prompt.txt -i file.jpg

# 3. Piped stdin
echo "Describe this" | vlmrun chat - -i file.jpg
cat prompt.txt | vlmrun chat - -i file.jpg

​
Using Skills

Pass a local skill directory with -k to apply skill instructions inline: 
# Use a local skill
vlmrun chat "Extract data from this invoice" -i invoice.pdf -k ./my-skill

# Multiple skills
vlmrun chat "Analyze this" -i photo.jpg -k ./skill-a -k ./skill-b
The -k flag sends the skill inline with the request (no server-side upload). To create a persistent server-side skill, use vlmrun skills upload.

​
Stateful Sessions

Use --session-id to persist chat history across multiple calls: 
# Start a session
vlmrun chat "What's in this image?" -i photo.jpg --session-id my-session

# Continue the conversation
vlmrun chat "Now describe the colors" --session-id my-session

​
Models

ModelDescription
vlmrun-orion-1:fastSpeed-optimized
vlmrun-orion-1:autoBalanced (default)
vlmrun-orion-1:proMost capable
vlmrun chat "Describe this" -i photo.jpg -m vlmrun-orion-1:pro

​
Output Formats

# Rich formatted output (default)
vlmrun chat "Describe this" -i photo.jpg

# JSON output for scripting
vlmrun chat "Describe this" -i photo.jpg --json

# Pipe JSON to jq
vlmrun chat "Describe this" -i photo.jpg --json | jq '.content'

# Disable streaming (wait for complete response)
vlmrun chat "Describe this" -i photo.jpg --no-stream

​
Artifact Handling

When Orion generates artifacts (images, videos, etc.), they are automatically downloaded: 
# Default: saved to ~/.vlm/cache/artifacts/<session_id>/
vlmrun chat "Generate a variation of this image" -i photo.jpg

# Custom output directory
vlmrun chat "Generate a variation" -i photo.jpg -o ./output/

# Skip artifact download
vlmrun chat "Generate a variation" -i photo.jpg --no-download

​
Supported File Types

CategoryExtensions
Images.jpg, .jpeg, .png, .gif, .webp, .bmp, .tiff
Videos.mp4, .mov, .avi, .mkv, .webm
Documents.pdf, .doc, .docx
Audio.mp3, .wav, .m4a, .flac, .ogg

​
Options Reference

OptionShortDescription
--prompt-pPrompt: text string, file path, or stdin
--input-iInput file (repeatable)
--skill-kPath to a skill directory (repeatable)
--output-oArtifact output directory
--model-mModel variant (default: vlmrun-orion-1:auto)
--json-jOutput JSON instead of formatted text
--no-stream-nsDisable streaming
--no-download-ndSkip artifact download
--session-id-sSession UUID for stateful conversations
--base-urlAPI base URL override