Configuration

Configuration Files

Gee-Code uses layered configuration. Rules from all layers combine, with more specific layers taking priority:

File	Scope	Purpose
~/.gee-code/gee.md	Global	Your personal preferences, coding style, workflow rules
.gee/gee.md	Project	Project-specific guidance and rules
.gee/continuity.md	Session	Session state ledger (auto-managed)
.gee-code/config.json	Global	Technical settings (model, provider, RLM, etc.)

User Preferences (~/.gee-code/gee.md)

This file is loaded for every session in every project. Use it for personal rules:

# User Preferences

## Coding Style
Always use TypeScript. Prefer functional style.

## Communication
Be concise. Explain reasoning only when asked.

## Workflow
Never auto-commit. Always run tests before suggesting done.

## Git Commit Conventions
All commits must use conventional commit format: type(scope): description

Gee-Code reads this as natural language instructions. Write it however makes sense to you.

Project Guidance (.gee/gee.md)

Per-project rules and context. Generated by /init or written manually:

# Project Guide

## Project Overview
A React + Express full-stack app with PostgreSQL.

## Key Rules
- Follow existing code patterns
- Ask before major architectural changes
- Never modify the migration system directly

Every project can have its own rules that override or extend your global preferences.

Models

Switch AI models during a session:

/model sonnet       # Claude Sonnet (fast, capable)
/model opus         # Claude Opus (most capable)
/model haiku        # Claude Haiku (fastest, cheapest)
/model gpt4o        # GPT-4o
/model gemini       # Gemini Pro

Add custom model aliases:

/model add my-model provider/model-id

Set a default model in config:

/model default sonnet

Providers

Gee-Code supports multiple AI providers. See Models & Providers for the full list and setup instructions.

Quick overview:

• Anthropic — Claude models (default)
• OpenAI — GPT models
• Google — Gemini models
• xAI — Grok
• Groq, Cerebras — Fast inference
• Ollama — Local models
• OpenRouter — Multi-model gateway
• Custom — Any OpenAI-compatible API

Manage providers with /provider.

Credentials

API keys are managed through a multi-source resolution system:

/credentials              # Show all credential status
/credentials set openai   # Set an API key

Credentials are resolved from (in order):

1. Credentials store (scoped to gee/team/endeavor)
2. Local file (~/.gee-code/credentials.json)
3. Environment variables
4. Gee backend (your stored credentials)

See Models & Providers for details.

Environment Variables

Variable	Purpose
GEE_CODE_MODEL	Default model
GEE_CODE_PROVIDER	Default provider
GEE_CODE_WEB_PORT	Web UI port (default: 7779)
GEE_CODE_DEBUG	Enable debug output
ANTHROPIC_API_KEY	Anthropic/Claude API key
OPENAI_API_KEY	OpenAI API key
GOOGLE_API_KEY	Google/Gemini API key
FAL_KEY	FAL image/video generation key

Next Steps

• How It Works — understand the execution model
• Models & Providers — full provider setup
• Guardrails — safety constraints for autonomous operation