​

Open Responses CLI

A CLI tool for setting up a self-hosted alternative to OpenAI’s Responses API. This API lets you create and manage open-ended AI responses for your applications, similar to OpenAI’s Responses API, but fully under your control.

​

Features

• Easy setup with Docker Compose
• Compatible API endpoints with OpenAI’s Responses API
• Management UI for creating, viewing, and managing responses
• Local data storage with PostgreSQL
• Customizable authentication and timeout settings

​

Installation

You can install this CLI using Go, npm, or Python:

​

Using Go

go install github.com/julep-ai/open-responses@latest
open-responses

​

Using npm

npx open-responses

Or install it globally:

npm install -g open-responses
open-responses

Or with bunx:

bunx open-responses

​

Using Python

pipx install open-responses
pipx open-responses

Or install with pip globally:

pip install open-responses
open-responses

Or with uv:

uvx open-responses

​

Usage

​

First-time Setup

Before using any commands, you must run the setup command:

open-responses setup

This will:

• Ask for configuration settings with default values:
  • Host (default: 127.0.0.1)
  • Port (default: 8080)
  • Docker tag (default: latest_responses)
  • Base Docker Compose URI (default: https://u.julep.ai/responses-compose.yaml)
  • Environment file location (default: .env in Git root or current directory)
  • API version (default: 0.0.1)
• Ask for API configuration values (port, authentication key, timeout)
• Create a .env file with your settings
• Download or generate a docker-compose.yml file with the necessary services:
  • API server
  • Database
  • Management UI
• Create a configuration file (open-responses.json) to track your settings

The CLI will automatically check for this configuration before running any other commands. If the configuration file doesn’t exist, it will prompt you to run the setup command first.

​

Configuration File

The CLI stores its configuration in open-responses.json, which can be located in:

• The current directory
• The parent directory
• The Git repository root directory

The configuration file tracks:

• All user-defined settings
• Environment variable values
• Creation and update timestamps (both camelCase and snake_case formats are supported)
• File locations and version information

When you run setup again with an existing configuration, it will let you update your settings while preserving your previous values as defaults. If timestamps are missing from an existing configuration, they’ll be added automatically when the configuration is updated.

​

API Configuration

The API service includes the following configuration options with sensible defaults:

​

Basic Settings

• HOST: Host address for the API (default: 127.0.0.1)
• PORT: Port for the UI service (default: 8080)
• RESPONSES_API_PORT: Port for the API service (default: 8080)
• DOCKER_TAG: Docker image tag (default: latest_responses)
• API_VERSION: API version (default: 0.0.1)

​

Performance & Limits

• NODE_ENV: Node.js environment (default: production)
• LOG_LEVEL: Logging level (default: info)
• REQUEST_TIMEOUT: API request timeout in ms (default: 120000 - 2 minutes)
• MAX_PAYLOAD_SIZE: Maximum request payload size (default: 10mb)
• RATE_LIMIT_WINDOW: Rate limit window in ms (default: 60000 - 1 minute)
• RATE_LIMIT_MAX: Maximum requests per rate limit window (default: 100)

​

Resource Allocation

The Docker Compose configuration also includes resource limits to ensure stable operation:

• API Service: 1 CPU, 2GB memory (min: 0.25 CPU, 512MB)
• Database: 1 CPU, 1GB memory (min: 0.1 CPU, 256MB)
• Redis: 0.5 CPU, 768MB memory (min: 0.1 CPU, 128MB)
• UI: 0.5 CPU, 512MB memory (min: 0.1 CPU, 128MB)

These settings provide a good balance for most deployments, but you can adjust them in the docker-compose.yml file if needed.

​

User-Friendly Commands

The CLI provides easy-to-use commands for common operations:

​

Starting the service

open-responses start

This user-friendly command:

• Pulls Docker images for your specific architecture
• Starts all services in foreground mode with log streaming
• Automatically stops all services when you press Ctrl+C
• Shows the status of services after startup
• Displays access URLs for the API and admin UI

To run in detached mode (background):

open-responses start --background

​

Stopping the service

open-responses stop

This command stops all services and performs cleanup (alias for open-responses compose down).

