Setup

Running `/eyeroll:init`

The init command does three things:

Installs the eyeroll CLI if missing
Configures your backend and API key
Generates .eyeroll/context.md with codebase context

In Claude Code

/eyeroll:init

The agent will:

Check if eyeroll is installed and install it if not
Ask which backend you want (Gemini, OpenAI, Ollama, OpenRouter, Groq, Grok, Cerebras, or openai-compat)
Run eyeroll init for interactive API key setup
Explore your codebase and generate .eyeroll/context.md

Standalone CLI

eyeroll init

Interactive flow:

Which backend do you want to use?

  1. gemini  -- Google Gemini Flash API (fast, cheap, best quality)
  2. openai  -- OpenAI GPT-4o (great vision, Whisper audio)
  3. ollama  -- Local models via Ollama (private, no API key)

Choose [1/2/3]: 1

How do you want to authenticate with Google?

  1. API key         -- get one at https://aistudio.google.com/apikey
  2. credentials.json -- service account for Vertex AI

Choose [1/2]: 1

Enter your Gemini API key
  (input is hidden): ****

Validating API key...
Setup complete. Run `eyeroll watch <video>` to get started.

What gets created

File	Purpose
`~/.eyeroll/.env`	Backend preference and API key
`.eyeroll/context.md`	Codebase context (generated by `/eyeroll:init` only)

The .env file stores:

EYEROLL_BACKEND=gemini
GEMINI_API_KEY=your-key-here

Codebase context

The .eyeroll/context.md file is a short summary of your project (under 80 lines) that eyeroll uses to ground its analysis in real file paths. Without it, all file paths in reports are labeled as hypotheses.

Structure:

# Project: my-app
## Stack
Python, FastAPI, PostgreSQL
## Structure
src/ -- application code
tests/ -- test suite
## Key Files
- src/api/routes.py: API route definitions
- src/models/user.py: User model
## Recent Activity
- abc1234 Fix login redirect
- def5678 Add rate limiting

Tip

Re-run /eyeroll:init after major refactors to update the codebase context.