CLI Reference

The eyeroll CLI provides direct access to video analysis without Claude Code.

Commands

`eyeroll init`

Set up eyeroll -- choose a backend and configure API key.

eyeroll init

Interactive. Saves configuration to ~/.eyeroll/.env.

`eyeroll watch`

Analyze a video or screenshot and produce structured notes.

eyeroll watch <source> [options]

Arguments:

Argument	Description
`source`	URL (YouTube, Loom, etc.) or local file path

Options:

Flag	Short	Default	Description
`--context`	`-c`	None	Additional context text
`--codebase-context`	`-cc`	None	Codebase context (inline text or path to file)
`--backend`	`-b`	`EYEROLL_BACKEND` or `gemini`	Backend: `gemini`, `openai`, `ollama`, `openrouter`, `groq`, `grok`, `cerebras`, `openai-compat`
`--base-url`		None	Base URL for `openai-compat` backend
`--model`	`-m`	Backend default	Model override
`--max-frames`		20	Maximum key frames to analyze
`--parallel`	`-p`	3 (API) / 1 (ollama)	Concurrent workers for frame analysis
`--no-cache`		false	Skip cache, force fresh analysis
`--output`	`-o`	stdout	Write report to file
`--verbose`	`-v`	false	Show progress details

Examples:

# Basic usage
eyeroll watch https://loom.com/share/abc123

# With context
eyeroll watch ./bug.mp4 --context "checkout broken after PR #432"

# With codebase context file
eyeroll watch ./bug.mp4 -cc .eyeroll/context.md

# Parallel frame analysis
eyeroll watch ./bug.mp4 -p 4 --verbose

# Use Ollama with a specific model
eyeroll watch ./demo.mp4 -b ollama -m qwen3-vl:2b

# Write report to file
eyeroll watch ./bug.mp4 -o report.md

# Force fresh analysis (skip cache)
eyeroll watch ./bug.mp4 --no-cache

Model inference

If you pass --model without --backend, eyeroll infers the backend from the model name. Models starting with gemini use the Gemini backend. Models starting with gpt, o1, or o3 use OpenAI. All others default to Ollama.

`eyeroll history`

List past analyses from the cache.

eyeroll history [options]

Options:

Flag	Short	Default	Description
`--limit`	`-n`	All	Show only the last N entries
`--json`		false	Output as JSON

Examples:

eyeroll history
eyeroll history --limit 5
eyeroll history --json

`eyeroll history clear`

Clear all cached analyses.

eyeroll history clear [options]

Options:

Flag	Short	Description
`--yes`	`-y`	Skip confirmation prompt

Examples:

eyeroll history clear
eyeroll history clear --yes

Environment variables

Variable	Description	Default
`EYEROLL_BACKEND`	Default backend	`gemini`
`GEMINI_API_KEY`	Gemini API key
`OPENAI_API_KEY`	OpenAI API key (also fallback for compatible providers)
`OPENROUTER_API_KEY`	OpenRouter API key
`GROQ_API_KEY`	Groq API key
`GROK_API_KEY`	xAI Grok API key
`CEREBRAS_API_KEY`	Cerebras API key
`OLLAMA_HOST`	Ollama server address	`http://localhost:11434`
`GOOGLE_APPLICATION_CREDENTIALS`	Path to Google service account JSON
`GOOGLE_CLOUD_PROJECT`	Google Cloud project ID (for service account)
`GOOGLE_CLOUD_LOCATION`	Google Cloud region (for service account)	`us-central1`

eyeroll loads environment variables from ~/.eyeroll/.env and the project's .env file.