​

Checking service status

open-responses status

Shows detailed information about all services, including:

• Running state and health status
• Uptime information
• Resource usage summary
• Access URLs

​

Viewing logs

open-responses logs [SERVICE]

Shows logs from services with sensible defaults:

• Follows logs in real-time
• Shows colorized output
• Displays last 100 lines by default
• Can target specific services

Examples:

# View logs from all services:
open-responses logs

# View logs from a specific service:
open-responses logs api

​

Initializing a new project

open-responses init

Creates a new project structure with guided setup:

• Creates directory structure (data, config, logs)
• Generates helpful documentation files
• Runs interactive configuration
• Sets up Docker Compose with best practices

​

Managing API keys

open-responses key <action>

Manages API keys for the Responses API service:

# List all API keys (masked):
open-responses key list

# Generate a new API key:
open-responses key generate [type]

# Update an API key:
open-responses key set <type> [value]

​

Updating components

open-responses update

Updates all components to the latest version:

• Updates Docker Compose configuration
• Pulls latest Docker images
• Backs up your configuration

​

Advanced Docker Compose Commands

For more advanced operations, use the compose command group:

open-responses compose <command> [args...]

Available commands include:

# Start services with additional options:
open-responses compose up [flags]

# Stop and clean up services:
open-responses compose down [flags]

# View logs with custom options:
open-responses compose logs [flags] [SERVICE...]

# List containers:
open-responses compose ps [flags]

# Build services:
open-responses compose build [flags] [SERVICE...]

# Restart services:
open-responses compose restart [flags] [SERVICE...]

# Pull service images:
open-responses compose pull [flags] [SERVICE...]

# Execute commands in containers:
open-responses compose exec [flags] SERVICE COMMAND [ARGS...]

# Run one-off commands:
open-responses compose run [flags] SERVICE COMMAND [ARGS...]

# Validate Docker Compose configuration:
open-responses compose config [flags]

# View processes in containers:
open-responses compose top [SERVICE...]

# Monitor resource usage:
open-responses compose stats [SERVICE...]

Each compose command is a direct proxy to the equivalent Docker Compose command and accepts all the same flags and arguments. This provides full access to Docker Compose functionality when needed.

For detailed examples of each command, use the --help flag:

open-responses compose up --help
open-responses compose logs --help

Or for general help:

open-responses --help

​

API Endpoints

Once your service is running, the following endpoints will be available:

• POST /v1/responses - Create a new response
• GET /v1/responses/{id} - Retrieve a response
• GET /v1/responses - List all responses
• DELETE /v1/responses/{id} - Delete a response

You can access the management UI at http://localhost:8080 (or your configured port).

​

Requirements

• Docker must be installed on your system
• Docker Compose must be installed (either as a standalone binary or integrated plugin)
  • Docker Compose V2 (≥ 2.21.0) is recommended for best compatibility
  • Docker Compose V1 is supported but with limited functionality
• No other runtime dependencies required (no Node.js or Python needed for running the service)

The CLI will check Docker and Docker Compose requirements and provide helpful instructions if they’re not met.

​

How it works

This CLI is built with Go and compiled to native binaries for Windows, macOS, and Linux. When installed via npm or pip, the appropriate binary for your platform is used automatically.

The service itself runs in Docker containers, providing a compatible alternative to OpenAI’s Responses API.

​

Development

​

Project Structure

• main.go: Core CLI functionality built with Go
• open_responses/__init__.py: Python wrapper for binary distribution
• scripts/postinstall.js: Node.js script for platform detection and setup
• bin/: Directory for compiled binaries

​

Building from Source

Build for your current platform:

npm run build

Build for all platforms:

npm run build:all

This will generate binaries in the bin/ directory:

• bin/open-responses-linux
• bin/open-responses-macos
• bin/open-responses-win.exe

​

Installing for Development

For Python:

pip install -e .

For npm:

npm link

​

Development Guidelines

This project follows strict formatting and linting guidelines to maintain code quality. We use:

• Go: Standard Go formatting with go fmt
• Python: Ruff for linting and formatting
• JavaScript: ESLint and Prettier for linting and formatting

​

Setting Up Development Environment

1. Install development dependencies:

# Install JavaScript dependencies
npm install

# Install Python dependencies
uv pip install ruff

# Install git hooks for automatic linting/formatting
npm run install:hooks

2. The git hooks will automatically run formatting and linting checks before each commit.

​

Code Formatting and Linting

You can manually run code formatting and linting using these commands:

# Format all code
npm run format:all

# Lint all code
npm run lint:all

# Format/lint individual languages
npm run format       # JavaScript/JSON/Markdown files
npm run py:format    # Python files
npm run go:format    # Go files
npm run lint         # JavaScript files
npm run py:lint      # Python files

​

License

Apache-2.0